Issuu on Google+

Special Issue on Cooperative Banks Volume 3, Issue 1, 2014 Edited by Silvio Goglio and Yiorgos Alexopoulos Available online at www.jeodonline.com

ISSN: 2281-8642


Special Issue on Cooperative Banks Volume 3, Issue 1, 2014 Edited by Silvio Goglio and Yiorgos Alexopoulos Available online at www.jeodonline.com


Jeod is governed by its Chief Editors, a Scientific Committee, and an Editorial Board. The composition of the Scientific Committee and Editorial Board reflects the international and multidisciplinary nature of the journal. This distinguished list of scholars helps to ensure the academic rigor of the journal. Co-Chief Editors Carlo Borzaga, University of Trento, Department of Economics* Giovanni Ferri, Holy Mary of the Assumption Free University, Department of Economics & Political Sciences & Modern Languages Fabio Sabatini, Sapienza University of Rome, Department of Economics and Law* Managing Editor Barbara Franchini, Euricse Editorial Staff Federica Silvestri, Euricse Scientific Committee - Masahiko Aoki, Stanford University - Partha Dasgupta, University of Cambridge, Department of Economics - Giovanni Dosi, Sant’Anna School of Advanced Studies, Institute of Economics and Management - Bruno Frey, University of Zurich, Department of Economics - Enrico Giovannini, University of Rome “Tor Vergata”, Department of Economic-Financial Studies and Quantitative Methods - Henry Hansmann, Yale University, Yale Law School - Geoffrey M. Hodgson, University of Hertfordshire Business School - Margit Osterloh, University of Zurich, Department of Business Administration and Warwick Business School, University of Warwick, UK - Stefano Zamagni, University of Bologna, Department of Economics - Vera Zamagni, University of Bologna, Department of Economics

Editorial Board - Yiorgos Alexopoulos, Agricultural University of Athens, Department of Agricultural Economics and Rural Development* - Michele Andreaus, University of Trento, Department of Business Studies* - Avner Ben-Ner, University of Minnesota, Center for Human Resources and Labor Studies - Johnston Birchall, University of Stirling, School of Applied Social Sciences - Simone Borghesi, University of Siena, Department of Law, Economics and Government - Maurizio Carpita, University of Brescia, Department of Economics and Management* - Marco Casari, University of Bologna, Department of Economics - Jacques Defourny, HEC-University of Liège, Department of Economics - Giacomo Degli Antoni, University of Parma, Department of Law - Emanuele Felice, Universitat Autònoma de Barcelona, Department of Economics and Economic History - Antonio Fici, University of Molise, Faculty of Economics* - John Field, University of Stirling, School of Education - Silvio Goglio, University of Trento, Department of Economics* - Benedetto Gui, University of Padua, Department of Economics - Lou Hammond Ketilson, University of Saskatchewan, Centre for the Study of Co-operatives, Edwards School of Business, Department of Management and Marketing - Derek Jones, Hamilton College, Department of Economics - Raffaele Miniaci, University of Brescia, Department of Economics - Maria Minniti, Syracuse University, Whitman School of Management - Jay Mitra, University of Essex, Centre for Entrepreneurship Research, Essex Business School - Pier Angelo Mori, University of Florence, Department of Economics* - Mario Morroni, University of Pisa, Department of Economics - Virginie Pérotin, Leeds University Business School - Victor Pestoff, Ersta Sköndal University College, Institute for Civil Society Studies - Maximo Rossi, Universidad de la Republica, Department of Economics - Lorenzo Sacconi, University of Trento, Department of Economics - Enrico Santarelli, University of Bologna, Department of Economics - Francesco Sarracino, National Institute of Statistics and Economic Studies of Luxembourg; LCSR-Higher School of Economics; and Leibniz Institute for the Social Sciences - Carlo Scarpa, University of Brescia, Department of Economics - Roger Spear, Co-operatives Research Unit, Open University - Alberto Zevi, Roma Tre University, Department of Economics

*also affiliated with Euricse

JEOD is a registered periodical in Trento, Italy, under the general direction of Fulvio Gardumi. Registration number: 34/2011/12-11-2011 ISSN: 2281-8642 The current issue and full text archive of this journal is available at: www.jeodonline.com JEOD articles are also distributed by SSRN and available at: www.ssrn.com/link/JEOD.html

AT T R I B U T I O N 3 . 0

You are free to share and to remix, you must attribute the work


Contents

Volume 3, Issue 1, 2014 Special Issue on Cooperative Banks

Editorial: Cooperative Banks at a Turning Point? Silvio Goglio and Yiorgos Alexopoulos

1

Part 1. Cooperative banks in the recent crisis Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Hans Groeneveld

11

Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Giovanni Ferri, Panu Kalmi and Eeva Kerola

35

Did the Extent of Their Hybridization Better Enable Cooperative Banking Groups to Face the Financial Crisis? Yasmina Lemzeri

57

Part 2. Cooperative banking: assumptions and evidence Soft Information and Default Prediction in Cooperative and Social Banks Simon CornĂŠe

89

Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence Matteo Alessi, Stefano Di Colli and Juan Sergio Lopez

111

Investigating Management Turnover in Italian Cooperative Banks Mitja Stefancic

131

Part 3. Modelling cooperative banks’ interest policy A Model for the Interest Margin of a Risk-Neutral Bank. The Role of the Bank Orientation Marco Pedrotti

167

Optimal Interest Rates in Co-operative Banks with Non-member Clients Ivana Catturani and Ragupathy Venkatachalam

181


AT T R I B U T I O N 3 . 0

You are free to share and to remix, you must attribute the work

?2014 3, Issue (2014) 1-91-8 56-85 Publication date: 17 June| Volume 2014 | Vol.3, Volume Issue 3,1 Issue 1 (2014) 1 (2014)

AUTHOR SILVIO GOGLIO University of Trento and Euricse silvio.goglio@unitn.it YIORGOS ALEXOPOULOS Agricultural University of Athens and Euricse galexop@aua.gr

Editorial: Cooperative Banks at a Turning Point?

KEY-WORDS STAKEHOLDER BANKS; CO-OPERATIVE GOVERNANCE; REGULATIONS; BIODIVERSITY IN BANKING

FINANCIAL

CRISIS;

JEL Classification: G2; G3; L2; P13 | DOI: http://dx.doi.org/10.5947/jeod.2014.001

1 JEOD - Vol.3, Issue 1 (2014)

BANKING


Editorial: Cooperative Banks at a Turning Point? Goglio, S.; Alexopoulos, Y.

Are cooperative banks at a turning point? Many changes in the economic and social framework over recent years make this question quite timely. The great financial crisis, which began in 2007, is clearly at the centre of the picture. The crisis stemmed from numerous financial innovations and reckless behaviours that attempted to create more elasticity in the credit system, to the detriment of transparency, which instead generated higher uncertainty and damaged trust toward and within the banking system. As a consequence the crisis has deeply affected the European banking system and its relationship with the productive system, increasing the cost of capital and depressing the economy (IMF, 2007; Banca d’Italia, 2008). However, we cannot ignore the fact that this financial crisis followed (and to a great extent was a consequence of ) twenty years of radical changes in the financial system’s size, operational procedures and organizational structure. Many of these changes resulted in clear symptoms of disease in the financial services industry. In particular, there was poor oversight during these changes, including incapability, and in some cases even unwillingness, to institute new forms of governance better fitted to the new context. Financial innovations, deregulation, the states’ decreasing role, as well as growing international openness all led to stronger competition; they also prompted a far-reaching consolidation process of banking institutions and an overall rationalization of the banks’ productive structures. On the one hand, the mainstream financial industry’s innovations concentrated on making high-margin profits while staying compliant and formally within the rules. The banks’ benchmark for success was the quarterly earnings report to shareholders. As Stiglitz (2009) outlined, creating earnings based on market values is easier than creating value in the real economy. As a consequence, the financial system had been trapped in an illusion of making money from money and in a short term profitability approach. On the other hand, a long trend of mergers and acquisitions increased banks’ average size and spread the practice of standardized loans procedures based on easily observable, verifiable and transmittable data existing within the more complex and structured forms of group organization. Moreover a myopic focus on short-term profit severely affected their customer-relation policies and methods (ECB, 2000, 2001; Belaisch et al., 2001). The re-engineering processes and cost-rationalization strategies widened the physical distance between the new merged banks and peripheral areas and activities, resulting in making small-scale customers more dependent on personal relations (Leyshon and Thrift, 1993, 1995; Berger et al., 2001). Because of physical or “informational” distance, mainstream banks lost efficiency in generating borrower-specific information, which, in addition, due to its “soft” characteristics, is difficult to transmit through large institutions’ communication channels1. Many potential borrowers were dismissed because they lacked credit records or sufficient collateral and needed small loans, which the banks did not view as profitable. Moreover, the crisis revealed the detrimental systemic effects of incentives based on greed, as one may describe the applied executive’s compensation policies. As Stiglitz maintains, banks commonly apply executive compensation schemes that encouraged and infused bankers’ irresponsible behaviours. Paid by stock options, executives have an incentive to increase their market value, which is easier by increasing reported income than by increasing true profits, which are bound to real business. The crisis has also proved the soundness of many researchers’ hypotheses regarding cooperative banks’ behaviour in general. The hypotheses included the cooperative banks’: i) tendency to adopt less risky strategies and to have much lower volatility of returns, with positive consequences for the financial stability of the territories where they operate (Groeneveld and de Vries, 2009; Fonteyne, 2007; Hesse and Cihak, 2007; Groeneveld and de Vries, 2009; Fonteyne, 2007); ii) solidity of network organization, higher order organs and mutual support mechanisms, that has lessened the negative impacts of small size (Oliver Wyman, 2008); iii) propensity to defend consumer interests and maximize consumer surplus, offering 1

For elaboration, as well as for cooperative bank development’s challenges brought along with opportunities, see discussions in Alexopoulos and Goglio, 2011. 2 JEOD - Vol.3, Issue 1 (2014)


Editorial: Cooperative Banks at a Turning Point? Goglio, S.; Alexopoulos, Y.

simple and transparent products (that are fairly priced and well-designed to meet local needs) in a manner ensuring that risks are well-understood and communicated (Alexopoulos, 2004); and iv) lower inclination, during a credit crunch, to ration credit to customers and to raise loan interest rates, thanks to better capitalization and more prudent lending (Fonteyne, 2007; Ferri, 2008). These potential effects all stem from cooperative banks’ governance, business model and specialization, which heavily rely on relationshipbased retail banking (Berger and Udell, 1990; Petersen and Rajan, 1994; Harhoff and Körtring, 1998; Di Salvo et al., 2004). Furthermore, while values such as prudence, responsiveness, empathy and transparency are strongly linked to risk management, coop banks tend to adopt a rather conservative approach toward risk. This is because their banking business model flows from their underlying principles and commitment to investing in the real economy and to creating benefits for members, customers and communities. Both structural changes and the financial crisis have fostered cooperative banks, at least at the local level (EACB, 2009; Goglio and Alexopoulos, 2013; Oliver Wyman, 2014). Combining community bonds and shared responsibility, the capacity to reduce information asymmetry and mobilize local savings (Ahrendsen et al., 1999; Berger et al., 2001) have allowed them to re-establish and even strengthen trust towards the banking system, in a period when this was strongly endangered by the crisis. As a first assessment we may therefore conclude that, dissimilar to some arguments in the mainstream literature, the very late developments show that cooperative banks are not disappearing in industrialized countries. Contrarily, they seem to be proving their flexibility by adapting to current conditions through innovation and redefining local conditions and, in fact, are among the fastest growing groups of financial institutions in some advanced nations. However, the crisis has modified the context as well as the relationships network in which cooperative banks operate. On the one hand, retail banking (the core of cooperative banks’ business) has shown an increase in competition, both from commercial banks trying to recover through relational finance and also from fellow cooperative banks. On the other hand, local identity (the traditional source of the strength of credit cooperatives) is losing importance in small firms’ and families’ financial decisions, vis-à-vis simple interest reasons. In this context, cooperative banks’ pyramidal organizational structure, historic reaction to the small scale weaknesses, and their recent process of “consolidating” or “defensive” mergers, that aim to cut costs and possibly also diversify risks, notwithstanding the synergies that they definitely create, are altering some fundamental characteristics of the grass-root initiative (Bonaccorsi di Patti and Gobbi, 2001; Di Salvo et al., 2002; Alexopoulos and Goglio, 2011; Alexopoulos and Goglio, 2013a). If we reflect on the above points, it stands to reason to conclude that despite their recent success, cooperative banks are at a turning point. They are being forced to deeply rethink their overall and local strategies, their daily activities with members and customers, as well as their loyalty to constitutive social principles. Re-evaluation must first ask: what are the cooperative banks’ main weaknesses and challenges? The first set of problems is inherent to the scale of business. Cooperative banks’ low diversification or territorial concentration in their loan portfolios (Barham et al., 1996) may exacerbate the impact of idiosyncratic shocks on the local financial system. This turns banking localism from an element of stabilization to one that amplifies crises. It also neutralizes the possible advantages that the bank may derive from local system externalities (closer relations with firms, informational advantages, accurate selection of debtors, peer monitoring and extra-economic sanctions on insolvent debtors). Personal knowledge and peer monitoring to be effective should be integrated into a clear vision of the territory’s strengths and opportunities: in other words, a development vision and consulting capacity are needed in order to improve the bank’s capacity to act as an agent of social change and development (Goglio, 2009; Alexopoulos and Goglio, 2013a). Both of these issues should be addressed and solved within the apex institutions and through modernizing the pyramidal structure. In other words, the pyramidal organization must be able to 3 JEOD - Vol.3, Issue 1 (2014)


Editorial: Cooperative Banks at a Turning Point? Goglio, S.; Alexopoulos, Y.

unify the pros of both dimensions, capitalize on inside information at the central level, and monitor from the decentralized network, all coherently with a shared social philosophy. The solutions to these problems clearly require modernizing both the government and the governance of the cooperative banks (Alexopoulos and Goglio, 2011). The old pattern of running these banks with relatively simple administrative practices, which straightforward management schemes could easily handle, is no longer adequate. It is inevitable that growth requires sophisticated professional management in order to deal with the new, more complex financial situations (Huppi and Feder, 1989; Poyo et al., 1993). On the other hand, the qualitative and quantitative reinforcement of management may lead to the separation of ownership and control, thus intensifying agency problems (Emmons and Schmid, 1999a, 1999b; Leggett and Strand, 2002). The risk is either a misappropriation of cooperative funds on behalf of the management for its own use, or a substantial misalignment between corporate philosophy and members’ needs and will. This leads us to consider the cooperative bank’s governance mechanisms and representation of membership to the Board of Directors. As cooperative banks grow, why members are motivated to be on the board becomes a more relevant question: it is more likely that directors follow their own interests, turning collective action away from its initial goals and giving rise to less efficient solutions (Goglio, 1999; Alexopoulos et al., 2013). But growth and increasing membership may also add difficulties in adequate internal control as it promotes free riding by members (Ferguson and McKillop, 1997; Ouattara et al., 1999) which may feel disempowered as the institution adds new members. This in turn makes it difficult to encourage existing members to exercise their ownership rights and responsibilities to oversee management, with declining participation in board elections (Caswell, 1987; Hariyoga, 2004; Osterberg and Nilsson, 2009). Other reasons of non-participation include insufficient knowledge of the subjects discussed and also the claim that the Board of Directors formulates the cooperative’s policies without taking into account member needs. In any case, member absenteeism from general meetings deprives them the possibility of realizing the reasoning behind the cooperative bank’s operations. As a result, members judge the bank’s performance mainly through their transactions with the cooperative bank, being ignorant of the true reasons that shape how transactions run and the consequences of the policy followed. When membership and assets extend beyond small numbers, common bonds lose their tight influence in maintaining a moral obligation to the cooperative (Alexopoulos, 2004; Goglio and Alexopoulos, 2013). The market can often ravage these bonds, as the latest financial turbulences have demonstrated. Therefore it is necessary to work either to restore or strengthen the bonds among cooperative values, members’ participation and business. This policy could also effectively solve the problems originating from reduced capacity of local enforcement and peer monitoring, which are most likely to arise when higher membership means looser links with its financial institution (Alexopoulos et al., 2013). Furthermore, it cannot be ignored that the turbulent banking scene created by the crisis has increased regulatory pressure toward strengthening both equity and profits (Alexopoulos and Goglio, 2011). As a consequence, there is the risk of augmented conflicts between the interests of member-depositors and member-borrowers (Smith et al., 1981; Smith, 1984; Patin and McNeil, 1991a, 1991b). Accommodating each group’s interests heavily influences how the financial intermediary operates, which in turn leads to policy issues that an experienced management may deal with more adequately. Insufficient development of appropriate participatory and monitoring procedures at the local level could lead to a large cooperative’s inability to detect and determine the members-customers’ (and the community’s) socio-economic needs as well as to provide solutions. This time however there is an additional risk due to the imposed stricter regulatory framework and the institutional steps taken toward introducing an EU banking union. As the costs of creating a union are 4 JEOD - Vol.3, Issue 1 (2014)


Editorial: Cooperative Banks at a Turning Point? Goglio, S.; Alexopoulos, Y.

expected to peak at a time when the banks will be required to build up their capital resources and finance economic growth, the underlying risk is that regulation will be seen again as a constraint to profit margins. Slowly but surely the financial system’s most opportunistic parts will return to lobbying against these legal constraints and argue in favour of further liberalization. Eventually they will find their way out and follow the letter but not the spirit of the law, once again infusing excessive risk into the system. So the real question yet again is: Cui bono from this one-size-fits-all regulatory framework? In conclusion, do cooperatives possess the appropriate characteristics to play a new and efficient role in the local banking market? Can they translate their constitutive features into a modern and competitive banking setting? And if they can, how would they successfully fulfill this role? Two conditions that are certainly necessary require the cooperative banks to have: i) a well-chosen, prepared and competent Board of Directors and management, possessing solid cooperative training and knowledge, and ii) a committed membership, bound by close everyday links, even when is increasing in both numbers and demand. But these conditions are not sufficient for change. In order to add potential to their competitive position within a territory, cooperative banks should also strive to become true “local banks.” This means not only in proximity or as a relational bank but as a financial institution truly rooted in the territory, with an intensive relationship with the territory and being able to support local economic activities evaluated inside a development pattern. For this reason, the bank’s decisional bodies must be “in the territory” in order to have both a thorough knowledge of the socio-economic reality (strengths, weaknesses and possible paths) as well as privileged relationships with local economic actors (Alexopoulos and Goglio, 2013b). In times of crisis, it is important to understand the power of alternative approaches to business and in this case, particularly to the financial system. As the papers collected in this issue extensively illustrate, cooperative banks are indeed a different “animal” in the banking “zoo”; however this animal is gradually adopting others’ behaviours, quite often through conscious decisions. Nevertheless, as we are in an era in which competition is blurring the lines between a pure commercial and a cooperative enterprise, if cooperative banks want to survive and flourish as an institutional pattern, they must re-establish the - lost in a market logic approach - link between cooperative values, members’ active participation and commercial strategy, as well as excellence in practices. Otherwise they will disappear, either physically or institutionally. However, their historical and cultural variety suggests that no simple or linear development path can be prescribed for all cooperative banks; their development varies under the influence of historically specific and contemporary economic and social conditions. Since cooperative credit is a flexible mechanism, not necessarily associated to simple or backward social and economic systems, its organization can and should coherently evolve with the development of the territories in which the bank operates. In the same way the institutional and legal framework should acknowledge their different approach to perform banking activities and help their evolution more than in a strictly business-like way, placing bonds and incentives consequently.

5 JEOD - Vol.3, Issue 1 (2014)


Editorial: Cooperative Banks at a Turning Point? Goglio, S.; Alexopoulos, Y.

References Ahrendsen, B., Dixon, B. and La Derrek, T.L. (1999). “Independent Commercial Bank Mergers and Agricultural Lending Concentration”, Journal of Agricultural and Applied Economics, 31(2): 215-227. Alexopoulos, Y. (2004). Financial Cooperatives and Rural Development in Greece. Ph.D Thesis, University of Leicester, UK. Alexopoulos, Y. and Goglio, S. (2011). “Financial Cooperatives: Problems and Challenges in the PostCrisis Era”, Journal of Rural Cooperation, 39(1): 34-47. Alexopoulos, Y. and Goglio, S. (2013a). “Introduction: Cooperative Finance and Sustainable Local Development”, in Goglio, S. and Alexopoulos, Y. (eds.), Financial Cooperatives and Local Development. London: Routledge, pp. 1-18. Alexopoulos, Y. and Goglio, S. (2013b). “Conclusion: a Tale of Two Models”, in Goglio, S. and Alexopoulos, Y. (eds.), Financial Cooperatives and Local Development. Routledge, London, pp. 227-232. Alexopoulos, Y., Catturani, I. and Goglio, S. (2013). “Searching for a Model of Governance in Cooperative Banking”, in Brazda, J., Dellinger, M. and Rössl, D. (Eds.), Genossenschaften im Fokus einer neuen Wirtschaftspolitik - Bericht der XVII. Internationalen Genossenschaftswissenschaftlichen Tagung (IGT) 2012 in Wien. Berlin et al.: LIT Verlag, pp. 707-731. Banca d’Italia (2008), Bollettino Economico della Banca d’Italia No. 51, gennaio. Available online at: http://www.bancaditalia.it/pubblicazioni/econo/bollec/2008/bolls51/bollec51 [accessed 10 April 2014]. Barham, B.L., Boucher, S. and Carter, M. (1996). “Credit Constraints, Credit Unions and Small-Scale Producers in Guatemala”, World Development, 24(5): 793-806. http://dx.doi.org/10.1016/0305750X(96)00001-0 Belaisch, A., Kodres, L., Levy, J. and Ubide, A. (2001). “Euro-Area Banking at the Crossroads”, IMF working paper series, WP/01/28, March, IMF. Berger, A., Goldberg, A. and White, L. (2001). “The Effects on Dynamic Changes in Bank Competition on the Supply of Small Business Credit”, European Finance Review, 5: 115-139. http://dx.doi. org/10.1023/A:1012776427175 Berger, A.N. and Udell, G.F. (1990). “Relationship Lending and Lines of Credit in Small Firm Finance”, Journal of Business, 25. Bonaccorsi di Patti, E. and Gobbi, G. (2001). “The Effects of Bank Consolidation and Market Entry on Small Business Lending”, Temi di discussione, Banca d’Italia, 404. Caswell, J.A. (1987). “Dominant Forms of Corporate Control in The U.S. Agribusiness Sector”, American Journal of Agricultural Economics, 69(1): 11-21. http://dx.doi.org/10.2307/1241301 Di Salvo, R., Guidi, A. and Mazzilis, M.C. (2002). “La riorganizzazione industriale del Credito Cooperativo: gli effetti delle concentrazioni sull’efficienza gestionale delle BCC e le implicazioni di policy”, Cooperazione di Credito. Di Salvo, R., Lopez, J.S., and Pezzotta, A. (2004). “L’evoluzione del relationship banking nei mercati locali: il ruolo delle Banche di Credito Cooperativo. Un’analisi degli spostamenti di clientela”, Cooperazione di Credito, 183/4. Emmons, W.R. and Schmid, F.R. (1999a). “Credit Unions and the Common Bond”, Federal Reserve Bank of St. Louis Review, September/October, pp. 41-64. Emmons, W.R. and Schmid, F.R. (1999b). “Wages and Risk-Taking in Occupational Credit Unions: Theory and Evidence”, Federal Reserve Bank of St. Louis Economic Review, March/April, pp. 13-32.

6 JEOD - Vol.3, Issue 1 (2014)


Editorial: Cooperative Banks at a Turning Point? Goglio, S.; Alexopoulos, Y.

European Association of Cooperative Banks (2009), Annual Report 2008. Available online at: http://www.eacb.coop/en/publications/annual_report_2008_cooperative_banks. html?BackC=4&BackP=2&BackMaxR=5 [accessed 10 April 2014]. European Central Bank (2000), Mergers and Acquisitions Involving the EU Banking Industry – Facts and Implications. Available online at: http://www.ecb.europa.eu/pub/pdf/other/eubkmergersen.pdf [accessed 10 April 2014]. European Central Bank (2001), “Bank Concentration and Retail Interest Rates”, ECB Working Paper Series, No. 72, July. Available online at: http://www.ecb.europa.eu/pub/pdf/scpwps/ecbwp72.pdf [accessed 10 April 2014]. Ferguson, C. and McKillop, D. (1997). The Strategic Development of Credit Unions”. London: John Wiley and sons. Ferri, G. (2008). Why Cooperative Banks are Particularly Important at a Time of Credit Crunch, mimeographed, University of Bari, Italy. Fonteyne, W. (2007). “Cooperative Banks in Europe – Policy Issues”, IMF Working Paper. Goglio, S. (1999). “Local Public Goods: Productive and Redistributive Aspects”, Economic Analysis, Vol. 2, No. 1, February, pp. 5-21. Goglio, S. (2009). “Local Credit and Territorial Development: General Aspects and the Italian experience”, in Bonoldi, A. and Leonardi, A. (eds.), Recovery and Development in European Periphery (19451960). Bologna: Il Mulino. Goglio, S. and Alexopoulos, Y. (eds.) (2013). Financial Cooperatives and Local Development. London: Routledge. Groeneveld, H. and de Vries, B. (2009). “European Cooperative Banks: First Lessons of the Subprime Crisis”, International Journal of Cooperative Management, 4(2): 8-21, September. Harhoff, D. and Körtring, T. (1998). “Lending Relationship in Germany: Empirical Results from Survey Data”, Journal of Banking and Finance, 22. Hariyoga, H. (2004). An Economic Analysis of Factors Affecting the Failure of an Agricultural Marketing Cooperative: The Bankruptcy of Tri Valley Growers. Davis: University of California. Hesse, H. and Cihak, M. (2007). “Cooperative Banks and Financial Stability”, IMF Working Paper, 07/2. Huppi, M. and Feder, G. (1989). “The Role of Groups and Credit Cooperatives in Rural Lending”, Agriculture and Rural Development Department Working Paper Series 284, The World Bank. International Monetary Fund (2007), Global Financial Stability Report, September. Leggett, K.J. and Strand, R.W. (2002). “Membership Growth, Multiple Membership Groups and AgencyControl at Credit Unions”, Review of Financial Economics, 11: 37-46. http://dx.doi.org/10.1016/ S1058-3300(01)00032-5 Leyshon, A. and Thrift, N. (1993). “The Restructuring of the UK Financial Services Industry in the 1990s: A Reversal of Fortune?”, Journal of Rural Studies, 9(3): 223-241. http://dx.doi.org/10.1016/07430167(93)90068-U Leyshon, A. and Thrift, N. (1995). “Geographies of Financial Exclusion-financial Abandonment in Britain and the United States”, Transactions of the Institute of British Geographers, 20 (3): 312-341. http:// dx.doi.org/10.2307/622654 Oliver Wyman (2008). Cooperative Bank: Customer Champion. Available online at: http://www. globalcube.net/clients/eacb/content/medias/publications/external_studies/CBank_Costumer_ Champion.pdf [accessed 10 April 2014].

7 JEOD - Vol.3, Issue 1 (2014)


Editorial: Cooperative Banks at a Turning Point? Goglio, S.; Alexopoulos, Y.

Oliver Wyman (2014). Cooperative Banking: Leveraging the Cooperative Difference to Adapt to a New Environment. Available online at: http://www.oliverwyman.com/content/dam/oliverwyman/global/en/files/insights/financial-services/2014/Mar/2014%20Oliver%20Wyman_ Cooperative%20Banking_web.pdf [accessed 10 April 2014]. Osterberg, P. and Nilsson, J. (2009). “Members’ Perception of Their Participation in the Governance of Cooperatives: The Key to Trust and Commitment in Agricultural Cooperatives”, Agribusiness, 25(2): 181-97, Spring. http://dx.doi.org/10.1002/agr.20200 Ouattara, K., Gonzalez-Vega, C. and Graham, D.H. (1999). “Village Banks, Caisses Villageoises, and Credit Unions: Lessons from Client-Owned Microfinance Organisations in West Africa”, Economics and Sociology Occasional Paper, No. 2523. Department of Agricultural, Environmental and Development Economics, The Ohio State University. Patin, R.P. and McNeil, D.W. (1991a). “Benefit imbalances among credit union members”, Applied Economics, 23: 769-780. http://dx.doi.org/10.1080/00036849108841034; http://dx.doi. org/10.1080/772858989; http://dx.doi.org/10.1080/00036849100000184 Patin, R.P. and McNeil, D.W. (1991b). “Member Group Orientation of Credit Unions and Total Member Benefits”, Review of Social Economy, December, pp. 37-61. http://dx.doi. org/10.1080/00346769100000003 Petersen, M. and Rajan, R. (1994). “The Benefits of Lending Relationship: Evidence from Small Business Data”, The Journal of Finance, 49. Poyo, J., Gonzalez-Vega, C. and Aguilera-Alfred, N. (1993). “The Depositor as a Principal in Public Development Banks and Credit Unions: Illustrations from the Dominican Republic”, Paper presented at the Conference Finance 2000 – Financial Markets and Institutions in Developing Countries: Reassessing Perspectives, 27-28 May, Washington D.C. Smith, D. (1984). “A Theoretic Framework for the Analysis of Credit Union Decision Making”, The Journal of Finance, XXXIX (4): 1155-1168. http://dx.doi.org/10.1111/j.1540-6261.1984.tb03899.x Smith, D., Cargill, T. and Meyer, R. (1981). “Credit Unions: An Economic Theory of a Credit Union”, The Journal of Finance, Vol. XXXVI(2): 519-528, May. http://dx.doi.org/10.2307/2327039; http://dx.doi.org/10.1111/j.1540-6261.1981.tb00470.x Stiglitz, J. (2009). “The Financial Crisis of 2007/2008 and its Macroeconomic Consequences’, available online at: http://www2.gsb.columbia.edu/faculty/jstiglitz/download/papers/2008_Financial_Crisis. pdf [accessed 10 April 2014].

8 JEOD - Vol.3, Issue 1 (2014)


Part 1 Cooperative banks in the recent crisis

9 JEOD - Vol.3, Issue 1 (2014)


10 JEOD - Vol.3, Issue 1 (2014)


AT T R I B U T I O N 3 . 0

You are free to share and to remix, you must attribute the work

AUTHOR

Publication date: 17 June2014 June 2014||Vol.3, Vol.3,Issue Issue11(2014) (2014)1-24 11-33

Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles1

HANS GROENEVELD Cooperative & Sustainable Business of Rabobank, Nederland j.m.groeneveld@rn.rabobank.nl

ABSTRACT This paper complements the existing scarce literature on financial cooperatives in various ways. First, we describe the background and evolution of European Cooperative Banking Groups (ECBGs). Second, we summarize the main reasons for the past disregard and recent revaluation of the cooperative banking model. Third, we empirically investigate to what extent the financial performance of ECBGs over recent business cycles is related to the original cooperative characteristics. To this end, we have constructed a new database with a broad range of financial variables for fifteen ECBGs in ten countries and collected similar indicators for entire banking systems of the countries in question. Our empirical findings suggest that many previous assertions and qualitative statements about ECBGs really hold in practice and not just in periods of financial distress. Furthermore, ECBGs do exhibit a different performance compared to all other banks throughout different stages in recent business cycles. Their corporate governance with members’ influence and specific decision making mechanisms seems to lead to a relatively low risk appetite and high capitalization, a high degree of stability and a predominant focus on retail banking. It must be emphasized that these conclusions cannot be extrapolated into the future. Indeed, an abundance of historical examples of successes and failures among all types of banks exists.

KEY-WORDS EUROPEAN BANKING; SHAREHOLDER CORPORATE GOVERNANCE

BANKS;

COOPERATIVE

BANKS;

PERFORMANCE;

JEL Classification: G2; G21; G3; G32; G34; L21; P13 | DOI: http://dx.doi.org/10.5947/jeod.2014.002

1

The views in this article are personal and do not necessarily reflect those of Rabobank Nederland. 11 1 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

1. Introduction The European banking sector is not homogeneous. Basically, one can distinguish between state banks, shareholder-value banks, e.g. mainly listed banks, and stakeholder-value banks. The latter category comprises savings banks, credit unions, mutuals and cooperative banks. There are indications that these stakeholder-value banks weathered the subsequent storms relatively well so far, without large scale state support (EACB, 2010; Birchall, 2013). At the same time, these types of banks did not receive much academic and policy attention before the financial crisis hit. Hence the question arose as to why these banks apparently have avoided great financial distress. This article tackles this question for the largest category within the family of stakeholdervalue banks: European cooperative banking groups (henceforth ECBGs). Cooperative banks are controlled by members who have voting rights. Acknowledging the heterogeneity of ECBGs (Ayadi et al., 2010), the possible connection between the common features and the relative performance of fifteen ECBGs in ten European countries over the latest business cycles is explored. More specifically, the article investigates whether longstanding assertions about the corporate governance and organizational features are reflected in differences between performance indicators of ECBGs and all other banks in the time span 1997-2011 or 2002-2011, depending on data availability. The text will be larded with concrete examples of individual ECBGs. In this respect, this paper complements existing scarce academic studies and policy reports on financial cooperatives in various ways. In particular, it analyzes the central issue in a concise historical perspective and in the context of organizational characteristics of ECBGs. Second, fifteen ECBGs are simultaneously examined over a similar and relatively long time span, which enables us to draw robust conclusions about the entire cooperative banking sector. Most recent articles are case studies of – specific aspects of – individual ECBGs in different times of financial distress and/or over relatively short time spans (e.g. Stefancic and Kathiziotis, 2011; Mooij and Boonstra, 2012; Bley, 2012; Goglio and Alexopoulos, 2012), which results in a diffuse picture and does not allow for general conclusions. Third, we shall empirically validate qualitative postulations about ECBGs from previous publications (e.g. EACB, 2007). Fourth, we do not only investigate the relative performance of ECBGs in (recent) times of crisis, but our sample period also incorporates times of economic prosperity. The paper is structured as follows. Section 2 sketches the roots, organizational structure and evolution of ECBGs. We briefly describe how they eventually emerged from small local credit cooperatives more than a century ago. This clarification provides useful starting points for understanding their recent performance. The reasons for the past disregard and recent appraisal of the cooperative banking model are discussed in sections 3 to 5. Section 6 formulates testable hypotheses, which are derived from the preceding sections. In section 7, we highlight our newly constructed and more comprehensive database, which covers a broad range of indicators for fifteen ECBGs in ten European countries and contains similar measures for the entire banking systems of the countries in question. The sample period runs from 1997-2011 or 20022011, encompassing more than one business cycle. Section 8 contains the empirical results and section 9 summarizes the main findings.

12 2 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

2. The transformation of local credit cooperatives into ECBGs The history and evolution of many ECBGs is extensively documented2. In short, most cooperative banks were established more than a century ago in response to the problems that small urban and rural businesses had in accessing affordable financial services (Guinnane, 2001). These banks were able to serve these “excluded” groups because members provided funding or stood bail and were therefore involved in the decision-making process. Local cooperatives did not aim at maximizing short term profits, but profits were necessary for further growth and were for the larger part retained and added to the capital base. Beginning in Germany, the cooperative banking concept gradually dispersed to other European countries. Not all cooperative banks managed to survive the ravages of time. Quite a few cooperatively organised banks were unable to adapt to technological, social or competitive changes and consequently disappeared or now just live a marginal existence3. Many countries never had a cooperative banking sector of any significance, because the cooperative ideas did not find fertile soil due to policy impediments or absence of a supportive regulatory framework, among other things. In other countries, cooperative banks chose to be acquired by other banks or have converted into investor-owned banks4. Over time, the cooperative banking model of the “survivors” evolved and differentiated into a multiplicity of European institutions with characteristics reflecting the needs of cooperative members on the one hand and the specificities of national legislative frameworks on the other (Alexopoulos and Goglio, 2009). The majority of local cooperatives developed into national (network) organizations and became active in other fields of financial services such as insurance or leasing. The increasingly high level of domestic integration was partly prompted by regulatory requirements or the necessary realisation of economies of scale and higher efficiency levels from a competitive point of view. Subsequently, quite a few national organizations transformed into internationally active banking groups. Some ECBGs have sold a part of their business activities to investors or became partly listed, thus gradually transforming into a hybrid type of financial cooperative5. Hence, the organizational structures are definitely not static, but are constantly evolving. Figure 1 presents total assets of the ECBGs included in this study from the smallest to the largest. The ratio of the largest (French Crédit Agricole Group) to the smallest (Portuguese Credito Agricola Group) is 144, which shows the great disparity in sizes. ECBGs also vary in terms of their attitudes to membership. Some banks strive to make every customer a member, while others are not actively recruiting members (Oliver Wyman, 2008). Other striking differences include the extent of centralization and integration within the networks (Desrochers and Fischer, 2005)6, the size and focus of international activities, and the design of the cooperative governance with member authority (see Ayadi et al., Chapter 3, 2010). In most cases, governance reform and pressures of competition have fostered an accentuated centralisation of 2

See for instance Bosseno (1994); Aschhoff and Hennigsen (1995); Brazda (2001); Werner (2005); Albert (2008) and Mooij (2009).

3

In Sweden, the Föreningsbanken Sverige was more or less forced by the government to convert in 1993 from a cooperative ownership structure to a stock corporation (Körnert, 2012).

4

A striking example is the wave of demutualization of mutual building societies in the United Kingdom in the 1990s (Llewellyn, 2012). Since then, these societies adopted riskier business models, faced severe losses in the latest financial crises, went bankrupt or were close to failure.

5

As an example, the French Crédit Agricole S.A, listed since 2006 on the Euronext Paris, was created to represent all of the Group’s business lines and components. As of December 2011, 56.2% of Crédit Agricole S.A was owned by the regional banks that make up the Federation of Crédit Agricole, and 38.7% was owned by institutional and individual investors.

6

The Dutch Rabobank Group is one of the most centralised systems, whereas the Italian cooperative banking sector is the most decentralised system. 13 3 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

strategic and operating functions and processes. This has led to the establishment of so-called higher-tier networks, which still vary from loose associations to cohesive groups (Di Salvo, 2003)7. In a few cases, the central institution has an important supervisory role over its local bank members8. A general feature of ECBGs is the existence of some form of internal solvency and liquidity safety nets, except for the important supervisory role over its local bank members8. A general feature of ECBGs is the Italian of Banche structures form core of safety internalnets, mutual support schemes ECBGs. existence somePopolari. form ofThese internal solvency andtheliquidity except for the ItalianinBanche Due to These all thesestructures factors, many observers find that cooperatives relatively complexDue governance Popolari. formoutside the core of internal mutual support have schemes in ECBGs. to all structures fragmentation ownership, decision rights member one vote” principle), these factors,given manytheir outside observersoffind that cooperatives have(“one relatively complex governance structures given their of (Oliver ownership, decision mutual guarantees andfragmentation multi-level boards Wyman, 2012). rights (“one member one vote” principle), mutual guarantees and multi-level boards (Oliver Wyman 2012). Figure 1. Asset size of ECBGs (in EUR millions) FIGURE 1. ASSET SIZE OF ECBGS (IN EUR MILLIONS) 2000000 1800000 1600000 1400000 1200000 1000000 800000 600000 400000 200000 0

Source: data data are provided by ECBGs and pertain to 2011 Source: are provided by ECBGs and pertain to 2011

3. The era of underexposed and fading cooperative banking features For a long time, the features and values of the cooperative banking model did not attract a lot 3. The erainofarticles, underexposed andreports fading cooperative banking features of attention the press, and scholarly research for various reasons (Kalmi 2007). Firstly, the original “mission” of cooperative banks seems to have been largely completed and the a longd’être time, of thecooperative features and banks values have of thebecome cooperative model everyone did not attract a lot of originalFor raisons less banking valid; almost in Western Europe has access to financial servicesand today. Moreover, comparative nonattention in articles, the press, reports scholarly researchthe for various reasonsdisadvantages (Kalmi, 2007).that Firstly, the cooperative banks faced in the past for servicing small farmers and small businesses have also original “mission” of cooperative banks seems to have been largely completed and the original raisons d’être largely disappeared. Legal frameworks now offer much stronger contract enforceability and verifiable information about potential borrowers is generally available. In other words, the traditional differentiators of the initial local cooperatives have become less pronounced and less 7 For instance, French Crédit Agricole Group has a three-tier network, comprising local, regional and central organizations. understood over the time. The Dutch Rabobank has a two-tier network, consisting of local member banks and the central organization. Some Italian Another reason is that very the loosely transformation ofCentrale local delle cooperatives into network Banche Popolari are connected with the Istituto Banche Popolari, but(inter)national the majority of Banche Popolari acts completely independently each other.in varying degrees of hybridisation with the “capitalist” organizations (ECBGs) has ofresulted 8 corporate the proclaimed multiple of ECBGs arethegenerally more Group. difficult to This model. is the caseBesides, for the Austrian Volksbanken, the Finnish goals OP-Pohjola Group and Dutch Rabobank In these countries, thethe supervisors have delegated the respective APEX organizations formal supervisory powers its listed member understand than theoretically moretoeasily interpretable goal of profit maximizing of over most banks. These central institutions themselves are supervised by the national supervisors.

8

This is the case for the Austrian Volksbanken, the Finnish OP-Pohjola Group and the Dutch Rabobank Group. In these 14 4 countries, the supervisors have delegated to the respective organizations formal supervisory powers over its JEOD - Vol.3,APEX Issue 1 (2014) member banks. These central institutions themselves are supervised by the national supervisors.   4


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

of cooperative banks have become less valid; almost everyone in Western Europe has access to financial services today. Moreover, the comparative disadvantages that non-cooperative banks faced in the past for servicing small farmers and small businesses have also largely disappeared. Legal frameworks now offer much stronger contract enforceability and verifiable information about potential borrowers is generally available. In other words, the traditional differentiators of the initial local cooperatives have become less pronounced and less understood over time. Another reason is that the transformation of local cooperatives into (inter)national network organizations (ECBGs) has resulted in varying degrees of hybridisation with the “capitalist” corporate model. Besides, the proclaimed multiple goals of ECBGs are generally more difficult to understand than the theoretically more easily interpretable goal of profit maximizing of most listed banks. As pointed out by Ayadi et al. (2010), cooperative banks can be categorized as “dual-bottom line” institutions. They claim to fulfill other equally important objectives than mere shareholder value creation. This suggests that financial performance and economic efficiency are neither the only nor the ultimate standard of assessment for ECBGs. These aspects are indisputably important, but they are not sufficient to assess the contributions of cooperative banks to society and the economy. Moreover, the dominance of the free market thinking and the associated Anglo-Saxon model aimed at profit and shareholder value maximization did not encourage great interest in ECBGs. In this shareholder value era, some subsidiaries of ECBGs actually got partly listed9 or adopted practices from banks with other organizational forms. Other ECBGs extensively debated whether or not to change from a cooperative organization into a listed financial company10, because this was sometimes considered to be an appropriate way to attract external capital for faster growth (Deloitte, 2012). Hence, ECBGs themselves were also partly responsible for confusion and contempt of their cooperative business model. Besides, some ECBGs do not report reliable empirical data or longer and consistent time series for key cooperative and financial indicators, partly because they do not have the same extensive reporting requirements as listed banks. This aspect obviously hampers an objective evaluation of their business model and impedes empirical and scholarly research. All these developments were not favorable for retaining a clearly visible cooperative identity and the collaboration with or adoption of elements of non-cooperative enterprises have been sometimes viewed as a capitulation to capitalism. In fact, ECBGs were sometimes forced into a defensive position prior to the crisis as their cooperative business model was considered to be rather misty, outdated or even detrimental for the entire banking sector (Kodres and Narain, 2010). Cooperative institutions were not considered the most efficient, vibrant, or innovative institutions for a long time. PA Consulting Group (2003) even accused cooperative banks of “spoiling” the market conditions for other banks. Others (e.g. Oliver Wyman, 2008) underscored the sluggishness and opacity of decision making processes or exaggerated the principal-agent problem inside ECBGs based on merely theoretical considerations (Groeneveld and Llewellyn, 2012)11.

9 10 11

A subsidiary of the Austrian Raiffeisenbanken, Raiffeisen Zentral Bank, is listed. The Dutch Rabobank pursued the Great Cooperative Debate in the years 1995 through 1997. This refers to potential conflicts of interest between managers and owners of a bank. Agency issues arise in any organization in which there is a separation of decision and risk-taking functions. In the case of Cooperative banks, these issues emerge between the management and the members. In the case of shareholder value companies, these issues occur between the management and shareholders. 15 5 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

4. Assertions about ECBGs The European Association of Cooperative Banks (EACB) and the International Cooperative Banking Association (ICBA) made efforts to emphasize the special nature of cooperative banks in various reports well before and during the crisis. One of the messages is that the customer has always been and still is at the core of their operations and, at a local level, members still “have a say” in the local member bank’s policy (EACB, 2007). It is also suggested that cooperative banks have an “impact presence” on society and the entire banking market. Here, some qualitative publications refer to their contribution to economic growth, employment and the creation of more favourable interest rate conditions for customers. Another claim is that the orientation of the domestic cooperative banking part inside ECBGs has remained relatively unaltered (EACB, 2005). The first-level cooperative banks are still predominantly targeted towards retail banking and servicing the real economy, i.e. private individuals and SME’s, because effective member influence would force them into this direction. This would also translate into a lasting engagement with local regions and the real economy, which would be visible in relatively dense branch networks, i.e. physical proximity to customers and members. It is also stated that proximity is further reinforced through the participation in numerous social networks and by actively supporting the local communities. Suppose that local banks really have strong local ties and networks, they would theoretically be better equipped to assess the creditworthiness and risks of customers at a local level. If that is true, this differentiator would be reflected in relatively stable, and possibly higher, lending to households and corporate customers in favourable and unfavourable times. Retail banking is mainly about relationship banking which goes hand in hand with a long term orientation (Boot, 2000). This would imply that especially local cooperative banks within ECBGs do not aim at short term benefits of their operations, services and products for members and customers and themselves, but champion a “dual bottom line” approach. They do seek profit, but also strive for economic and social welfare in local communities. Consequently, their returns on equity or assets are expected to be lower but more stable. Their risk profile should be also comparatively moderate. Despite all the changes in the financial structures and composition of the balance sheets, it is also stated that ECBGs still add a considerable part of their net profits to their capital and reserves, which would lead to a solid capitalization. For ECBGs in our sample with a credit rating at the group level, this would be visible in relatively high ratings12. These comparatively high ratings would also stem from existing legallybinding cross guarantees to connect different entities of the group as a risk management tool. Rating agencies tend to view this type of arrangement as less risky since the entire organization is viewed as a single consolidated risk unit. Until the breakout of the credit crisis, many position papers and background documents had a predominantly qualitative, or even ideological, character and lacked “empirical” proof. It is undeniable, however, that cooperative banks stand out regarding their history, structure, organizational form and original business objectives from other banks. But these aspects were often ill understood and misinterpreted as elaborated in Section 3. The main observable differentiator of ECBGs is their specific corporate governance with some degree of member control. Member influence surely cannot rule out policy mistakes, but can basically bridge the distance between executives and policy makers and many different stakeholders. Theoretically, this intrinsic feature is “only” a precondition for ECBGs to be able to operate or position themselves differently in the market.

12

Seven ECBGs in our sample obtain a credit rating for the entire group, i.e. on a consolidated basis. 16 6 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

5. The appraisal of cooperative banking features The subsequent crises have positively changed the opinions and views about cooperative banks over the last five years. Preliminary evidence indicated that the cooperative organizational form in general had performed significantly better than other organizational forms after the global financial crisis of 2007/8 and the following recession (Birchall and Hammond Ketilson, 2009). Policy makers, regulators and academics started to wonder whether these achievements could indeed be related to asserted specificities of the cooperative banking model. Furthermore, the interest in cooperative banks was boosted by the United Nations which declared 2012 as the International Year of Cooperatives (UN, 2011). In addition, international consultancy firms (Deloitte, 2012; McKinsey, 2012; Oliver Wyman, 2012) and The Economist (2010) started to pay attention to the merits and characteristics of the cooperative banking model. The financial crisis disputed the alleged shortcomings of the cooperative banking model and the perceived superiority of the shareholder value banking model. For a long time, comparisons of the pros and cons of corporate governance structures between cooperative banks and investor-owned banks were sometimes misleading as they were based on incorrect starting points. The issue is that it is not always clear on what basis the comparison was being made: (i) the ideal investor-owned bank, (ii) the ideal cooperative bank, (iii) the actual investor-owned bank, and (iv) the actual cooperative bank. In other words, it is necessary to distinguish between how institutions behave in some abstract, theoretical or ideal state, and the way they operate in practice. The ideal investor-owned bank has clear-cut principles defining objectives, accountability and control. Therefore, the corporate governance of these banks was often deemed to be more straightforward than that of cooperative banks where many theoretical flaws of any corporate governance were thought to apply in practice (Kalmi, 2007; Fonteyne, 2007). However, recent experience unambiguously points to ill-functioning aspects of corporate governance mechanisms in investor-owned banks: the investor-owned model has shortcomings in practice as well. At the same time, the theoretical shortcomings of corporate governance arrangements in cooperative banks were magnified and exaggerated for a long time (Groeneveld and Llewellyn, 2012). Be that as it may, one can equally well assert that the management of quite some investor-owned banks has visibly failed to operate in the interests of their shareholders by following strategies to maximize shareholder value, which caused huge losses and write downs and necessitated large-scale government intervention in the last few years13. In conclusion, it is tendentious to compare the actual behavior of a cooperative bank model with some mythical ideal form of investor-owned model. It must be acknowledged that in practice both forms operate imperfectly and no safe conclusions can be drawn regarding the superiority of one form over the other. Another viewpoint regarding cooperative banking has also changed recently. It is increasingly realized that cooperative banking is not synonym to some kind of “philanthropic� banking which mainly exists to achieve social objectives (Bonin, 2012). Cooperative banks need to have adequate and innovative products and services at fair prices and state of the art distribution concepts in the first place. If not, they cannot survive and operate on banking markets and will not be chosen by customers.

13

Northern Rock, Fortis, UBS and Royal Bank of Scotland are clear examples of this. 17 7 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

6. Hypotheses The takeaway of recent articles and reports is that ECBGs still have internal characteristics and a business orientation which can be traced back to the key features of the former credit cooperatives. In short, member ownership is believed to contribute to continuity and a cautious course of the entire ECBGs via specific internal governance mechanisms. Here it should be noted that in some ECBGs local or regional local banks still constitute by far the greatest part of the group, whereas other ECBGs undertake sizeable activities outside their “traditional” cooperative core. Having said this, the asserted specificities should show up in a divergent performance of ECBGs compared to other banks. We shall test which of the proclaimed differentiators and assertions discussed in the previous sections are valid and visible throughout recent business cycles. Concretely, we have inferred five main interrelated hypotheses: a. Hypothesis 1: ECBGs have a strong customer focus and client proximity. The alleged engagement with local communities and the real economy as well as member influence should imply relatively dense branch networks. If ECBGs really put the customer interests first, are not risk seekers or profit maximizers, this should also be visible in recent data, especially in times of crisis. Indeed, many customers lost confidence in their financial institutions and financial advisors and were not satisfied with their behavior and performance in recent years. Moreover, the absence of explicit profit targets due to the proclaimed focus on customers’ interests, member influence and the emphasis on retail banking is expected to show up in lower average returns on assets (and equity) than investor-owned banks. b. Hypothesis 2: ECBGs aim at austerity and efficiency in operations. Austerity and efficiency in business operations were important characteristics of local credit cooperatives, which were set up with members’ money. Since member ownership still exists, frugality and efficiency should ideally be virtues of present ECBGs as well. Among other things, this implies that the absence of a profit objective, or a lower profit requirement, may not lead to inefficient operations. The stated focus on customer value cannot be an excuse for more relaxed cost control and inefficient operations. We shall test this hypothesis by comparing cost-income indicators of ECBGs with those of other banks. c. Hypothesis 3: ECBGs are relatively stable institutions with focus on retail banking. Because of member ownership, ECBGs are believed to be mainly focused on retail, commercial and SME banking. Consequently, they would have a limited appetite for non-core add-ons and a bias towards serving and financing “real economy” activities. This would be accompanied by a long-term view of relationships with local businesses and municipalities and an innate focus on customers. This area of banking is associated with relatively stable income streams across business cycles and a moderate risk profile. Hence, ECBGs are assumed to be fairly stable organizations with moderate returns on assets/ equity and a relatively large retail banking business. d. Hypothesis 4: ECBGs have a strong capitalization and low risk profile. A natural conservatism should be created by distributed, independent governance with member influence and ownership and relatively limited access to third party capital. This could mean that ECBGs steered away from riskier activities and practices, for example operating at relatively high levels of tier 1 capital (Laeven and Levine, 2009). The higher capitalization should in turn result in lower returns on equity compared to banks with another business orientation. e. Hypothesis 5: ECBGs have an impact presence. It is stated that ECBGs have an impact presence on the macro and local level, the banking market structure and banking conditions for customers. First, from an economic perspective, it has been argued that they create jobs, contribute to economic growth by granting loans and credits and aim at a sustainable development of local communities (EACB, 18 8 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

2005). Second, they are assumed to stimulate stability, diversity and competition in banking (Ayadi et al., 2010). Third, the presence of ECBGs is believed to lead to better conditions for customers, e.g. higher interest rates on savings and lower interest rates on loans.

7. Sample description The main objective of this article is to test these hypotheses by investigating the performance of fifteen ECBGs vis-à-vis entire banking sectors in eleven countries over the last turbulent decade14. Because of their specific nature, different reporting requirements and heterogeneity, it is inappropriate to use databases like Bankscope to collect data on cooperative banks. These databases contain inconsistencies and many caveats regarding cooperative banks. For some cooperative banks, consolidated data for the entire banking group are reported, whereas in other cases unconsolidated data for – small – individual local cooperative banks are given. If these differences are ignored, one easily arrives at misleading conclusions. Actually, data on individual local cooperative banks cannot be compared with those of other types of banks, which often pertain to consolidated group figures. Besides, individual cooperative banks usually obtain all kinds of support from a central institution (APEX), e.g. products, IT systems and HR services, to reach economies of scale inside the entire cooperative banking group. For our empirical investigation, we combine several data sources. We use consolidated data for ECBGs which are composed by these groups themselves15. If possible and appropriate, we have corrected the figures for major breaks in the time series caused by sizeable mergers and/or acquisitions to be able to make sensible comparisons between ECBGs and entire banking sectors. In countries with more than one cooperative banking group, we have constructed aggregated indicators by using total assets of individual cooperative groups as weights. Data on entire banking sectors in the countries under review are collected from national central banks or supervisory agencies as well as from the IMF and European Central Bank. The period of analysis is determined by the availability of good quality data and spans either 1997-2011 or 2002-11. Both periods encompass years of strong economic growth and financial stability as well as years of economic slack and financial instability. This feature offers the opportunity to test whether the asserted specificities of ECBGs really lead to different performances compared to those of entire banking systems both in economically and financially prosperous and in difficult times.

14

We have restricted our empirical analysis to European cooperative banking groups for two main reasons. The first one is that reliable data on cooperative banks in other parts of the world are hardly available. Second, cooperative banks in other parts of the world operate in totally different economic, regulatory and social circumstances and differ regarding their development phase and maturity. So, the overall analysis would be obscured by situations that differ considerably across continents.

15

In some cases, the consolidated figures were constructed upon request by the author. The data for the Italian Banche Popolari are an example. 19 9 JEOD - Vol.3, Issue 1 (2014)


acquisitions to be able to make sensible comparisons between ECBGs and entire banking sectors. In countries with more than one cooperative banking group, we have constructed aggregated indicators by using total assets of individual cooperative groups as weights. Data on entire banking sectors in the countries under review are collected from national Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles central banks or supervisory agencies as well as from Groeneveld, H. the IMF and European Central Bank. The period of analysis is determined by the availability of good quality data and spans either 1997-2011 or 2002-11. Both periods encompass years of strong economic growth and financial stability as well as years of economic slack and financial instability. This feature offers the opportunity to test whether the asserted specificities of ECBGs really lead to different performances compared to those 8. Empirical analysis of ECBGs of entire banking systems both in economically and financially prosperous and in difficult times.

8.1. Members analysis of ECBGs 8. Empirical 8.1.AsMembers stated before, ECBGs frequently publicly assert that they do not aim at maximising profits but customer value (EACB, 2005). Ideally, one would like to verify this assertion with direct insights and As stated before, ECBGs frequently publicly assert that they do not aim at maximising profits opinions from customers, i.e. “hard data”Ideally, or empirical Basically, it comes to the perception but customer value (EACB 2005). one evidence. would like to verify this down assertion with direct ofinsights customers banks “walki.e. their talk”. Or in words, keep their promisesitand treat andwhether opinionsECBGs from customers, “hard data” or other empirical evidence. Basically, comes down to the perception of customers whether ECBGs banks “walk and theirappreciation talk”. Or inofother words,of their customers fairly. Unfortunately, information about the perception customers keep their promises and focus treat their customers fairly. Unfortunately, information aboutfor themany perception this proclaimed customer and the maximization of customer value is not available banks, and appreciation of customers of this proclaimed customer focus and the maximization of customer including ECBGs. A more accurate indicator would be the level of “customer advocacy”: the perception by value is not available for many banks, including ECBGs. A more accurate indicator would be the customers that their financial institution does what is right for that theirtheir clients, not just what is right the level of “customer advocacy”: the perception by customers financial institution doesfor what bottom line. Trust and confidence are the key words in this respect. Some recent surveys and reports seem is right for their clients, not just what is right for the bottom line. Trust and confidence are the key in that this cooperative respect. Some recent seemfinancial to suggest that cooperative banks havein towords suggest banks havesurveys sufferedand lessreports than other institutions from a loss of trust suffered less than other financial institutions from a loss of trust in recent years, but the empirical recent years, but the empirical evidence remains flimsy (Michie, 2010; Ensor, 2012; Oliver Wyman, 2012). evidence remains flimsy (Michie 2010; Ensor 2012; Oliver Wyman, 2012). Table 1. Branches and members in individual countries TABLE 1. BRANCHES AND MEMBERS IN INDIVIDUAL COUNTRIES Branches (1997 = 100) Member to population ratio # Members (1997 = 100) ECBGs TBS1 2004 2011 2004 2011 1997 2004 2011 2011 Austria 73 73 129 135 29.8 28.2 28.7 102 Denmark 98 119 70 56 10.4 7.7 5.3 53 Finland 91 72 120 117 12.6 21.1 24.7 205 France 125 141 84 67 25.2 29.4 34.0 147 Germany 76 70 70 56 17.3 18.8 20.8 120 Italy 139 178 113 111 3.0 3.0 4.0 140 The Netherlands 71 48 50 38 3.4 8.9 11.1 355 Portugal 121 135 112 135 2.6 2.9 3.8 148 Spain 132 141 104 102 2.8 3.9 5.3 220 Switzerland 92 83 78 78 10.0 16.9 22.1 246 Total average 104 112 89 80 12.9 14.8 16.9 140 Source: ECBGs and ECB Note: Data of French and total ECBG branch offices are adjusted for major breaks caused by the acquisition of Crédit Lynnois by Source: ECBGs and ECB Crédit Agricole in 2006 and the ECBG merger of Banque Populaire and Caisse d’Eparges in caused 2009 by the acquisition of Crédit Lynnois by Note: Data of French and total branch offices are adjusted for major breaks Countries

Crédit Agricole in 2006 and the merger of Banque Populaire and Caisse d’Eparges in 2009 1

Number of branches of all other banks, i.e. excluding branches of local cooperative banks In some cases, the consolidated figures were constructed upon request by the author. The data for the Italian Banche Popolari are an example.   Hence, we confine ourselves to indirect proxies9for customer satisfaction and advocacy. We look at 15

member to population ratios and market shares which contain some implicit information about the attractiveness and popularity of the domestic cooperative banking part of ECBGs. Table 1 shows the development of the number of members and member-population ratio of the included ECBGs in their domestic markets. Strikingly, the number of members has increased every individual year, i.e. also in the era of underexposed cooperative banking features (see Section 3). Total number of members rose from around 37 million in 1997 to approximately 52 million in 2011, which equals a growth of about 40 per cent. On average, the member base grew at an annual growth rate of almost 2.5 per cent since 1997. In relative terms, the average member to population ratio showed an upward trend; the ratio rose from 12.9 in 1997 to 16.9 in 2011. Every ECBG attracted more members, with the notable exception of Denmark. The divergences in the level of this ratio can be explained by differences in the market position of individual ECBGs as well as variations in the attitude towards membership policy. The Dutch Rabobank witnessed by

20 10 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

far the largest inflow of members (plus 250 per cent)16, followed by considerable expansions in Switzerland, Spain and Finland. Implicitly, the absolute and relative rises in members point to an increasing popularity of the cooperative banking model in the home countries of ECBGs. The underlying reasons for the absolute and relative surge in members are hard to isolate and will probably be of a financial and immaterial nature17. It merely indicates that ECBGs have succeeded in attracting new members with their products, advisory services, client approach, business models or other features. The increase also signals customers confidence in ECBGs and corroborates tentative results of some fragmented surveys (Ensor, 2012; Oliver Wyman, 2012). Indeed, clients are presumably not very eager to become a member of local cooperative banks if the level of trust and satisfaction is low. 8.2. Domestic loan and deposit market shares The increase in the number of members has translated into rising market shares in national retail banking markets. Since 1997, ECBGs succeeded to increase their domestic market shares in mortgages and consumer loans as well as in private savings steadily and continuously throughout economic cycles. On average, both retail market shares rose by about 10 percentage points to 26 per cent in 2011. In the turbulent years 2007-11, ECBGs also strengthened their domestic market positions, but the increase did not differ significantly from that in the other sub-periods. These rises imply shifts of many billions of euros in loans and deposits towards ECBGs. The annual increases were mostly caused by endogenous growth, though in some years acquisitions of or mergers with non-cooperative players were also partly responsible for the rise in overall market shares18. The underlying data show that on balance no individual ECBG lost domestic market share over this period. Two thirds of all ECBGs increased their market shares, whereas the market position of other ECBGs remained stable. Like the substantial increase in the number of members, rising market shares are just signs that customers felt relatively more attracted to ECBGs for a myriad of different reason.

16

This can be attributed to a very active membership policy after the finalization of the Great Cooperative Debate in 1998 (see Mooij, 2009).

17

Reasons to become a member are manifold (EACB, 2007). It all starts with trust and confidence in the bank. When these elements are present, marketing and brand research shows that customers attach great importance to both material and immaterial aspects. For instance, the extent to which customers feel that the bank acts in their interests, the identification with the brand, access to the bank’s networks and knowledge, the stability/duration of relationships, the way banks deal with environmental and sustainability issues, the degree of product and price transparency, etc.

18

In France, ECBGs acquired several private banks over the time sample. The rise in market shares in 2009 was partly due to the merger between CrÊdit Mutuel and Caisse d’Epargne. 21 11 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

FIGURE 2. AVERAGE MARKET SHARE OF DEPOSITS, LOANS AND BRANCHES OF ECBGS (AS %) 40

40

Figure 2. Average market share of deposits, loans and branches of ECBGs (as %) FIGURE35 2. AVERAGE MARKET SHARE OF DEPOSITS, LOANS AND BRANCHES OF ECBGS (AS %) 30 25

40

40

35

35

30

30

25

25

20

20

15

15

10

10

35 30 25

20 15 10 5 0

20

5

5

0

0

1996

1996

1998

1998

2000

2000

2002 Deposits

2002

2004 Loans

2006

2004

2008

Branches

2006

2010

2008

2012

2010

15 10 5

2012

0

Deposits Branches Source: calculations based on data from individual ECBGs and theLoans ECB Source: calculations based on data from individual ECBGs and the ECB Note: The unweighted market shares pertain to domestic loans to private households (mortgages and/or consumer loans) and Note: The unweighted market shares pertain to domestic loans to private households (mortgages and/or consumer loans) and domestic retailofdeposits of households. The market shareofofbranches branches isisdefined as the of the local cooperative banks as a banks as a domestic retail deposits households. The market share defined as branches the branches of the local cooperative percentage of total bank officesfrom individual ECBGs and the ECB Source: calculations based on data percentage of total bank offices

Note: The unweighted market shares pertain to domestic loans to private households (mortgages and/or consumer loans) and 8.3. deposits Total loan deposit growth domestic retail of and households. The market share of branches is defined as the branches of the local cooperative banks as a percentage of total bank offices 8.3. Total loan and deposit growth Total loan and deposit growth rates shed additional light on the performance and specifics of Figure 3 and Table 2 provide visual and statistical information about total (inter)national 8.3. TotalECBGs. loan and deposit growth private sector since 1997 for ECBGs (CG ) and entire banking creditand growth to thegrowth non-financial ECBG Total loan deposit rates shed additional light on the performance and specifics of ECBGs. sectors (CGTBS). CGECBG is fairly stable and equals 8.3 per cent in every sub-period considered. Figure 3Total and Table 2 surpassed provide visual and statistical about totalthe (inter)national credit growth of CG sub-period. Hence, ECBGs are more stable loan providers to specifics CGECBG loanalsoand deposit growth rates shedinformation additional light on performance and TBS in every is generally much higher as the real economy than all other banks. The standard deviation of CG TBS and entire banking sectors(inter)national (CGTBS). toECBGs. the non-financial sector2 since 1997visual for ECBGs (CGECBG) information Figure 3 private and Table provide and statistical about total Table 2 demonstrates.

) and banking credit growth the non-financial sector sincesub-period 1997 for considered. ECBGs (CG is fairly to stable and equals 8.3private per cent in every CG alsoentire surpassed CG ECBG ECBG ECBG TABLE 2. A LOAN AND DEPOSIT GROWTH LOAN TO DEPOSIT RATIO ).VERAGE CGECBG is fairly stableareAND and equals 8.3 per cent intoevery sub-period considered. sectors (CG TBS in every sub-period. Hence, ECBGs more stable loan providers the real economy than all CG TBS Loan growth Deposit growthECBGs are more stable loan providers to CG also surpassed CG in every sub-period. Hence, ECBG TBS Loan to ratio muchin parentheses) higher as Table 2deposit demonstrates. other banks. The standard deviation CGTBS is generally (standard deviation in of parentheses) (standard deviation Period much higher as the real economy than all other banks. TBS is generally ECBGs TBS The standard ECBGs deviation TBS of CGECBGs TBS 1997-2004 8.3* (2.6 ) 5.8 (2.6) 5.7* (2.4#) 4.0 (2.9) 0.92* 1.31 Table 2 demonstrates. 2005-2011 8.3* (1.8*) 4.7 (5.3) 6.1* (1.4*) 8.1 (6.1) 1.11# 1.18 1997-2011 8.3* (3.4*) (4.0) 5.9 (1.9*) 6.1 (5.0) 1.01* 1.25 Table 2. Average loan and deposit growth and 5.3 loan to deposit ratio Source: own calculations based on figures from ECBGs, ECB and national statistics TABLE 2. A VERAGE LOAN AND DEPOSIT GROWTH AND LOAN TO DEPOSIT RATIO Note: time series are adjusted for major breaks caused by mergers and acquisitions. ECBGs stand for European cooperative banking

Period

groups and TBS stand for total banking sectors. Fifteen ECBGs from ten countries are included in the sample. An asterisk (*) and Loan growth Deposit growth hatch (#) denotes that the variable for European cooperative banking groups is statistically different from that for total banking Loan to deposit sectors at(standard the 1% and 5% significance respectively deviation in level parentheses) (standard deviation in parentheses)

ratio

ECBGs TBS ECBGs TBS ECBGs TBS Figure 3 shows a considerable CG(2.4 to4.0CG which 1.31 TBS #compared ECBG after 2006 1997-2004 8.3* (2.6 ) 5.8 (2.6)deceleration of 5.7* ) (2.9) 0.92* # but ECBGs below zero in 2009 and 2011. CGECBG 2005-2011 even dropped 8.3* (1.8*) 4.7 (5.3) 6.1* also (1.4*)slowed down 8.1remarkably, (6.1) 1.11 1.18 were still in a position to expand their credit portfolios in sub-period 2005-11, which 1997-2011 8.3* (3.4*) 5.3 (4.0) 5.9 (1.9*) 6.1 (5.0) 1.01* was 1.25 by economically difficult Thisnational can presumably be largely ascribed to a Source: owncharacterized calculations based on figures from ECBGs,times. ECB and statistics relatively good capitalization of ECBGs section 7.5 acquisitions. below), which allowed them meet the Source: own series calculations based on ECBGs, ECB and national statistics Note: time are adjusted forfigures major from breaks caused(see by mergers and ECBGs stand for to European cooperative banking credit demand of their customers for a longer period of time. Indeed, quite a few other banks neededAn groups and TBS stand for total banking sectors. Fifteen ECBGs from ten countries are included in the sample. asterisk (*) and Note: time series are adjusted for major breaks caused by mergers and acquisitions. ECBGs stand for European cooperative state support to variable survive and consequently had much less room to grant loans in different their deleveraging hatch (#) denotes that the for European cooperative banking groups is statistically from that for total banking banking groups and TBS stand for total banking sectors. Fifteen ECBGs from ten countries are included in the sample. An asterisk   11 * sectors at the 1% and 5% significance level respectively

( ) and hatch (#) denotes that the variable for European cooperative banking groups is statistically different from that for total banking sectors at the 1% and 5% significance level respectively

Figure 3 shows a considerable deceleration of CGTBS compared to CGECBG after 2006 which even dropped below zero in 2009 and 2011. CGECBG also slowed down remarkably, but ECBGs Figure 3 shows a considerable deceleration of CG compared to CGECBG after 2006 which even were still in a position to expand their creditTBSportfolios in sub-period 2005-11, which was also slowed were still in a to a dropped below zero 2009 and 2011. CGECBG times. characterized by ineconomically difficult Thisdown can remarkably, presumablybutbeECBGs largely ascribed position to expand their credit portfolios in sub-period 2005-11, waswhich characterized economically relatively good capitalization of ECBGs (see section 7.5 which below), allowedbythem to meet the difficult times. This ascribed to of a relatively good capitalization of ECBGs credit demand of can theirpresumably customersbe forlargely a longer period time. Indeed, quite a few other banks(see needed state support to survive and consequently much demand less room to grant loansfor in atheir deleveraging section 7.5 below), which allowed them to meethad the credit of their customers longer period of   time. Indeed, quite a few other banks needed state support to survive and consequently had much less 11

22 12 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

room to grant loans in their deleveraging process. Hence, loan data illustrate the relatively close ties of ECBGs to the real economy as well as their focus on retail lending. Regarding deposit one can also observe some striking last decade. Like process. Hence, loangrowth, dataillustrate illustrate therelatively relatively close tiesdevelopments ofECBGs ECBGstotoover thethe real economy wellasas process. Hence, loan data the close ties of the real economy asaswell credit growth, growth at ECBGs (DGECBG) shows a smooth development compared to that of all their focus focus ondeposit retaillending. lending. their on retail Regarding growth, onea can can also also observe some strikingfunding developments over the the last last ).deposit ECBGsgrowth, experienced fairly stableobserve growth ofsome an important source (deposits); other banks (DGTBSdeposit Regarding one striking developments over decade. Like credit growth, deposit growth ECBGsof (DG ) shows shows smooth development aa smooth development decade. Like credit growth, deposit growth atat ECBGs (DG ECBG was significantly lower than the variance DGECBG . )The large swings in DG are the variance of DG ECBG TBS TBS compared to that of all other banks (DG ). ECBGs experienced a fairly stable growth an compared to that of all other banks (DG ). ECBGs experienced a fairly stable growth TBS of DG TBS from around 4 per cent in 2005 to about ofof an remarkable. First, we can witness a sharp acceleration TBS was significantly significantly lower than the the important funding source (deposits); theprivate variance ofpresumably DGECBG than important source (deposits); the variance DG ECBG was 10 per cent funding in 2006-08. During this period, banksof needed funding for lower the strong variance of DG The large swings in DG are remarkable. First, we can witness a sharp variance of DG large swings in DG remarkable. First, we can witness a sharp TBS.. The TBS are TBS TBS expansion of their loan portfolios as well as for other investments with higher returns, which appeared from around around 44 per percent centinin2005 2005totoabout about10 10per percent centinin2006-08. 2006-08.During Duringthis this acceleration of of DG DGTBS from acceleration to be relatively risky TBS afterwards. Immediately after the initial credit crisis broke out, DGTBS decelerated period, private private banks bankspresumably presumablyneeded neededfunding fundingfor forthe thestrong strongexpansion expansionofoftheir theirloan loanportfolios portfoliosasas period, sharply, continued in the with subsequent whenwhich a deepappeared economictoto recession and banking crisis well as aswhich for other other investments with higher higheryears returns, which appeared berelatively relatively riskyafterwards. afterwards. well for investments returns, be risky unfolded in Europe. deceleratedsharply, sharply,which whichcontinued continuedinin Immediately after after the the initial initial credit credit crisis crisis broke brokeout, out,DG DGTBS Immediately TBSdecelerated

the the subsequent subsequent years yearswhen whenaadeep deepeconomic economicrecession recessionand andbanking bankingcrisis crisisunfolded unfoldedininEurope. Europe. Figure 3. Average credit and deposit growth

FFIGURE IGURE3. 3.AAVERAGE VERAGECREDIT CREDITAND ANDDEPOSIT DEPOSITGROWTH GROWTH Credit Creditgrowth growth

Deposit growth Depositgrowth

16 16

18 16 16 18

1818

14 14

16 14 14 16

1616

12 12

14 12 12 14

1414

10 10

12 10 10 12

1212

88

8 8 10 10

1010

66

66

88

88

44

44

66

66

22

22

44

44

00

00

22

22

-2 -2

-2-2

00

1997 1997

1999 1999

2001 2001 ECBGs ECBGs

2003 2003

2005 2005

2007 2007

2009 2009

2011 2011

Entire Entirebanking bankingsystem system(excluding (excludingECBGs) ECBGs)

1997 1997

1999 1999

2001 2001 ECBGs ECBGs

2003 2003

2005 2005

2007 2007

2009 2009

2011 2011

00

Entire banking system (excluding ECBGs) Entire banking system (excluding ECBGs)

Source: ECBGs, ECB and national statistics Source: ECBGs, national statistics Source: ECBGs, ECB and national statisticscooperative banking groups and total banking sectors, respectively. The credit data Note: ECBGs andECB TBSand stand for European Note: ECBGs and TBS stand for European cooperative banking groups and total sectors, respectively. The credit data refer Note: ECBGs and TBS stand for European cooperative banking groups totalbanking banking respectively. Thedata credit refer to all (inter)national credits and loans to the non-financial private sectorand of ECBGs and all sectors, other banks. The deposit referdata refer to to all all (inter)national (inter)nationalcredits creditsand andloans loanstotothe thenon-financial non-financialprivate privatesector sectorofofECBGs ECBGsand andall allother otherbanks. banks.The Thedeposit depositdata datarefer refertotoallall to all (inter)national savings and deposits of the non-financial private sector at ECBGs and other banks (inter)national (inter)nationalsavings savingsand anddeposits depositsof ofthe thenon-financial non-financialprivate privatesector sectoratatECBGs ECBGsand andother otherbanks banks

Dividing totaltotal loansloans by total deposits yields yields so-called loan to deposit (LDRs). These ratiosThese Dividing by loan deposit ratios (LDRs). Dividing total loans by total total deposits deposits yields so-called so-called loan totoratios deposit ratios (LDRs). These ratios the which banks on market indicate the extent which to banks depend on depend capital market funding. Overfunding. the entireOver timethe span and time ratios indicate indicate thetoextent extent to which banks depend on capital capital market funding. Over theentire entire time was significantly lower than LDR . ECBGs are on average span and first sub-period, LDR ECBG significantly lower thanare LDR . ECBGs are on average spansub-period, and first LDR sub-period, was LDR significantly lower than LDRTBS . ECBGs onTBS average less dependent first ECBG was TBS ECBG less dependent on and uncertain funding TBS. inin the lessvolatile dependent on volatile volatile andfunding uncertain external funding than TBS. However, However, the turbulent turbulent on and uncertain external thanexternal TBS. However, inthan the turbulent second sub-period, a increased, second sub-period, a remarkable convergence between both LDRs occurred; LDR ECBG increased, second sub-period, a remarkable convergence between both LDRs occurred; LDR ECBG remarkable convergence between both LDRs occurred; LDRECBG increased, whereas LDRTBS decreased. decreased. However, the difference remained significant atat the 5% confidence whereas LDR TBS decreased. However, the difference remained significant the 5% whereas LDR TBS However, the difference remained significant at the 5% confidence level in 2005-11. When the crisisconfidence hit, level level in in 2005-11. 2005-11. When When the the crisis crisis hit, hit, the thehigh highLDR LDRTBS provedtotobe beunsustainable unsustainableand andnecessitated necessitated TBSproved the high LDRTBS proved to be unsustainable and necessitated large scale government intervention and a large large scale scale government government intervention interventionand andaasubsequent subsequentcut cutback backinincredit creditgrowth growthon onbehalf behalfofofprivate private subsequent cut back in credit growth on behalf of private banks in Europe (CEPS, 2010). LDRs are yet banks banks in in Europe Europe (CEPS (CEPS 2010). 2010). LDRs LDRs are are yet yet another anotherexpression expressionof ofthe thedifferent differentnature natureofofECBGs ECBGs another expression of the different nature ECBGs with their predominant focus on retail banking. with predominant focus retail with their their predominant focuson on retailofbanking. banking. 8.4. and dense branch networks 8.4. Proximity and dense branch networks 8.4.Proximity Proximity and dense branch networks Local cooperative banksbanks have historically maintained extensive extensive branch networks tonetworks support strong Local Local cooperative cooperative banks have have historically historically maintained maintained extensive branch branch networkstotosupport support links to their members and communities. Although the urgency to focus on efficiency improvements in strong links to their members and communities. Although the urgency to focus on efficiency strong links to their members and communities. Although the urgency to focus on efficiency physical networksin as aphysical result of mobile banking, integrated cash management is and improvements networks as result mobile banking, contactless payments improvements in physical networks as aacontactless result of ofpayments mobileand banking, contactless payments and

integrated integrated cash cash management management isis obvious, obvious, local local cooperative cooperativebanks banksstill stilloperate operatewith withrelatively relativelydense dense networks. The average market share for branch offices even shows an upward trend since 1997. networks. The average market share for branch offices even shows an upward trend since 1997.ItItisis 23 13 approximately approximately 10 10 percentage percentage points points higher higher than than that that for for loans loans and and deposits. deposits. This This fact fact supports supports JEOD - Vol.3, Issue 1 (2014) hypothesis 1 that local cooperative banks as part of the ECBGs usually have relatively dense branch hypothesis 1 that local cooperative banks as part of the ECBGs usually have relatively densebranch


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

obvious, local cooperative banks still operate with relatively dense networks. The average market share for branch offices even shows an upward trend since 1997. It is approximately 10 percentage points higher than that for loans and deposits. This fact supports hypothesis 1 that local cooperative banks as part of the ECBGs usually have relatively dense branch networks in their home markets. On balance, the number of branches increased from around 54,000 in 1997 to more than 60,000 in 2011, whereas total bank branches decreased from 191,000 to 170,000 over this period. As a result, ECBGs have strengthened their local presence. As a further example of the heterogeneity of ECBGs, Table 1 contains national branch data of local cooperative banks and all other banks. The table reveals that the market share increase of cooperative banks was predominantly due to the expansion of branch office networks in Italy, Spain, France, Portugal and Denmark, respectively. The rise in branch market share is to a lesser extent caused by the fact that ECBGs have slimmed down the number of branch offices to a somewhat lesser extent than their competitors in respectively Switzerland, Germany and the Netherlands. Here, the strong consolidation in both the Netherlands and Germany catches the eye. On the other hand, Austrian and Finnish ECBGs lost branch market share, because they closed down branches whereas all other banks actually opened new bank offices. 8.5. Capitalization Figure 4 shows the average tier-1 ratio for ECBGs (tier-1ECBG) and national banking sectors. This ratio reflects the amount of equity relative to the risk-weighted assets of ECBGs and national banking sectors. It can be concluded that ECBGs maintain a comparatively high level of capital, e.g. the risk profile of ECBGs is more conservative than that of all other banks. There are a number of explanations for this (Oliver Wyman, 2008). Firstly, high capitalisation is connected with the strong focus of ECBGs on retail operations, for which relatively high capital requirements prevail. Secondly, ECBGs add a major portion of their profit to the capital reserves each year19. In effect, they build the core of their equity base the hard way, through increasing retained earnings. Thirdly, solid capitalisation is simply necessary for ECBGs with a view to continuity. ECBGs have less additional options to raise capital – after sizeable losses – than investor-owned banks, as most of them cannot issue shares20. Besides, this fact could mitigate the risk appetite of executives, because they know that capital cannot be easily replenished after incurring considerable losses.

19

However, some ECBGs do pay limited dividends to members.

20

This impossibility to issue shares on the stock exchange is not a feature exclusive to most non-listed ECBGs, though. The recent financial crisis has demonstrated that some (or quite a few) listed banks were unable to issue shares, when their capital vanished into thin air as a result of substantial losses and write-downs. Instead, quite a few listed banks had to be rescued by some form of state support. Moreover, without a certain profit level, investors will not be inclined to buy additional shares. Consequently, the bank in question will be unable to expand its capital buffer by issuing new shares. 24 14 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

Figure 4. Tier-1 ratio 14

14

12

12

10

10

8

8

6

6

4

4

2

2

0

2002

2003

2004

2005

2006

ECBGs

2007

2008

2009

2010

2011

2012

0

Entire banking system

Source: ECBGs, ECB, IMF and national supervisory agencies Source: ECBGs, ECB, IMF and national supervisory agencies

Figure44shows shows that that ECBGs ECBGs entered entered the crisis period strong Figure the crisis periodstarting startinginin2007 2007with witha arelatively relatively strong capitalization andeven evenstrengthened strengthened their position up toup2010 independently. In 2008 In and2008 2009,and capitalization and theircapital capital position to 2010 independently. quite a few private banks improved their battered positions with government aidgovernment or acquired fresh 2009, quite a few private banks improved theircapital battered capital positions with aid or acquired capital. 2011, tier-1 declined somewhat, whereas tier-1 continued to declined somewhat, whereas tier-1 continued to improve slightly. This capital.fresh In 2011, tier-1In ECBG TBS ECBG TBS improve slightly. This development is again a reflection of the strong focus of ECBGs on serving development is again a reflection of the strong focus of ECBGs on serving the real economy. At that time, the real economy. that time, European just had or gone a major many EuropeanAt countries just many had gone throughcountries a major recession eventhrough re-entered into arecession new one or evenfollowing re-entered into a new one following the credit crisis. Given the emphasis on retail banking, the credit crisis. Given the emphasis on retail banking, the rising number of failures in the SMEthe rising number of failuresbanks in the SME sector sector hit cooperative relatively hard. hit cooperative banks relatively hard.

8.6. Efficiency

8.6. Efficiency

If the claims regarding the business orientation and principles of ECBGs are true, If the claims regardingand the business andagainst principles of ECBGs are true, benchmarking of benchmarking of expenses revenuesorientation of ECBGs banking sector standards is somewhat expenses and sector standards is somewhat misleading. Be thatbanks as it may, misleading. Be revenues that as ofit ECBGs may, itagainst is a banking fact that ECBGs face competition from other with it is a fact that ECBGs face competition from other banks with increasingly sophisticated social agendas increasingly sophisticated social agendas and less emphasis on profit maximization. Hence, ECBGs less scale emphasis profit efficiently maximization. Hence, ECBGs must build scale and operate cost-to-income efficiently to mustand build and on operate to withstand competition. Figure 5 displays ) and the entire banking sectors in ECBG individual Over ratios for ECBGs (CIECBG ) and the countries. entire banking withstand competition. Figure 5 displays cost-to-income ratios for(CI ECBGs TBS) (CI different sub-periods, CI ratios of individual ECBGs do not deviate significantly from CI ratios sectors (CITBS) in individual countries. Over different sub-periods, CI ratios of individual ECBGs do not of the entire sector. This is of in the lineentire withbanking other sector. preliminary deviate banking significantly from CI ratios This is inand lineless withcomprehensive other preliminarystudies and (Moody’s 2003; Čihák and Hesse 2007; Oliver Wyman, 2008). Moreover, the higher costs less comprehensive studies (Moody’s, 2003; Čihák and Hesse, 2007; Oliver Wyman, 2008). Moreover, the of relatively branch networks of networks ECBGs of were more than offset higher revenues. This higher extensive costs of relatively extensive branch ECBGs were more than by offset by higher revenues. outcome suggests that they use their assets and capital base in an efficient way. This outcome suggests that they use their assets and capital base in an efficient way.

FIGURE 5. AVERAGE COST-INCOME RATIOS (2002-2011)

25 15 JEOD - Vol.3, Issue 1 (2014)

 

14


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

Figure 5. Average cost-income ratios (2002-2011) 90 80 70 60 50 40 30 20 10 0

ECBGs

Entire banking system

Source: ECBGs, ECB and national supervisory authorities Source: ECBGs, ECB and national supervisory authorities

8.7. Stability

8.7. Stability

measurethe thestability stability of entire banking systems by using Z-score. The Z-score a WeWe measure ofECBGs ECBGsand and entire banking systems by the using the Z-score. TheisZused measure of bank’sof distance default (Laeven and Levine, et al., 2007) that et is scorewidely is a widely used measure bank’stodistance to default (Laeven2009; and Mercieca Levine 2009; Mercieca monotonically associated with the bank’s probability of failure (thus bank risk is defined as the inverse of al. 2007) that is monotonically associated with the bank’s probability of failure (thus bank risk is the as Z-score). This variable is defined as: defined the inverse of the Z-score). This variable is defined as: Z-score Z-scorei = (ROAi+Ei/Ai) / σ(ROA ), i = (ROAi+Ei/Ai) / σ(ROAi), i

where: where: ROA is the Return on Assets. ROA is the Return on Assets. E/A stands for equity capital over total assets. E/A stands for equity capital over total assets. σ (ROA) is the standard deviation (volatility) of ROA calculated as a four-year rolling time 21 window21σ. (ROA) is the standard deviation (volatility) of ROA calculated as a four-year rolling time window . i denotes Europeancooperative cooperative banking or total banking systems (TBS).(TBS). i denotes European bankinggroups groups(ECBGs) (ECBGs) or total banking systems FIGURE 6. AVERAGE Z-SCORES

21

21

While in large parts of the literature the volatility of ROA is computed over the full sample period, we use the average σ(ROAi) for the period 2002-05 and a four-year rolling time window for σ(ROAi) to allow for time variation in the denominator of the Z-score starting in 2006. This approach avoids that the variation in Z-scores over time is exclusively driven by variation in the levels of capital and profitability.

While in large parts of the literature the volatility of ROA is computed over the full sample period, we use the average σ(ROAi) for the period 2002-05 and a four-year rolling time window for σ(ROAi) to allow for time variation in 26 16 the denominator of the Z-score starting in 2006. This approach avoids that the variation in Z-scores over time is JEOD Vol.3, Issue 1 (2014) exclusively driven by variation in the levels of capital and profitability.   15


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

Figure 6. Average Z-scores 120

120

100

100

80

80

60

60

40

40

20

20

0

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 ECBGs

0

Entire banking system

Source: calculations based on data from ECBGs, European Central Bank, International Monetary Fund and national supervisory Source: calculations based on data from ECBGs, European Central Bank, International Monetary Fund and national supervisory authorities. authorities Note: the figure displays the average Z-score of fifteen ECBGs in countries ten countries Z-score of the entire banking sector Note: the figure displays the average Z-score of fifteen ECBGs in ten andand the the Z-score of the entire banking sector in in these these countries countries

AA higher impliesa lower a lower probability of insolvency. 6 shows that Z-score the average higherZ-score Z-score implies probability of insolvency. Figure 6 Figure shows that the average for Z-score for ECBGs (Zalways been than muchthat higher than that of total(Zbanking sectors (Z ECBG) has TBS). beenalways much higher of total banking sectors ). This finding is in ECBGs (ZECBG) has TBS This line finding is in line with earlier studies (Čihák and Hesse 2007). Formal tests confirm that Z ECBG with earlier studies (Čihák and Hesse, 2007). Formal tests confirm that ZECBG is always significantly is always higher than ZTBSlevel. at the levelthe . One can also observe that the 1% confidence One1% canconfidence also observe that stability of ECBGs was negatihighersignificantly than ZTBS at the stability of ECBGs was negatively impacted by the financial turbulences after 2007. ZECBG dropped vely impacted by the financial turbulences after 2007. ZECBG dropped from almost 120 in 2007 to less than from almost 120 in 2007 to less than 60 in 2008, but remained well above ZTBS. Entire banking 60 in 2008, but remained well above ZTBS. Entire banking systems were fairly unstable with a ZTBS of less systems were fairly unstable with a ZTBS of less than 20 in 2008/9. During these years, quite a few than 20 in 2008/9. years, quite a few investor-owned had to be to supported with state investor-owned banks During had to these be supported with state aid or werebanks nationalized maintain financial aid orand wereconfidence nationalizedamong to maintain stability and confidence among publicnational (CEPS, 2010). stability the financial public (CEPS 2010). In 2010 andthe 2011, banking In 2010 and 2011, national banking systems showed a fragile recovery with a slight improvement ZTBS.for . This picture does notinhold systems showed a fragile recovery with a slight improvement in ZTBS exhibited a strong recovery in This picture does not hold for ECBGs. After reaching its low in 2009, Z ECBGs. After reaching its low in 2009, ZECBG exhibited a strong ECBG recovery in the last two years, last two which points to the resilience of ECBGs. whichthepoints to years, the resilience of ECBGs. Looking componentsofof Z-score, we that findthe thatfirst thecomponent, first component, ratio of Lookingatatthe thethree three components thethe Z-score, we find “the ratio“the of equity/ equity/total assets” (E/A), is systematically higher for ECBGs (E/A , see Figure 7). This ECBG total assets” (E/A), is systematically higher for ECBGs (E/AECBG, see Figure 7). This supports hypothesis supports 4 thatlarger ECBGs maintain larger capital buffers, on average. E/A remained remained fairly stable upECBG to 2007, but 4 thathypothesis ECBGs maintain capital buffers, on average. E/AECBG fairlydropped stable in up2008. to 2007, but dropped in 2008. This decline stayed well behind the decrease This decline stayed well behind the decrease of E/ATBS, which already began in 2005. of E/ATBS , which already began in 2005. Anyway, ECBGs banks entered the crisis with larger buffers, Anyway, ECBGs banks entered the crisis with larger buffers, which calls for the qualification that in good which calls for the qualification that in good times high buffers are viewed as “non-productive” as times high buffers are viewed as “non-productive” as voiced by some earlier critical analyses of ECBGs (PA voiced by some earlier critical analyses of ECBGs (PA Consulting 2003). On the contrary, the Consulting 2003). On the contrary, the pendulum has swung to the other side. Improving the resilience pendulum has swung to the other side. Improving the resilience of financial institutions by raising of financial institutions by raising capital (andof liquidity) is one of the keythe reforms that follocapital (and liquidity) requirements is one the keyrequirements reforms that followed financial crisis. occurred in 2009, partly due to capital injections by wed the financial crisis. Some improvement in E/A Some improvement in E/ATBS occurred in 2009, TBS partly due to capital injections by national national governments and deleveraging many This banks.rise Thisdid risenot didinaugurate not inaugurate a clear trendreversal, reversal, as governments and deleveraging by manyby banks. a clear trend dropped again below 5% in 2011. as E/A E/ATBS dropped again below 5% in 2011. TBS FIGURE 7. AVERAGE EQUITY TO ASSETS RATIO 27 17 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

Figure 7. Average Equity to Assets ratio 7,0%

7,0%

6,0%

6,0%

5,0%

5,0%

4,0%

4,0%

3,0%

3,0%

2,0%

2,0%

1,0%

1,0%

0,0%

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 ECBGs

0,0%

Entire banking system

Source: calculations based on data from ECBGs, European Central Bank, International Monetary Fund and national supervisory Source: calculations based on data from ECBGs, European Central Bank, International Monetary Fund and national supervisory authorities authorities Note: the figure displays the average E/A ratio of fifteen ECBGs in ten countries and the average E/A ratio of the entire banking Note: the figure displays the average E/A ratio of fifteen ECBGs in ten countries and the average E/A ratio of the entire banking sector in these countries sector in these countries

Thesecond second component , the, the return on assets (ROAECBG ), is a widely used proxy for profitability. return on assets (ROA The componentofofZECBG ZECBG ECBG), is a widely used proxy for Earlier assertions fuel the expectation that ECBGs belowhave average profitability, as they target profitability. Earlier assertions fuel the expectation thathave ECBGs below average profitability, as customer-value maximisation instead of profit maximisation and operate with higher levels of equity. they target customer-value maximisation instead of profit maximisation and operate with higher is not from the return on assets of the totalreturn banking Ourofcalculations show that ROAECBG is not statistically different from on levels equity. Our calculations show thatstatistically ROAECBGdifferent assets of total banking systems (ROA ) over the whole period and in 2002-06. This picture and in 2002-06. This picture changes in the time span 2007-11, systems (ROATBS) over the whole periodTBS was significantly higherbythan changes in the time span 2007-11, when the average ROA. ECBG than ROA ECBGs were obviously affected the when the average ROAECBG was significantly higher TBS ROA . ECBGs were obviously affected by the subsequent crises, but ROA fell less sharply 22 TBS ECBG fell less sharply than ROA . subsequent crises, but ROA ECBG TBS than ROATBS22. On the face of it, this finding does not seem to be in line with hypothesis 1 that ECGBs would have a lower ROA due to theirOlower profit requirements stemming from their member influence and focus on FIGURE 8. AVERAGE RETURN N ASSETS retail banking. However, this finding can be plausibly explained by the fact that ECBGs were to a lesser extent involved in riskier wholesale operations and expanded their credit portfolios rather moderately in the years before the crisis. Hence, ECBGs experienced fairly limited losses and write downs. Groeneveld (2011) estimates that the ECBGs share of the total losses and write-downs of all European banks during the first years of the crises was around 8 per cent, which is much smaller than their overall market share. In other words, the divergent development of ROAECBG and ROATBS confirms hypothesis 3. It should be stressed that the general situation in banking remains rather troublesome as illustrated by the sharp drop in ROA in 2011. It is generally expected that profitability in banking will definitely not return to the levels prevailing before 2007. There is general agreement that the situation in banking was not sustainable at that time. Moreover, due to their close ties with the real economy, ECBGs probably suffer more from economic slack in local economies and declining industries in the regions where they operate.

A similar pattern emerges for the return on equity (ROE), with one notable exception. In the sub-period 2002-06, ROEECBG was significantly lower than ROETBS. The opposite is true for the time span 2007-11. Over the entire period, ROEECBG and ROETBSpattern were exactly the same (7.8return per cent). volatility of ROE consistently lower in every A similar emerges for the on The equity (ROE), withECBG oneis notable exception. In thesub-period. sub-period 2002-06, 22

22

ROEECBG was significantly lower than ROETBS. The opposite is true for the time span 2007-11. Over the entire period, ROEECBG and ROETBS were exactly the same (7.8 per cent). The volatility of ROEECBG is consistently lower in every 28 18 sub-period. JEOD - Vol.3, Issue 1 (2014)   17


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

Figure 8. Average Return On Assets 0,80%

0,80%

0,70%

0,70%

0,60%

0,60%

0,50%

0,50%

0,40%

0,40%

0,30%

0,30%

0,20%

0,20%

0,10%

0,10%

0,00%

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 ECBGs

0,00%

Entire banking system

Source: calculations based datafrom fromECBGs, ECBGs,European European Central Bank, Fund andand national supervisory Source: calculations based on on data Bank,International InternationalMonetary Monetary Fund national supervisory authorities authorities Note:Note: the figure displays the the average ROA ten countries countriesand andthe theROA ROAofof entire banking sector in these the figure displays average ROAofoffifteen fifteen ECBGs ECBGs in in ten thethe entire banking sector in these countries countries

On the face of it, this finding does not seem to be in line with hypothesis 1 that ECGBs would The third component of the Z-score, the volatility of returns, is significantly lower at ECBGs in all have a lower ROA due to their lower profit requirements stemming from their member influence sub-periods, again in line withHowever, hypothesis 3. Thisfinding can be largely explained by the relatively by extensive retailthat and focus on retail banking. this can be plausibly explained the fact operations banks,involved which on in theriskier whole wholesale generate more stable profits. have credit to ECBGs were of to cooperative a lesser extent operations and However, expandedwetheir stress that the moderately standard deviation in the second time ECBGs span for ECBGs as wellfairly as forlimited the portfolios rather in thealmost years doubled before the crisis. Hence, experienced banking systems as a whole. losses and write downs. Groeneveld (2011) estimates that the ECBGs share of the total losses and write-downs of all European banks during the first years of the crises was around 8 per cent, which is much smaller than their overall market share. In other words, the divergent development of ROAECBG and ROATBS confirms hypothesis 3. It should be stressed that the general situation in banking remains rather troublesome as illustrated by the sharp drop in ROA in 2011. It is generally 9. Conclusions expected that profitability in banking will definitely not return to the levels prevailing before 2007. There is general agreement that the situation in banking was not sustainable at that time. Moreover, This article is contributing to a balanced view of ECBGs by describing their historical characteristics due to their close ties with the real economy, ECBGs probably suffer more from economic slack in and investigating empirically to what extent their recent performance is connected with their proclaimed local economies and declining industries in the regions where they operate. specific and historical features. In Z-score, this respect, may be considered as one of the first statistically-based The third component of the theitvolatility of returns, is significantly lower at ECBGs evidence on the relationship between the original specificities of ECBGs and their performance in in all sub-periods, again in line with hypothesis 3. This can be largely explained by the relatively economically and badoftimes over the past fifteen years.on Thisthe article stresses that cooperative banking extensive retail good operations cooperative banks, which whole generate more stable profits. 23 . It for is not better or worse than other modelsdeviation and not a almost panacea doubled for post-crisis banking in general However, we have to stress that banking the standard in the second time span can only be considered a viable and parallelasalternative ECBGs as well as for theasbanking systems a whole. particularly to investor-owned banks which have been in the spotlight for most of the time in recent decades. 9. Conclusions The main message is that the overall performance is still largely explicable by the original features and roots of ECBGs. These characteristics still knock about in ECBGs that eventually emerged from local

This article is contributing to a balanced view of ECBGs by describing their historical characteristics and investigating empirically to what extent their recent performance is connected GutiĂŠrrez (2008) points to the factand that there have beenfeatures. at least 20 cases of Italian cooperative banksbe with special administration with23their proclaimed specific historical In this respect, it may considered as one of procedures since 2000. the first statistically-based evidence on the relationship between the original specificities of ECBGs and their performance in economically good and bad times over the past fifteen years. This article 29 19

JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

credit cooperatives established more than a century ago. Using a new comprehensive data base, we find that this conclusion holds in recent times of economic distress as well as in those of prosperity. This also implies that we cannot reject many unfounded assertions from earlier studies and reports about the impact of the characteristics and business orientation on the financial performance of ECBGs in economic recessions and financial crises. The specific ownership structure at the local level still appears to result in a focus on retail banking (see also Oliver Wyman, 2012), a moderate risk appetite, stable operations and solid capitalization for ECBGs. Indisputably, the economic and financial performance of ECBGs has deviated from that of all other banks in different phases of the latest business cycles. From different angles, we find statistical support for most of the formulated hypotheses which were derived from unverified or poorly substantiated statements and rather partial analyses with deficient data material in previous policy documents and research papers. Table 3 summarizes the various sub-hypotheses and records whether they can be accepted or must be rejected on the basis of our empirical analysis. It should be noted that the postulations are stated in relative terms, i.e. ECBGs are compared to all other banks in the countries under review. From a policy point of view, it is important to acknowledge that the specific governance and ownership structure of ECBGs apparently leads to relatively stable institutions and a relatively stable performance (López-Puertas Lamy, 2012). This result has important implications for academics and policy makers alike, since it indicates that ignoring this ownership structure can lead to erroneous banking regulations which may eventually undermine the positive impact of the specific governance on ECBG’s stability and hence the stability of entire national financial systems. As final remarks, we have to make some qualifications. Firstly, the performance and stability of ECBGs have been assessed in relative terms, i.e. vis-à-vis all other banking groups. In absolute terms, the performance and stability of all ECBGs have deteriorated in recent years. The crises had a negative impact on ECBGs, proving that they are not immune to economic and financial shocks. Nowadays, ECBGs are confronted with increased volatility in results, a surging number of bankruptcies of local SME firms, a damaged reputation of the entire banking industry and an explosion of regulatory and compliance measures and costs. At the same time, access to external funding and accumulation of capital via retained profits have become more difficult. They cannot hide from cost reductions and efficiency improvements to remain competitive, financially solid and hence viable. In addition, they face an important internal challenge. ECBGs have to safeguard or improve internal governance structures to enable members to preserve the cooperative nature of their local banks and to determine the strategic course of the entire organization24. In short, it will always remain an open question whether ECBGs will manage other future economic and financial crises equally well or will succeed in keeping their overall course and operations closely aligned with member interests in the future.

24

It is impossible to prove whether the functioning of the cooperative governance and strategic decisions represent members’ will and interests. 30 20 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

Table 3. Assessment of formulated (sub-)hypotheses about ECBGs (1997-2011) TABLE 4. ASSESSMENT OF FORMULATED (SUB-)HYPOTHESES ABOUT ECBGS (1997-2011) Hypothesis

Assessment

Explanation

Empirical evidence

Undecided

Absolute and relative increases in members and rising domestic loan and deposit market shares are no “hard” empirical proof that local cooperative banks have a strong customer orientation. However, loan growth of entire ECBGs is less cyclical than that of all other banks. Besides, deposit growth is higher in economically difficult times, pointing to some safe haven effects. It is unknown whether the level of customer satisfaction and/or advocacy at ECBGs differs significantly from that of other banks.

None or implicit at best (rising domestic market shares and numbers of members)

Trivial

This assertion is awkward. Every bank creates jobs and contributes to economic growth due to its intermediary role. It is impossible to investigate whether ECBGs perform better in this respect.

No adequate data or indicators are available

Partly investigated and accepted

It is difficult to demonstrate a noticeable causal relationship between ECBGs and structural characteristics of banking markets. Such a causality is hard to demonstrate empirically, because it also works the other way around: the market environment influences cooperative banks. However, ECBGs do contribute to the stability of national banking systems and diversity in banking and have a focus on retail banking and serving the real economy. It has not been investigated how ECBGs influence the overall competitive environment in banking.

Z-scores point to positive impact on stability and diversity Impact on competitive conditions is not investigated

3. Impact on banking conditions for customers

Not investigated

The impact of ECBGs on (the quality of) products & services, distribution methods, innovation and price conditions in banking is not investigated (and would be very difficult).

No adequate data or indicators are available

Dual bottom line approach

Not investigated

The cooperative local banks within ECBGs are also believed to aim at contributing to a sustainable development of their members’ local communities and to be engaged in many local networks. This statement is very difficult to substantiate with empirical data.

No adequate data or indicators are available

Accepted

ECBGs have relatively dense branch networks in the domestic cooperative banking part.

Market share for branches

Austerity and efficiency in operations

Accepted

Over the entire period, ECBGs have operated with similar efficiency ratios as other banks (despite relatively expensive distribution methods in accordance with their historical roots). ECBGs were even significantly more efficient in the period 2008-11, where many other banks witnessed a larger drop in (volatile) revenues and a greater surge in (funding) costs following the initial credit crisis.

Cost to income ratios

Focus on retail banking and the real economy

Accepted

ECBGs are more stable loan providers to the real economy. They had a better loan to deposit ratio before the crisis hit. In 2010 and 2011, the impact of the economic recession is visible in declining profits (due to rising bankruptcies of SME’s).

Loan and deposit growth and loan to deposit ratio

Moderate/lower returns on assets and equity

Rejected

Despite the absence of profit targets, ROA/ROEECBG is similar to ROA/ROETBS in 2002-11, and even significantly higher in 2005-11.This is partly due to relatively large losses and write-downs at other banks.

ROA and ROE

Stable organizations

Accepted

ZECBG is significantly higher than ZTBS. The volatility of ROAECBG and ROEECBG is consistently lower. ECBGs have a lower risk appetite in booming times and less risk aversion in bad times. Deposit growth (DGECBG) exhibits a more stable pattern.

Z-scores, ROA and ROE, loan and deposit growth

High capitalization

Accepted

Most capital has been built up via retained earnings. Tier 1ECBG and E/AECBG consistently surpass Tier 1TBS and E/ATBS.

Tier 1 and E/A ratios

Moderate risk profile

Accepted

Focus is on retail banking which is a less risky activity. Besides, ECBGs did not need large scale state support in recent years.

Loan to deposit ratio

Low cost of capital

Not investigated

The high capitalization and the high deposit base could make it cheaper for ECBGs to obtain external funding.

Information is not readily available

High ratings

Not investigated

Some ECBGs are not supervised on a consolidated basis. Hence, no overall ratings exist for these ECBGs.

No overall ratings exist for many ECBGs

Customer focus, customer interests’ first, long-term relationships

Presence value 1. Impact on macro and local economy

2. Impact on banking market structures

Physical proximity

20

 

References Albert, E. (2008). Les banques populaires en France, 1878-2009; 130 ans de coopération. Paris: Eyrolles. Alexopoulos, Y., Goglio, S. (2009). Financial deregulation and economic distress: Is there a future for financial co-operatives?. Euricse Working Papers, No. 001.   21 Aschhoff, G., Hennigsen, E. (1995). The German cooperative system - Its history, structure and strength. Frankfurt a.M.: Fritz Knapp Verlag. Ayadi, R., Llewellyn, D., Schmidt, R.H., Arbak, E., de Groen, W.P. (2010). Investigating diversity in the banking sector in Europe: Key developments, performance and role of cooperative banks. Brussels: Centre for European Policy Studies. Birchall, J. (2013). Resilience in a downturn: The power of financial cooperatives. Geneva: International Labour Office, International Labour Organization. 31 21 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

Birchall, J., Hammond Ketilson, L. (2009). Resilience of the co-operative business model in times of crisis. Geneva: International Labour Office, Sustainable Enterprise Programme, International Labour Organization. Bley, A. (2012). Lending stabilizer – German cooperative banks during the financial crisis, Paper presented at the conference “Cooperative responses to global challenges”, March 21-23 in Berlin, Germany. Bonin, H. (2012). French cooperative banks across crises in the 1930s and in 2007-2012. In Mooij, J., Boonstra, W.W. (Eds.). Raiffeisen’s footprint: The cooperative way of banking. Amsterdam, The Netherlands: VU University Press, pp. 19-36. Boot, A.W.A. (2000). Relationship banking: What do we know?. Journal of Financial Intermediation vol. 9, pp. 7-25. http://dx.doi.org/10.1006/jfin.2000.0282. Bosseno, C. (1994). Crédit agricole, un siècle au présent, 1894-1994, Tome 1: des origines aux années 1950. Paris: Hervas. Brazda, J. (Ed.) (2001). 150 Jahre Volksbanken in Österreich. Schulze-Delitzsch Schriftenreihe Vol. 23. CEPS (2010). Bank state aid in the financial crisis: Fragmentation or level playing field?. Brussels: CEPS Task Force Report. Čihák, M., Hesse, H. (2007). Cooperative banks and financial stability. IMF Working Papers, No. WP/07/02, SSRN, International Monetary Fund, Washington. D.C. Deloitte (2012). Funding the future: Emerging strategies in cooperative financing and capitalization. Report commissioned for the 2012 International Summit of Cooperatives, held in Quebec City from October 8-11 in 2012. Desrochers, M., Fischer, K.P. (2005). The power of networks: Integration and financial cooperative performance. CIRPÉE, Working Paper 05-14, Quebec, Canada. Di Salvo, R. (2003). The governance of mutual and cooperative bank systems in Europe. Cooperative Studies. Edizioni del Credito Cooperativo. EACB (2005). Cooperative banks in Europe: Values and practices to promote development. Brussels. EACB (2007). 60 million members in co-operative banks: What does it mean?. Brussels. EACB (2010). European cooperative banks in the financial and economic turmoil: First assessments. Research Paper. Brussels. Economist, the (2010), Europe’s cooperative banks: Mutual respect, 21st of January, 2010. Ensor, B. (2012). Customer advocacy 2011: How customers rate European banks. Forrester Report for eBusiness & Channel Strategy Professionals. Fonteyne, W. (2007). Cooperative banks in Europe – Policy issues. IMF Working Papers, WP 07/159, International Monetary Fund, Washington. D.C. Goglio, S., Alexopoulos, Y. (Eds.) (2012). Financial cooperatives and local development. Routledge Studies in Development Economies. Groeneveld, J.M. (2011). Morality and integrity in cooperative banking. Ethical Perspectives Vol. 18, No. 4, pp. 515-540. Groeneveld, J.M., Llewellyn, D. T. (2012). Corporate governance in cooperative banks. In Mooij, J., Boonstra, W.W. (Eds.). Raiffeisen’s footprint: The cooperative way of banking. Amsterdam, The Netherlands: VU University Press, pp. 19-36. Guinnane, T.W. (2001). Co-operatives as information machines: German rural credit co-operatives, 1883-1914. The Journal of Economic History Vol. 61(2), pp. 366-389. http://dx.doi.org/10.1017/ S0022050701028042. Gutiérrez, E. (2008). The reform of Italian cooperative banks: Discussion proposals. IMF Working Paper, WP/08/84, International Monetary Fund, Washington. D.C. 32 22 JEOD - Vol.3, Issue 1 (2014)


Features, Facts and Figures of European Cooperative Banking Groups over Recent Business Cycles Groeneveld, H.

Kalmi, P. (2007). The disappearance of co-operatives from economics textbooks. Cambridge Journal of Economics 31, pp. 625-47. http://dx.doi.org/10.1093/cje/bem005. Kodres, L., Narain, A. (2010). Redesigning the contours of the future financial system. IMF Staff Position Note, August 16, SPN/10/10. Körnert, J. (2012). Swedish cooperative banking in the 1990s: A decade of crisis and transition. In Mooij, J., Boonstra, W.W. (Eds.). Raiffeisen’s footprint: The cooperative way of banking. Amsterdam, The Netherlands: VU University Press, pp. 217-230. Laeven, L., Levine, R. (2009). Bank governance, regulation and risk taking. Journal of Financial Economics 93, pp. 259-275. http://dx.doi.org/10.1016/j.jfineco.2008.09.003. Llewellyn, D.T. (2012). UK building societies: The strengths of mutuality. In Mooij, J., Boonstra, W.W. (Eds.). Raiffeisen’s footprint: The cooperative way of banking. Amsterdam, The Netherlands: VU University Press, pp. 231-246. López-Puertas Lamy, M. (2012). Commercial banks versus stakeholder banks: Same business, same risks, same rules?. Working Paper prepared for EACB award for young researchers, December. McKinsey (2012). McKinsey on cooperatives. Autumn 2012, McKinsey & Company. Mercieca, S., Schaeck, K., Wolfe, S. (2007). Small European banks: Benefits from diversification?. Journal of Banking and Finance 31, pp. 1975-98. http://dx.doi.org/10.1016/j.jbankfin.2007.01.004. Michie, J. (2010). Promoting corporate diversity in the financial services sector. Paper Kellogg College, University of Oxford. Moody’s (2003). European cooperative banks: Moving beyond issues of costs and efficiency. Moody’s Investors service, Special Comment by A. Cunningham, London. Mooij, J. (Ed) (2009). Rabobank matters. Utrecht, The Netherlands: Rabobank. Mooij, J., Boonstra, W.W. (Eds.) (2012). Raiffeisen’s footprint: The cooperative way of banking. Amsterdam, The Netherlands: VU University Press. Oliver Wyman (2008). Co-operative bank, customer champion. Financial Services, United Kingdom. Oliver Wyman (2012). The outlook for co-operative banking in Europe 2012: Banking on values, building on agility. Financial Services, United Kingdom. PA Consulting Group (2003). Mutually assured destruction?. London: PA Consulting. Stefancic, M., Kathiziotis, N. (2011). An evaluation of Italian banks in the period of financial distress. International Business & Economics Research Journal Vol. 10(10), pp. 103-113. United Nations (2011). Cooperatives in social development and implementation of the International Year of Cooperatives. Agenda item A/66/136, New York. Werner, W. (2005). Auf der Straße des Erfolges – Zur Geschichte der österreichischen Raiffeisenbewegung von kleinen Ortsgenossenschaften zu international tätigen Netzwerken. Kooperation und Wettbewerb Vol. 4. München: Mering.

33 23 JEOD - Vol.3, Issue 1 (2014)


AT T R I B U T I O N 3 . 0

You are free to share and to remix, you must attribute the work

AUTHOR

Publication date: 17 2014 | Vol.3, Issue34-55 1 (2014) 35-55 ??? |June Vol.3, Issue 1 (2014)

GIOVANNI FERRI LUMSA University Rome g.ferri@lumsa.it

Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes

PANU KALMI University of Vaasa, Finland panu.kalmi@uwasa.fi EEVA KEROLA Dept. of Economics, AALTO University School of Business, Helsinki eeva.kerola@aalto.fi

ABSTRACT The Great Crisis that started in 2007 deeply affected banks throughout Europe. Using the assessments of the two global agencies that publish ratings of the financial strength of individual banks, we study whether the crisis hit European banks differently depending on their organizational structure. We analyze the changes in the ratings during the crisis and how they are related to bank ownership. Our results lend support to the hypothesis that stakeholder banks, especially cooperative banks, were downgraded to a lesser degree than shareholder banks. However, the results differ somewhat across the rating agencies. We also discuss the sources of ratings disagreements in the paper. Our paper is among the first presenting statistically based evidence on the relative merits of different organizational structures during the recent financial and economic crisis.

KEY-WORDS EUROPEAN BANKING; SHAREHOLDER BANKS; COOPERATIVE BANKS; SAVINGS BANKS; PERFORMANCE; CREDIT RATING AGENCIES

Acknowledgements The authors would like to thank the Academy of Finland (project numbers 120234 and 12140952), OP-Pohjola Research Foundation, and Yrjรถ Jahnsson Foundation for generous financial support. Earlier versions of the paper have been presented at the 2nd Euricse Conference of Cooperative Finance and Sustainable Development at the University of Trento in June 2011, ICA Research Conference in Mikkeli in August 2011, IGT-ICCS Conference in Vienna in September 2012, International Conference on the Global Financial Crisis and European Markets and Institutions in Southampton in April 2013, and in departmental research seminars at the University of Edinburgh Business School and at University Wisconsin-Madison. The authors warmly thank the audience of these events for useful feedback. JEL Classification: G2; G3; L2; P1 | DOI: http://dx.doi.org/10.5947/jeod.2014.003

35 34 JEOD - Vol.3, Issue 1 (2014)


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

1. Introduction Stakeholder-value oriented banks (cooperative and savings banks) are prevalent in European banking markets. In many countries, stakeholder-oriented banks have a larger market share than profit-maximizing banks. Although these banks are common in many markets, their ownership structure prevents takeovers and has often been viewed as an impediment to economic efficiency. The financial and economic crisis that started in 2007 has instigated a lively debate about the merits of different ownership structures within the banking industry. However, it is yet unclear what general lessons there are on the comparative performance of different types of banking organizations, as there are successes and failures among both stakeholder-value oriented and profit-maximizing banks. The novel question that we tackle here is examining the rating differences across different ownership structures and whether and to what extent the main credit ratings agencies changed their views across ownership structures in response to the crisis. The evidence comes from the bank financial strength ratings from two of the three major rating agencies (Moody’s and Fitch). This type of data is less susceptible to certain types of measurement error than standard financial statements, and it is more readily available as the crisis unfolds. Bank financial strength ratings are well suited for analyzing bank strength because they are supposedly not confounded by the likelihood of government support to financial organizations. Ratings are also relevant because they are a major determinant of the cost of raising funds for banks. However, there are certain concerns regarding ratings (rater subjectivity, potential bias towards profit-maximization objectives) that we discuss in the paper as well. In addition, in this paper, we build upon previous work (Ferri et al., 2013) in making a clear differentiation between different types of stakeholder-value oriented banks. We divide the stakeholder-value banks into cooperative groups, individual cooperative banks, private savings banks, and public savings banks, where the term “public” refers to the fact that these banks are often sponsored by public sector entities such as a region, county or municipality (Ayadi et al., 2009; Bülbül et al., 2013). Our findings suggest that ownership structure does influence the development of ratings and particularly that cooperative groups were downgraded less during the crisis than other banks. However, clear differences also appear across the rating agencies. Therefore, we proceed to analyze the determinants of ratings disagreements. We find evidence that one of the agencies gives systematically higher ratings to certain types of stakeholder banks than the other, while the evidence that random disagreements depend on ownership is more limited. This study may be the first to address the question of how rating disagreements are related to the ownership structures of banks. The rest of the paper is structured as follows. In the next section, we summarize the debate on the relative performance of shareholder versus stakeholder banks and discuss how we can use the explicit assessments of the credit rating agencies to evaluate the relative performance of these bank categories. In section 3, we present our empirical approach and data. In section 4, we present descriptive statistics and explain how ratings have changed over time. In section 5, we present the regressions concerning the rating changes and in section 6, we present the analysis concerning rating disagreements. Section 7 concludes the paper.

36 35 JEOD - Vol.3, Issue 1 (2014)


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

2. The performance of stakeholder-value oriented banks in the crisis The viability of different organizational forms is most clearly tested in times of crises, when the mortality of banking organizations far exceeds that of normal times. There are several reasons why ownership structure could be a determinant of performance in a crisis situation. First of all, profit-maximizing banks face a problem that has variously been called “risk shifting” or “asset substitution”: when management acts as a faithful agent of the owners (shareholders), they have incentives to choose investment policies that are excessively risky from the perspective of the depositors (John et al., 1991; Hermalin and Wallace, 1994; Esty, 1997). Customer-owned or non-profit financial institutions internalize the interests of depositors and thus do not have incentives for excessive risk taking; this should be counted as an advantage in a crisis situation (e.g., Rasmusen, 1988; Alexopoulos and Goglio, 2009; Coco and Ferri, 2010). An additional advantage of stakeholder-value oriented institutions is that they typically have unusually loyal customers, and a large part of their liabilities is composed of deposits (e.g., Amess, 2002). This feature is useful in a crisis situation, where alternative sources of short-term funding often become unavailable. However, a clear disadvantage of stakeholder-value oriented banks in crises is their inherent difficulty in raising equity capital, a handicap that has sometimes led to conversions to shareholder ownership (Fonteyne, 2007). There is evidence, for instance from the US during the Great Depression of the 1930s or the Savings & Loan Crisis of the 1980s, that stakeholder-value oriented organizations may show greater resilience in crises (Rasmusen, 1988; Hermalin and Wallace, 1994). However, there are also examples where organizational diversity has been reduced in a crisis due to the collapse of stakeholder-value oriented banks; examples include Swedish cooperative banks and Finnish savings banks in the Nordic banking crises of the early 1990s (Kalmi, 2012; Körnert, 2012). Similarly, the recent crisis that started in 2007 has provided a fairly mixed response regarding the performance of stakeholder banking vis-à-vis profit-maximizing banks. While there is not yet much academic literature, journalists have been quick to comment on the issue. Thus, on September 2, 2009, the Financial Times ran a long article entitled “Mutual Suspect”, expressing doubts regarding the ability of cooperative banks, especially building societies, to address the crisis. Almost as a response to that article, The Economist published an article on January 21, 2010 entitled “Mutual Respect”, highlighting the good performance of European cooperative banks in the crisis. In addition, Birchall and Hammond Ketilson (2009), Birchall (2013) and several articles by Mooij and Boonstra (2012) have stressed the resilience of cooperative banks during the crisis. Other stakeholder-value oriented organizations, especially the German central public savings banks (Landesbanken) and the Spanish private savings banks (cajas), which have been subsequently forced to convert into joint-stock ownership, have been the focus of much critical discussion (e.g., Hau and Thum, 2009; IMF, 2012). Because it is always possible to find examples of both success stories and failures among all types of organizations in a crisis, it is essential to move from case examples to statistical analysis. In this paper, we propose that ratings data may provide interesting evidence on the comparative performance of different organizational types. There are several advantages in using bank ratings to compare bank performance, which we may best appreciate as contrasted to the main alternative, financial statement analysis. When analyzing data coming from a large number of observations, the analyst often does not have all of the relevant information that affects the balance sheet. For instance, she may have difficulty taking into account the impact of major acquisitions or divestments. Additionally, banks may strategically manipulate the timing of reported writedowns and inventory valuations. Because analysis is typically based on annual data, income statements also reflect historical development with a lag. 37 36 JEOD - Vol.3, Issue 1 (2014)


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

Bank ratings are based on publicly available information, such as financial statements, and also on a host of additional information, some of which is collected from non-public sources. In essence, ratings consist of a large number of case studies on performance. In this way, ratings embody much more than reported financial information alone. For this reason, from an information-theoretic perspective, ratings should be preferred over financial statements as a performance measure. Furthermore, they reflect the current, rather than a historical, situation1. The best known ratings for banks are those issued by the three largest credit rating agencies (CRAs: S&P’s, Moody’s and Fitch). Interpreting bank ratings requires some attention. As argued convincingly by Alessandri and Haldane (2009), the recent crisis has shown the large extent to which the survival of banks depends on implicit or explicit government guarantees providing their bailouts. Thus, even the ratings issued on banks by the CRAs may incorporate the extent of government support. Because the eagerness of the government to provide support is supposedly skewed in favor of the large financial institutions – as some form of Too Big (or Too Interconnected) to Fail has proven to hold (with the exception of Lehman Brothers) – it suggests that one should use a measure that separates state support to tell the true underlying strength of each bank. Being aware of this problem, Moody’s and Fitch issue bank financial strength ratings along with bank overall credit ratings. Since 1995, a new rating scale, named bank financial strength ratings (BFSRs), has been published by Moody’s to grade the financial strength of a bank as a stand-alone concern, thus disregarding any external support. BFSRs are published in addition to overall bank deposit credit ratings (BDCRs), where BDCRs take into account not only the banks’ own financial performance but also other institutional factors such as the macroeconomic environment, the quality of supervision, and the implicit or explicit deposit insurance setup2. To be sure, Poon et al. (1999) already showed that the determinants of BFSRs do differ from those of BDCRs. Using a logit regression, they find that BFSRs may be correctly classified using bank-specific accounting and financial data alike (in decreasing order of importance): loan provision ratios, the dimensions of risk, and profitability. In addition, while sovereign ratings do not figure as a significant determinant, BDCRs help to correctly classify almost 70 per cent of the BFSRs. Later on, Fitch also started publishing its individual ratings (IR), assessing the likelihood that a bank will survive as a going concern without any outside assistance, and support ratings that indicate the likelihood of receiving such assistance in the case of need. Caporale et al. (2012) studied the determinants of Fitch’s IR and found that, at least prior to the banking crisis, they were strongly related to country effects, bank capitalization, bank size, profitability and non-performing loans. We will take these factors into account in our analysis, in which we focus on Moody’s BFSRs and on Fitch’s bank individual ratings3. Unfortunately, this positive view of the rating data may not be the entire story, as the use of ratings may lead to new problems. As with any subjective data, subjectivity raises fears of measurement error due to idiosyncratic errors potentially introduced by the fact that different individuals rate different companies. The rating firms use algorithms to standardize the ratings; however, they also retain some discretion in assigning ratings, and they do not make the algorithms public. The discretion can be justified by the fact that not all relevant information may be easily captured by standard algorithms and the secrecy by the need to protect the source of business income. Depending on the point of view chosen, this residual subjectivity is both a strength and a cause for concern. Recently, the accuracy of credit ratings has been under severe

1

However, agencies may be somewhat slow in revising ratings, which means that in a situation where the economic environment deteriorates rapidly (in a crisis situation), ratings tend to be excessively high. Although the rating agencies aimto incorporate forward-looking indicators into ratings, it may still be that they mostly reflect past development.

2

Following this intuition, Ferri and Liu (2007) estimate large potential government liabilities behind banking systems.

3

S & P does not have this measure available. 38 37 JEOD - Vol.3, Issue 1 (2014)


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

criticism (e.g., Blöchlinger et al., 2012; Hau et al., 2013). For instance, analyzing bank credit ratings, Hau et al. (2013) argue that larger banks obtain systematically better ratings than smaller banks with similar economic fundamentals, thereby exacerbating the “too-big-to-fail” problem. However, this criticism has been targeted to bond issuer ratings rather than BFSRs; furthermore, it is not clear whether this type of measurement error is related in any systematic way to ownership structure. There may be even stronger concerns that ratings may introduce measurement error that is correlated with ownership structures. Among the factors that raters take into account when evaluating bank strength is the governance structure. There is a danger that the governance structures of stakeholder-value oriented banks are viewed as being inferior simply because they are different from those of profit-maximizing banks. Another case in point is profitability. Stakeholder-value oriented banks may record somewhat lower profitability than profit-maximizing banks, simply because the former do not attempt to maximize profits by all means but have other objectives as well. In this case, stakeholder-value oriented banks may receive a lower score not for efficiency reasons but simply because they have a different objective function. There is some previous literature related to the ratings of bond issues of cooperative and commercial banks. Fischer and Mahfoudhi (2002) study whether the rating agencies take into account the systematically lower risk of bonds issued by cooperative banks by comparing ratings to the market prices of the bonds. They find that, controlling for a variety of characteristics for bonds and issuing banks, the market charges a lower spread for bonds issued by cooperative banks, suggesting that the rating agencies ignore the lower default risks of the bonds issued by cooperative banks. Flageole and Roy (2005) find that different factors contribute to the ratings of the bonds of cooperative banks than to those of commercial banks, suggesting (somewhat in contradiction to Fischer and Mahfoudhi’s findings) that the rating process differs. Again, these analyses relate to bond issues, and their relevance to bank financial strength ratings may be limited. We are not aware of much literature comparing the relationship between ownership structure and organizational performance using ratings data. The only exception we are aware of is the study by Iannotta et al. (2013), which examines differences in ratings between government-owned banks and private commercial banks. These authors find that government-owned banks have systematically lower bank ratings and higher issuer ratings than banks in private ownership, and they interpret this to be a result of the government protection of government-owned banks. However, they do not divide private banks further into cooperative, savings and shareholder banks; neither do they examine how the ratings change over different periods. In this paper, we are interested in three questions: first, controlling for a number of factors – including various bank-specific economic variables and country controls – have the rating agencies treated stakeholder-value oriented banks differently than profit-maximizing ones since the beginning of the crisis (i.e., comparing the end of 2011 vis-à-vis that of 2006)? Second, was there variance in the change across finer breakdowns of stakeholder banks (cooperatives vs. savings; tightly federated vs. non-federated cooperatives and public vs. private savings)? Third, do the different rating agencies treat banks in different ownership classes in a similar way, or do their opinions about bank strength systematically differ? With this study, we aim to compare the rating performances of banks with different ownership forms by concentrating on the Fitch Bank Individual Rating (Fitch IR) and Moody’s Bank Financial Strength (Moody’s BFS). Both ratings measure the ability of banks or bank groups to survive without outside support. In their ratings, both agencies take into account issues such as financial fundamentals, branch names, risk positions, bank management and overall operating environment4.

4

See Appendix for more information on Fitch Individual Ratings and Moody’s Financial Strength Ratings. 39 38 JEOD - Vol.3, Issue 1 (2014)


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

3. Data and empirical strategy We are interested in rating changes during the crisis and how this is related to ownership. The time period we use is from December 2006 to December 2011. We convert the ratings into a numerical scale to use them in a regression analysis. One fundamental difficulty in transforming ratings into a numerical scale is that, strictly speaking, they are ordinal, rather than cardinal, measures of credit risk. A simple linear transformation assuming equivalent differences between the ratings notches all across the distribution may be too simple, as it can be argued that the distances between the notches depend on the location within the ratings scale. Related to bond issuer ratings, it is known that the default probabilities increase exponentially when the bond rating deteriorates (e.g., Fitch, 2012), and the same applies to interest rate spreads (e.g., Langohr and Langohr, 2009). With issuer ratings, one can estimate the distances between the notches using default probabilities or interest rate spreads. Unfortunately, no comparable measure exists for BFS ratings. Nevertheless, we can utilize the known pattern that distances in the bottom of the scale are larger than the top the scale by taking the natural logarithm of our numerical rating scale The (lograting). The of theinscale by of taking the natural logarithm of our numerical rating scale (lograting). dependent of the scale by taking the natural logarithm of our numerical rating scale (lograting). The dependent variable Δ lograting is defined as lograting – lograting . When using the log scale, differences – lograting dependent variable D lograting is defined as lograting t t – 1 t – 1. When using the log scale, differences t Δ lograting isscale defined as lograting the logbetween scale, differences variable t – lograting tequivalent – 1. When using at the bottom of the become more important than differences notches at at the bottom ofofthe scale become more important thanthan equivalent differences between notchesnotches at the top at the bottom the scale become more important equivalent differences between at the top of the scale. of scale. thethe topWe of the scale.two types of regressions: a parsimonious regression and an extended regression. present We present twotwo types of regressions: a parsimonious regression and anand extended regression. In the We present types of regressions: a for parsimonious regression ancountry extended In the parsimonious regression, we control ownership categories, the of regression. origin, the In the parsimonious we control for ownership of origin, parsimonious regression, we control for ownership categories, thecategories, country origin, the (similarly difference in difference in the (log)regression, ratings of the sovereign debt of the home countryofofthe thecountry bank to the difference ratings of the sovereign debt ofof thethe home country of to thethe bank (similarly tot –the (log) ratingsin ofthe the(log) sovereign debt ofisthe home as country bank (similarly dependent variable, rating dependent variable, this variable defined log(sovereign rating t) – log(sovereign 1), 5 t) – log(sovereign rating t – 1), dependent variable, this variable is defined as log(sovereign rating and variable level ofisrating in the rating previous (lograting . In the extended regression, we )t – period log(sovereign rating tt-1– 1)), and level of rating (in logs) in the this defined(inas logs) log(sovereign )5. In theofextended regression,latter we and of rating (in logs) in the previous period (lograting t-1purpose also level control for various bank-level economic variables. The introducing 5 ) . In the extended regression, we also control for various bank-level the economic previous period (lograting t-1 bank-level economic variables. The purpose of introducing the latter also control for various variables is to control for differences in the banks’ business strategies. Following the previous variables. The purpose of introducing the latter variables is business to controlstrategies. for differences in the banks’ business variables is to control for differences in the the loan previous literature (especially Caporale et al. 2012), webanks’ include equity per assets, Following log of assets, loss strategies. Following the previous literature (especially Caporale et al., 2012), we include equity per assets, literature (especially Caporale et al. 2012), we include equity per assets, log of assets, loan loss reserves per total loans, profitability (ROA), and deposits per assets. Although many of these reserves per total profitability (ROA), and per assets. of many these log of assets, loan lossloans, reserves per loans, profitability (ROA), andcoefficients depositsAlthough peras assets. Although variables might be collinear, ourtotal primary interest liesdeposits not in their such,many but rather we 6 coefficients as such, but rather we variables might be collinear, our primary interest lies not in their Thecoefficients first type ofasregression is: we arethese interested in might how they influence theprimary ownership coefficients of variables be collinear, our interest lies not in6 .their such, but rather are interested in how they influence the ownership coefficients . The first type of regression is: are interested in how they influence the ownership coefficients6. The first type of regression is: Δ lograting i; t, t-1 = α + β ∗ ownershipi + χ ∗ countryi + δ ∗ lograting i, t-1 + φ ∗ Δ log(sovereign rating) i; t, t-1 + εi, t Δ lograting i; t, t-1 = α + β ∗ ownershipi + χ ∗ countryi + δ ∗ lograting i, t-1 + φ ∗ Δ log(sovereign rating) i; t, t-1 + εi, t

(1) (1)

where the dependent variable is the difference in lograting between years t and t-1, α is the where thethe dependent variable is theisdifference in lograting betweenbetween years t and t-1,t aand is the where the lograting years t-1,intercept, α is the β is thedependent vector of variable coefficients fordifference ownershipincategories, χ is the vector of coefficients for intercept, bintercept, is the vector of coefficients for ownership categories, c is the vector of coefficients for the country β is the vector of coefficients for ownership categories, χ is the vector of coefficients for the country dummies, δ is the coefficient of the lagged rating, φ is the coefficient for the change in 7 the country dummies, of the flagged rating, isfor thethe coefficient for the change in dummies, dratings, is the coefficient of coefficient the lagged the coefficient change in sovereign ratings, is the the error term,rating, which isis given by theφ random effects model . Ownership sovereign and δε is 7 7 ε is the error term, which is given by the random effects model . Ownership sovereign ratings, and . Ownership remains and ecountry is the error term, which is given by the random effects model remains constant for all banks throughout the period. Next,and wecountry augment this and country so remains constant for all banks throughout the period. Next, we augment this specification that (2) becomes constant for all banks throughout the period. Next, we augment this specification so that (2) becomes specification so that (2) becomes Δ lograting i; t, t-1 = α + β ∗ ownershipi, +  χ ∗ countryi + δ ∗ lograting i, t-1 + φ ∗ Δ log(sovereign rating) i; t, t-1 + γ controlsi, Δ lograting i; t, t-1 = α + β ∗ ownershipi, +  χ ∗ countryi + δ ∗ lograting i, t-1 + φ ∗ Δ log(sovereign rating) i; t, t-1 + γ controls (2)i, t-1 + εi, t (2) t-1 + εi, t

is the of vector of coefficients for bank-level control all variables, allone lagged where where g is theγvector coefficients for bank-level control variables, lagged by year.by one year. where the vector bank-level control allrating laggedasby one year. We runγ is different setsofofcoefficients regressions for using either the Fitch variables, or Moody’s our dependent We run different sets of regressions using either the Fitch or Moody’s rating as our dependent variable. Using data from two different sources also allows us to analyze the ratings disagreements. variable. Using data from two different sources also allows us to analyze the ratings disagreements. 5We analyze ratings disagreements in the following way: because the scales used by Fitch and For sovereign ratings, we use the Fitch sovereign ratings as an explanatory variable for both Fitch’s and Moody’s bank rating We analyze disagreements in indicating the way: the scales used bynotches, Fitch and changes. However, results are very that the choice the sovereign ratings variable does not influence Moody’s are ratings not the necessarily the similar, same, andfollowing in any case, theofbecause Moody’s rating has more we Moody’s are not necessarily the same, and in any case, the Moody’s rating has more notches, we results. need first to convert the data into a comparable scale. We convert the data by transforming both 6need first to convert the data into a comparable scale. We convert the data by transforming both The results of Iannotta et al.new (2007)standardized and Ferri et al. (2013) on European banks indicate thatvalue the values variables numerical scales into variables that have a mean ofof0these andcontrol a standard numerical intowe new standardized variables a of mean value of 0 biased and a construct standard vary substantially depending on ownership structures, so that thethat failure to control for them might yield deviation ofscales 1. Then, take the difference between the have values this new variable. Wecoefficients. 7deviation of 1. Then, we take the difference between the values of this new variable. We construct A fixed model issystematic not feasible because there are no changes explanatory variable, ownership.We alsogiven estimated two neweffects variables, disagreement, which inisour thekeydifference between the ratings by two systematic disagreement, which ischoice thegives difference between givenand by thenew pooled OLS models, and our results are not dependent on theFitch of the model. Fitch andvariables, Moody’s (positive values indicating that a better rating the thanratings Moody’s), Fitch anddisagreement, Moody’s (positive values indicating that gives a better rating than Moody’s), and random which is the absolute value ofFitch said difference. random disagreement, which is the absolute value of said difference. We use an unbalanced panel for the years 39 2006 – 2011. Using the unbalanced panel, we can 40 unbalanced panel for theJEOD years 2006 – 2011. Using thethe unbalanced panel, we can - Vol.3, Issue make We sureuse thatanwe do not exclude observations that1 (2014) disappear during period due to business make sure that we do not exclude observations that disappear during the period due to business failure. Given our interest in focusing especially on distressed banks, it makes sense to use an failure. Given our interest in focusing especially on distressed banks, it makes sense to use an


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

We run different sets of regressions using either the Fitch or Moody’s rating as our dependent variable. Using data from two different sources also allows us to analyze the ratings disagreements. We analyze ratings disagreements in the following way: because the scales used by Fitch and Moody’s are not necessarily the same, and in any case, the Moody’s rating has more notches, we need first to convert the data into a comparable scale. We convert the data by transforming both numerical scales into new standardized variables that have a mean value of 0 and a standard deviation of 1. Then, we take the difference between the values of this new variable. We construct two new variables, systematic disagreement, which is the difference between the ratings given by Fitch and Moody’s (positive values indicating that Fitch gives a better rating than Moody’s), and random disagreement, which is the absolute value of said difference. We use an unbalanced panel for the years 2006-2011. Using the unbalanced panel, we can make sure that we do not exclude observations that disappear during the period due to business failure. Given our interest in focusing especially on distressed banks, it makes sense to use an unbalanced rather than a balanced panel. In total, there are 218 banks in our dataset, of which 152 are in the Fitch dataset, 193 in the Moody’s dataset, and 127 in both datasets. The source for all data was Bankscope, from which we derived the ownership data, ratings data, and financial data. The main constraint to the sample size was the ratings data: we collected ratings data for as many banks we could find. However, we included only one observation per banking group (i.e., excluded subsidiaries). We also included only bank groups headquartered in Europe (i.e., no subsidiaries of non-European banks included). Table 1 gives the distribution of observations according to country and ownership. The ownership categories are similar to those in Ferri et al. (2013), which were in turn influenced by Desrochers and Fischer (2005) on cooperative banks and Gardener et al. (1997) on savings banks8. Cooperative banks have been classified either as cooperative groups (in which case the same rating applies to all banks within the group) or independent cooperatives (mostly building societies or Italian Banche Popolari). Savings banks have been regarded as either in private ownership if foundation-owned (mostly in Spain and Norway, and, in Moody’s data, also Italy) or public (this category consists mostly of German Landesbanken). As is evident from Table 1, in both datasets, there is a slight majority of observations from shareholder banks rather than stakeholder banks; and for stakeholder banks, there are somewhat more savings banks than cooperative banks. All in all, moving from general to specific, we use three breakdowns: i) a “mission-based” breakdown of shareholder (profit maximizing commercial banks) banks versus stakeholder banks (catering not just to their shareholders; the other four categories are grouped); ii) an “ownership-based” categorization of the stakeholder banks differentiating cooperative banks from savings banks; iii) an “organizational/ownershipbased” breakdown of the stakeholder banks in which cooperative banks are further subdivided into groups versus independent banks (called solo cooperatives in the tables) and savings banks are also split into private versus public.

8

The ownership assignments are based on ready-made classifications from Bankscope, but some of these were corrected and the classifications were developed further. See Ferri et al. (2013) for more details. 41 40 JEOD - Vol.3, Issue 1 (2014)


was Bankscope, from which we derived the ownership data, ratings data, and financial data. The main constraint to the sample size was the ratings data: we collected ratings data for as many banks we per banking banking group group (i.e., (i.e., excluded excluded we could could find. find. However, However, we we included included only only one one observation observation per subsidiaries). We also included only bank groups headquartered in Europe (i.e., no subsidiaries of subsidiaries). We also included only bank groups headquartered in Europe (i.e., no subsidiaries of Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes non-European banks included). Ferri, G.; Kalmi, P.; Kerola, E. non-European banks included). Table according to to country country and and ownership. ownership. The The Table 11 gives gives the the distribution distribution of of observations observations according ownership categories are similar to those in Ferri et al. (2013), which were in turn influenced by ownership categories are similar to those in Ferri et al. (2013), which were in turn influenced 8by 8. Desrochers and Fischer (2005) on cooperative banks and Gardener et al. (1997) on savings banks Desrochers and Fischer (2005) on cooperative banks and Gardener et al. (1997) on savings banks . Table 1. Observation per country and ownership structure TTABLE ABLE1. 1. O OBSERVATION BSERVATION PER PER COUNTRY COUNTRY AND AND OWNERSHIP OWNERSHIP STRUCTURE STRUCTURE Panel Data Panel A A :Fitch :Fitch Data

Cooperative Cooperative groups groups

Solo Solo cooperatives cooperatives

Austria Austria Belgium Belgium Cyprus Cyprus Denmark Denmark Finland Finland France France Germany Germany Greece Greece Iceland Iceland Ireland Ireland Italy Italy Luxembourg Luxembourg Netherlands Netherlands Norway Norway Portugal Portugal Spain Spain Sweden Sweden Switzerland Switzerland United UnitedKingdom Kingdom Total Total

11 00 00 00 11 22 11 00 00 00 00 00 11 00 00 11 00 00 00 77

00 00 0 0 0 0 0 0 0 2 10 10 0 0 0 0 3 0 0 10 10 25 25

Austria Austria Belgium Belgium Cyprus Cyprus Denmark Denmark Finland Finland France France Germany Germany Greece Greece Iceland Iceland Ireland Ireland Italy Italy Luxembourg Luxembourg Netherlands Netherlands Norway Norway Portugal Portugal Spain Spain Sweden Sweden Switzerland Switzerland UnitedKingdom Kingdom United Total Total

22 00 00 00 11 44 22 00 00 00 00 00 11 00 00 11 00 11 00 12 12

0 0 0 1 0 0 1 0 0 2 8 0 0 0 0 44 00 00 11 11 27 27

Country Countryname name

GovernmentGovernmentowned savings savings owned banks banks 00 11 00 00 0 0 0 0 0 0 0 0 0 9 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 2 20 0 0 0 0 1 0 0 13 24 Panel B: Moody’s Data 0 1 0 0 0 0 0 0 0 0 0 0 12 0 0 0 0 0 0 0 0 5 0 1 0 0 5 0 2 0 14 00 14 00 00 33 00 00 00 24 19 24 19 Private Private savings savings banks banks

Shareholder Shareholder banks banks

Total Total

11 44 33 22 11 66 55 66 00 55 10 10 00 77 22 34 99 44 33 11 11 83 83

33 44 33 22 22 88 15 15 66 00 77 20 20 00 88 66 66 33 33 44 44 21 21 152 152

22 44 33 99 22 55 77 77 22 66 17 17 22 55 33 44 11 11 55 44 13 13 111 111

55 44 33 10 10 33 99 22 22 77 22 88 30 30 33 66 88 66 30 30 55 88 24 24 193 193

In Cooperative Table 2, we show ratings numerical counterparts we assigned to (in them. Fitch IR’sthe numeric Cooperative banks have and beenthe classified either as cooperative groups which case same banks have been classified either as cooperative groups (in which case the same values fromto to 10 (10 within being the and or 1orthe worst), while Moody’s FS’ numeric values go from 1 ratingspan applies to1all all banks within thebest group) independent cooperatives (mostly building societies rating applies banks the group) independent cooperatives (mostly building societies or13 Italian Banche Popolari). Savings banks9. have have been regarded as We also show the equivalents scale. ownership to (againBanche 13 beingPopolari). the best and 1 the worst) or Italian Savings banks been regarded as either eitherinin inlogprivate private ownership ifif foundation-owned (mostly in Spain and Norway, and, in Moody’s data, also Italy) or public (this 8 category consists mostly ofareGerman As is evident from Table in both datasets, The ownership assignments based onLandesbanken). ready-made classifications from Bankscope, but1,some of these were corrected and the classifications were developed further. See Ferri et al. (2013) for more details.

8

7

The ownership assignments are based on ready-made classifications from Bankscope, but some of these were corrected and the classifications were developed further. See Ferri et al. (2013) for more details.

7

9

Fitch IR ceased to exist after December 2011, when Fitch changed bank Individual Ratings to a different scale (with 19 points) and renamed it Viability Ratings. However, we use only data up to December 2011, and therefore our estimations are not affected by this data change. 42 41 JEOD - Vol.3, Issue 1 (2014)


based” categorization of the stakeholder banks differentiating cooperative banks from savings banks; iii) an “organizational/ownership-based” breakdown of the stakeholder banks in which cooperative banks are further subdivided into groups versus independent banks (called solo cooperatives in the tables)Structure and savings banks areamong alsoEuropean split into versus Organizational and Exposure to Crisis Banks:private Evidence from Ratingpublic. Changes Ferri,numerical G.; Kalmi, P.; Kerola, E. In Table 2, we show ratings and the counterparts we assigned to them. Fitch IR’s numeric values span from 1 to 10 (10 being the best and 1 the worst), while Moody’s FS’ numeric values go from 1 to 13 (again 13 being the best and 1 the worst)9. We also show the equivalents in log scale. Table 2. Ratings and their numerical equivalents TABLE 2. RATINGS AND THEIR NUMERICAL EQUIVALENTS Log equivalent 2.30 2.20 2.08 1.95

Linear equivalent

Fitch Individual

Explanation

10 9 8 7

A A/B B B/C

Very Strong

1.79 1.61

6 5

C C/D

Adequate

1.39 1.10

4 3

D D/E

Problematic

0.69 0

2 1

E F

Serious problems or default

Strong

Moody’s Financial Strength A AB+ B BC+ C CD+ D DE+ E

Linear equivalent 13 12 11 10 9 8 7 6 5 4 3 2 1

Log equivalent 2.57 2.48 2.40 2.30 2.20 2.08 1.95 1.79 1.61 1.39 1.10 0.69 0

4. Descriptive statistics

B9 2.20 since the onset of the crisis? A slightly different story is told How have bank ratings evolved 1.79 4.depending Descriptive onstatistics which of the two rating agencies we examine. In Figure 1, we present the 6 development of the average (linear) C numerical ratings using data from Fitch. The figure presents the average numerical ratingsevolved for Adequate thesince end the of onset the years for shareholder banks, cooperative How have bank ratings of the2006-2011 crisis? A slightly different story is told depending C+ banks, savings banks, and stakeholder banks. It can be observed that shareholder banks start with 8 on which of the two rating agencies we examine. In Figure 1, we present the development of the average somewhat higher average ratings2.08 than stakeholder banks, but after 2009, their ratings deteriorate (linear) from The figure thestakeholder average numerical 1.95 faster, numerical and by theratings end of using 2011,data they are atFitch. a clearly lower presents level than banks. ratings for the 1.61 end of the years 2006-2011 for shareholder banks, cooperative banks, savings banks, and stakeholder banks. 5 It can be observed that shareholderC/D banks start with somewhat higher average ratings than stakeholder

banks, but after 2009, their ratings deteriorate faster, and by the end of 2011, they are at a clearly lower C 7 level than stakeholder banks. 1.95

Figure 1. Shareholder banks vs. stakeholder, cooperative and savings banks, Fitch FIGURE 1. SHAREHOLDER BANKS VS. STAKEHOLDER, COOPERATIVE AND SAVINGS BANKS, FITCH

C6 1.79 1.39 9 Fitch IR ceased to exist after December4 2011, when Fitch changed bank Individual Ratings to a different scale (with 19 points) and renamed it Viability Ratings. However, we use only data up to December 2011, and therefore our D estimations are not affected by this data change. Problematic D+ 8 5 1.61 1.10 3 D/E D

Figure 2 tells the same story for Moody’s. In the early period, all types of banks are hit roughly equally and shareholder banks have on average higher ratings than stakeholder banks, but in 2011, situation is reversed. tells thethesame story for Moody’s. In the early period, all types of banks are hit roughly

Figure 2 equally and shareholder banks have onBANKS average higher ratings than AND stakeholder banks, but FIGURE 2. SHAREHOLDER VS. STAKEHOLDER , COOPERATIVE SAVINGS BANKS , MOODY ’S in 2011, the situation is reversed.

43 42 JEOD - Vol.3, Issue 1 (2014)


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

Figure 2 tells the same story for Moody’s. In the early period, all types of banks are hit roughly equally and shareholder banks have on average higher ratings than stakeholder banks, but in 2011, the situation is reversed. Figure 2. Shareholder banks vs. stakeholder, cooperative and savings banks, Moody’s FIGURE 2. SHAREHOLDER BANKS VS. STAKEHOLDER, COOPERATIVE AND SAVINGS BANKS, MOODY’S

In Figures 3 and 4, we utilize the more disaggregated classes taken from Ferri et al. (2013)

In Figures and 4,cooperative we utilizebanks the more disaggregated taken from Ferri et al. and3 divide into cooperative groupsclasses and independent cooperatives, and(2013) savingsand divide banks into private and public savings banks. Starting from Figure 3, which displays the Fitch cooperative banks into cooperative groups and independent cooperatives, and savings banks into private ratings, we note that there is initially a large difference between the ratings of different types of and public savings from banks Figureactually 3, which displays theratings Fitch ratings, we note savings banks. banks, Starting private savings having the best in 2006-2007 and that publicthere is inisavings banks, the worst. In contrast, cooperative groups and independent cooperative banks have tially a large difference between the ratings of different types of savings banks, private savings banks actually 9 having the best ratings in 2006-2007 and public savings banks, the worst. In contrast, cooperative groups and independent cooperative banks have fairly similar ratings from Fitch in 2006-2007, but this changes in 2008, when the ratings of independent cooperative banks deteriorate substantially. In 2009, cooperative groups are clearly downgraded less vis-à-vis other types of banks. Moreover, public savings banks, while having deteriorating ratings between 2007 and 2009, start to actually improve their ratings after 2009, thus reducing the gap between this and other categories. Figure 3. The performance of shareholder banks and various types of stakeholder banks, Fitch

When looking at Moody’s ratings (Figure 4), it is notable that cooperative groups start clearly above other groups, and, though their ratings generally deteriorate during the period, they remain clearly above the other groups. Private savings banks and independent cooperatives experience a large decline in 2009, after which point their ratings stabilize. The ratings of shareholder banks and public savings banks decline 44 43 JEOD - Vol.3, Issue 1 (2014)


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

consistently throughout the period. Figure 4. The performance of shareholder banks and various types of stakeholder banks, Moody’s FIGURE 4. THE PERFORMANCE OF SHAREHOLDER BANKS AND VARIOUS TYPES OF STAKEHOLDER BANKS, MOODY’S

An alternative view of the same process can be obtained by looking at downgrades and

An alternative view of thethesame process be obtained looking at downgrades upgrades throughout process; Table 3can presents these for theby Fitch and Moody’s data. Initially,and upgrades and Moody’s rather these different on theand developing crisis: Moody’s throughout theFitch process; Table 3took presents forviews the Fitch Moody’ssubprime data. Initially, Fitch and Moody’s downgraded 47% of banks in the sample during 2007, while Fitch did so only for 5% of the banks. took rather different views on the developing Moody’s downgraded of banks in the The situation reversed somewhat in 2008,subprime when Fitchcrisis: downgraded 34% of the banks and 47% Moody’s only 23%. 2009 was the worst year in terms of ratings development: Fitch downgraded 51% of the sample duringbanks, 2007, Fitch did so 63%. only Only for 5% of the (by banks. andwhile Moody’s downgraded one upgrade Fitch)The took situation place duringreversed the years somewhat in 2008-2009. In 2010, things no longer looked as bad: Fitch downgraded 15% of the banks and year in terms 2008, when Fitch downgraded 34% of the banks and Moody’s only 23%. 2009 was the worst upgraded 7%, whereas for Moody’s, the respective figures were 15% and 3%. In 2011, Fitch of ratings development: Fitchofdowngraded 51% of 10%; the banks, and Moody’s downgraded 63%. Only one downgraded 17% the banks and upgraded Moody’s, more pessimistically, downgraded 25% of the banks and upgraded only 5%. During the entire period, 68% of the banks were upgrade (by Fitch) took place during the years 2008-2009. In 2010, things no longer looked as bad: Fitch downgraded by Fitch, and only 2% were upgraded. For Moody’s, the figures are much bleaker: downgraded 15% of the banks and upgraded 7%, whereas for Moody’s, the respective figures were 15% 87% were downgraded, and 3% were upgraded. and 3%. In 2011, Fitch downgraded 17% of the banks and upgraded 10%; Moody’s, more pessimistically, TABLE 3. THE PROPORTION OF DOWNGRADES AND UPGRADES AMONG BANKS, BY OWNERSHIP: PERCENTAGE OF OBSERVATIONS downgraded 25% of the banks and upgraded only 5%. During the entire period, 68% of the banks were Panel A: Fitch Data downgraded by Fitch, and only upgraded.2009 For Moody’s, bleaker: 87% were 2007 2% were 2008 2010 the figures 2011are much 2006-2011 Down Up Down Up Down Up Down Up Down Up Down Up downgraded, and 3% were4.6upgraded. All 3.9 33.8 0 50.7 0.7 15.4 7.4 17.2 9.7 68.4 2.3 Shareholders Stakeholders Cooperatives Savings Coop group Coop solo Private savings Public savings

2.6 7.4 4.1 10 0 5.6 10 10

5.3 1.9 0 3.3 0 0 5 0

28.8 40 37.9 41.7 0 47.8 41.7 41.7

0 0 0 0 0 0 0 0

All Shareholders Stakeholders Cooperatives

47.1 46.3 48.4 28.1

11.5 15.8 4.8 3.1

23.1 21.7 25 29.7

0 0 0 0

48.2 1.2 22.5 53.7 0 7.3 35.5 0 9.1 69.4 0 5.6 0 0 25 44 0 4 75 0 8.7 58.3 0 0 Panel B: Moody’s Data 63 0 14.7 65.7 0 17.7 59.2 0 10.7 56.8 0 16.67

11

45 44 JEOD - Vol.3, Issue 1 (2014)

8.8 5.8 3 8.3 12.5 0 0 23.1

21.8 11.9 15.6 8.6 14.3 16 4.6 15.4

11.5 7.5 0 14.3 0 0 0 38.5

69.3 67.2 60 72.7 33.3 68.4 80.1 58.3

1.3 3.5 4 3 0 5.3 0 8.3

2.8 2.9 2.7 0

24.5 28.4 18.5 15.2

5.4 2.9 9.2 12.1

86.7 87.2 85.7 82.1

3.3 5.3 0 0


only 23%. 2009 was the worst year in terms of ratings development: Fitch downgraded 51% of the banks, and Moody’s downgraded 63%. Only one upgrade (by Fitch) took place during the years 2008-2009. In 2010, things no longer looked as bad: Fitch downgraded 15% of the banks and upgraded 7%, whereas for Moody’s, the respective figures were 15% and 3%. In 2011, Fitch Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes downgraded 17% of the banks and upgraded 10%; Moody’s, more pessimistically, downgraded Ferri, G.; Kalmi, P.; Kerola, E. 25% of the banks and upgraded only 5%. During the entire period, 68% of the banks were downgraded by Fitch, and only 2% were upgraded. For Moody’s, the figures are much bleaker: 87% were downgraded, and 3% were upgraded. Table 3. The of downgrades and upgrades among banks, AMONG by ownership: of observations TABLE 3. Tproportion HE PROPORTION OF DOWNGRADES AND UPGRADES BANKSpercentage , BY OWNERSHIP : PERCENTAGE OF OBSERVATIONS

All Shareholders Stakeholders Cooperatives Savings Coop group Coop solo Private savings Public savings All Shareholders Stakeholders Cooperatives Savings Coop group Coop solo Private savings Public savings

2007 Down Up 4.6 3.9 2.6 5.3 7.4 1.9 4.1 0 10 3.3 0 0 5.6 0 10 5 10 0 47.1 46.3 48.4 28.1 70 55.6 17.4 76.5

61.5

11.5 15.8 4.8 3.1 6.7 0 4.4 5.9

7.7

2008 Down Up 33.8 0 28.8 0 40 0 37.9 0 41.7 0 0 0 47.8 0 41.7 0 41.7 0 23.1 21.7 25 29.7 20.5 27.3 30.8 23.8

16.7

0 0 0 0 0 0 0 0

0

Panel A: Fitch Data 2009 2010 Down Up Down Up 50.7 0.7 15.4 7.4 48.2 1.2 22.5 8.8 53.7 0 7.3 5.8 35.5 0 9.1 3 69.4 0 5.6 8.3 0 0 25 12.5 44 0 4 0 75 0 8.7 0 58.3 0 0 23.1 Panel B: Moody’s Data 63 0 14.7 2.8 65.7 0 17.7 2.9 59.2 0 10.7 2.7 56.8 0 16.67 0 61.5 0 5.1 5.1 40 11 0 10 0 63 0 19.2 0 85.7 0 0 9.5

33.3

0

11.1

0

2011 Down Up 17.2 9.7 21.8 11.5 11.9 7.5 15.6 0 8.6 14.3 14.3 0 16 0 4.6 0 15.4 38.5 24.5 28.4 18.5 15.2 21.9 30 8.7 7.1

33.3

5.4 2.9 9.2 12.1 6.25 0 17.4 7.1

5.6

2006-2011 Down Up 68.4 2.3 69.3 1.3 67.2 3.5 60 4 72.7 3 33.3 0 68.4 5.3 80.1 0 58.3 8.3 86.7 87.2 85.7 82.1 89.3 77.8 84.2 83.3

93.8

3.3 5.3 0 0 0 0 0 0

0

Examining the rating changes by ownership, we note that in the earliest stage of the crisis (2007-2009) both agencies downgraded stakeholder and shareholder banks in roughly equal measure. After 2009, clearly more shareholder banks than stakeholder banks were downgraded. Examining the ratingthat changes by ownership, we agencies note that in stage of the crisis (2007-2009) Another observation is visible with both is the theearliest massive downgrading of private both agencies downgraded stakeholder and shareholder banks in roughly equal measure. After 2009, clearly savings banks in the year 2009. Table 4 presents the stakeholder average changes ownership category by observation both agencies more shareholder banks than banks by were downgraded. Another thatinis linear visible and with logarithmic scale. The results for Fitch indicate that the ratings for stakeholder banks have both agencies is the massive downgrading of private savings banks in the year 2009. deteriorated less on average, and the difference is statistically significant. Moreover, in the linear Table 4 presents the average changes by ownership category by both agencies in linear and logarithmic scale, differences are statistically significant for cooperative groups and public savings banks. In log scale. results Fitch indicate that5% thelevel ratingsoffor stakeholder banks havefor deteriorated lessbanks, on average, scale,The there is afordifference at the statistical significance stakeholder for and the difference is statistically linear scale, statistically cooperatives in general, and for significant. cooperativeMoreover, groups. IninthetheMoody’s data, differences there are noarestatistically significantfordifferences, giving and a first indication that there a difference in how significant cooperative groups public savings banks. In logmay scale,bethere is a difference at thethe 5%two level agencies treat shareholder and stakeholder banks. of statistical significance for stakeholder banks, for cooperatives in general, and for cooperative groups. In In any case, the differences found with the Fitch data may also be attributed to various factors. the Moody’s data, there are no statistically significant differences, giving a first indication that there may be Stakeholder banks had initially lower ratings and thus had less room to deteriorate. The differences amay difference in how some the twostrategic agenciesdifferences treat shareholder and the stakeholder also reflect between differentbanks. types of banks. Perhaps most In any case, differencesmay found withthe thefact Fitchthat datathose may countries also be attributed to various factors. importantly, thethedifference reflect where stakeholder banks (especially banks cooperatives) are lower strongratings were less hit by 2013). the last Stakeholder had initially and thus hadthelesscrisis room(Leogrande to deteriorate. The While differences may point maysome suggest that stakeholder banks create a positive externality forPerhaps other banks in the country, also reflect strategic differences between the different types of banks. most importantly, the we control for this effect by including country dummies as well as initial ratings and bank business difference may reflect the fact that those countries where stakeholder banks (especially cooperatives) are strategies on the basis of bank financial variables. strong were less hit by the crisis (Leogrande, 2013). While the last point may suggest that stakeholder banks TABLEa 4. MEAN RATING CHANGES DURING THE ENTIRE PERIOD we control for this effect by including country create positive externality for other banks in the country, dummies as well as initial ratings and bank business strategies on the basis of bank financial variables. Fitch Moody’s All Shareholder Stakeholder Cooperative Savings Coop group Coop solo Private savings Public savings

Annual change in notches -0.36 (1.03) -0.42 (1.22) -0.28* (0.72) -0.28 (0.7) -0.28 (0.74) -0.06* (0.35) -0.34 (0.76) -0.35 (0.7) -0.17*

Annual change in log scale -0.07 (0.26) -0.09 (0.32) -0.05** (0.17) -0.05* (0.15) -0.05 (0.19) -0.01* (0.06) -0.06 46 45 (0.17) -0.06 JEOD - Vol.3, Issue 1 (2014) (0.19) -0.03

Annual change in notches -0.55 (1.10) -0.58 (1.15) -0.50 (1.02) -0.48 (1.01) -0.53 (1.02) -0.44 (0.78) -0.49 (1.09) -0.63 (1.09) -0.42

Annual change in log scale -0.09 (0.24) -0.10 (0.26) -0.08 (0.2) -0.08 (0.22) -0.08 (0.2) -0.07 (0.17) -0.09 (0.24) -0.09 (0.20) -0.08


In any case, the differences found with the Fitch data may also be attributed to various factors. Stakeholder banks had initially lower ratings and thus had less room to deteriorate. The differences may also reflect some strategic differences between the different types of banks. Perhaps most Structure and Exposure Crisis among Banks: Evidence from Rating Changes importantly, theOrganizational difference may reflectto the fact European that those countries where stakeholder banks Ferri, G.; Kalmi, P.; Kerola, E. (especially cooperatives) are strong were less hit by the crisis (Leogrande 2013). While the last point may suggest that stakeholder banks create a positive externality for other banks in the country, we control for this effect by including country dummies as well as initial ratings and bank business strategies on the basis of bank financial variables.

Table 4. Mean rating changes during the entire period TABLE 4. MEAN RATING CHANGES DURING THE ENTIRE PERIOD

Fitch Moody’s Annual change in Annual change in log Annual change in Annual change in log notches scale notches scale -0.36 -0.07 -0.55 -0.09 All (1.03) (0.26) (1.10) (0.24) -0.42 -0.09 -0.58 -0.10 Shareholder (1.22) (0.32) (1.15) (0.26) -0.28* -0.05** -0.50 -0.08 Stakeholder (0.72) (0.17) (1.02) (0.2) -0.28 -0.05* -0.48 -0.08 Cooperative (0.7) (0.15) (1.01) (0.22) -0.28 -0.05 -0.53 -0.08 Savings (0.74) (0.19) (1.02) (0.2) -0.06* -0.01* -0.44 -0.07 Coop group (0.35) (0.06) (0.78) (0.17) -0.34 -0.06 -0.49 -0.09 Coop solo (0.76) (0.17) (1.09) (0.24) -0.35 -0.06 -0.63 -0.09 Private savings (0.7) (0.19) (1.09) (0.20) -0.17* -0.03 -0.42 -0.08 Public savings (0.78) (0.19) (0.93) (0.19) Note: the asterisks indicate (unconditional) statistically different means with shareholder banks. Significance levels: *** 10%; ** 5%; *the 1%asterisks indicate (unconditional) statistically different means with shareholder banks. Significance levels: *** 10%; ** Note:

5%; * 1%

5. Regression analysis of rating changes

We begin the analysis by presenting the results from regressions that use Fitch data and logarithmic changes in ratings as their dependent variable. The models (I)-(III) in Table 5 present 5. Regression analysis of rating changes 12 We begin the analysis by presenting the results from regressions that use Fitch data and logarithmic changes in ratings as their dependent variable. The models (I)-(III) in Table 5 present results that include ownership and country dummies, lagged level of (log) ratings, and the difference in the (log of ) sovereign ratings. Models (IV)-(VI) additionally include the bank-specific control variables. All reported standard errors are heteroskedasticity and autocorrelation robust. The R-squares vary between 0.16 (models I-III) to 0.26 (models IV-VI). For the control variables, lagged ratings are always statistically significant (at a level of at least 5%) and negative, indicating that the higher the rating initially, the larger the downgrade. This result was rather expected given that the banks with higher ratings have also more room to deteriorate. The coefficient of change in the sovereign rating is positive and statistically significant at the 1% level (when additional bankspecific variables have not been added), which means that the bank ratings and sovereign ratings for the country where the bank is located move into the same direction. This result may be surprising in light of Poon et al.’s (1999) results, which indicate that sovereign ratings do not influence bank ratings, but it may be understood in the context of a crisis situation in which the macroeconomic environment that the bank operates within becomes a crucial determinant of bank viability. The evidence presented in column 3 suggests that cooperative banking groups have weathered the crisis better than other types of banks, as the coefficient it receives is positive and statistically significant (at the 5% level). The coefficient for independent cooperative banks is also positive, but it is not statistically significant. For saving banks, there is no evidence that they would have performed better than shareholder banks during the crisis.

47 46 JEOD - Vol.3, Issue 1 (2014)


situation in which the macroeconomic environment that the bank operates within becomes a crucial determinant of bank viability. The evidence presented in column 3 suggests that cooperative banking groups have weathered the crisis better than other types ofandbanks, asCrisis theamong coefficient it receives positive Organizational Structure Exposure to European Banks: Evidence fromis Rating Changes and statistically Ferri, G.; Kalmi, P.; Kerola, E. significant (at the 5% level). The coefficient for independent cooperative banks is also positive, but it is not statistically significant. For saving banks, there is no evidence that they would have performed better than shareholder banks during the crisis. TABLE 5. RANDOM EFFECTS REGRESSION RESULTS FITCH LOG SCALE:errors COEFFICIENTS AND Table 5. Random effects regression results for Fitch data, FOR ratings in logDATA scale:, RATINGS coefficientsINand standard STANDARD ERRORS

Model Stakeholder Cooperative Savings Coop group

I 0.0127 (0.020)

II 0.0347 (0.022) -0.0162 (0.027)

Difference in log ratings (Fitch) III IV 0.0316 (0.024)

0.0408** (0.019) 0.0331 (0.029) -0.0116 (0.040) -0.0234 (0.023) 0.166** (0.077) -0.155** (0.067)

V

VI

0.0543* (0.028) 0.00787 (0.029)

0.0846*** (0.025) Solo coop 0.0459 (0.035) Private savings 0.0208 (0.041) Public savings -0.0130 (0.033) Change in sovereign rating t-1 0.165** 0.166** 0.0766 0.0762 0.0775 (0.076) (0.076) (0.084) (0.084) (0.084) Log (rating t-1) -0.149** -0.154** -0.309*** -0.318*** -0.323*** (0.067) (0.066) (0.052) (0.051) (0.051) (equity/assets) t-1 -0.353 -0.331 -0.312 (0.434) (0.440) (0.438) log(assets) t-1 0.0115* 0.0125** 0.0128** (0.006) (0.006) (0.006) (loanloss -2.594 -2.371 -2.543 provisions/loans) t-1 (2.603) (2.710) (2.728) ROAAt-1 0.143*** 0.146*** 0.145*** (0.028) (0.029) (0.029) (deposits/assets) t-1 0.0413 0.0260 0.0321 (0.100) (0.096) (0.094) Country and year dummies YES YES YES YES YES YES # of observations 711 711 711 590 590 590 R2 0.163 0.166 0.166 0.260 0.262 0.263 # of banks 152 152 152 146 146 146 Notes: 1) Standard errors are heteroskedasticity and autocorrelation robust; 2) Significance levels: ***<1%; **<5%; *<10%

Note: 1) Standard errors are heteroskedasticity and autocorrelation robust; 2) Significance levels: ***<1%; **<5%; *<10%

Models IV – VI introduce the bank-specific control variables, which turn out to be significant as a 13 group. Banks that have better profitability and that are larger were downgraded less during the crisis10. Instead, loan loss provisions, capitalization and deposits per assets are not related to ratings changes in any statistically significant way. Of the ownership dummies, an important difference compared to the previous model is that now the dummy for cooperatives in general is positive and significant at a 10% level. The dummy for cooperative groups remains significant, now at a 1% level. Inclusion of the bank-specific controls has increased the cooperative group dummy’s coefficient twofold compared to column 3. Table 6 reports the results for Moody’s. The results concerning the lagged rating (negative) and the change in sovereign ratings (positive) are consistent with the Fitch results. For the other control variables, size is again a statistically significant factor slowing the downgrade, as with Fitch. Unlike with Fitch, profitability loses its significance, whereas now loan loss provisions per total loans is negatively and deposits per assets is positively related to ratings change. However, the ownership dummies are never statistically significant.

10

The latter result offers some indication that the favorable treatment of large banks in assigning ratings (identified by Hau et al. 2013) may also apply to BFSRs. 48 47 JEOD - Vol.3, Issue 1 (2014)


coefficient twofold compared to column 3. Table 6 reports the results for Moody’s. The results concerning the lagged rating (negative) and the change in sovereign ratings (positive) are consistent with the Fitch results. For the other control variables,Organizational size is again a and statistically significant factor slowing theRating downgrade, as with Fitch. Structure Exposure to Crisis among European Banks: Evidence from Changes G.; Kalmi, P.; Kerola, E. Unlike with Fitch, profitability loses itsFerri, significance, whereas now loan loss provisions per total loans is negatively and deposits per assets is positively related to ratings change. However, the ownership dummies are never statistically significant. TABLE 6. RANDOM EFFECTS REGRESSIONS RESULTS FOR MOODY ’S scale: DATAcoefficients RATINGS IN LOG SCALEerrors : COEFFICIENTS Table 6. Random effects regressions results for Moody’s data ratings in log and standard AND STANDARD ERRORS

Model Stakeholder Cooperative Savings Coop group

I -0.00896 (0.016)

II -0.00308 (0.020) -0.0161 (0.018)

Difference in log ratings (Moody’s) III IV -0.0179 (0.016)

0.00116 (0.029) -0.00496 (0.026) -0.0180 (0.024) -0.0133 (0.025) 1.047*** (0.197) -0.145*** (0.035)

V

VI

-0.0178 (0.020) -0.0159 (0.021)

0.00869 (0.034) Solo coop -0.0312 (0.023) Private savings -0.0261 (0.030) Public savings -0.00624 (0.028) Change in sovereign rating t-1 1.046*** 1.047*** 1.097*** 1.097*** 1.098*** (0.197) (0.197) (0.245) (0.245) (0.246) Log (rating t-1) -0.145*** -0.145*** -0.240*** -0.240*** -0.204*** (0.035) (0.035) (0.054) (0.055) (0.054) (equity/assets) t-1 0.543 0.543 0.591 (0.400) (0.400) (0.392) log(assets) t-1 0.0118* 0.0118* 0.0109* (0.005) (0.006) (0.006) (loanloss -5.757*** -5.753*** -5.853*** provisions/loans) t-1 (1.789) (1.797) (1.799) ROAAt-1 0.0240 0.0240 0.0234 (0.019) (0.019) (0.019) (deposits/assets) t-1 0.177*** 0.177*** 0.177*** (0.065) (0.066) (0.068) Country and year dummies YES YES YES YES YES YES # of observations 861 861 861 717 717 717 R2 0.324 0.324 0.324 0.377 0.377 0.377 # of banks 193 193 193 183 183 183 Notes: 1) Standard errors are heteroskedasticity and autocorrelation robust; 2) Significance levels: ***<1%; **<5%; *<10%

Note: 1) Standard errors are heteroskedasticity and autocorrelation robust; 2) Significance levels: ***<1%; **<5%; *<10%

For a robustness check, we performed the analysis using the OLS estimator and also use the linear for ratings instead of the logarithmic In the all cases, the results similar. For scale a robustness check, we performed the analysisone. using OLS estimator andremained also use the linear scale In sum, somewhat mixed findings emerge from the analysis. The findings from Fitch data for ratings instead of the logarithmic one. In all cases, the results remained similar. indicate that ownership structure matters: cooperative banking groups experienced smaller In sum, somewhat mixed findings emerge from the analysis. The findings from Fitch data indicate that ownership structure matters: cooperative banking groups experienced smaller downgrades than shareholder 10 The latter result offers some indication that the favorable treatment of large banks in assigning ratings (identified by banks, theremay is some evidence that this holds also for cooperatives more generally. From Moody’s data, Hau et and al. 2013) also apply to BFSRs. 14 to the four distinct ownership groups, they are even though the results are not in conflict with respect much more muted, and ownership dummies never emerge as statistically significant. What leads to this discrepancy? There are two distinct possibilities: either there are significant differences in the samples, or the agencies in fact do rate the banks differently. The first of these possibilities is easily tested: we can repeat the regression by using only the observations that are common to both datasets. Table 7 presents some of these results for this limited dataset providing evidence that can be compared to Models I – III in Tables 5 and 611. The results for the Fitch dataset remain very similar to the previous results, although the result for cooperative groups is now even stronger (statistically significant at a 1% level). An interesting finding relates to Moody’s dataset, where (in column II) the coefficient for savings banks as a group is negative and now statistically significant (at a 5% level), indicating that savings banks have experienced more downgrades than shareholder banks. However, the dummies for cooperatives in general, and cooperative groups in particular, continue to be 11

We omit from the presentation the results including economic variables, but the results remain similar if they are included. 49 48 JEOD - Vol.3, Issue 1 (2014)


the agencies in fact do rate the banks differently. The first of these possibilities is easily tested: we can repeat the regression by using only the observations that are common to both datasets. Table 7 presents some of these results for this limited dataset providing evidence that can be compared to Models I – III in Tables 5 and 611. The results for the Fitch dataset remain very similar to the Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes previous results, although the result for groups is now even stronger (statistically Ferri,cooperative G.; Kalmi, P.; Kerola, E. significant at a 1% level). An interesting finding relates to Moody’s dataset, where (in column II) the coefficient for savings banks as a group is negative and now statistically significant (at a 5% level), indicating that savings banks have experienced more downgrades than shareholder banks. However, the dummies for cooperatives in general, and cooperative groups in particular, continue to insignificant. These results suggest that the differences are, by and large, not driven by differences in the be insignificant. These results suggest that the differences are, by and large, not driven by two samples. in the two samples. differences Table 7. Changes in ratings, log scale: FitchSCALE and Moody’s TABLE 7. CHANGES IN RATINGS , LOG : FITCHcompared AND MOODY’S COMPARED

Stakeholder Cooperative

(I) 0.0187 (0.023)

Fitch Difference in ratings (II)

Savings Coop group

(III)

0.0353 (0.025) -0.00282 (0.031)

(I) -0.0215 (0.021)

Moody’s Difference in ratings (II)

(III)

-0.0000531 (0.026) -0.0491** (0.025)

0.0683*** 0.0322 (0.022) (0.029) Solo coop 0.0270 -0.0120 (0.033) (0.033) Private savings 0.0326 -0.0516 (0.050) (0.035) Public savings -0.0410 -0.0432 (0.030) (0.033) Change in sovereign rating t-1 0.229*** 0.230*** 0.232*** 0.935*** 0.936*** 0.937*** (0.079) (0.079) (0.079) (0.226) (0.226) (0.226) Log (rating t-1) -0.241** -0.246*** -0.253*** -0.148*** -0.148*** -0.150*** (0.094) (0.094) (0.095) (0.040) (0.039) (0.040) Country and year dummies YES YES YES YES YES YES # of observations 566 566 566 566 566 566 R2 0.208 0.210 0.213 0.349 0.352 0.352 # of banks 127 127 127 127 127 127 Notes: 1) Standard errors are heteroskedasticity and autocorrelation robust; 2) Significance levels: ***<1%; **<5%; *<10%

Note: 1) Standard errors are heteroskedasticity and autocorrelation robust; 2) Significance levels: ***<1%; **<5%; *<10%

6. Analysis of rating disagreements

Because the differences in the two samples are not sufficient to explain the divergent results, it is instructive to look at rating disagreements. As explained above, we make this examination by 6.creating Analysisnew of rating disagreements standardized variables out of the end-of-year ratings for those observations for which we have both a Fitch Individual Rating and a Moody’s Financial Strength Rating. We have a dataset consisting of differences 586 observations. Both Moody’s Fitch’sto ratings are divergent transformed intoit ais Because the in the two samples are notand sufficient explain the results, standardized scale having mean 0 and standard deviation 1. We define the systematic disagreement instructive to look at rating disagreements. As explained above, we make this examination by creating as a difference between the standardized value of Fitch’s rating and the standardized value of new standardized variables out of the end-of-year ratings for those observations for which we have both a 11Fitch Individual Rating and a Moody’s Financial Strength Rating. We have a dataset consisting of 586 We omit from the presentation the results including economic variables, but the results remain similar if they are observations. Both Moody’s and Fitch’s ratings are transformed into a standardized scale having mean 0 included. and standard deviation 1. We define the systematic disagreement as a difference between the standardized 15 value of Fitch’s rating and the standardized value of Moody’s rating. A positive value for this difference means that Fitch has given the bank a higher standardized rating than Moody’s, whereas a negative value means the opposite. We also calculate the random rater disagreement, defined as the absolute value of the rating disagreement. This measure is similar to the measures of business opacity developed by Morgan (2002)12.

12

In the work by Morgan (2002), the measure of opacity was drawn using split (bond) ratings. When using bank financial strength ratings where the scales are different, the concept of split ratings does not apply, but we believe our method is a close equivalent. 50 49 JEOD - Vol.3, Issue 1 (2014)


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

Table 8. Mean values of systematic and absolute rater disagreement, by ownership TABLE 8. MEAN VALUES OF SYSTEMATIC AND ABSOLUTE RATER DISAGREEMENT, BY OWNERSHIP Ownership Shareholder Stakeholder Cooperative Savings Cooperative Group Solo Cooperative Private Savings Public Savings

Systematic disagreement -0.08 (0.66) 0.23 (0.62) 0.32 (0.67) 0.15 (0.56) 0.08 (0.95) 0.38 (0.55) 0.37 (0.53) -0.07 (0.49)

Absolute disagreement 0.50 (0.44) 0.50 (0.43) 0.56 (0.48) 0.44 (0.37) 0.72 (0.62) 0.52 (0.43) 0.49 (0.43) 0.39 (0.30)

Number of observations 347 239 119 120 26 93 61 59

summary statistics related to systematic and absolute rating disagreements are displayed TheThe summary statistics related to systematic and absolute rating disagreements are displayed in Table 8. in Table 8. There appears to be a slight bias in Fitch’s ratings in favor of stakeholder banks, which There appears to be a slight bias in Fitch’scooperative ratings in favor of stakeholder which is most pronounced is most pronounced for independent banks and privatebanks, savings banks. In contrast, in for independent cooperative banks and private savings banks. In contrast, in absolute disagreements, absolute disagreements, there is no difference between shareholder and stakeholder banks. Ofthere all isownership no difference between shareholder and stakeholder banks.toOf ownershipdifferent groups, only public savings groups, only public savings banks appear beallsomewhat by having lower values of absolute disagreements (and no systematic disagreement at all). banks appear to be somewhat different by having lower values of absolute disagreements (and no systematic Table at9 all). reports the regressions on rater disagreements, where in addition to ownership, disagreement country and year effects are controlled for. The reported standard errors are heteroskedasticity and Table 9 reports the regressions on rater disagreements, where in addition to ownership, country and autocorrelation robust. The reported coefficients tend to confirm the findings from the summary year effects There are controlled for. The standard are higher heteroskedasticity andFitch autocorrelation statistics. is evidence that reported stakeholder bankserrors receive ratings from than from robust. The(Column reportedI);coefficients to confirm the findings from the summary statistics. Moody’s this effecttend is visible especially in cooperative banks (Column II), andThere in theis more disaggregated ratings, it is more pronounced for independent cooperative banks and private evidence that stakeholder banks receive higher ratings from Fitch than from Moody’s (Column I); this savings banks especially (Column in III). There arebanks few differences in and absolute the significant effect is visible cooperative (Column II), in thedisagreements: more disaggregated ratings, it ownership dummy is that of savings banks (Column V), and that of public savings banks in is more pronounced for independent cooperative banks and private savings banks (Column III). There particular (Column VI); the negative sign indicates that there is more rater agreement regarding are few differences in absolute disagreements: the significant dummy is that of savings banks public savings banks than other types of banks. In other ownership words, there is approximately the same (Column V),overall and thatdisagreement of public savings banks in particular (Column VI);regarding the negative sign indicates that degree of regarding stakeholder banks than shareholder banks. However, in shareholder banks, the savings disagreement between rating agencies tendswords, to be there is morewhereas rater agreement regarding public banks than otherthe types of banks. In other moreisrandom, amongthe stakeholder banks, Fitchdisagreement appears to give higherstakeholder ratings to atbanks leastthan certain types there approximately same degree of overall regarding regarding of stakeholder banks than it gives to shareholder banks. The underlying source of this shareholder banks. However, whereas in shareholder banks, the disagreement between the rating agencies methodological difference is not known in the absence of more details regarding rating tends to be more random, among stakeholder banks, Fitch appears to give higher ratings to at least certain methodologies. types of stakeholder banks than it gives to shareholder banks. The underlying source of this methodological TABLE 9. SisYSTEMATIC AND RATER DISAGREEMENT AND OWNERSHIP : REGRESSION ANALYSIS difference not known inABSOLUTE the absence of more details regarding rating methodologies. Stakeholder Cooperative Savings Coop group

(I) 0.285*** (0.088)

Systematic disagreement (II) (III) 0.372*** (0.141) 0.170 (0.143)

0.285 (0.425) 0.417*** (0.109) 0.352** (0.177) -0.0467

Solo coop Private savings Public savings

16

51 50 JEOD - Vol.3, Issue 1 (2014)

(I) -0.00688 (0.055)

Absolute disagreement (II) 0.109 (0.085) -0.161* (0.090)

(III)

0.256 (0.206) 0.0715 (0.076) 0.0115 (0.125) -0.339***


particular (Column VI); the negative sign indicates that there is more rater agreement regarding public savings banks than other types of banks. In other words, there is approximately the same degree of overall disagreement regarding stakeholder banks than regarding shareholder banks. However, whereas in shareholder banks, the disagreement between the rating agencies tends to be Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes more random, among stakeholder banks,Ferri, Fitch appears to E.give higher ratings to at least certain types G.; Kalmi, P.; Kerola, of stakeholder banks than it gives to shareholder banks. The underlying source of this methodological difference is not known in the absence of more details regarding rating methodologies. Table 9. Systematic and absolute rater disagreement and ownership: regression analysis TABLE 9. SYSTEMATIC AND ABSOLUTE RATER DISAGREEMENT AND OWNERSHIP: REGRESSION ANALYSIS

Stakeholder

Cooperative 12

(I) 0.285*** (0.088)

Systematic disagreement (II) (III)

0.372***

(I) -0.00688 (0.055)

Absolute disagreement (II)

(III)

0.109

In the work by Morgan (2002), the measure of opacity was drawn using split (bond) ratings. When using bank (0.141) (0.085) financial strength ratings where the scales are different, the concept of split ratings does not apply, but we believe our Savings 0.170 -0.161* method is a close equivalent. (0.143) (0.090) Coop group

16

0.285 0.256 (0.425) (0.206) Solo coop 0.417*** 0.0715 (0.109) (0.076) Private savings 0.352** 0.0115 (0.177) (0.125) Public savings -0.0467 -0.339*** (0.231) (0.113) Country and year dummies YES YES YES YES YES YES # of observations 586 586 586 586 586 586 R2 0.288 0.294 0.304 0.138 0.164 0.187 # of banks 129 129 129 129 129 129 Notes: 1) Standard errors are heteroskedasticity and autocorrelation robust; 2) Significance levels: ***<1%; **<5%; *<10%

Note: 1) Standard errors are heteroskedasticity and autocorrelation robust; 2) Significance levels: ***<1%; **<5%; *<10%

7. Conclusions

In this study, we addressed the relative performance of European banks during the recent crisis by examining bank individual rating changes from two of the most prominent credit rating 7.agencies: Conclusions Moody’s and Fitch. We divided the banks into different classes according to their ownership structure, making a distinction between shareholder and stakeholder banks, and dividing stakeholder bankswefurther into cooperative savings banks. Finally,banks we divide cooperative In this study, addressed the relative and performance of European during the recentbanks crisis into groups and independent cooperatives, and savings banks into private and public. by examining bank individual rating changes from two of the most prominent credit rating agencies: It turns out that there are important differences across ownership categories both in the levels Moody’s and Fitch. We divided the banks into different classes according to their ownership structure, of ratings when the crisis started and in the subsequent ratings changes between 2006 and 2011. For making a distinction shareholderbanks and stakeholder banks, higher and dividing banks further instance, before thebetween crisis, shareholder had, on average, ratingsstakeholder than stakeholder banks, into cooperative and savings banks. Finally, we divide cooperative banks into groups and independent and public savings banks had much lower ratings than other banks. During the crisis, the ratings of shareholder and banks deteriorated those of stakeholder banks, especially after 2009. cooperatives, savings banks into more privatethan and public. Cooperative groups, in particular, appear more to the crisis. It turns out that there are important differencesresilient across ownership categories both in the levels of ratings When analyzing the changes in a regression framework, where we and control forFor changes in when the crisis started and in the subsequent ratings changes between 2006 2011. instance, sovereign ratings, lagged rating, country effects and bank-specific control variables, we find that before crisis, shareholder banks had, average, higher ratingsgroups than stakeholder banks, andthose public when the using ratings data from Fitch, theonratings of cooperative deteriorate less than of savings banks had muchHowever, lower ratings thanusing other ratings banks. During the crisis, the ratings shareholderacross banks shareholder banks. when data from Moody’s, the ofdifferences deteriorated than those stakeholder banks, especially afterfrom 2009.the Cooperative particular, ownership more categories, whileofmostly similar to those deriving Fitch data,groups, are notinstatistically significant. This result prompted us to analyze the rater disagreements. Our results indicate that appear more resilient to the crisis. there are systematic rating disagreements between the two agencies, Fitch giving more positive When analyzing the changes in a regression framework, where we control for changes in sovereign ratings to independent cooperatives and private savings banks than Moody’s. There is less evidence ratings, lagged rating, country effects and bank-specific control variables, we find that when using ratings for random rating disagreements between the agencies, although there is evidence that there are data from Fitch, the ratings of cooperative groups deteriorate less thanbanks those compared of shareholder banks. However, fewer disagreements on the financial strength of public savings to other banks. when using ratingshas data from Moody’s, thedebate differences across while mostly similar The study implications for the between theownership merits of categories, different ownership structures economic Overall, support the interpretation that cooperative have toduring those an deriving fromcrisis. the Fitch data, the are results not statistically significant. This result prompted usbanks to analyze been more resilient during the financial crisis than shareholder banks. This result is consistent the rater disagreements. Our results indicate that there are systematic rating disagreements betweenwith the the claims that the cooperative ownership structure induces less risk taking than a profittwo agencies, Fitch giving more positive ratings to independent cooperatives and private savings banks than maximizing structure. Moody’s. There is less evidence random rating disagreements between the agencies, although there is evidence Our analysis also hasfor interesting implications in terms of rating disagreements. It appears that that there are fewer disagreements on the financial strength of public savings banks compared to other banks.to two major rating agencies, Fitch and Moody’s, apply somewhat different rating methodologies stakeholder banks. Further for research on between the specific mechanisms behind thesestructures differences is The study has implications the debate the merits of different ownership during recommended. an economic crisis. Overall, the results support the interpretation that cooperative banks have been more

52 51 JEOD - Vol.3, Issue 1 (2014)

17


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

resilient during the financial crisis than shareholder banks. This result is consistent with the claims that the cooperative ownership structure induces less risk taking than a profit-maximizing structure. Our analysis also has interesting implications in terms of rating disagreements. It appears that two major rating agencies, Fitch and Moody’s, apply somewhat different rating methodologies to stakeholder banks. Further research on the specific mechanisms behind these differences is recommended.

Appendix Fitch Ratings: Bank Individual Ratings Fitch IR attempts to assess how a bank would be viewed if it were entirely independent and could not rely on external support. Ratings are designed to assess the bank’s exposure to, appetite for, and management of risk, and thus represent the agency’s view on the likelihood that the bank would run into significant financial difficulties such that it would require support. A: A very strong bank – Characteristics may include outstanding profitability and balance sheet integrity, franchise, management, operating environment or prospects. B: A strong bank – There are no major concerns regarding the bank. Characteristics may include strong profitability and balance sheet integrity, franchise, management, operating environment or prospects. C: An adequate bank that, however, possesses one or more troublesome aspects – there may be some concerns regarding its profitability and balance sheet integrity, franchise, management, operating environment or prospects. D: A bank that has weaknesses of internal and/or external origin – there are concerns regarding its profitability and balance sheet integrity, franchise, management, operating environment or prospects. Banks in emerging markets are necessarily faced with a greater number of potential deficiencies of external origin. E: A bank with very serious problems, which either require or are likely to require external support. F: A bank that has either defaulted or, in Fitch Ratings’ opinion, would have defaulted if it had not received external support. Examples of such support include state or local government support, (deposit) insurance funds, acquisition by some other corporate entity or an injection of new funds from its shareholders or equivalent. Moody’s Investor Services: Bank Financial Strength Ratings Moody’s FS ratings represent Moody’s opinion of a bank’s intrinsic safety and soundness. Ratings do not address either the probability of timely payment (i.e., default risk) or the loss that an investor may suffer in the event of a missed payment. Instead, FS is a measure of the likelihood that a bank will require assistance from third parties such as its owners, its industry group, or official institutions, to avoid default. FS ratings do not take into account the probability that the bank will receive such external support nor do they address the external risk that sovereign actions may interfere with a bank’s ability to honor its domestic or foreign currency obligations. Factors considered in the assignment of FS ratings include bank-specific elements such as financial fundamentals, franchise value, and business and asset diversification as well as risk factors in the bank’s operating environment, such as the strength and prospective performance of the economy, the structure and relative fragility of the financial system, and the quality of banking regulation and supervision. Moody’s FS ratings range from A to E, with A for banks with the greatest intrinsic financial strength and E for banks with the least intrinsic financial strength. A “+” modifier may be appended to ratings below a category and a “-” modifier may be appended to ratings above a category to identify those banks 53 52 JEOD - Vol.3, Issue 1 (2014)


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

that are placed higher or lower in a rating category13.

References Alessandri, P., Haldane, A.G. (2009). Banking on the State. Bank of England, November, Paper from material presented at the Federal Reserve Bank of Chicago/World Bank event - International Financial Crisis: Have the Rules of Finance Changed? 24 September. Alexopoulos, Y., Goglio, S. (2009). Financial deregulation and economic distress: Is there a future for financial co-operatives? Euricse Working Papers n. 01 / 09. Amess, K. (2002). Financial institutions, the theory of the firm and organizational form. The Service Industries Journal 22(2), pp. 129-148. http://dx.doi.org/10.1080/714005069. Ayadi, R., Schmidt, R.H., Carbó Valverde, S. (2009). Investigating diversity in the banking sector in Europe: The performance and role of savings banks. Brussels: Centre for European Policy Studies. Birchall, J. (2013). The potential of co-operatives during the current recession; theorizing comparative advantage. Journal of Entrepreneurial and Organizational Diversity 2(1), pp. 1-22. http://dx.doi. org/10.5947/jeod.2013.001. Birchall, J., Hammond Ketilson, L. (2009). Resilience of the co-operative business model in times of crisis. Geneva: International Labour Organization. Blöchlinger, A., Leippold, M., Maire, B. (2012). Are ratings the worst form of credit assessment apart from all others. Working Paper, University of St. Gallen. Bülbül, D., Schmidt, R.H., Schüwer, U. (2013). Caisses d’epargne et banques cooperatives en Europe. Revue d’Economie Financiere 111, pp. 159-187. Caporale, G.M., Matousek, R., Stewart, C. (2012). Ratings assigments: Lessons from international banks. Journal of International Money and Finance 31, pp. 1593-1606. http://dx.doi.org/10.1016/j.jimonfin.2012.02.018. Coco, G., Ferri, G. (2010). From shareholders to stakeholders finance: A more sustainable lending model. International Journal of Sustainable Economy 2(3), pp. 352-364. http://dx.doi. org/10.1504/IJSE.2010.033399. Desrochers, M., Fischer, K.P. (2005). The power of networks: Integration and financial cooperative performance. Annals of Public and Cooperative Economics 76(3), pp. 307-354. http://dx.doi. org/10.1111/j.1370-4788.2005.00281.x. Esty, B.C. (1997). A case study of organizational form and risk shifting in the savings and loans industry. Journal of Financial Economics 44(1), pp. 57-76. http://dx.doi.org/10.1016/S0304405X(96)00008-6. Ferri, G., Kalmi, P., Kerola, E. (2013). Governance and performance: Reassessing the pre-crisis situation of European Banks. In Goglio, S., Alexopoulos, Y. (Eds.). Financial cooperatives and local development. Abingdon, UK: Routledge, pp. 37-54. Ferri, G., Liu, L-G. (2007). Government guarantees behind banks: International evidence from the ratings. Unpublished manuscript, University of Bari and Hong Kong Monetary Authority. Fischer, K.P., Mahfoudhi, R.M. (2002). Corporate governance and ratings: Do agencies rate mutual bank bonds fairly. Unpublished manuscript, University of Laval.

13

For a full methodology for Moody’s Financial strength factors, see Moody’s (2012) and references therein. 54 53 JEOD - Vol.3, Issue 1 (2014)


Organizational Structure and Exposure to Crisis among European Banks: Evidence from Rating Changes Ferri, G.; Kalmi, P.; Kerola, E.

Fitch (2012). Fitch ratings global corporate finance 2011 transition and default study. Flageole, M-A., Roy, J. (2005). Rating cooperative and commercial bank bonds: A comparative approach. Annals of Public and Cooperative Economics 76(3), pp. 407-35. http://dx.doi.org/10.1111/j.13704788.2005.00284.x. Fonteyne, W. (2007). Cooperative banks in Europe – Policy issues. IMF Working Paper WP/07/159, Washington DC. Gardener, E.P.M., Molyneux, P., Williams, J., Carbo, S. (1997). European savings banks: Facing up to the new environment. International Journal of Bank Marketing 15, pp. 243-54. http://dx.doi. org/10.1108/02652329710194937. Hau, H., Langfield, S., Marquez-Ibanez, D. (2013). Bank ratings: What determines their quality. Economic Policy 28, pp. 289-333. http://dx.doi.org/10.1111/1468-0327.12009. Hau, H., Thum, M. (2009). Subprime crisis and board (In-)competence: Private vs. public banks in Germany. Economic Policy 24(60), pp. 701-52. http://dx.doi.org/10.1111/j.1468-0327.2009.00232.x. Hermalin, B.E., Wallace, N.E. (1994). The determinants of efficiency and solvency in savings and loans. RAND Journal of Economics 25(3), pp. 361-381. http://dx.doi.org/10.2307/2555767. Iannotta, G., Nocera, G., Sironi, A. (2007). Ownership structure, risk and performance in the European Banking Industry. Journal of Banking and Finance 31, pp. 2127-2149. http://dx.doi.org/10.1016/j. jbankfin.2006.07.013. Iannotta, G., Nocera, G., Sironi, A. (2013). The impact of government ownership on bank risk. Journal of Financial Intermediation 22, pp. 152-176. http://dx.doi.org/10.1016/j.jfi.2012.11.002. IMF (2012). Spain: The reform of Spanish savings banks technical notes. IMF Country Report 12 / 141, International Monetary Fund, Washington DC. John, K., John, T.A., Sembet, L.W. (1991). Risk shifting incentives of depository institutions: A new perspective on federal deposit insurance reform. Journal of Banking & Finance 15(4-5), pp. 895915. http://dx.doi.org/10.1016/0378-4266(91)90105-U. Kalmi, P. (2012). Finnish cooperative banks and the crisis of early 1990s. In Mooij, J., Boonstra, W.W. (Eds.). Raiffeisen’s footprint: The cooperative way of banking. Amsterdam: VU University Press, pp. 181-96. Körnert, J. (2012). Swedish cooperative banking in the early 1990s: A decade of crisis and transition. In Mooij, J., Boonstra, W.W. (Eds.). Raiffeisen’s footprint: The cooperative way of banking. Amsterdam: VU University Press, pp. 217-30. Langohr, H., Langohr, P. (2009). The rating agencies and their credit ratings: What they are, how they work, and why they are relevant. Chichester, UK: Wiley. Leogrande, A. (2013). Cooperative banks vs. financial crisis: An application of stakeholder vs. shareholder debate. Unpublished manuscript, University of Bari. Moody’s (2012). Bank financial strength rating. http://v2.moodys.com/cust/prodserv/prodserv. aspx?source=StaticContent/Free%20Pages/Products%20and%20Services/Static%20Projects/ GBRM/bfsr.html&viewtemplate=/templates/mdcHeaderFooter.xml Accessed last on 26 May 2013. Mooij, J., Boonstra, W.W. (2012, Eds.). Raiffeisen’s Footprint: The cooperative way of banking. Amsterdam: VU University Press. Morgan, D.P. (2002). Rating banks: Risk and uncertainty in an opaque industry. American Economic Review 92(4), pp. 874-88. http://dx.doi.org/10.1257/00028280260344506. Poon, W.P.H., Firth, M., Fung, H.G. (1999). A multivariate analysis of the determinants of Moody’s Bank financial strength ratings. Journal of International Financial Markets, Institutions and Money 9(3), pp. 267-83. http://dx.doi.org/10.1016/S1042-4431(99)00011-6. Rasmusen, E. (1988). Stock banks and mutual banks. Journal of Law and Economics, 31, pp. 395-422. http://dx.doi.org/10.1086/467162. 55 54 JEOD - Vol.3, Issue 1 (2014)


AT T R I B U T I O N 3 . 0

You are free to share and to remix, you must attribute the work

Vol.3,Issue Issue (2014) 57-85 Publication date: 17 June2014 June 2014||Vol.3, Volume 3, Issue 11(2014) 1 (2014) 1-8 56-85

AUTHOR YASMINA LEMZERI CEREFIGE UniversitĂŠ de Lorraine yasmina.lemzeri@univ-lorraine.fr

Did the Extent of Hybridization Better Enable Cooperative Banking Groups to Face the Financial Crisis? ABSTRACT The 2008 financial crisis affected both cooperative and joint-stock banking groups. But since these groups had adopted different forms and modes of governance, cooperative banks might have suffered less. Cooperative banking groups are seen as more risk-averse than jointstock banking groups. One possible explanation is that they are owned by their members and unlisted; another reason could be the extent of their presence in a local area, which enables them to reduce information asymmetry. Joint-stock banking groups are seen as more ready to take risks. As they are held by stockholders requiring high-returns, they are more motivated to undertake risky projects. As cooperative banking groups have evolved, some have adopted joint-stock banking group features. This evolution can have more important consequences on their management style. To study whether cooperative banking groups faced the financial crisis better than jointstock groups, we compared their sensibility to the financial crisis and their contribution to financial stability. We built a sample composed of European cooperative and joint-stock banks and computed a z-score indicator, reflecting the probability of bankruptcy. A dummy variable set for the governance criteria distinguishes between the different types of cooperative banking groups. We used a data panel treatment to highlight the potential differences due to governance factors over the entire period studied (2002-2011); we then divided this period into three sub-periods to determine whether some banks, according to the extent of hybridization, showed on the one hand more resistance, and on the other more resilience. Our principal conclusion is that cooperative banking groups that have retained the main features of their original model while diversifying their activities have contributed most to financial stability.

KEY-WORDS COOPERATIVE BANKS; HYBRIDIZATION; FINANCIAL STABILITY; FINANCIAL CRISIS; RESISTANCE; RESILIENCE

JEL Classification: G320 | DOI: http://dx.doi.org/10.5947/jeod.2014.004

57 56 JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

1. Introduction The financial crisis that began in mid-20071 with the subprime lending scandal in the United States does not seem to have spared any bank. Regardless of legal statuses, all of them suffered as the crisis unfolded. As each bank pursues different objectives, one may presume that the crisis did not have the same impact on all banks. The banking area is mainly composed of cooperative and joint-stock banking groups. Cooperative banking groups are network constructs composed of local and/or regional banks. The owners are also customers. These banks may be affiliated to a structure that offers them logistical support, or alternatively can hold (or be held, even partially by) other unifying activities. So, if we consider the whole structure, we conceive the notion of a group. Cooperative banking groups have evolved since their creation: some have listed subsidiaries, others are partially listed, and some have retained the main features of the original cooperatives. Even within one country, cooperative banking groups can behave differently. These notable differences in organizational models enable us to report that some are now built on a hybrid model that confers the properties of both joint-stock and cooperative banking groups. In this paper we discuss the degree of hybridization of the cooperative model and examine whether the extent of hybridization affected the ability of cooperative banking groups to face the financial crisis. We begin by conducting a theoretical analysis that explains why cooperative banking groups are expected to contribute more to financial stability than joint-stock groups. We also describe the evolution and process of hybridization that brings some cooperative banking groups closer to joint-stock groups. Then, we explore empirically whether the extent of a banking groupâ&#x20AC;&#x2122;s hybridization played a role in its contribution to financial stability. This is represented by a number of endogenous variables as the z-score characterizes the probability of bank bankruptcy, the loans to assets ratio shows a bankâ&#x20AC;&#x2122;s capacity to maintain its lending level even in times of crisis, and the return on equity shows the evolution of financial performance during three different sub-periods. Following a descriptive statistics analysis of the endogenous variables retained, we present an econometric data panel model in order to reinforce the initial results, and, finally, we conclude.

2. Theoretical analysis 2.1. Cooperative banking groups benefit from a governance model that plays a stabilizing role in periods of crisis Cooperative banking groups pursue low-risk goals. Cooperative banks are held by their member-owners, who are also their customers. To preserve the continuity of their bank, they are not inclined to take risks that could lead to bankruptcy. Being a shareholder does not confer the right to a dividend, so whatever the level of profit realized by the bank, it has no effect on shareholder wealth (Gurtner et al., 2006). This is an important asset compared to joint-stock banks. The latter are listed and their stockholders want to maximize their wealth, so they exercise pressure on management to ensure that the maximum profit will be realized. To achieve this, the management is motivated to undertake excessively risky projects. This difference in ownership shows that cooperative banking groups are more risk-averse than joint-stock banks.

1

The crisis began in mid-2007 but the effects were really perceptible in the 2008 balance sheets and income statements of the banks. 58 57 JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

Cooperative banking groups are protected from hostile takeovers. Cooperative banks are not listed, so they cannot be easily bought by another bank (Gurtner et al., 2002). Obligations to reach a minimal level of profitability do not worry cooperative banks. Even if they discover inefficiencies, they are not afraid of a hostile purchase and so can continue to focus on objectives that do not feed instability. Members in cooperative banks are focused on intergenerational inheritance. Members want to preserve their inheritance in order to transmit it to the next generation. Indeed, another specificity of this structure is, generally, the impossibility of members’ sharing the accumulated profit (Soulage, 2000). It must remain the property of the cooperative and be transmitted to succeeding shareholders. This obligation to consolidate the profit to the banks’ reserves helps reinforce cooperative banks’ powerful financial stability. Networked cooperative banks are able to reduce information asymmetry. Cooperative banks are built as networks. They are settled throughout an entire area. This strength of deployment enables them to be close to their customers. The latter includes not only households, but also small and medium-sized firms. This proximity enables the banks to know their customers better and reduce the information asymmetry that appears before the beginning of a contractual relation (Amess, 2002). The territorial establishment of independent local banks is a substantial advantage compared to joint-stock banks. Cooperative banks are better able to determine their customers’ real-risk profiles and may adjust the risk premium according to the specific ability to refund a loan. Cooperative banks meet the economic needs of small and medium sized-firms. Joint-stock banks are less attracted to small and medium-sized firms, which have a reputation for being more vulnerable to bankruptcy than larger firms. By contrast, thanks to their territorial distribution and closeness to customers, cooperative banks are more likely to actively contribute to the development and maintenance of smaller firms (Angelini et al., 1998). They grant loans even to firms that are perceived as weak; by diversifying their lending portfolio among the activity areas of small and medium-sized firms, they minimize the risk undertaken. Cooperative banks do not grant loans to firms within a single business sector; however, they do grant loans to firms in a variety of loosely correlated business sectors. A welldiversified lending portfolio means that if one particular business sector is hit by a crisis, it will have little or no impact on the bank. Cooperative banking groups are seen as important participants in local economic development and can help limit the effects of a credit crunch in times of crisis. Their implication is also justified by the fact that the managers of the local banks are also entrepreneurs. They know how difficult it is for small firms to obtain loans, they are familiar with the environment in which small and mediumsized firms operate, and they understand their needs. In return, entrepreneurs among the shareholders will expect the bank to continue to support them in times of crisis through granting loans. Cooperative banks aim to develop long-term valorization Banks obviously need to realize a profit. Since this is not the main concern of cooperative banks, perceived as “resource wasters”, select authors have stated that they are doomed to fail (Hansmann and Krackmann, 2001). They have stressed that the costs registered by these banks are too high (Akella and Greenbaum, 1988); and as these banks are not motivated to lower their costs in order to improve profits, they are considered as inefficient banks. However, the aim of cooperative banks is to develop long-term valorization (Allen and Gale, 2004). They do make a profit, a share of which is forwarded into reserves. 59 58 JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

Shareholder members aim to stabilize their bank over the long term; this point increases the difference between cooperative and joint-stock banks. Cooperative banks’ compensation policy is less performance-related. Cooperative banks practice different compensation policies compared to joint-stock banks. They do not use motivating wage tools related to market exchange value, for example, stock options. These appeared in joint-stock banks to ensure that managers would have the same interests as stockholders in the firm; to increase wealth, they must maximize firm value, and in order to achieve that goal, they undertake high-risk projects. If they are successful, the firm will do well. But this compensation policy has a devastating effect when projects fail, as they can lead to bankruptcy of both the bank and its stakeholders (Beltratti and Stulz, 2011). All these factors seem to confer a status that cooperative banking groups have more financial stability than joint-stock banks. While they do not enable a bank to maximize its profit and achieve high performance in times of favorable growth, in times of financial crisis they enable the bank to amortize its losses. Those factors can lead us to expect that cooperative banking groups: - will be more stable and resistant than joint-stock banking groups in times of financial crisis as they are more capitalized and own fewer risky assets (hypothesis 1, tested in the first econometrical model); - will continue to maintain their superior level of lending to small and medium-sized firms (hypothesis 2, tested in the second econometrical model); - will have weaker financial performance than joint-stock banking groups in crisis-free periods. However, this difference will disappear in periods of crisis as joint-stock banking groups that are focused on financial performance may realize poor or negative profits (hypothesis 3, tested in the third econometrical model). Before we embark on the empirical study, we need to classify cooperative banking groups. Since their creation, they have evolved differently, enjoying a particularly strong growth period in the 1990s when they took advantage of the financial crisis that affected the joint-stock banks. During that period, some cooperative banks used their huge reserves to acquire joint-stock banks. Some cooperatives now have characteristics that are close to or farther from the original cooperative banking group model. Here, we use the term “hybridization degree” to describe the extent to which they have adopted the features of joint-stock banks or retained the cooperative ones. 2.2. Cooperative banking groups’ diversity and hybridization degree Some cooperative banks have adopted features that bring them closer to joint-stock banks, while retaining some features of cooperative banks. These modifications lead us to use the term “hybridization of the cooperative model”. Among cooperative banking groups, we find some that can be described as banking groups under cooperative control, as regards the capital link2. These are networks composed of local and/ or regional banks that are strongly integrated in the group, but the group also holds listed subsidiaries that can either be placed under the control of a holding, or under one of the listed apexes. The apex loses its cooperative status once it is listed. Any customer of this type of bank can access the bank’s entire range of banking products, including the most sophisticated, just like in a joint-stock bank. They belong to the most hybrid category of cooperative banking groups. In this category we find the French Crédit Agricole and BPCE groups, the Austrian Raiffeisen group, the Italian Istituto Centrale delle Banche Popolari group, 2

The capital link deals with the fact that cooperative local or regional banks are generally the main owners of the shares of the apex; and so, they own the majority of capital and voting rights. 60 59 JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

and the Finnish OP-Pohjola group. Among cooperative banks, we also find those that intervene in a particular area and do not truly form a group. They are composed of local banks that freely affiliate themselves with a group that provides logistical support. Those banks could not afford to finance their own IT platform; for example, as the costs of functioning would exceed the benefits for an insufficient number of operations. So, it is in their interest to share a common platform by affiliating with a group, so as to achieve economies of scale. The group does not own any listed subsidiary. This is the case for Spanish Cajas Rurales, UK building societies, the Portuguese CCCAM, Austrian Volksbanken and the Italian BCC. These belong to the least hybrid category in this research paper. Finally, some cooperative banking groups form an intermediate category of hybridization degree. They are composed of numerous decentralized local banks and unlisted subsidiaries that operate in various areas (corporate and investment banking, market finance, etc.). In this category, we find the German DZ Bank group, the Dutch Rabobank group, and the French Crédit Mutuel and Crédit Coopératif groups. We decided to include the French Crédit Mutuel group in this category, even though it holds a listed subsidiary (CIC). There is a very small share of stocks that do not belong to the Crédit Mutuel and are in free exchange. This free float cannot have a negative effect on the stability of CIC; if all investors decided to sell their stock at the same time, neither CIC nor Crédit Mutuel would suffer. As there are important differences in governance amongst cooperative banking groups on the one hand, and between cooperative and joint-stock banking groups on the other, it is important to examine and measure the impact of these differences. We approached this by first conducting a descriptive statistics analysis to see how three representative indicators of financial stability evolved during (a) the whole period studied and (b) during three discrete periods within that time span: before the crisis, the crisis year, and after the crisis. The target was to measure the relative levels of resistance and resilience between cooperative and joint-stock banking groups. In order to increase the descriptive statistics analysis robustness, we conducted an econometric analysis using the data panel method.

3. Empirical analysis 3.1. Methodology To analyze how hybridization degree affected bank stability during the financial crisis, we began by first defining the different concepts, then leading a descriptive statistics analysis, and finally, strengthening the first observations stemming from this analysis by carrying out an econometric study. Concept definitions. In this analysis, we divide the whole period (2002-2011) into three periods in order to examine the behaviour of each category before the crisis (2002-2007), during the crisis (2008) and after the most important phase of the crisis (2009-2011), using the concepts of resistance and resilience. Thus, a bank is considered resistant if it was able to maintain its level of management and financial indicators during the crisis, or if it sustained a lower deterioration of its indicators compared with other banks. A bank is said to be resilient if it was able to overcome the crisis. If it suffered from a deterioration of its indicators during the crisis, it will show more resilience if, post-crisis, it manages to present indicator levels close to pre-crisis ones. 61 60 JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

The concepts of resistance and resilience are used to describe a bankâ&#x20AC;&#x2122;s financial stability. According to the European Central Bank, financial stability is the ability to resist shocks and absorb financial imbalances. In this paper, we classify the cooperative banking groups according to their hybridization degree but put joint-stock banking groups under a single category since they all function the same way. A cooperative banking group that belongs to a defined hybridization category will contribute more to financial stability if it shows greater resistance during the year of the crisis than the other banking groups in the period of reference (2002-2007). This will be measured in terms of its results, which will be better than those of the other banking groups in the period of reference. A banking group that has a defined hybridization degree will contribute to financial stability if it is resilient, that is to say, if it performs better in the post-crisis period than in the crisis period. To test the difference in financial contribution according to hybridization degree, we used three endogenous indicators: a computed z-score, the loans to assets ratio, and the return on equity. Building the database. To create our database, we retained all the cooperative banking groups within the European Union. In order to compare the contribution of cooperative banks and joint-stock banks to the financial stability, we collected the same data for the joint-stock banking groups established in the same countries with the largest total assets. This last criterion should support this study; as cooperative banking groups have rather high total assets profiles, it is important to compare them with joint-stock banks with the same profiles. Here, we consider that a group comprises the local bankâ&#x20AC;&#x2122;s activities plus the activities of the subsidiaries held by the group. In this way, we consider the whole activities of the group. The data were obtained from the Bankscope database. For some cooperative banking groups, we had to proceed to an aggregation of data. Indeed, some groups, such as DZ Bank in Germany, do not take into account data produced by local banks. In this case, we had to aggregate group data with data from local banks to obtain data that truly reflect the groupâ&#x20AC;&#x2122;s performance. Our database is composed of 15 cooperative banking groups and 49 jointstock banks in Austria, Finland, France, Germany, Italy, the Netherlands, Portugal, Spain, and the United Kingdom (the countries in which cooperative banking groups are found). The number of observations for the whole period was 601. Classification according to hybridization degree. Once the data were collected, we classified the cooperative banking groups according to their hybridization degree. We consider that a cooperative banking group has a low level of hybridization if it is composed of independent and decentralized local banks. The latter have a free affiliation to the group, which has no listed subsidiaries. We called this category LHCoop (least hybrid cooperative banking groups). The second, intermediate category comprises local or regional banks, with no listed subsidiaries except where the public share is very low. We named this category IHCoop (intermediate hybridization cooperative banking groups). The final category refers to cooperative banking groups that have adopted some features of joint-stock banking groups. They are partly listed and hold a listed subsidiary that they have built ex-nihilo or have acquired from a commercial rival; they are also subject to stockholder pressure, unlike traditional cooperative banking groups, which have none of these features, and are likely to be penalized more heavily during a financial crisis. We named them MHCoop (most hybrid cooperative banking groups). We refer to the joint-stock banks in our sample as JSB.

62 61 JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

3.2. Descriptive statistics analysis (Appendices 1, 2 and 3) This analysis precedes the econometrical analysis. It enables us to see how banks in the four categories behave during each of the three periods in terms of financial stability, ability to maintain their level of lending and financial performance. The different periods are: - 2002-2007: the period before the crisis. This is the period of reference, as presented earlier. - 2008: the year of the financial crisis. We compare changes in the indicators before and during the crisis to see if banks, according to their hybridization degree, were more resistant to the crisis and therefore contributed more to financial stability. - 2009-2011: the post-crisis period. We will see whether the banks show resilience by examining the improvement in their results following the crisis. Definitions of the endogenous variables used as proxies of contributions to financial stability. Z-score: we used a z-score indicator to determine whether cooperative banking groups, according to their hybridization degree, contribute more to financial stability. There are different definitions of the z-score in the literature. Although Altman (1968) pioneered the definition, we focus our analysis on the definition used in the IMF studies (Hesse and Cihak, 2007) and by Laeven and Levine (2008). Our preference is explained by the presence of the return on assets standard deviation as a denominator, which enables risk measurement. The z-score formula we use is: z-score = (K+µ)/ σ, where K is the capital funds to total assets ratio, µ represents the return on assets ratio and σ represents the return on assets standard deviation. The z-score indicator takes into account the size of the bank’s capitalization, its assets return performance and the risk contained in the latter. It can be considered as a solvency probability estimator as it measures the number of standard deviations the return on assets must lose to reach bankruptcy; the higher this indicator is, the more the bank contributes to financial stability. The advantages of this indicator are the ease of collecting data to build it, the ease of interpretation and the ability to measure the probability of bankruptcy. Loans to assets: we use this ratio to examine a bank’s lending level. If this ratio is raised, a bank allocates an important part of its assets to financing the economy. Return on equity: this ratio measures the financial performance realized by a bank. If it is high, it indicates a bank makes a substantial profit. It could also indicate that a bank invests in risky projects that offer high returns. Cooperative banking groups present higher indicators before the crisis. Before the crisis, cooperative banking groups with low and intermediate degrees of hybridization present higher z-score means (Appendix 1) than joint-stock banks. The z-score mean of the most hybrid banks is inferior to that of the joint-stock banks. At this stage, cooperative banking groups that have retained the main features of the original cooperatives contribute more to financial stability than joint-stock banks. Moreover, all the cooperative banking groups have a loans to assets ratio (Appendix 2), which is on average higher than that of joint-stock banks. Cooperative banks that have the lowest hybridization are those that lend most to the economy. As regards to financial performance (Appendix 3), joint-stock banks performed best (12%). We notice that cooperative banking groups benefit from an average return indicator that is close to that of joint-stock banks. The most hybrid cooperative banking groups have an indicator of 11 per cent; the least hybrid banks have an indicator 63 62 JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

of 9.7 percent. So, before the crisis, the situation was advantageous for cooperative banking groups, and particularly for those that had a low or intermediate hybridization degree. The least and intermediate hybrid cooperative groups are more resistant in time of crisis. In 2008, during the crisis, we notice that the intermediate hybridization degree banks saw an improvement in the z-score mean. The least hybrid banks suffered from a deteriorating z-score mean; nonetheless, it remained higher than that of the joint-stock banks. The latter suffered from an important deterioration of their z-score mean, as did the most hybrid degree cooperative banks. But, the z-score of the joint-stock banks was higher than the most hybrid cooperative banks. Regarding the granted loansâ&#x20AC;&#x2122; preservation level, banks in all categories slightly improved, on average, their ratio, with the exception of banks belonging to the most hybrid degree. The highest improvement is observed among the least hybrid banks, followed by intermediate hybrid banks and then by the joint-stock banks. Regarding financial performance evolution, we first notice that all the bank categories suffered from a drop in their indicator. But the largest declines were registered by the joint-stock banks and the most hybrid banks. On average, the least and intermediate hybrid banks presented the highest indicators. We can thus say that the cooperative banks belonging to the intermediate and least hybridization degree showed higher resistance than banks in the other categories. The intermediate and most hybrid cooperative groups seem more resilient. During the 2009-2011 period, almost all the banks improved their z-score mean. The exception was those in the least hybrid category; their z-score mean continued to deteriorate, becoming the lowest of all categories. Intermediate hybrid cooperative banks showed the highest improvement in this indicator, followed by the most hybrid cooperative banks and lastly by the joint-stock banks. As regards to the level of lending, we note that only the banks in the intermediate and most hybrid categories improved their ratios. The joint-stock banks were the lowest but held a steady lending profile. The least hybrid banks suffered from a slight drop in their indicator. Finally, the evolution of the financial performance indicator benefited the intermediate hybrid banks, which improved their average indicator (the highest indicator in this period). They were followed by the most hybrid banks, which greatly improved their average ratio. The other bank categories realized negative performances. We conclude that cooperative banks in the intermediate and most hybrid categories show more resilience in term of contribution to financial stability. The initial results show that the least and intermediate hybrid banks show more resistance during a financial crisis, while the intermediate and the most hybrid banks show more resilience during the post-crisis period. As these results are drawn from the descriptive statistics analysis, we attempt to reinforce them through an econometrical analysis. This will enable us to determine if there is a different contribution to financial stability during the various periods, according to the hybridization degree. 3.3. Econometric analysis Data panel treatment choice. The purpose of this study was not to follow the evolution of financial stability over a period of time, but to examine whether, during distinct periods, cooperative and joint-stock banking groups behaved in different ways in terms of financial stability. To reach this goal, the econometric tool which meets our needs is the data panel method. The variation of the z-score, the loans to assets and the return on equity are thus not only studied over determined periods but also for a sample of individuals. The database from which we work is an unbalanced panel containing 601 observations. As we work on data panel treatment, we had to treat different points as the unitary root process presence. Our database does not contain sufficient data to use (augmented) Dickey-Fuller or Philips Perron tests. To avoid obtaining biased 64 63 JEOD - Vol.3, Issue 1 (2014)


to treat different points as the unitary root process presence. Our database does not contain sufficient data to use (augmented) Dickey-Fuller or Philips Perron tests. To avoid obtaining unbalanced panel 601Levin, observations. we test, workwhich on data treatment, we had biased results, wecontaining applied the Lin andAs Chu is panel specifically adapted to small Didcontaining the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? unbalanced panel 601 observations. As we work on data panel treatment, we had to treat different points the found unitarythat root Lemzeri, process presence. indicators Our database doeswas not contain samples (Appendix 5).asWe the endogenous series stationary on Y. to treat different points as the unitary root processorpresence. Our database does notobtaining contain sufficient data to use (augmented) Dickey-Fuller Philips Perron tests. To avoid level. We have also taken care to determine whether we were in the case of random or fixed sufficient data to (augmented) Dickey-Fuller Philips tests. Toadapted avoid obtaining biased results, weuse applied the Levin, Lin and Chuor test, whichPerron is specifically to small with effects. We supposed that there was a random effect on the coefficient associated biased results, we applied the Levin, Lin and Chu test, which is specifically adapted to small samples (Appendix 5). We foundby that the endogenous indicators series stationary on individuals. This was Hausman test (Appendix 4). was samples (Appendix 5).confirmed We found thatthethe endogenous indicators series was stationary on level. We have also taken care to determine whether we were in the case of random or fixed

results, we the Lin and to Chu test, whichwhether is specifically small samples (Appendix We level. Weapplied have alsoLevin, taken we adapted were intothe case of random or 5). fixed effects. We supposed thatcare there determine was a random effect on the coefficient associated with Description of the models found thatWe theThis endogenous indicators stationary on level. also takenassociated care to determine effects. supposed that therebyseries was a random effect on We thehave coefficient with individuals. was confirmed the was Hausman test (Appendix 4). Inwethe first model, usebythe z-score as We a proxy forthat financial stability. z-score whether were in was the case ofwe random or effects. supposed there was a randomThe effect on the is the individuals. This confirmed thefixed Hausman test (Appendix 4). endogenous we attempt toconfirmed explain its variability a set of 4). dummy variables coefficient associated with and individuals. This was by the Hausman with test (Appendix Description ofvariable the models

that represent level of the hybridization degree, along with otherTheexplanatory Description thethe models In the of first model, we use z-score as a proxy for financial stability. z-score is thecontrol In theThe first model, we useattempt themodel, z-score proxy forasfinancial stability. z-score is the variables. model is: endogenous variable and to explain itsz-score variability with afinancial set of The dummy variables Inwe the first weas usea the a proxy for stability. The z-score Description of the models. endogenous variable and we attempt to explain its variability with a set of dummy variables that represent the level of hybridization degree, along with other explanatory control is the endogenous variable and we attempt to explain its variability with a set of dummy variables that 2 that represent the+ β1 level of hybridization degree, along with other explanatory variables. is: Z-Score = model αi β3 + control β5i Loans i, t,The clevel i LHCoop i,t + β2 i IHCoop i,t + explanatory i MHCoop i,t + β4 i Loans i,t + i,t model represent the of hybridization degree, along with other control variables. The is: variables. The model is: β6i NIMi,t + β7i Feesi,t + ∑ φt Crisist + β8c Ratec,t + β9i CTIRi,t + ɛi,t ∀ i =2 1,..,64 Z-Scorei, t, c = αi + β1i LHCoopi,t + β2i IHCoopi,t + β3i MHCoopi,t + β4i Loans i,t + β5i Loansi,t + 2 Z-Score αii Fees + β1i,ti LHCoop i,t + β2 i,t + β5i Loansi,t + β6which: + ∑ φt Crisis β8c Ratei,tc,t++β3 β9i MHCoop ∀ ii Loans = 1,..,64 i NIMi,i,tt, c+=β7 t +i IHCoop i CTIRi,t +i,tɛ+ i,t β4 In β6i NIMi,t + β7i Feesi,t + ∑ φt Crisist + β8c Ratec,t + β9i CTIRi,t + ɛi,t ∀ i = 1,..,64 - which: LHCoop, IHCoop, MHCoop are a set of dummy variables that describe the degree of In In which: hybridization. LHCoop the least hybrid banks, IHCoop contains In - which: LHCoop, IHCoop, MHCoop are includes a set variables that cooperative describe degree ofthe hybridization. - LHCoop, IHCoop, MHCoop areof adummy set of dummy variables thatthedescribe degree of cooperative banks that have an intermediate level of hybridization, and MHCoop LHCoop includes the least hybrid cooperative banks, IHCoop contains cooperative banks that have - LHCoop, IHCoop, MHCoop are a set of dummy variables that describe the of hybridization. LHCoop includes the least hybrid cooperative banks, IHCoopdegree contains represents the most hybrid cooperative banks. In this set of dummies, the omitted hybridization. LHCoop the least hybrid cooperative contains cooperativelevel banks that includes have an intermediate level ofthehybridization, and MHCoop an intermediate of hybridization, and MHCoop represents mostbanks, hybridIHCoop cooperative banks. dummy isthe JSB, which represents thebanks. joint-stock and is the reference cooperative banks that have an intermediate level ofbanks hybridization, and MHCoop represents most hybrid cooperative In this set dummies, the omitted In this set of dummies, the omitted dummy is JSB, which represents theofjoint-stock banks and is thein our model; represents the hybrid cooperative banks. In banks this set the omitted dummy ismodel; JSB,most which represents the joint-stock andof isdummies, the reference in our reference in our 2 dummy is the JSB,quadratic which represents the joint-stock banks and is loans the reference inratio; our model; loans to assets ratio and Loans is the to assets Loans 2 - Loans is the2 quadratic loans to assets ratio and Loans is the loans to assets ratio; model; isthe thenet quadratic to assets ratio and Loans is the loans to assets ratio; - Loans NIM interest margin; - NIM is the 2is net interest margin;loans is the quadratic loans to assets ratio and Loans is the loans to assets ratio; --- Loans NIM the net interest margin;income Fees isrepresents theofamount fees divided and income - fees represents the amount fees and of by totaldivided assets; by total assets; -- NIM is the net interest margin; represents the amount of fees and income divided by total assets; - Fees Crisis is a temporal dummy variable; - crisis is a temporal dummy variable; -- Fees represents the amount fees and income divided by total assets; a temporal dummyof variable; - rate the long-term interest rate; - isCrisis Rate isis the long-term interest rate; -- Crisis is a temporal dummy variable; Rate is the long-term interest rate; - CTIR is the cost to cost income - CTIR is the to ratio. income ratio. -- Rate interestratio. rate; CTIRisisthe thelong-term cost to income - CTIR is the cost to income ratio. In the secondmodel, model, the endogenous variable is the return onreturn equity.on Weequity. want to We examine theto evolution In second theendogenous endogenous variable is the want examine the In the the second model, the variable is the return on equity. We want to examine the of financial performance over the entire period, as well as within separate periods, according to the degree evolution of financial performance over the entire period, as well as within separate periods, evolution of financial performance the entire as well as We within separate periods, In the second model, the endogenousover variable is theperiod, return on equity. want to examine the according to the degree of hybridization. As control variables, we introduce the long-term of hybridization. As control variables, we introduce the long-term interest rate and the net interest margin. accordingoftofinancial the degree of hybridization. control variables, the long-term evolution performance over theAs entire period, as wellwe as introduce within separate periods, interest rate and thenet netinterest interest margin. model is: The modelrate is:toand interest margin. TheThe model is: variables, according the the degree of hybridization. As control we introduce the long-term interest rate and the net interest margin. The model is: αi++β1 β1i LHCoop + β2 +i MHCoop β3i MHCoop β4i i,tNIM +β5c,t ROE +β5i,tc Rate + c,t + ROE i, t,t, cc == αi i LHCoop i,t β2 i IHCoop i,t i+NIM c Rate i,t + i IHCoop i,t +i,tβ3 i,t + β4 ∑ φ Crisis + ɛ ∀ i = 1,..,64 = αi + β1 LHCoop + β2 IHCoop + β3 MHCoop + β4 NIM +β5 Rate ROE ∑ t i,Crisis tt + ɛ i,ti,t ∀ t, c i i = 1,..,64 i,t i i,t i i,t i i,t c c,t + In third we use to to assets ratioratio as an variable to see to if the ∑ φthe ɛi,t ∀ i = 1,..,64 In thirdt +model, model, we usethetheloans loans assets asendogenous an endogenous variable see if the t Crisis banks, according to their hybridization degree, have maintained their level of lending. As In the third model, we use the loans assets ratio as an endogenous variable to see if the banks, In the third model, we use the loans to assets ratio as an endogenous variable to see if thea As a banks, according to their hybridization degree, have maintained their level of lending. control variable, we keep only the long-term interest rate. The model is: banks, to degree, havetheir maintained their level of lending. As we a according to their hybridization degree, maintained level of lending. Asis: a control variable, controlaccording variable, wetheir keephybridization only thehave long-term interest rate. The model control variable, we keep only the long-term interest rate. The model is: keep only the long-term interest rate. The model is: = αi + β1 LHCoopi,t + β2i IHCoopi,t + β3i MHCoopi,t + β4c Ratec,t + ∑ φt Crisist + ɛi,t Loans Loans i,i, t,t, cc = αi + β1i i LHCoop i,t + β2i IHCoopi,t + β3i MHCoopi,t + β4c Ratec,t + ∑ φt Crisist + ɛi,t ∀ i = 1,..,64 Loans i, t, c = αi + β1i LHCoopi,t + β2i IHCoopi,t + β3i MHCoopi,t + β4c Ratec,t + ∑ φt Crisist + ɛi,t ∀ i = 1,..,64 ∀ i = 1,..,64 3.3.1 Definition of the variables and expected signs 3.3.1 Definitionofofthe thevariables variables and expected signs 3.3.1 Definition Definition expected 3.3.1. of the variables and and expected signs signs To differentiate the three categories, we had to create a set of dummy variables. We determined the first 9 category as LHCoop: this dummy takes the value 1 if the cooperative banking group belongs to the least 9 hybrid category, 0 otherwise. As the banks that belong to this category have no diversified activities, we do not expect them to contribute a great deal to financial stability. If the bank is specialized in financing a single activity, for example real estate, if a crisis arises in this sector the bank will be annihilated. 65 64 JEOD - Vol.3, Issue 1 (2014)

9


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

IHCoop: this dummy takes the value 1 if the cooperative banking group has an intermediate degree of hybridization, 0 otherwise. These banks have diversified activities through unlisted subsidiaries. They retain the main features of a cooperative bank and are unlisted. So, the profile they benefit from should help them to contribute the most to financial stability. MHCoop: this dummy takes the value 1 if the cooperative banking group belongs to the most hybrid category, 0 otherwise. These banks are close to joint-stock banks, as they have listed subsidiaries or are themselves partially listed. They retain some features of the original cooperative banks but their proximity to joint-stock banks allows us to assume that there is no difference between their financial stability contributions. Finally, in JSB, we included all the joint-stock banking groups. In our model, JSB is the reference. This enables us to compare the financial stability of each category of cooperative banks relative to the joint-stock banking groups. To give our model more robustness, we decided to include other variables that could explain the endogenous indicators’ volatility. As each country benefits from specificities that can have some impact on the endogenous indicators’ variation, we also decided to introduce an indicator characterizing financial data appropriate to each country. Net interest margin. The net interest margin represents the net interest income expressed as a percentage of earning assets. If this ratio is high, it means the bank is able to obtain funding at low prices or is able to impose high margins on its customers. If this ratio is low, it means a bank obtains extensive funding or is unable to calibrate fair and sufficient margins for its customers according to their real risk profile. We expect it to have a positive effect on the z-score. As the z-score takes into account the return on assets, if the net interest margin increases, the realized return will also increase and the z-score will be higher. Cost to income ratio. Cost to income ratio measures the costs of running of a bank (for example staff salaries) divided by the income before provisions generated by the bank and thus represents an efficiency measure. We expect that this ratio will have a negative effect on the z-score. Indeed, costs and returns are highly and negatively correlated. If costs increase, the realized return will decrease and the z-score will drop. Fees and commissions to total assets. We divide a bank’s generated fees and commissions by its total assets. We thus try to measure the element of fees and commissions that are unrelated to lending, in order to see if the bank’s income sources are diversified. We expect a positive impact on the z-score, as an increase in fees and commissions will increase the level of returns. Loans to assets. We also want to know if the bank prioritizes granting loans or if it has other priorities besides lending. We divide the loans granted by each bank by its total assets. The higher the ratio, the greater the bank’s lending. We expect a quadratic specification and convexity. As we suppose that cooperative banking groups have a greater lending profile, we expect banks that lend the most will contribute more to financial stability. Long-term interest rates by country. This indicator measures the rate at which a country issues its 10-year bonds. If this indicator is high, it means a bank is located in a poorly rated country. If the rate is raised, the interest rate of the long-term bonds will be low, despite some exceptions, such as France. Since France lost its AAA rating, it has been borrowing money at lower rates. An increase in this rate will have negative consequences on the financial stability of banks in France. Indeed, an increase will contribute to the depreciation of bank assets. As the z-score is built on capital funds, an increase in rates will cause a decrease 66 65 JEOD - Vol.3, Issue 1 (2014)


indicator is high, it means a bank is located in a poorly rated country. If the rate is raised, the interest rate of the long-term bonds will be low, despite some exceptions, such as France. Since France lost its AAA rating, it has been borrowing money at lower rates. An increase in Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? this rate will have negative consequences onLemzeri, the financial stability of banks in France. Indeed, Y. an increase will contribute to the depreciation of bank assets. As the z-score is built on capital funds, an increase in rates will cause a decrease in asset value and consequently a lower zscore. Finally, when a country issues long-term bonds, banks are expected to participate in this operation by subscribing to these bonds. If the country is poorly rated, the risk undertaken in asset value and consequently a lower z-score. Finally, when a country issues long-term bonds, banks are by the bank will reduce its asset value and thus lower its z-score value. expected to participate in this operation by subscribing to these bonds. If the country is poorly rated, the risk undertaken by thevariable bank will reduce its asset value and thus lower its z-score value. 2008 crisis dummy

To appreciate the effect of the financial crisis on the z-score variation, we included a appreciate effect of financial crisis on z-score variation, we 2008 crisis dummyvariable, variable. To temporal dummy named 2008the Crisis. It the takes the value 1 the if the year is 2008; otherwise, the value is 0.variable, If it is anamed significant variable, then a role in theotherwise, z-score included a temporal dummy 2008 Crisis. It takes the2008 value played 1 if the year is 2008; variation. the value is 0. If it is a significant variable, then 2008 played a role in the z-score variation. 3.3.2 Results Results 3.3.2.

We conducted regressions under data panel treatment over four periods: the entire period under study (2002–2011), the pre-crisis period (2002–2007), the crisis year (2008) and conductedperiod regressions under data panel over four periods: thetoentire periodthe under study theWe post-crisis (2009–2011). The treatment main regression attempts explain z-score (2002-2011), the pre-crisis period (2002-2007), the crisisfinancial year (2008) and the post-crisis period (2009variability. The other regressions aim to explain performance and variability in 2011). The main regression attempts tothe explain z-score variability. The other regressions aim to explain lending levels. This demonstrates levelthe and evolution of contribution to financial stability financial performance and variability in lending levels. This demonstrates the level and evolution contribution according to the degree of hybridization throughout the different periods. In thisofway, we can establish banks ofto atheparticular hybridization degree show more periods. resilience and/or to financial whether stability according degree of hybridization throughout the different In this way, resistance during times of financial crisis. We introduced other control variables that we can establish whether banks of a particular hybridization degree show more resilience and/or resistance influenced thefinancial z-scorecrisis. variability. In each wevariables explainthatwhether thethecontrol during times of We introduced othercase, control influenced z-score variables variability. introduced in each regression are significant and whether their coefficient has the expected In each case, we explain whether the control variables introduced in each regression are significant and sign. whether their coefficient has the expected sign.

Model 1: Z-score as a measure of bank solvency

Model 1: Z-score as a measure of bank solvency Do cooperative banking groups present more robust z-scores according to their Do cooperative banking groups present more robust z-scores according to their hybridization degree? hybridization degree?

by explaining results of thewhich first has model, whichashas the z-scorevariable. as an WeWe first first beginbegin by explaining the resultsthe of the first model, the z-score an endogenous endogenous variable. The model was (see Appendix 6.1): The model was (see Appendix 6.1): Z-Scorei, t, c = αi + β1i LHCoopi,t + β2i IHCoopi,t + β3i MHCoopi,t + β4i Loans2i,t + β5i Loansi,t + β6i NIMi,t + β7i Feesi,t + ∑ φt Crisist + β8c Ratec,t + β9i CTIRi,t + ɛi,t ∀ i = 1,..,64 In this model, regardless of the period, we see that the control variables have the expected influence on z-score variability. Thethesign is variables coherenthave with hypothesis In this model, regardless of the period, we see that control the the expected influencewe on assessed earlier. As regards to the loans to assets ratio with a quadratic specification, the z-score variability. The sign is coherent with the hypothesis we assessed earlier. As regards to the loans to curveratio is convex. The minimum α on whole period 53.3 perαcent. assets with a quadratic specification, thethe curve is convex. Theisminimum on theThis wholeindicates period is that 53.3 banks that have a ratio superior to α contribute more to financial stability. The coefficient of per cent. This indicates that banks that have a ratio superior to α contribute more to financial stability. The this quadratic ratio is also significant in the post-crisis period. The curve remains convex and

coefficient of this quadratic ratio is also significant in the post-crisis period. The curve remains convex and the minimum α is 36.5 percent. As such we can corroborate our explanation of the results on the dummy variables that characterize the degree of hybridization.

11

Intermediate hybrid cooperative banks contribute more to financial stability. First of all, when we consider the whole period (2002-2011) that relies on 591 observations3, we note that the explanatory power is 21.6 percent. The variables introduced in the model explain 21.6 percent of the z-score variance, which is an interesting level in the social sciences area. Regarding the dummy variables of the hybridization degree, we note that the coefficient related to the intermediate hybrid banks is positive (53.36) and significant. This signifies that, on average, cooperative banking groups that have an intermediate hybridization degree have

3

We used 591 observations instead of 601 because of the missing Fees data for Goldman Sachs International. 67 66 JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

higher z-scores than the joint-stock banks that are the reference in our model. As regards the other hybridization degree categories, we cannot deduce any difference between said categories and joint-stock banks, as the coefficients related to the dummies are insignificant. So, over the whole period, intermediate hybrid cooperative banks contribute more to financial stability. Looking at the situation before the crisis period (2002-2007), we can see that once again, cooperative banking groups with an intermediate hybridization degree are affected by a positive and significant coefficient of 49.83. In 2008, the crisis year, this coefficient rises to 55.96. Finally, in the post-crisis period (2009-2011), it continues to rise, reaching a value of 58.72. In all three periods, these coefficients remain significant at a one percent level. Thus, compared to the jointstock banks, intermediate hybrid cooperative groups contribute more to financial stability, even during the crisis period. As Rabobank belongs to this category, and as the joint-stock banks include Dexia, we decided to run those regressions omitting both cases4. The coefficient related to the intermediate hybrid banks is lower (22.43) but it remains positive and significant at a five percent level. We note that this coefficient evolves as before. It is worth more during the crisis period (23.61) than in the pre-crisis period (20.13), and it is worth more in the post-crisis period (24.70) than during the crisis. Therefore, we cannot say there is a Rabobank and Dexia effect (although in the case of Rabobank the coefficient is not as high as before). Thus we may conclude that even during a period of financial crisis, cooperative banking groups that have an intermediate hybridization degree contribute more to financial stability than the joint-stock banks. This can be explained by: their diversification, the fact that they and the subsidiaries they hold are not listed, they are well established in various countries, and they benefit from a stable structure that better enables them to face the crisis. Regarding the banks that have lower (LHCoop) and greater (MHCoop) degrees of hybridization, even in this restricted model, we cannot observe any difference between them and the joint-stock banks since the coefficients remain insignificant. This can be explained by the fact that the least hybrid cooperative banks are often specialized in a single business area. Their lack of diversification makes them vulnerable, particularly in times of crisis; if they are specialized in mortgages, for example, a housing crisis will lead them to bankruptcy. The most hybrid cooperative banks have the features of joint-stock banks. As they are partially listed or hold listed subsidiaries, they receive pressure from stockholders. As they can act as jointstock banks, the lack of difference between them is not surprising. To deepen our analysis, we tested the hybridization degree effect of banks on To deepen our analysis, we tested the hybridization degree effect of banks on other endogenous variables, other endogenous variables, particularly those we examined in the descriptive statistics particularly we examined descriptive statistics analysis. We want see if financial analysis. those We want to see inifthe financial performance remains thetosame before, performance during and after remains the same before, during and after the crisis. To do this, we examine the link between financial by the the crisis. To do this, we examine the link between financial performance (represented performance (represented thehybridization return on equity ratio) and hybridization degreeperiods. in the three return on equity ratio)by and degree in the three different Wedifferent also attempt periods. We also attempt to see if the hybridization degree played a role in supporting the economy. To for a to see if the hybridization degree played a role in supporting the economy. To search search for arelation, possible relation, useloans the loans to assets ratioasas the the endogenous variable. possible we usewethe to assets ratio endogenous variable. Model 2: How does financial performance evolve throughout each period? Model 2: How does financial performance evolve throughout each period? TheThe model we ranwe to explain the variation financial performance (Appendix 6.2): model ran to explain the in variation in financialwas performance was (Appendix 6.2):

ROE i, t, c = αi + β1i LHCoopi,t + β2i IHCoopi,t + β3i MHCoopi,t + β4i NIMi,t +β5c Ratec,t + ∑ φt Crisist + ɛi,t ∀ i = 1,..,64 In this model, the control variables show different effects on financial performance. The 4 long-term interest rate indicator has an insignificant effect on the return on equity, except in We decided to withdraw Rabobank as this cooperative banking group has always presented strong results and performances throughout the banking area. Until the end of 2011, itwhen was an AAA-rated group and its results could havethe biasedsample. the real performances the post-crisis period (2009–2011) we withdrew Dexia from The effect is of other cooperative banking groups. As regards to Dexia, it was so affected by the debt and financial crisis that its presence in significant and negative, as expected. Regardless of the period, the net interest margin always the sample could have underestimated the performances of the other joint-stock banking groups. has a positive and significant effect on financial performance, except in 2008, where the effect is insignificant. Over the whole period, whatever the sample, the temporal dummy always has 68 67 a negative and significant effect on financial performance. JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

In this model, the control variables show different effects on financial performance. The long-term interest rate indicator has an insignificant effect on the return on equity, except in the post-crisis period (2009-2011) when we withdrew Dexia from the sample. The effect is significant and negative, as expected. Regardless of the period, the net interest margin always has a positive and significant effect on financial performance, except in 2008, where the effect is insignificant. Over the whole period, whatever the sample, the temporal dummy always has a negative and significant effect on financial performance. The least and intermediate hybrid cooperative banking groups are less profitable in the pre-crisis period. Over the whole period (2002-2011), we note that there is no difference in the financial performance of cooperative banking groups and joint-stock banks, regardless of their hybridization degree. Indeed, all coefficients related to the hybridization degree are insignificant. If we focus on the pre-crisis period (2002-2007), we see a negative and significant coefficient related to the least and intermediate hybrid banks. So, a bank in the least hybrid degree category has on average a financial performance inferior to 3.29 points compared to that of jointstock banks. A bank that has an intermediate degree of hybridization has on average a financial performance inferior to 3.91 points compared to that of joint-stock banks. We note that there is no difference between the financial performance of cooperative banking groups with the highest degree of hybridization and that of joint-stock banks, as the coefficient is insignificant. So in the pre-crisis period, the least and intermediate hybrid cooperative banking groups had a lower financial performance than joint-stock banks. This is not unexpected; as we described earlier, cooperative banks are not under stockholder pressure, they are not listed and do not pursue profit maximization. In times of crisis, there is no difference in profitability between cooperative banks and joint-stock banks. In 2008, the crisis year, and during the post-crisis period (2009-2011), we cannot report any difference between the financial performance of cooperative banks, whatever their hybridization degree, and that of joint-stock banks. This signifies that cooperative banking groups that had lower and significant financial performance in the pre-crisis period showed higher resistance; there was no difference between the return realized by the different categories. We repeated the analysis omitting Rabobank and Dexia. The results are the same in all three sub-periods, but not over the whole period. In this latter case, the coefficient related to we run the model without Dexia, the least hybrid banks have lower financial performance the least hybrid banks is negative and significant. So over the entire period, when we run the model without than joint-stock banks. Dexia, the least hybrid banks have lower financial performance than joint-stock banks.

Model 3: Do cooperative banks lend more in times of crisis?

Model 3: Do cooperative banks lend more in times of crisis?

We ran some models with the loans to assets as the endogenous variable. The control variables long-term temporalvariable. dummy. model usedarewas We ran are somethe models with theinterest loans to rate assetsand as thethe endogenous TheThe control variables the (Appendix 6.3): rate and the temporal dummy. The model used was (Appendix 6.3): long-term interest Loans i, t, c = αi + β1i LHCoopi,t + β2i IHCoopi,t + β3i MHCoopi,t + β4c Ratec,t + ∑ φt Crisist + ɛi,t ∀ i = 1,..,64 As a rise in interest rates leads to lower asset value, we expected a positive sign related As a rise in interest rates leads to lower asset value, we expected a positive sign related to its coefficient. to its coefficient. Indeed, if the asset value decreases, the loans to assets ratio will increase. Indeed, if the assetisvalue decreases, the loans to assets ratio willregardless increase. This hypothesis is confirmed. This hypothesis confirmed. Nevertheless, we note that of the period and sample, Nevertheless, note that thethe period and level, sample,asthe dummy has no effect on the the temporalwe dummy hasregardless no effectofon lending itstemporal coefficient is insignificant. lending level, as its coefficient is insignificant.

The least and intermediate hybrid banks lent more than the joint-stock banks during the period 2002-2011 In the first model, which covered the entire period, we noted that the least hybrid cooperative banking groups lent more than the joint-stock banks. The coefficient related to the least hybrid category was positive and highly significant. Their ratio was on average 11.49 69 68 points higher than that of joint-stock banks. Intermediate hybrid cooperative banks also lent JEOD - Vol.3, Issue 1 (2014) more than joint-stock banks. On average, their ratio was 5.44 points higher and significant at the 10 percent level. A Rabobank effect was noted in this case. Indeed, if we withdraw


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

The least and intermediate hybrid banks lent more than the joint-stock banks during the period 2002-2011. In the first model, which covered the entire period, we noted that the least hybrid cooperative banking groups lent more than the joint-stock banks. The coefficient related to the least hybrid category was positive and highly significant. Their ratio was on average 11.49 points higher than that of joint-stock banks. Intermediate hybrid cooperative banks also lent more than joint-stock banks. On average, their ratio was 5.44 points higher and significant at the 10 percent level. A Rabobank effect was noted in this case. Indeed, if we withdraw Rabobank, the coefficient becomes insignificant. There is no difference in lending behaviour between the more hybrid and the joint-stock banks. These results are consistent with the earlier theoretical analysis, where we noted that cooperative banks are established throughout a territory; they are close to their customers and are willing to contribute to community and economic development. The least and intermediate hybrid banks also lent more during the crisis and post-crisis periods. If we focus on the banksâ&#x20AC;&#x2122; behaviour during the different periods as regards to their lending offers, we note that during the pre-crisis period as well as in the crisis year, there is no difference between the different categories of banks, as no coefficient appears significant. Because the crisis year gives poor results, and the post-crisis period also gives insufficient results compared to the whole period, which shows that there is a difference in behaviour between the banks, we added a post-crisis period covering 2008-2011. It shows that the least hybrid cooperative banking groups lent more on average than joint-stock banks. Their coefficient is 9.34 and is significant at the 5 percent level. It remains significant and positive even if we withdraw Rabobank and Dexia. Intermediate hybrid cooperative banking groups also lent more on average than joint-stock banks. Their loans to assets ratio is on average 8.91 points higher and is significant at the 10 percent level. As in the previous analysis over the whole period, there seems to be a Rabobank effect here as well. When we withdraw this bank, the coefficient becomes insignificant. We stated earlier that cooperative banking groups lend more even in times of crisis. As they are held by shareholders who are also customers - and as their customers include managers of small and medium-sized firms who need loans so as to enable their firm to grow - it seems normal for cooperative banks to contribute to the economic development of the area where they are settled. This hypothesis was verified in the descriptive statistics analysis and is corroborated by the econometric analysis. Since they continue to lend even during the crisis period, they show higher resistance than joint-stock banks.

4. Conclusion In this paper, we have shown that there is a link between a cooperative banking groupâ&#x20AC;&#x2122;s hybridization degree and its financial stability. We tested levels of resistance and resilience through various indicators: a computed z-score in order to measure the probability of bankruptcy; the loans to assets ratio to measure a bankâ&#x20AC;&#x2122;s capacity to maintain its level of lending in times of financial crisis; and the return on equity to see how financial performance evolves according to their hybridization degree over the different periods considered. The first conclusion we draw is that intermediate hybrid cooperative banks contribute significantly to financial stability. They benefit from higher z-scores, which show that they are well-capitalized and realize steady returns. Their bankruptcy probability is lower than that of joint-stock banks, even in times of crisis. Rabobank plays an important role in this. Moreover, we showed that whatever the period even when we withdraw Rabobank and Dexia from the sample - intermediate hybrid cooperative banks continue to 70 69 JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

provide higher z-scores than joint-stock banks. These banks have diversified activities that enable them to mitigate economic fluctuations and their firm establishment in an area (local, regional or country-wide) gives them a better understanding of their customers. As local and/or regional banks benefit from this proximity, and as group activities are diversified and unlisted, they accumulate different assets that enable them to contribute more to financial stability, as shown through their higher z-score. According to this indicator, these banks show more resistance and resilience. The results for the least and most hybrid banks do not differ from those for joint-stock banks. The least hybrid banks suffer from their lack of diversification and are more vulnerable to macroeconomic shock. The most hybrid banks have features that also characterize joint-stock banks, which explains the lack of difference in their results. The second conclusion concerns the ability of cooperative banking groups to make a profit. During the pre-crisis period, their return is lower than that of joint-stock banks, with the exception of the most hybrid cooperative banking groups, whose level of returns is not much different than the joint-stock banking groups. But in times of crisis and during the post-crisis period, we find that there is no difference between the financial performances of cooperative and joint-stock banking groups. As the least and intermediate hybrid cooperative banks had weaker return on equity ratios before the crisis, and as there are no differences between the cooperative and joint-stock banks during the other periods, it demonstrates that they showed more resistance during the crisis. Because the least and intermediate hybrid banks are unlisted or do not hold listed subsidiaries, they do not suffer from a fall in stock market prices. Additionally, they do not undertake high-risk projects that can have negative consequences in times of crisis if they fail. The third and final conclusion we draw is the capacity of the least and intermediate hybrid cooperative banking groups to maintain their lending level in times of crisis. As these banks are built as local and/or regional networks throughout an entire country, they are close to their customers. These banks are leading players in regards to the economic development of their area. As they are often managed by entrepreneurs of small and medium-sized firms, who themselves know the difficulties of obtaining a loan, the results we found are consistent with their willingness to develop the local economy. While drawing conclusions in this paper, we found that, globally, cooperative banking groups with an intermediate hybridization degree have faced the financial crisis better. They contribute more to financial stability than joint-stock banking groups. They show more resistance and resilience before, during and after the crisis. Even if they do not differentiate themselves from joint-stock banking groups in terms of the z-score indicator, the least hybrid banks also showed more resistance and resilience as far as lending is concerned and they showed more resistance than joint-stock banks in regards to financial performance. They faced the financial crisis more successfully and contributed more to financial stability with respect to those indicators. Nevertheless, these models - particularly those that use the loans to assets and the return on equity as endogenous variables - need to be enhanced, as their explanatory power is rather limited. Even if the variablesâ&#x20AC;&#x2122; coefficients are often significant, the introduction of further control variables that influence the endogenous variable would improve the models. Collecting more data for the following years would also be useful to check if the least and intermediate cooperative banks continue to succeed in maintaining their lending levels.

71 70 JEOD - Vol.3, Issue 1 (2014)


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

Appendix 1: Z-score descriptive statistics Appendix 1: Z-score descriptive statistics Z-score descriptive statistics sorted by periods, countries and hybridization degree where LHCoop is Z-score descriptive statistics sorted by periods, countries and hybridization degree where LHCoop is the least hybrid cooperative banks, IHCoop the the cooperative banks thatthat have anan intermediate Appendix 1: Z-score descriptive statistics the least hybrid cooperative banks, IHCoop cooperative banks have intermediatedegree of degree of hybridization, MHCoop the most hybrid cooperative banks and JSB the joint-stock hybridization, MHCoop the most hybrid 1: cooperative banks and JSB the joint-stock banks. Appendix Z-score descriptive statistics LHCoop is Z-score descriptive statistics sorted by periods, countries and hybridization degree where banks. the least hybrid cooperative banks, IHCoop the cooperative banks that have an intermediate Z-score descriptive statistics sortedthe by crisis periods, countries and hybridization degree where LHCoop is Table 1-a: Z-score descriptive statistics before Table 1-a: descriptive statisticsthe before thehybrid crisis cooperative banks and JSB the joint-stock degree of Z-score hybridization, MHCoop most the least hybrid cooperative banks, IHCoop the cooperative banks that have an intermediate banks. degree of hybridization, MHCoop the most hybrid cooperative banks and JSB the joint-stock LHCOOP IHCOOP MHCOOP JSB 2002-07 (2) (1) banks. Meandescriptive Std(1) Obs Mean Obs(2) Mean Std(1) Obs(2) Mean Std(1) Obs(2) Table 1-a: Z-score statistics before Std the crisis Austria 8.23 0.62 6 Table 1-a: Z-score descriptive statistics before 6.60 the crisis Germany 49.02 6 LHCOOP IHCOOP 2002-07 Spain 31.25 1.65(1) 6 (2) (1) Mean Std Obs Mean Std Obs(2) France 40.05 IHCOOP 9.31 12 LHCOOP Austria 8.23 0.62 6 2002-07 Netherlands 156.77 8.82 6 (2) Mean Std(1) Obs(2) 49.02 Mean Std(1) Obs Germany 6.60 6 UK 23.95 4.13 6 Austria 8.23 6 Spain 31.25 0.62 1.65 Italy 39.48 0.56 6 Germany 49.02 6.60 612 France 40.05 9.31 Portugal 25.89 1.77 Spain 66 Netherlands 31.25 1.65 156.77 8.82 6 Finland France 40.05 9.31 12 UK 23.95 4.13 6 Total 25.76 10.65 30 71.47 51.11 Netherlands 156.77 8.82 624 Italy 39.48 0.56 6 Note: (1) standard-deviation (2) number of observations UK 23.95 6 Portugal 25.89 4.13 1.77 Italy 39.48 (2) 0.56 6 of observations Note: (1) standard-deviation number Finland Table 1-b: Z-score the year of the24 crisis Portugal 25.89 1.77 Total 25.76 descriptive 10.65 630statistics 71.47 51.11 Finland Note: (1) standard-deviation (2) number of observations TotalZ-score descriptive 25.76 LHCOOP 10.65 30 51.11 24 IHCOOP Table 1-b: statistics the (2)year71.47 of the crisis 2008 (1) (2) Note: (1) standard-deviation (2) number of observations Mean Std Obs Mean Stdof(1)theObs Table 1-b: Z-score descriptive statistics the year crisis Austria 6.96 NA 1 Germany 52.05 NAof the1crisis Table 1-b: Z-score descriptive LHCOOP statistics the year IHCOOP 2008 Spain 28.57 NA (1) 1 (2) Mean Std Obs Mean Std(1) Obs(2) France 39.21 IHCOOP 16.18 2 LHCOOP Austria 6.96 NA 1 2008 Netherlands 163.42 NA 1 (2) Mean Std(1) Obs(2) 52.05 Mean Std(1) Obs Germany NA 1 UK 15.88 NA NA 1 Austria 6.96 1 Spain 28.57 NA 1 Italy 37.75 NA 1 Germany 52.05 NA France 39.21 16.18 12 Portugal 27.86 NA 1 Spain 28.57 NA 1 Netherlands 163.42 NA 1 Finland France 39.21 16.18 2 UK 15.88 NA 1 Total 23.40 12.03 5 73.47 60.98 Netherlands 163.42 NA 14 Italy 37.75 NA 1 Note: (1) standard-deviation (2) number of observations UK 15.88 NA 1 Portugal 27.86 NA 1 Italy 37.75 NA 1 Finland Table 1-c: Z-score descriptive1 statistics after the crisis year Portugal 27.86 Total 23.40 NA 12.03 5 73.47 60.98 4 Finland Note: (1) standard-deviation (2) number of observations Note: (1) standard-deviation (2) number 2009-11 Total 23.40 LHCOOP 12.03 5 of observations 73.47 IHCOOP 60.98 4 (1) (2) (1) (2) Meandescriptive Std(2) Obs Mean Stdcrisis Obs Note: standard-deviation number of observations Table(1) 1-c: Z-score statistics after the year Austria 5.12 2.85 3 Table 1-c: Z-score descriptive statistics after the56.87 crisis Table 1-c: Z-score descriptive statistics afteryear the crisis 3year Germany 0.53 2009-11 LHCOOP IHCOOP Spain 27.37 1.62(1) 3 (2) Mean Std Obs Mean Std(1) Obs(2) France 48.37 IHCOOP 11.87 6 2009-11 LHCOOP Austria 5.12 2.85 3 Netherlands Mean Std(1) Obs(2) 182.22 2.29 3 (2) Mean Std(1) Obs Germany 56.87 0.53 3 UK 15.51 1.77 3 Austria 5.12 33 Spain 27.37 2.85 1.62 Italy 32.11 4.07 3 Germany 56.87 0.53 France 48.37 11.87 36 Portugal 24.89 1.53 Spain 33 Netherlands 27.37 1.62 182.22 2.29 3 Finland France 48.37 11.87 6 UK 15.51 1.77 3 Total 21.00 10.17 15 83.96 59.91 312 Netherlands 182.22 2.29 Italy 32.11 4.07 3 Note: (1) standard-deviation (2) number of observations UK 15.51 1.77 3 Portugal 24.89 1.53 3 Italy 32.11 4.07 3 Finland Table 1-d: Z-score on the 59.91 whole period Portugal 24.89 1.53 Total 21.00 descriptive 10.17 315statistics 83.96 12 Finland Note: (1) standard-deviation (2) number of observations Total 21.00 LHCOOP 10.17 15 83.96 IHCOOP 59.91 12 2002-11

15.63

2002-11

LHCOOP Std(1) Obs(2) LHCOOP Mean Std(1) Obs(2) Mean

IHCOOP Std(1) Obs(2) IHCOOP Mean Std(1) Obs(2) Mean

6

Mean 15.63 22.12 19.54

Std 3.12 1.62 7.44

Obs 6 6 18

32.11 19.54 21.41 22.12

1.83 7.44 7.51 1.62

6 18 36 6

22.12 32.11 1.62 1.83 66 21.41 7.51 36 32.11 1.83 6 21.41 MHCOOP 7.51 36 Mean Std(1) Obs(2) 13.97 NA 1

MHCOOP Mean Std(1) Obs(2) 14.60 7.60 3 13.97 MHCOOP NA 1 Mean Std(1) Obs(2) 13.97 NA 1 21.66 NA 1 14.60 7.60 3

21.48 14.60 16.82 21.66

NA 7.60 6.06 NA

31 6 1

21.66 11 21.48 NA NA 16.82 6.06 6 21.48 NA 1 16.82 MHCOOP 6.06 6 Mean Std(1) Obs(2) 18.24 2.18 3

MHCOOP Mean Std(1) Obs(2) 24.44 MHCOOP 1.99 6 18.24 2.18 3 Mean Std(1) Obs(2) 18.24 2.18 3 21.34 1.97 3 24.44 1.99 6

24.11 24.44 22.52 21.34

1.46 1.99 3.05 1.97

63 15 3

21.34 24.11 1.97 1.46 33 22.52 3.05 15 24.11 1.46 3 22.52 MHCOOP 3.05 15

Note: (1) standard-deviation (2) number(2)of observations (1) Table 1-d: Z-score statistics on the Std whole period Mean descriptive Std(1) Obs Mean Obs(2) Mean Note: (1) standard-deviation (2) number of observations Table 1-d: Z-score descriptive statistics on the whole period

2002-11

3.12

MHCOOP Mean Std(1) Obs(2) 19.54 7.44 18 15.63MHCOOP 3.12 6 (2) (1)

12.60 7.37 JSB 12.53 Std(1) 11.95 JSB 12.60 7.02(1) Std 7.37 31.10 12.60 12.53 28.72 7.37 11.95 9.20 12.53 7.02 13.68 11.95 31.10 21.11 7.02 28.72 31.10 9.20 28.72 13.68 9.20 21.11 13.68 21.11 JSB Std(1) 20.37 6.51 JSB 11.64 Std(1) 13.89 JSB 20.37 6.16(1) Std 6.51 22.41 20.37 11.64 35.05 6.51 13.89 9.93 11.64 6.16 6.47 13.89 22.41 20.39 6.16 35.05 22.41 9.93 35.05 6.47 9.93 20.39 6.47 JSB 20.39 Std(1) 20.90 6.02 JSB 13.89 Std(1) 15.23 JSB 20.90 7.02(1) Std 6.02 27.00 20.90 13.89 27.26 6.02 15.23 13.09 13.89 7.02 2.61 15.23 27.00 20.34 7.02 27.26 27.00 13.09 27.26 2.61 13.09 20.34 2.61 20.34 JSB

23 38 42 (2) Obs 18 23 (2) 24 Obs 38 59 23 42 42 38 18 12 42 24 12 18 59 270 24 42 59 12 42 12 12 270 12 270 Obs(2) 4 8 7 (2) Obs 3 4 (2) 4 Obs 8 4710 837 742 2 310 4747 10 2 72 247 2 47 Obs(2) 12 20 21 (2) Obs 9 12 (2) 12 Obs 20 30 12 21 21 20 9 6 21 12 6 930 137 12 21 30 6 21 6 6137 6 137

Obs(2)

Mean

Std(1)

Obs(2)

MHCOOP Mean Std(1) Obs(2) MHCOOP Mean Std(1) Obs(2)

Mean

JSB Std(1) JSB Std(1)

Obs(2)

72 71 JEOD - Vol.3, Issue 1 (2014)

Std(1)

30.32 13.51 26.28 Mean 19.74 30.32 15.52 Mean 13.51 22.55 30.32 26.28 33.71 13.51 19.74 21.05 26.28 15.52 22.24 19.74 22.55 23.36 15.52 33.71 22.55 21.05 33.71 22.24 21.05 23.36 22.24 23.36 Mean 29.04 9.32 22.34 Mean 13.13 29.04 10.38 Mean 9.32 14.28 29.04 22.34 37.40 9.32 13.13 19.44 22.34 10.38 15.67 13.13 14.28 19.21 10.38 37.40 14.28 19.44 37.40 15.67 19.44 19.21 15.67 19.21 Mean 35.82 13.70 24.10 Mean 19.30 35.82 13.88 Mean 13.70 20.49 35.82 24.10 33.03 13.70 19.30 23.09 24.10 13.88 13.94 19.30 20.49 22.49 13.88 33.03 20.49 23.09 33.03 13.94 23.09 22.49 13.94 22.49

Mean

17

Obs(2)

17

17


Germany Spain France Netherlands UK Italy Portugal Finland Total

13.70 24.10 48.37 11.87 6 24.44 1.99 6 19.30 182.22 2.29 3 13.88 15.51 1.77 3 20.49 Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? 32.11 4.07 3 21.34 1.97 3 33.03 Lemzeri, Y. 24.89 1.53 3 23.09 24.11 1.46 3 13.94 21.00 10.17 15 83.96 59.91 12 22.52 3.05 15 22.49 27.37

1.62

56.87

3

0.53

3

Note: (1) standard-deviation (2) number of observations

6.02 13.89 15.23 7.02 27.00 27.26 13.09 2.61 20.34

20 21 9 12 30 21 6 6 137

Table 1-d: Z-score descriptive statistics on the whole Table 1-d: Z-score descriptive statistics on period the whole period 2002-11 Mean 7.17

LHCOOP Std(1) Obs(2) 2.04 10

Mean

IHCOOP Std(1) Obs(2)

Austria Germany 51.68 6.16 10 Spain 29.82 2.37 10 France 42.46 10.81 20 Netherlands 165.07 13.73 10 Austria 7.17 10 UK 20.61 2.04 5.36 10 Germany 51.68 6.16 10 Italy 37.10 3.99 10 Spain 29.82 2.37 10 Portugal 25.78 1.74 10 France 42.46 10.81 20 Finland Netherlands 165.07 13.73 10 Total 24.10 10.64 50 75.42 53.58 40 UK 20.61 5.36(2) number 10 of observations Note: (1) standard-deviation Italy 37.10 3.99 10 Note: (1) standard-deviation number Portugal 25.78 (2) 1.74 10 of observations Finland Total 24.10 10.64 50 75.42 40 Appendix 2: Loans53.58 to assets Note: (1) standard-deviation (2) number of observations

Mean 16.24

20.08

MHCOOP Std(1) Obs(2) 2.94 10

7.04

27

16.24

2.94

10

21.84

1.57

10

20.08 28.65 21.22

7.04 4.78 6.59

27 10 57

21.84

1.57

10

28.65 4.78 10 21.22 6.59 57 descriptive statistics

Mean 31.88 13.06 25.23 18.95 14.51 31.88 21.09 13.06 33.88 25.23 21.50 18.95 19.09 14.51 22.67 21.09 33.88 21.50 19.09 22.67

JSB Std(1) Obs(2) 16.07 39 6.93 66 17 12.76 70 12.84 30 6.95 40 16.07 39 28.99 99 6.93 66 28.53 70 12.76 70 10.03 20 12.84 30 11.32 20 6.95 40 20.80 454 28.99 99 28.53 70 10.03 20 11.32 20 20.80 454

Loans to assets descriptive statistics classified by periods and hybridization degree, where LHCoop is Appendix 2: Loans assets descriptive statisticsthe cooperative banks that have an intermediate the least hybridtocooperative banks, IHCoop Appendix 2: Loans assetscooperative descriptivebanks statistics degree of hybridization, MHCoop the mosttohybrid and JSB the joint-stock banks. Loans statisticsclassified classifiedbybyperiods periods hybridization degree, LHCoop is LHCoop is Loansto to assets assets descriptive descriptive statistics andand hybridization degree, wherewhere the least hybrid cooperative banks, IHCoop the cooperative banks thatthat have an an intermediate degree of the least cooperative banks, IHCoop banks have intermediate Table 2-a:hybrid Loans to assets descriptive statistics on the the cooperative whole period degree ofMHCoop hybridization, MHCoop most hybrid cooperative and JSB the joint-stock hybridization, the most hybridthecooperative banks and JSB banks the joint-stock banks. banks. 2002-11 LHCOOP IHCOOP MHCOOP JSB In % [30, 40) [40, 50) 2002-11 [50, 60)

Obs(2)

Mean

NA 1 55.58 LHCOOP 2.66 13 (2) Mean Std(1) Obs 65.78 2.33 19

47.02

74.75 43.22

2.62 NA

17 1

55.58

2.66

13

72.31 47.02 84.35 54.21

1.90 3.27 NA 2.15

6 110

Std(1) Obs(2) 2.49 12 2.50 19 50.91 MHCOOP 0.69 3 Mean 1.70 Std(1) 13 Obs(2) 66.71 37.31 2.05 2.49 75.17 912 44.71 2.50 81.10 NA 119

50.91

0.69

3

65.78

2.33

19

64.14

3.61

9

66.71

1.70

13

74.75

2.62

17

72.31

1.90

6

75.17

2.05

9

84.35

NA

1

81.10

NA

1

40

53.95

14.71

57

Mean

Std(1)

Std(1)

Obs(2)

Mean 37.31 44.71

Table 2-a: Loans assets to descriptive statistics onstatistics the wholeonperiod Table 2-a:toLoans assets descriptive the whole period

In 70) % [60, [30, 40) [70, 80) [40, [80, 50) 90) [50, 60) [0, 20) [60, [20, 70) 40) [70, [40, 80) 60) [80, [60, 90) 80)

43.22

3.27 10 54.21 IHCOOP 2.15 14 Mean 3.61 Std(1) 9Obs(2) 64.14

[0, 20) [80, 100) [20, 40)

65.72 8.47 50 58.12 10.06 [40, Note:60) (1) standard-deviation (2) number of observations

14

[60, 80) Table 2-b: Loans to assets descriptive statistics before the crisis [80, 100) 65.72

8.47 50 58.12 10.06 40 2002-07 LHCOOP IHCOOP Note: (1) standard-deviation (2) number of observations Std Obs Std Obs In % Mean Mean (1) (2) (1) (2)

53.95 Mean

Note: (1) standard-deviation (2) number of observations Table 2-b: Loans to assets descriptive statistics before the crisis [40, 50) [50, 60) 2002-07

[60, 70) In % [70, 80) [40, [80, 50) 90) [50, [30, 60) 40)

65.73 Mean 74.22

2.83 Std (1) 2.82

12 Obs (2) 10

55.81

3.29

8

45.81 3.86 54.49 IHCOOP 2.17 62.31 2.30 Std Mean (1) 73.94 2.72 45.81 84.35 3.86 NA 54.49 2.17

65.73

2.83

12

62.31

2.30

5

74.22

2.82

10

73.94

2.72

2

84.35

NA

1

55.81 LHCOOP 3.29 8

6 10

5Obs (2) 2 6 1 10

Mean

14.71 57 MHCOOP Std Obs (1)

(2)

MHCOOP Std Obs (1)

(2)

Mean

Std(1)

Mean

Std(1)

Obs(2)

JSB Obs(2)

11.52

5.01

41

31.75

5.45

74

52.47

6.00

143

67.05 11.52 85.94 31.75 53.52 52.47

6.16 5.01 3.42 5.45 20.81 6.00

152 41 44 74 454 143

67.05

6.16

152

85.94

3.42

44

53.52

20.81 JSB Std

454

Mean

Mean

(1)

JSB Std (1)

Obs (2)

Obs (2)

36.90

2.53

10

43.99

2.40

9

12.00

5.28

22

51.13

0.82

2

31.92

5.44

42

66.68 36.90

1.90 2.53

9 10

52.63

5.31

101

[0, 20)

43.99

2.40

9

12.00

5.28

22

[20, 40)

51.13

0.82

2

31.92

5.44

42 18

66.68

1.90

9

52.63

5.31

101

[60, 70) [0, 20) [70, [20, 80) 40) [80, [40, 90) 60) [30, 40)

[40, 60)

73 72

JEOD - Vol.3, Issue 1 (2014)

18


[80, 90)

84.35

NA

1

81.10

NA

1

[0, 20)

11.52

5.01

41

[20, 40)

31.75

5.45

74

[40, 60) [60, 80)

52.47 6.00 Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

[80, 100) 65.72

8.47

50

58.12

10.06

40

53.95

14.71

57

143

67.05

6.16

152

85.94

3.42

44

53.52

20.81

454

Note: (1) standard-deviation (2) number of observations

Table 2-b: Loans assets to descriptive statistics before the crisis Table 2-b:toLoans assets descriptive statistics before the crisis 2002-07

LHCOOP Std Obs

In %

Mean

[40, 50) [50, 60)

55.81

3.29

8

45.81 54.49

3.86 2.17

6 10

[60, 70)

65.73

2.83

12

62.31

2.30

5

[70, 80)

74.22

2.82

10

73.94

2.72

2

84.35

NA

1

(1)

(2)

[80, 90)

Mean

IHCOOP Std Obs (1)

(2)

Mean

MHCOOP Std Obs (1)

(2)

Mean

JSB Std

Obs

(1)

(2)

[30, 40)

36.90

2.53

10

[0, 20)

43.99

2.40

9

12.00

5.28

[20, 40)

51.13

0.82

2

31.92

5.44

42

[40, 60) [60, 80)

66.68 76.32

1.90 1.76

9 5

52.63 66.19

5.31 5.81

101 75

[80, 100)

81.10

NA

1

85.64

3.34

30

53.61 76.32

15.72 1.76

36 5

53.53 66.19

20.04 5.81

270 75

81.10

NA

1

85.64

3.34

30

36

53.53

20.04

270

Obs(2)

Mean

JSB Std(1)

Obs(2)

MHCOOP Std(1) Obs(2)

Mean

JSB Std(1)

Obs(2)

[60, 80)

65.92

7.75

30

56.82

10.25

24

Note: (1) standard-deviation (2) number of observations [80, 100)

Note: (1) standard-deviation (2) number observations 65.92 7.75 30 of 56.82 10.25 24 53.61 15.72 Table 2-c: Loans to assets descriptive statistics the year of the crisis

22

18

Note: (1) standard-deviation (2) number of observations

2008 LHCOOP statistics the year IHCOOP MHCOOP Table 2-c: Loans assets to descriptive of the Table 2-c:toLoans assets descriptive statistics the crisis year of the crisis In % [45, 50) 2008 [50, 55) In % [60,50) 65) [45, [65, 70) [50, 55) [75,65) 80) [60, [30, 40) [65, 70) [40,80) 50) [75, [60, 70) [30, 40)

[70,50) 80) [40, [0, 20) [60, 70) [20,80) 40) [70, [40, 60) [0, 20) [60, 80) [20, 40) [80,100) [40, 60)

[60, 80)

Mean

52.87 Mean

Std(1)

Obs(2)

LHCOOP NA 1 Std(1) Obs(2)

65.85 52.87 76.72

0.65 NA 1.42

12 2

65.85

0.65

2

76.72

1.42

2

67.60

9.90

5

Mean Std(1) Obs(2) 48.55 NA 1 IHCOOP 50.86 NA(1) 1 (2) Mean Std Obs 62.28 NA NA 48.55 11 69.64 NA NA 50.86 11

62.28

NA

1

69.64

NA

1

57.83

9.89

4

Mean

Mean

Std(1)

39.08

NA

1

45.49

0.32

3

68.27 39.08 71.82 45.49

NA NA NA 0.32

11 13

68.27

NA

1

71.82

NA

1

52.60

13.78

6

Note: (1) standard-deviation (2) number of observations

[80,100)

67.60 to 9.90 5 57.83 9.89after4the crisis 52.60 Table 2-d: Loans assets descriptive statistics year 13.78

6

10.39

3.22

5

30.32

4.98

8

52.87 10.39 67.94 30.32 87.73 52.87 54.32 67.94

7.33 3.22 6.86 4.98 3.63 7.33 23.30 6.86

10 5 19 8 5 10 47 19

87.73

3.63

5

54.32

23.30

47

Mean

JSB Std(1)

Obs(2)

Mean

JSB Std(1)

Obs(2)

11.17

5.28

14

31.94

5.75

24

51.84 11.17 67.88 31.94 85.94 51.84 53.23 67.88

7.63 5.28 6.30 5.75 3.69 7.63 21.53 6.30

32 14 58 24 9 32 137 58

Note: (1) standard-deviation (2) number of observations

2009-11 LHCOOP IHCOOP MHCOOP Note: (1) standard-deviation (2) number of observations (1) (2) crisis year Table assets descriptive statistics the In %2-d: Loans Mean toStd Obs(2) Mean Std(1)after Obs Mean Std(1) Obs(2) [45,50) 2009-11 [50, 55) In % [55, 60) [45,50) [60, 65) [50, 55) [65, 70) [55, 60) [70, 75) [60, 65) [30, 40) [65, 70) [40, 50) [70, 75) [50, 60) [30, 40) [60, 70) [40, 50) [70, 80) [50, 60) [0, 20) [60, 70) [20, 40) [70, 80) [40, 60) [0, 20) [60, 80) [20, 40) [80,100) [40, 60)

[60, 80)

Mean

LHCOOP Std(1) Obs(2)

48.92 0.16 3 IHCOOP 53.61 1.44 2 Mean Std(1) Obs(2) 55.92 NA 1 48.92 0.16 3 63.98 NA 1 53.61 1.44 2 69.87 NA 1 55.92 NA 1 71.50 0.96 4 63.98 NA 1

Mean

MHCOOP Std(1) Obs(2)

39.67

NA

1

45.30

3.103099

7

5 1 5 4

50.47 39.67 66.31 45.30 74.39 50.47

NA NA 1.21 3.103099 0.91 NA

1 1 3 7 3 1

1.54

5

66.31

1.21

3

2.51

5

74.39

0.91

3

13.32

15

43.22

NA

1

55.79

0.61

4

65.86 43.22 75.01 55.79

1.54 NA 2.51 0.61

65.86 75.01

69.87

NA

1

71.50

0.96

4

74 73 JEOD - Vol.3, Issue 1 (2014)

64.71

9.83

15

60.81

10.02

12

55.29


[60, 70)

68.27

NA

1

[70, 80)

71.82

NA

1

[0, 20)

10.39

3.22

5

[20, 40)

30.32

4.98

8

[40, 60) [60, 80)

Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis?7.33 52.87 Lemzeri, Y.

[80,100) 67.60

9.90

5

57.83

9.89

4

52.60

13.78

6

10

67.94

6.86

19

87.73

3.63

5

54.32

23.30

47

Note: (1) standard-deviation (2) number of observations

Table 2-d: Loans to assets descriptive statistics after the crisis year Table 2-d: Loans to assets descriptive statistics after the crisis year 2009-11 In % [45,50)

Mean

LHCOOP Std(1) Obs(2)

IHCOOP Std(1) Obs(2) 0.16 3 1.44 2

[50, 55)

Mean 48.92 53.61

[55, 60)

55.92

NA

[60, 65)

63.98

NA

1

[65, 70)

69.87

NA

1

[70, 75)

71.50

0.96

4

Mean

MHCOOP Std(1) Obs(2)

Mean

JSB Std(1)

Obs(2)

1

[30, 40)

39.67

NA

1

[40, 50)

43.22

NA

1

45.30

3.103099

7

[50, 60)

55.79

0.61

4

50.47

NA

1

[60, 70)

65.86

1.54

5

66.31

1.21

3

[70, 80)

75.01

2.51

5

74.39

0.91

3

[0, 20)

11.17

5.28

14

[20, 40)

31.94

5.75

24

[40, 60)

51.84

7.63

32

[60, 80)

67.88

6.30

58

[80,100) 64.71

9.83

15

60.81

10.02

12

55.29

13.32

15

85.94

3.69

9

53.23

21.53

137

Note: (1) standard-deviation (2) number of observations

Note: (1) standard-deviation (2) number of observations

19

Appendix 3: Return on equity descriptive statistics

Appendix 3: Return on equity descriptive statistics

Return onon equity by periods periodsand andhybridization hybridization degree level, where LHCoop Return equitydescriptive descriptivestatistics statistics classified classified by degree level, where LHCoop is the least hybrid cooperative banks, IHCoop the cooperative banks that have an degree of is the least hybrid cooperative banks, IHCoop the cooperative banks that have an intermediate intermediate degree of hybridization, MHCoop the most hybrid cooperative banks and JSB hybridization, MHCoop the most hybrid cooperative banks and JSB the joint-stock banks. the joint-stock banks.

Table 3-a:Table Return onReturn equity on descriptive statistics before the crisis 3-a: equity descriptive statistics before the crisis 2002-07 In %

LHCOOP Mean Std(1) Obs

[4, 6) [6, 8) [8, 10) [10, 12) [12, 14) [14, 16) [0, 5) [5, 10) [10, 15) [15, 20) [20, 25) [25, 30) [-40, -0) [-20, 0) [0, 20) [20, 40)

5.53 6.88 8.96 11.03 12.96 14.43

Mean

(2)

9.70

0.39 0.74 0.72 0.47 0.68 0.34

2 5 10 8 3 2

2.49

IHCOOP Std

4.72 7.56 11.94 15.20

30

8.30

(1)

NA 1.28 1.24 NA

2.49

Obs (2)

1 19 3 1

24

Mean

MHCOOP Std (1)

Obs (2)

2.02 7.94 11.61 17.63 22.30 26.04

NA 1.26 0.95 1.72 0.14 NA

1 18 11 3 2 1

11.01

5.09

36

Mean

JSB Std

-22.41 -10.94 11.52 24.30 12.01

0.74 5.34 4.31 4.57 7.56

(1)

Obs (2)

2 7 233 28 270

Note: (1) standard-deviation (2) number of observations

Note: (1) standard-deviation (2) number of observations Table 3-b: Return on equity descriptive statistics the year of the crisis 2008 In % [-200, -50) [-100, -50)

LHCOOP Mean Std Obs (1)

(2)

Mean

IHCOOP 75 Std 74 Obs

Mean

(2) JEOD(1)- Vol.3, Issue 1 (2014)

MHCOOP Std (1)

Obs (2)

Mean

JSB Std

-178.94 -56.52

NA NA

(1)

Obs (2)

1 1


[14, 16) 14.43 0.34 2 [15,[0, 20) 15.20 NA 11 17.63 1.72 5) 4.72 NA 2.02 NA 13 [20,[5, 25) 22.30 0.14 2 10) 7.56 1.28 19 7.94 1.26 18 [25,[10, 30)15) 26.04 NA 1 11.94 1.24 3 11.61 0.95 11 [-40, -0)20) -22.41 0.74 [15, 15.20 NA 1 17.63 1.72 3 [-20, 0) 25) Did the Extent of Hybridization better Enable Cooperative22.30 -10.94Crisis? 5.34 [20, 0.14 to Face 2 the Financial Banking Groups Lemzeri, Y.26.04 [0, [25, 20) 30) 11.52 4.31 NA 1 -22.41 [20,[-40, 40) -0) 24.30 0.74 4.57 [-20, 0) 9.70 -10.94 2.49 30 8.30 2.49 24 11.01 5.09 36 12.01 5.34 7.56 [0,(1) 20)standard-deviation (2) number of observations 11.52 4.31 Note: [20, 40) 24.30 4.57 9.70 2.49 30 8.30 2.49 24 11.01 5.09 36 12.01 7.56

Table 3-b: Table Return on standard-deviation equity descriptive statistics the year of the theyear crisisof the crisis Note:3-b: (1) (2)descriptive number of observations Return on equity statistics

2008 LHCOOP IHCOOP MHCOOP Table 3-b: Return on equity descriptive statistics the year of the crisis In % Mean Std Obs Mean Std Obs Mean Std (1)

[-200, 2008 -50) % [-100, In -50)

(2)

(1)

LHCOOP Mean Std Obs (1)

(2)

(2)

IHCOOP Mean Std

(1)

(2)

Obs

(1)

(2)

NA NA 0.08 NA NA NA

1 1 3 1 1 1

0.80

6.29

6

1.60

0.80

6.29

6

1.60

0.08 NA

JSB Std (1)

-178.94 JSB NA Mean Std -56.52 NA (1) -19.82 20.15 -178.94 NA 9.90 5.76 -56.52 NA 53.17 NA -19.82 20.15 9.90 5.76 53.17 NA

-10.84 -2.09 4.14 -10.84 5.33 -2.09 4.14 5.33

2 1 12 41 1 4

Mean

(2)

MHCOOP Mean Std

Obs

(1)

[-50, 0) [0, [-200, 50) -50) [-100, [-15, -10) -50) [-50, 0) [-5, 0) -1.48 NA 1 [0, 50) [0, 5) 4.54 NA 1 [-15, -10) [5, [-5, 10) 0) 6.84 0.79 21 -1.48 NA [10,[0, 15) 13.05 5) 4.54 NA NA 1 1 [0, [5, 2) 10) 1.36 0.47 6.84 0.79 2 [2, [10, 4) 15) 2.52 NA 13.05 NA 1 [8, [0, 10)2) 8.50 NA 1.36 0.47 5.95 5.23 5 3.43 3.42 [2, 4) 2.52 NA [8,(1) 10)standard-deviation (2) number of observations 8.50 NA Note: 5.95 5.23 5 3.43 Note: (1) standard-deviation (2) number of observations3.42

Obs

2 7 233 2 28 7 270 233 28 270

Obs (2)

1 Obs 1(2) 6 1 38 11 6 38 1

3 1

32.05 32.05

47 47

Note: (1) standard-deviation (2) number of observations

Table 3-c: Return on equity descriptive statistics after the crisis year

Table 3-c:2009-11 Return on equityLHCOOP descriptive statistics after the crisis year IHCOOP

MHCOOP Table 3-c: Return on equity descriptive statistics after the crisis year In % Mean Std Obs Mean Std Obs Mean Std Obs (1) (2) (1) (2) (1) (2) 2009-11 LHCOOP IHCOOP MHCOOP In % Mean Std Obs Mean Std Obs Mean Std Obs [-400, -300) (1)

[-400, -300) [-200, -100) [-100, [-200,0) -100) [0, 100)0) [-100, [-10, -5) [0, 100) [-5, 0)-5) [-10, [0, [-5,5)0) [5, [0,10) 5) [10, 15) [5, 10) [2, 4) [10, 15) [4, [2,6) 4) [6, [4,8) 6) [8, [6,10) 8) [-40, -30) [8, 10) [-30, [-40,-20) -30) [0, 10)-20) [-30,

[0, 10)

-36.68 -21.64 -36.68 3.35 -21.64 -0.98 3.35

-0.98

NA NA NA 1.21 NA 11.84 1.21

11.84

(2)

(1)

1 11 13 1 15 13

2.75 4.85 2.75 6.56 4.85 9.46 6.56

0.72 0.67 0.72 0.34 0.67 NA 0.34

2 42 54 15

5.59

1.91

12

9.46

15

(2)

5.59

NA

1.91

(1)

Mean -374.8100

(2)

-5.23 -0.18 -5.23 3.34 -0.18 6.54 3.34 13.79 6.54

NA NA NA 1.27 NA 0.82 1.27 NA 0.82

1 11 41 84 18

4.94

4.21

15

13.79

Mean

NA

JSB Std JSB (1) NAStd

Obs (2)

Obs 1(2)

(1)

-374.8100 -100.7050 -17.79913 -100.7050

NA 1 0.120208 20 2 22.46528 232 0.120208 20 22.46528 23

8.18 -0.56 8.18

4.99 37.30 4.99

-17.79913

1

1

12

4.94

4.21

15

-0.56

111 137 111

37.30

137

Table 3-d: Return on equity descriptive statistics on the whole period

Table 3-d: Return on equity descriptive statistics on the whole period

Table 3-d: Return on equity descriptive statistics on the whole period

2002-11 In % 2002-11

In % [-400, -300) [-200, -100) [-400, -300) [-100, 0) [-200, -100) [0, 5) [-100, 0) [5, 10) [0, 5) [10, 15) [5, 10) [15, 20) [10, 15) [20, 30) [15, 20) [-40, -20) [20, 30) [-20, 0) [-40, -20) [-20, -10) [-20, 0) [-10, 0) [-20, -10) [0, 10) [-10, 0) [10, 20) [0, 10) [0, 20) [10, 20) 20, 30) [0, 20) [20, 40) 20, 30) [40, 60) [20, 40) [40, 60)

LHCOOP Mean LHCOOP Std Obs (1) (2) Mean Std Obs (1)

(2)

-29.16 -1.48 -29.16 -1.48

10.63 NA 10.63 NA

2 1 2 1

7.78

3.62

47

7.78 6.12

3.62 8.32

IHCOOP Mean IHCOOP Std Obs (1) (2) Mean Std Obs (1)

3.23 7.41 3.23 11.94 7.41 15.20 11.94 15.20

1.43 1.27 1.43 1.24 1.27 NA 1.24 NA

8.32

50

50

7.00

7.00

2.92

2.92

Note: (1) standard-deviation (2) number of observations

(1)

(2)

9 27 9 3 27 1 3 1

47

Note: (1) standard-deviation (2) number of observations

6.12

(2)

MHCOOP Mean MHCOOP Std Obs (1) (2) Mean Std Obs

40

40

-10.84 -2.50 -10.84 6.53 -2.50 12.96 6.53 12.96 23.54

NA 2.54 NA 2.11 2.54 2.68 2.11 2.68 2.16

1 3 1 35 3 15 35 15 3

23.54

2.16

3

8.33

6.16

8.33

6.16

57

57

Mean

JSB (1) Std JSB

Obs(2)

Mean -374.81 -126.78 -374.81 -18.11 -126.78 -18.11

Std NA 45.16 NA 19.99 45.16 19.99

Obs(2) 1 3 1 39 3 39

10.21

4.66

377

10.21 24.19 53.17 24.19 7.14 53.17

4.66 4.50 NA 4.50 24.32 NA

377 33 1 33 1 454

7.14

(1)

24.32

454

Note: (1) standard-deviation (2) number of observations Appendix 4: Hausman test results

Appendix 4: Hausman test results The Hausman test is a specification test. It allows us to determine whether both estimations’ (fixed and random coefficientstest. are statistically Under the null hypothesis H0 The Hausman test effect) is a specification It allows usdifferent. to determine whether both estimations’ 76 75 of orthogonality between explanatory variables and the errorUnder term the of the random effect (fixed and random effect)the coefficients are statistically different. null hypothesis H0 JEOD - Vol.3, Issue 1 (2014) model, both estimators within (LSDV) and generalized leasterror squares are unbiased of orthogonality between the explanatory variables and the term (GLS) of the random effect estimators. this case, within there is(LSDV) no significant differenceleast between the (GLS) coefficients in the model, bothIn estimators and generalized squares are unbiased


[0, 10) [10, 20) [0, 20) 20, 30) [20, 40) [40, 60)

7.78

3.62

47

6.53 12.96

2.11 2.68

35 15

23.54

2.16

3

10.21

4.66

377

24.19

4.50 NA

33 1

Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y. 53.17

6.12

8.32

50

7.00

2.92

Note: (1) standard-deviation (2) number of observations

40

8.33

6.16

57

7.14

24.32

454

Appendix 4: Hausman test results Appendix 4: Hausman test results The Hausman test is a specification test. It allows us to determine whether both estimations’ (fixed and The Hausman test is a specification test. It allows us to determine whether both estimations’ random effect) coefficients are statistically different. Under the null hypothesis H0 of between (fixed and random effect) coefficients are statistically different. Under theorthogonality null hypothesis H0 the explanatory variables and the term of the random effect model, both estimators (LSDV) of orthogonality between theerror explanatory variables and the error term of the within random effect and generalized least squareswithin (GLS) are unbiasedand estimators. In thisleast case, there is no(GLS) significant model, both estimators (LSDV) generalized squares are difference unbiased between the coefficients in thethere withinisand leastdifference square estimations. use the generalized estimators. In this case, nogeneralized significant betweenThus thewecoefficients in the least square as ours least is a random model. within andmethod, generalized squareeffect estimations. Thus we use the generalized least square method, as ours is a random effect model. The hypothesis test is:

The hypothesis test is: a random effect model H0: thus,the themodel modelis is a random effect model H0: âLSDV – âGLS = 0 thus, H1: âLSDV – âGLS ≠ 0 thus, thus,the themodel modelis is a fixed effect model a fixed effect model H1: H statisticsformula formulais: is: TheThe H statistics H = (âLSDV – âGLS)’[Var(âLSDV) – Var(âGLS)]-1(âLSDV – âGLS)

21

The H statistics formula is:

a αthreshold, fixed threshold, H statistics follow withk kdegree degree of freedom. -1 freedom. H statistics followaachi-square chi-square with of If H>If H> (k) for(k) a αfor fixed then we H = (â LSDV – âGLS)’[Var(âLSDV) – Var(âGLS)] (âLSDV – âGLS) thenreject we will reject H0. estimator The within is so we can effect rejectspecification the random will H0. The within is soestimator unbiased, we canunbiased, reject the random andeffect use H statistics follow a individual chi-square with k degree of freedom. If H> (k) for a α fixed threshold, specification and use an fixed effects model. an individual fixed effects model. then we will reject H0. The within estimator is so unbiased, we can reject the random effect

In our model, weand ranuse theanHausman results indicate that for a degree of freedom of 7, the specification individual test. fixedThe effects model. In our model, we ran the Hausman test. The results indicate (and that even for athe degree of freedom of 7, chi-square statistics are widely inferior to the five percent threshold 10 percent threshold). the chi-square statistics widely test. inferior to theindicate five that percent threshold (and ofeven our model, we ran are Hausman The results for a degree of freedom 7, the 10 ThereforeInwe cannot reject thetheH0 hypothesis that the model is a random effect model. chi-squareTherefore statistics arewe widely inferior to the threshold the is10a random percent the threshold). cannot reject the five H0 percent hypothesis that(and the even model percent threshold). Therefore we cannot reject the H0 hypothesis that the model is a random effect model.

effect model. Effects – Hausman Test Correlated Random Equation: EQZSCORE Correlated Random Effects – Hausman Test Test cross-section Correlated random Random effects Effects – Hausman Test Equation: EQZSCORE

Equation: EQZSCORE Test cross-section random effects Test cross-section random effects

Chi-sq. Chi-sq.Chi-Sq. d.f. statistic

Test summary

Test summary

Chi-Sq. d.f.

Prob.

2.882528

7

0.8957

2.882528

Cross-section random

Cross-section random

Prob.

statistic

7

0.8957

Cross-section random effects test comparisons: Variable Fixed Cross-section random effects test comparisons: Loans to assets^2 Variable Loans to assets

Net interest margin Loans assets^2 on assets Fees andto commissions 2008 Loans CRISIS to assets Long-term interest Net interest margin rate Cost to income ratio

0.003242 Fixed -0.343761 3.659333 0.003242 4.951734 -2.119291 -0.343761 -0.776844 3.659333 -0.024749

Random

Var(Diff.)

0.003037 Random -0.320551 3.739466 0.003037 4.873370 -2.108077 -0.320551 -0.776473 3.739466 -0.025341

Prob.

0.000000 0.2910 Var(Diff.) 0.000473 0.2858 0.003532 0.1775 0.000000 0.045381 0.7130 0.000534 0.6274 0.000473 0.000102 0.9707 0.003532 0.000001 0.4663

Fees and commissions on assets 4.951734 4.873370 CRISIS 2008 -2.119291 -2.108077 Long-term interest rate -0.776844 -0.776473 Cost to income Appendix ratio -0.024749 5: Research of unit-0.025341 root process

0.045381 0.000534 0.000102 0.000001

Prob. 0.2910 0.2858 0.1775 0.7130 0.6274 0.9707 0.4663

In order to check the unit root test presence, we use the specific Levin, Lin and Chu test adapted for the panel data unit root test search. This fits the panel that contains the least observations (few individuals and/or periods).of unit root process Appendix 5: few Research Here, the test indicates that we can reject the null hypothesis that assumes a common unit root

In orderprocess. to check thez-score unit series root istest presence, Levin, Lin and Chu test So, the stationary and wewe canuse workthe withspecific it on level. adapted for the panel data unit root test search. This fits the panel that contains the least 77 76 Panel unit root test: summary JEODperiods). - Vol.3, Issue 1 (2014) observations (few individuals Series: ZSCOREand/or few Date: 04/28/13 Time: 18:11 Sample: 2002 2011


Loans to assets^2 0.003242 0.003037 0.000000 Loans to assets -0.343761 -0.320551 0.000473 Net interest margin 3.659333 3.739466 0.003532 Fees and Did commissions on assets 4.951734 4.873370 0.045381 the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y. CRISIS 2008 -2.119291 -2.108077 0.000534 Long-term interest rate -0.776844 -0.776473 0.000102 Cost to income ratio -0.024749 -0.025341 0.000001

0.2910 0.2858 0.1775 0.7130 0.6274 0.9707 0.4663

Appendix 5: Research of unit root process

Appendix 5: Research unit rootLevin, process In order to check the unit root test presence, we useofthe specific Lin and Chu test adapted for the panel data unit root test search. This fits the panel that contains the least observations (few individuals In and/or orderfew to periods). check the unit root test presence, we use the specific Levin, Lin and Chu test adapted for the panel data unit root test search. This fits the panel that contains the least Here, the test indicates that we can reject the null hypothesis that assumes a common unit root process. observations (few individuals and/or few periods). So, the z-score series is stationary and we can work with it on level. Here, the test indicates that we can reject the null hypothesis that assumes a common unit root Panel unit root test: summary process. So, the z-score series is stationary and we can work with it on level. Series: ZSCORE Panel unit18:11 root test: summary Date: 04/28/13 Time: Series: ZSCORE Sample: 2002 2011 Date: 04/28/13 Time: 18:11 Exogenous variables: individual effects Sample: 2002 2011 User-specifiedExogenous lags: 1 variables: individual effects Newey-West automatic bandwidth User-specified lags: 1 selection and Bartlett kernel

Newey-West automatic bandwidth selection and Bartlett kernel CrossMethod Statistic Prob.** sections Levin, &(assumes Chu t* -4.41185 0.0000 61 61 Null: unit Lin root root process) Levin, Lin & Chu t* common unit -4.41185 0.0000 Levin, Lin & Chu t*

-4.41185

0.0000

61

Obs 471 471 471

22

Appendix6.1: 6.1:Results Results from from the regressions where the endogenous variable is Appendix thedata datapanel panel regressions where the endogenous variable is Appendix 6.1: Results from the data panel regressions where the endogenous variable is the Z-score the theZ-score Z-score

Appendix 6.1: Results from the data panel regressions where the endogenous variable is 2 2 i Loansi,t + αi++β1 β1iLHCoop LHCoopi,t ++β2 Z-Scorei, t, = c =αi i,t + β3 i MHCoopi,t + β4i Loans i,t + β5 the Z-score Z-Score β2i IHCoop i, t, c i i,t i IHCoop i,t + β3i MHCoopi,t + β4i Loans i,t + β5i Loansi,t + β6i NIMi,t + β7i Feesi,t + ∑ φt Crisist + β8c Ratec,t + β9i CTIRi,t + ɛi,t ∀ i = 1,..,64, where β6i NIM i,t + β7i Feesi,t + ∑ φt Crisist + β8c Ratec,t + β9i CTIRi,t + ɛi,t ∀ i = 1,..,64, where 2 that have an LHCoop the β1 least hybrid cooperative banks,+ IHCoop the cooperative banks =isαi LHCoop + β2i IHCoop β3IHCoop + β4i Loans + β5i Loans Z-Score i, t, c is i,tcooperative i,t i MHCoop i,t + an i,tbanks LHCoop the+ least hybrid banks, thei,tcooperative cooperative that have where LHCoop is thei least hybrid cooperative banks, IHCoop the banks thatand have an intermediate degree of hybridization, MHCoop the most hybrid cooperative banks JSB + hybridization, ∑ φt Crisist +MHCoop β8c Ratec,tthe+ most β9 CTIR ɛi,t ∀ i = 1,..,64, where β6intermediate i NIMi,t + β7degree i Feesi,tof i,t +cooperative hybrid and JSB the joint-stock banks. intermediate degree of hybridization, MHCoop the most hybrid icooperative banks and JSBbanks the jointLHCoop is the least hybrid cooperative banks, IHCoop the cooperative banks that have an the joint-stock banks. stock banks. intermediate degree of hybridization, MHCoop the most hybrid cooperative banks and JSB Table 6.1 a: Results the joint-stock banks.for the data panel regression using the full sample Table 6.1 a: Results for the data panel regression using the full sample Full sample

Z-score Z-score Z-score Z-score 2002–11 2002–7 2008 2009–11 Table 6.1 a: Results for the data panel regression using the full Z-score sample Full sample Z-score Z-score Z-score 25.59 28.16 2.37 15.71 C 2002–11 2002–7 2008 2009–11 (5.388***) (5.248***) (0.033) (2.160**) 25.59 Full sample Z-score -0.78 Z-score 2.4128.16 Z-score Z-score -10.36 2.37 -2.913420 15.71 C LHCOOP (5.388***) (5.248***) (-0.905) (-0.071) (0.219843) 2002–11 2002–7 2008 (0.033) (-0.252346) 2009–11(2.160**) 53.36 49.83 55.96 58.72 -0.78 -10.36 -2.913420 25.59 28.162.41 2.37 15.71 IHCOOP C LHCOOP (4.368***) (4.106***) (4.608***) (-0.071) (0.219843)(4.649***) (-0.905) (-0.252346) (5.388***) (5.248***) (0.033) (2.160**) -1.825 -2.56 -1.80 -3.313 58.72 53.36 49.83 55.96 -0.78 2.41 -10.36 -2.913420 MHCOOP IHCOOP LHCOOP (-0.191) (-0.254) (-0.183) (-0.287) (4.368***) (4.106***) (4.649***) (4.608***) (-0.071) (0.219843) (-0.905) (-0.252346) 0.003 0.0012 0.004 0.0043 -1.825 -2.56 -1.80 -3.313 LOANS_TOASSETS2 53.36 49.83 55.96 58.72 (2.679***) (0.839) (0.654) (2.019**) MHCOOP IHCOOP (-0.191) (-0.254) (-0.183) (-0.287) (4.368***) (4.106***) (4.649***) (4.608***) -0.32 -0.225 -0.477 -0.314 LOANS_TOASSETS2 0.003 0.0012 (-0.734) 0.004 0.0043 -1.825 -2.56 -1.80 -3.313 (-2.632***) (-1.473) (-1.367) LOANS_TOASSETS MHCOOP (2.679***) (0.839) (0.654) (2.019**) (-0.191) (-0.254) (-0.183) (-0.287) 3.73 2.18 14.05 4.162 Net Interest Margin 0.003 0.0012 0.004 0.0043 -0.32 -0.225 (2.480***) -0.477 -0.314 (6.872***) (3.318***) (2.470**) 2 LOANS_TOASSETS LOANS_TOASSETS 4.873 8.26 -6.769 15.76 (-1.367) (2.679***) (0.839) (0.654) (2.019**) (-2.632***) (-1.473) (-0.734) FEES_AND_COMM_ONASSETS (2.985***) (3.976***) (3.127***) -0.323.73 -0.2252.18 (-0.424) -0.477 -0.314 14.05 4.162 LOANS_TOASSETS Net Interest Margin -2.10 (-2.632***) (-1.473) (-0.734) (-1.367) (6.872***) (3.318***) (2.480***) (2.470**) CRISE2008 (-3.028***) Not Included 3.73 2.188.26 14.05 4.162 4.873 -6.769 15.76 NetFEES_AND_COMM_ONASSETS Interest Margin -0.77 -0.969 3.336 -1.135 (6.872***) (3.318***) (2.480***) (2.470**) (2.985***) (3.976***) (-0.424) (3.127***) Long-Term Interest rate (-2.549***)77 (-1.761*) (0.210) (-3.671***) 78 4.873 8.26 -6.769 15.76 -2.10 FEES_AND_COMM_ONASSETS -0.036 -0.049 -0.011 CRISE2008 JEOD-0.025 - Vol.3, Issue 1 (2014) Cost To Income Ratio (2.985***) (3.976***) (-0.424) (3.127***) (-3.028***) Not Included (-1.753*) (-1.379) (-0.498) (-0.552) -2.10 -0.77 -0.969 3.336 -1.135 R² CRISE2008 Long-Term Interest rate


the Z-score Appendix 6.1: Results from the data panel regressions where the endogenous variable is 2 the Z-score Z-Scorei, t, c = αi + β1i LHCoopi,t + β2i IHCoop i,t + β3i MHCoopi,t + β4i Loans i,t + β5i Loansi,t + β6i NIMi,t + β7i Feesi,t + ∑ φt Crisist + β8c Ratec,t + β9i CTIRi,t + ɛi,t ∀2 i = 1,..,64, where Extent of iHybridization Cooperative Groups Face the Financial Crisis? Z-Scorei,Did = αi + β1 LHCoop + β2Enable β3iBanking MHCoop +toβ4 i,tbetter i IHCoop i,t + IHCoop i,t i Loans i Loans i,t +an i,t + β5 LHCoop ist, cthe the least hybrid cooperative banks, the cooperative banks that have Lemzeri, Y. β6i NIMi,t + degree β7i Fees + ∑ φt Crisist MHCoop + β8c Rate + β9i hybrid CTIRi,t cooperative + ɛi,t ∀ i = banks 1,..,64,and where c,t most intermediate ofi,t hybridization, the JSB LHCoop is thebanks. least hybrid cooperative banks, IHCoop the cooperative banks that have an the joint-stock intermediate degree of hybridization, MHCoop the most hybrid cooperative banks and JSB the joint-stock banks. Table 6.1 a: Results for the data panel regression using the full sample

Table 6.1 a: Results for the data panel regression using the full sample

Table 6.1 Full a: Results sample samplefor the data panel regression Z-score using the full Z-score Z-score 2002–11 2002–7 2008 Full sample Z-score 25.59 Z-score 28.16 Z-score2.37 C 2002–11 2002–7 2008 (5.388***) (5.248***) (0.033) 25.59 28.16 2.37 -0.78 2.41 -10.36 C LHCOOP (5.388***) (5.248***) (0.033) (-0.071) (0.219843) (-0.905) -0.78 2.41 -10.36 53.36 49.83 55.96 LHCOOP IHCOOP (-0.071) (0.219843) (-0.905) (4.368***) (4.106***) (4.649***) 53.36 49.83 55.96 -1.825 -2.56 -1.80 IHCOOP MHCOOP (4.368***) (4.106***) (4.649***) (-0.191) (-0.254) (-0.183) -1.825 -2.56 -1.80 0.003 0.0012 0.004 2 MHCOOP LOANS_TOASSETS (-0.191) (-0.254) (-0.183) (2.679***) (0.839) (0.654) 0.003 0.0012 0.004 -0.32 -0.225 -0.477 LOANS_TOASSETS2 LOANS_TOASSETS (2.679***) (0.839) (0.654) (-2.632***) (-1.473) (-0.734) -0.32 -0.225 -0.477 3.73 2.18 14.05 LOANS_TOASSETS Net Interest Margin (-2.632***) (-1.473) (-0.734) (6.872***) (3.318***) (2.480***) 3.73 2.18 14.05 Net Interest Margin 4.873 8.26 -6.769 (6.872***) (3.318***) (2.480***) FEES_AND_COMM_ONASSETS (2.985***) (3.976***) (-0.424) 4.873 8.26 -6.769 FEES_AND_COMM_ONASSETS -2.10 (2.985***) (3.976***) (-0.424) CRISE2008 (-3.028***) Not Included -2.10 CRISE2008 -0.77 -0.969 Not Included 3.336 (-3.028***) Long-Term Interest rate (-2.549***) (-1.761*) (0.210) -0.77 -0.969 3.336 Long-Term Interest rate -0.025 -0.036 -0.049 (-2.549***) (-1.761*) (0.210) Cost To Income Ratio (-1.753*) (-1.379) (-0.498) -0.025 -0.036 -0.049 R²Cost To Income Ratio (-1.753*) (-1.379) (-0.498) 0.216 0.2157 0.4246 R² Adjusted R² 0.216 0.2157 0.4246 0.202 0.195264 0.3231 Adjusted R² F 0.202 0.195264 0.3231 15.989*** 10.544*** 4.182*** F 15.989*** 10.544*** 4.182*** Number of observations 591 355 61 Number of observations 591 355 61 Note: Into brackets, the Student T. Note: Into brackets, Student T. *** significant at 1%the level; ** significant at 5% level; * significant at 10% level. *** significant at 1% level; ** significant at 5% level; * significant at 10% level.

Z-score 2009–11 Z-score 15.71 2009–11 (2.160**) 15.71 -2.913420 (2.160**) (-0.252346) -2.913420 58.72 (-0.252346) (4.608***) 58.72 -3.313 (4.608***) (-0.287) -3.313 0.0043 (-0.287) (2.019**) 0.0043 -0.314 (2.019**) (-1.367) -0.314 4.162 (-1.367) (2.470**) 4.162 15.76 (2.470**) (3.127***) 15.76 (3.127***) -1.135 (-3.671***) -1.135 -0.011 (-3.671***) (-0.552) -0.011 (-0.552) 0.3126 0.3126 0.2751 0.2751 8.340*** 8.340*** 175 175

Note: Into brackets, the Student T. *** significant at 1% level; ** significant at 5% level; * significant at 10% level.

Table 6.1 b: Results for the data panel regression using the full sample except Dexia Table 6.1 b: Results for the data panel regression using the full sample except Dexia Table 6.1 b: Results for the data panel regression using the full sample except Dexia Full sample except Dexia Full sample except Dexia C LHCOOP IHCOOP MHCOOP LOANS_TOASSETS2 LOANS_TOASSETS Net Interest Margin FEES_AND_COMM_ONASSETS CRISE2008 Long-Term Interest rate Cost To Income Ratio R² Adjusted R² F Number of observations

Z-score Z-score 2002-11 2002-11 25.71 (5.323***) -1.055 (-0.094) 53.10 (4.308***) -2.078 (-0.2163) 0.00305 (2.674***) -0.3194 (-2.603***) 3.725 (6.805***) 4.908 (2.987***) -2.034 (-2.868***) -0.792 (-2.574***) -0.024 (-1.487)

Z-score Z-score 2002-7 2002-7 28.67

(5.257***) 2.189 (0.196) 49.58 (4.044***) -2.814 (-0.275) 0.0012 (0.832) -0.2272 (-1.469) 2.18 (3.288***) 8.238 (3.928***)

(0.043) -10.619 (-0.919262) 55.49 (4.560***) -1.98 (-0.199) 0.0033 (0.523) -0.4033 (-0.599) 13.91 (2.432**) -8.215 (-0.502)

Z-score Z-score 2009-11 2009-1116.82 (2.178**) -2.947 2323 (-0.252) 58.62 (4.558***) -3.473551 (-0.298) 0.0042 (1.948*) -0.3050 (-1.316) 3.955 (2.266**) 15.65 (3.070***)

-1.003 (-1.790*) -0.036 (-1.381)

Not Included 3.139 (0.196) -0.053 (-0.531)

-1.134 (-3.651***) -0.022 (-0.612)

0.2139

0.2155

0.4178

0.3102

0.2002

0.1947

0.3130

0.2722

15.545***

10.351***

3.987***

8.148***

582

349

60

173

Note: the Student T is in brackets. *** significant at 1% level; ** significant at 5% level; * significant at 10% level.

Z-score Z-score 2008 2008 3.096

Note: the Student T is in brackets. Table c: results for the data panel regression using at the10% full sample *** significant at 1%6.1 level; ** significant at 5% level; * significant level. except rabobank Full sample except Rabobank C LHCOOP IHCOOP

Z-score Z-score 2002-11 2002-7 24.16 28.20 79 (6.016***) 78 (5.881***) JEOD --0.87 Vol.3, Issue 1 (2014) 2.15 (-0.104) (0.249) 22.70 20.39

Z-score 2008 18.98 (0.357) -10.21 (-1.193) 24.049

Z-score 2009-11 15.667 (2.415**) -2.91 (-0.347) 24.709


Long-Term Interest rate Cost To Income Ratio R²

-0.792 (-2.574***) -0.024 (-1.487)

-1.003 (-1.790*) -0.036 (-1.381)

3.139 (0.196) -0.053 (-0.531)

-1.134 (-3.651***) -0.022 (-0.612)

0.2139 0.2155 0.4178 0.3102 Adjusted R² 0.2002 Did the Extent of Hybridization better Enable Cooperative Banking0.1947 Groups to Face0.3130 the Financial Crisis? 0.2722 Lemzeri, Y. F 15.545*** 10.351*** 3.987*** 8.148*** Number of observations

582

349

Note: the Student T is in brackets. *** significant at 1% level; ** significant at 5% level; * significant at 10% level.

60

173

Tablefor 6.1the c: results for the data panel regression the full sample except rabobank Table 6.1 c: results data panel regression using the fullusing sample except rabobank Full sample except Rabobank

28.20 (5.881***) 2.15 (0.249) 20.39 (1.878*) -2.64 (-0.334) 0.00078 (0.560) -0.183 (-1.249) 2.35 (3.714***) 7.791308 (3.900***)

Z-score 2008 18.98 (0.357) -10.21 (-1.193) 24.049 (2.343156**) -3.048 (-0.413) 0.00239 (0.518) -0.378 (-0.780037) 14.91 (3.524***) -5.616804 (-0.471)

Z-score 2009-11 15.667 (2.415**) -2.91 (-0.347) 24.709 (2.338**) -3.657 (-0.436) 0.00388 (1.838*) -0.312 (-1.400) 4.98 (3.031***) 16.38264 (3.435***)

-1.16 (-2.167**) -0.036 (-1.430)

Not Included -0.72 (-0.061) -0.053 (-0.728)

-1.15 (-3.719***) -0.011 (-0.541)

0.2167

0.2072

0.3986

0.2778

0.2029

0.1861

0.2903

0.2377

15.769***

9.844***

3.682***

6.925***

581 15.769***

349 9.844***

60 3.682***

24 172 6.925***

*** significant at 1% level; ** significant at 5% level; * significant at 10% level.349 581

60

172

C LHCOOP IHCOOP MHCOOP LOANS_TOASSETS2 LOANS_TOASSETS Net Interest Margin FEES_AND_COMM_ONASSETS CRISE2008 Long-Term Interest rate Cost To Income Ratio R² Adjusted R² F

FNumber of observations Note: theof Student T is in brackets. Number observations

Z-score 2002-11 24.16 (6.016***) -0.87 (-0.104) 22.70 (2.155**) -1.969 (-0.274) 0.00239 (2.248**) -0.265 (-2.323**) 3.79 (7.404***) 4.69 (3.059***) -2.11 (-3.196***) -0.66 (-2.289**) -0.023 (-1.693*)

Z-score 2002-7

Note: the Student T is in brackets. *** significant

Note: the Student T is in brackets. ***1% significant at 1% level; ** significant at 5% *level; * significant at 10%level. level. at level; ** significant at 5% level; significant at 10%

Table 6.1 d: Results for the data panel regression using the full sample except Dexia and Rabobank

Table for 6.1 the d: Results for the data panel regression the full sample except Dexia and Rabobank Table 6.1 d: Results data panel using the fullusing sample except Rabobank Full sample except Dexiaregression and Z-score Z-scoreDexia andZ-score Z-score Rabobank

C Full sample except Dexia and Rabobank LHCOOP C IHCOOP LHCOOP MHCOOP IHCOOP LOANS_TOASSETS2 MHCOOP LOANS_TOASSETS2 LOANS_TOASSETS Net Interest Margin LOANS_TOASSETS FEES_AND_COMM_ONASSETS Net Interest Margin CRISE2008 FEES_AND_COMM_ONASSETS Long-Term Interest rate CRISE2008 Cost To Income Ratio Long-Term Interest rate R² Cost To Income Ratio Adjusted R² R² F Adjusted R²

FNumber of observations Note: theof Student T is in brackets. Number observations

2002-11 Z-score 24.29 (5.948***) 2002-11 -1.14 24.29 (-0.135) (5.948***) 22.43 -1.14 (2.109**) (-0.135) -2.22 22.43 (-0.306) (2.109**) 0.0024 -2.22 (2.241**) (-0.306) -0.263 0.0024 (-2.290**) (2.241**) 3.78 -0.263 (7.331***) (-2.290**) 4.71 3.78 (3.051***) (7.331***) -2.041 4.71 (-3.026***) (3.051***) -0.676 -2.041 (-2.316**) (-3.026***) -0.0223 -0.676 (-1.439) (-2.316**)

2002-7

Z-score 28.74 (5.899***) 2002-7 1.91 28.74 (0.219) (5.899***) 20.13 1.91 (1.834*) (0.219) -2.89 20.13 (-0.362) (1.834*) 0.000774 -2.89 (0.545) (-0.362) -0.183 0.000774 (-1.239) (0.545) 2.35 -0.183 (3.683***) (-1.239) 7.74 2.35 (3.839***) (3.683***)

2008 19.66 Z-score (0.368) 2008 -10.45 19.66 (-1.213) (0.368) 23.61 -10.45 (2.280**) (-1.213) -3.21 23.61 (-0.433) (2.280**) 0.001727 -3.21 (0.361) (-0.433) -0.307 0.001727 (-0.612) (0.361) 14.78 -0.307 (3.465***) (-0.612) -7.00 14.78 (-0.574) (3.465***) Not Included -7.00

2009-11 Z-score 16.70 (2.417**) 2009-11 -2.91 16.70 (-0.343) (2.417**) 24.70 -2.91 (2.315**) (-0.343) -3.77 24.70 (-0.445) (2.315**) 0.003755 -3.77 (1.761*) (-0.445) -0.299 0.003755 (-1.330) (1.761*) 4.74 -0.299 (2.779***) (-1.330) 16.22 4.74 (3.350***) (2.779***)

-0.0574 0.3875 (-0.771610)

-0.0225 0.2736 (-0.623)

7.74 (3.839***) -1.202 (-2.196**) -0.0367 -1.202 (-1.441) (-2.196**)

(-0.574) -0.913 Not Included (-0.076) -0.0574 -0.913 (-0.771610) (-0.076)

16.22 (3.350***) -1.152 (-3.693***) -0.0225 -1.152 (-0.623) (-3.693***)

0.2000 0.2140

0.1857 0.2072

0.2750 0.3875

0.2327 0.2736

15.279*** 0.2000

9.670*** 0.1857

3.445*** 0.2750

6.696*** 0.2327

572 15.279***

343 9.670***

59 3.445***

170 6.696***

59

170

-0.0223 0.2140 (-1.439)

-0.0367 0.2072 (-1.441)

*** significant at 1% level; ** significant at 5% level; * significant at 10% level.

572

Note: the Student T is in brackets. Student T is in brackets. *** significant at 1% level; ** significant at 5% level; * significant at 10% level.

343

Note: the *** significant at 1% level; ** significant at 5% level; * significant at 10% level. Appendix 6.2: Results of the data panel regressions where the endogenous variable is the return on equity Appendix 6.2: Results of the data panel regressions where the endogenous variable is the return on equity ROEi, t, c = αi + β1i LHCoopi,t + β2i IHCoop i,t + i MHCoopi,t + β4c Ratec,t + β5i NIMi,t + ∑ φt 80 79β3 Crisist + ɛi,t ∀ i = 1,..,64, JEOD - Vol.3, Issue 1 (2014) β1the β3i MHCoop + β4cooperative ∑ φt ROE c = αi + is i LHCoop i,t + β2cooperative i IHCoopi,t +banks, i,t the c Ratec,t + β5 i NIMthat i,t + have wherei, t,LHCoop least hybrid IHCoop banks 1,..,64,of hybridization, MHCoop the most hybrid cooperative banks and JSB Crisis t + ɛi,t ∀ i =degree an intermediate


F F Number of observations Number of observations

0.2000 0.2000

0.1857 0.1857

0.2750 0.2750

0.2327 0.2327

15.279*** 15.279***

9.670*** 9.670***

3.445*** 3.445***

6.696*** 6.696***

572

343

59

170 170

Did the Extent of Hybridization better Enable Cooperative Banking Groups 343 to Face the Financial Crisis? 572 59

Lemzeri, Y. Note: the Student T is in brackets. Note: the Student T is in brackets. *** significant at 1% level; ** significant at 5% level; * significant at 10% level. *** significant at 1% level; ** significant at 5% level; * significant at 10% level.

Appendix of of thethe data panel regressions wherewhere the endogenous variablevariable is the return on Appendix 6.2: Results data panel regressions the is Appendix6.2: 6.2:Results Results of the data panel regressions where the endogenous endogenous variable is the the return on equity equity return on equity

ROE = αi αi + + β1 β1i LHCoop i,t + β2i IHCoopi,t + β3i MHCoopi,t + β4c Ratec,t + β5i NIMi,t + ∑ φt ROEi,i, t,t, cc = i LHCoopi,t + β2i IHCoopi,t + β3i MHCoopi,t + β4c Ratec,t + β5i NIMi,t + ∑ φt + ɛ ∀ i = 1,..,64, Crisis i,t ∀ i = 1,..,64, where LHCoop is the least hybrid cooperative banks, IHCoop the cooperative Crisistt + ɛi,t where LHCoop is the the least least hybrid hybrid cooperative cooperative banks, banks, IHCoop IHCoop the cooperative cooperative banks banks that that have have where LHCoop banks that have anisintermediate degree of hybridization, MHCoop thethe most hybrid cooperative banks and an intermediate degree of hybridization, MHCoop the most hybrid cooperative banks and JSB an intermediate degree of hybridization, MHCoop the most hybrid cooperative banks and JSB JSB the joint-stock banks. the joint-stock joint-stock banks. banks. the Table 6.2 a: Results for the data panel regression using the full sample

Table 6.2 a: Results for the data panel regression using the full sample Table 6.2 a: Results for the data panel regression using the full sample Full sample Full sample

C C LHCOOP LHCOOP IHCOOP IHCOOP MHCOOP MHCOOP Long-term interest rate Long-term interest rate Net Interest Margin Net Interest Margin 2008CRISIS R² 2008CRISIS R² Adjusted R² Adjusted R² F F Number of observations Number of observations

Note: the Student T is in brackets.

ROAE 2002-11 ROAE 2002-11 -0.893 (-0.175) -0.893 -6.116 (-0.175) (-1.158) -6.116 -1.181 (-1.158) (-0.208) -1.181 0.219 (-0.208) (0.047) 0.219 0.190 (0.047) (0.165) 0.190 4.982 (0.165) (3.700***) 4.982 -5.364 (3.700***) (-2.051**) -5.364 (-2.051**) 0.03185 0.03185 0.02207 0.02207 3.257*** 3.257*** 601 601

ROAE 2002-7 ROAE 2002-7 13.673 (4.279***) 13.673 -3.294 (4.279***) (-2.348**) -3.294 -3.918 (-2.348**) (-2.688***) -3.918 -1.239 (-2.688***) (-1.020) -1.239 -0.718 (-1.020) (-0.956) -0.718 0.834 (-0.956) (2.008**) 0.834 (2.008**)

ROAE 2002-11 ROAE 2002-11 7.509 (2.897***) 7.509 -5.260 (2.897***) (-2.217***) -5.260 -2.251 (-2.217***) (-0.992) -2.251 (-0.992) -1.078 (-0.517) -1.078 (-0.517) -0.708 -0.708 (-1.184) 2.839 (-1.184) 2.839 (4.323***) -3.347 (4.323***) -3.347 (-2.432**) (-2.432**) 0.045 0.045 0.035 0.035 4.660*** 4.660*** 591 591

ROAE 2002-7 ROAE 2002-7 13.327 (4.104***) 13.327 -3.318 (4.104***) (-2.349**) -3.318 -3.854 (-2.349**) (-2.625***) -3.854 (-2.625***) -1.182 (-0.966) -1.182 (-0.966) -0.682 -0.682 (-0.894) 0.907 (-0.894) 0.907 (2.139**) (2.139**)

0.037 0.037 0.023 0.023 2.720** 2.720** 360 360

ROAE 2008 ROAE 2008 -74.205 (-0.919) -74.205 -1.732 (-0.919) (-0.122) -1.732 4.744 (-0.122) (0.307) 4.744 -0.903 (0.307) (-0.072) -0.903 16.065 (-0.072) (0.841) 16.065 3.702 (0.841) (0.767) 3.702 (0.767) Not included Not included 0.043 0.043 -0.043 -0.043 0.502 0.502 62 62

ROAE 2009-11 ROAE 2009-11 -8.478 (-0.830) -8.478 -5.913 (-0.830) (-0.456) -5.913 -2.127 (-0.456) (-0.149) -2.127 1.948 (-0.149) (0.153) 1.948 -3.318 (0.153) (-1.565) -3.318 14.503 (-1.565) (3.117***) 14.503 (3.117***)

ROAE 2008 ROAE 2008 -54.000 (-1.205) -54.000 -1.167 (-1.205) (-0.148) -1.167 0.786 (-0.148) (0.092) 0.786 (0.092) -4.141 (-0.598) -4.141 (-0.598) 13.606 13.606 (1.283) -0.053 (1.283) -0.053 (-0.020) (-0.020) Not included Not included 0.045 0.045 -0.042 -0.042 0.514 0.514 61 61

ROAE 2009-11 ROAE 2009-11 7.758 (2.106**) 7.758 -4.749 (2.106**) (-1.303) -4.749 -0.978 (-1.303) (-0.241) -0.978 (-0.241) 0.162 (0.045) 0.162 (0.045) -2.393 -2.393 (-2.503**) 3.504 (-2.503**) 3.504 (2.360**) (2.360**)

*** 1% ; ** significant at 5% level ; * significant at 10% level. Note:significant the T islevel in brackets. Note: the Student T Student is inatbrackets. *** significant at 1% level ; ** significant at 5% level ; * significant at 10% level. *** significant at 1% level ; ** significant at 5% level ; * significant at 10% level. Table 6.2 b: Results for the data panel regression using the full sample except dexia Table 6.2 b: Results for the data panel regression using the full sample except dexia Table 6.2 b: Results for the data panel regression using the full sample except dexia

Full sample except Dexia Full sample except Dexia

C C LHCOOP LHCOOP IHCOOP IHCOOP MHCOOP MHCOOP Long-term interest rate Long-term interest rate Net Interest Margin Net Interest Margin 2008CRISIS 2008CRISIS R² R² Adjusted R² Adjusted R² F F Number of observations Number of observations

0.038 0.038 0.024 0.024 2.723** 2.723** 354 354

Note: the Student T is in brackets. *** at 1% ** significant at 5% level; * significant at 10% level. Note:significant the Student T islevel; in brackets. *** significant 1% level; ** significant at 5% level; * significant at 10% level. Student T is in atbrackets.

0.060 0.060 0.032 0.032 2.192* 2.192* 179 179

0.061 0.061 0.033 0.033 2.195* 2.195* 176 176

Note: the *** significant at 1% level; ** significant at 5% level; * significant at 10% level. Table 6.2 c: Results for the data panel regression using the full sample except rabobank Table 6.2 c: Results for the data panel regression using the full sample except rabobank Full sample except Rabobank Full sample except Rabobank

ROAE ROAE 2002-11 81 2002-11 80

ROAE ROAE 2002-7 2002-7

JEOD - Vol.3, Issue 1 (2014)

ROAE ROAE 2008 2008

ROAE ROAE 2009-11 2009-11

26 26

25 25


R² Adjusted R² F

0.045

0.038

0.045

0.061

0.035

0.024

-0.042

0.033

4.660***

2.723**

0.514

2.195*

591

354

61

176

Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

Number of observations

Note: the Student T is in brackets. *** significant at 1% level; ** significant at 5% level; * significant at 10% level.

Table 6.2 c:for Results theregression data panel regression using the rabobank full sample except rabobank Table 6.2 c: Results the datafor panel using the full sample except Full sample except Rabobank C C LHCOOP LHCOOP IHCOOP IHCOOP MHCOOP MHCOOP Long-term interest rate Long-term interest rate Net Interest Margin Net Interest Margin 2008CRISIS 2008CRISIS R² R² Adjusted R² Adjusted R² F F Number of observations Number of observations

ROAE 2002-11 -0.891 -0.891 (-0.173) (-0.173) -6.138 -6.138 (-1.152) (-1.152) -1.976 -1.976 (-0.301) (-0.301) 0.215 0.215 (0.046) (0.046) -0.184 -0.184 (-0.157) (-0.157) 5.004 5.004 (3.686***) (3.686***) -5.456 -5.456 (-2.052**) (-2.052**) 0.032 0.032 0.022 0.022 3.232*** 3.232*** 591 591

ROAE 2002-7 13.665 13.665 (4.208***) (4.208***) -3.297 -3.297 (-2.330**) (-2.330**) -3.946 -3.946 (-2.347**) (-2.347**) -1.239 -1.239 (-1.012) (-1.012) -0.718 -0.718 (-0.940) (-0.940) 0.836 0.836 (1.992**) (1.992**) 0.033 0.033 0.019 0.019 2.377** 2.377** 354 354

Note: the Student T is in brackets. Note: the Student T is in brackets. *** significant 1% level; ** significant at 5 % level; * significant at 10% level. Note: the Student T isat in brackets. *** significant at 1% level; ** significant at 5 % level; * significant at 10% level.

ROAE 2008 -73.468 -73.468 (-0.901) (-0.901) -1.746 -1.746 (-0.122) (-0.122) 3.208 3.208 (0.180) (0.180) -0.917 -0.917 (-0.073) (-0.073) 15.885 15.885 (0.823) (0.823) 3.734 3.734 (0.767) (0.767) Not included Not included 0.043 0.043 -0.044 -0.044 0.490 0.490 61 61

ROAE 2009-11 -8.563 -8.563 (-0.830) 26 (-0.830) -5.964 -5.964 (-0.455) (-0.455) -4.492 -4.492 (-0.272) (-0.272) 1.918 1.918 (0.149) (0.149) -3.343 -3.343 (-1.561) (-1.561) 14.629 14.629 (3.105***) (3.105***) 0.060 0.060 0.032 0.032 2.165* 2.165* 176 176

*** significant at 1% level; ** significant at 5 % level; * significant at 10% level.

Table 6.2 d: Results for the data panel regression using the full sample except Dexia and Rabobank Table 6.2 d: Results for the data panel regression using the full sample except Dexia and Rabobank Table 6.2 d: Results for the data panel regression using the full sample except Dexia and Rabobank Full sample except Dexia and Full sample except Dexia and Rabobank Rabobank C C LHCOOP LHCOOP IHCOOP IHCOOP MHCOOP MHCOOP Long-term interest rate Long-term interest rate Net Interest Margin Net Interest Margin 2008CRISIS 2008CRISIS R² R² Adjusted R² Adjusted R² F F Number of observations Number of observations

ROAE ROAE 2002-11 2002-11 7.556 7.556 (2.877***) (2.877***) -5.275 -5.275 (-2.203***) (-2.203***) -3.078 -3.078 (-1.049) (-1.049) -1.083 -1.083 (-0.515) (-0.515) -0.725 -0.725 (-1.196) (-1.196) 2.859 2.859 (4.307***) (4.307***) -3.413 -3.413 (-2.440**) (-2.440**) 0.046 0.046 0.036 0.036 4.646*** 4.646*** 581 581

ROAE ROAE 2002-7 2002-7 13.316 13.316 (4.034***) (4.034***) -3.321 -3.321 (-2.331**) (-2.331**) -3.890 -3.890 (-2.297**) (-2.297**) -1.183 -1.183 (-0.958) (-0.958) -0.680 -0.680 (-0.878) (-0.878) 0.910 0.910 (2.122**) (2.122**) 0.034 0.034 0.020 0.020 2.391** 2.391** 348 348

Note: the Student T is in brackets. Note: the Student T islevel; in brackets. *** significant at 1% ** significant at 5% level; * significant at 10% level. *** significant atin 1%brackets. level; ** significant at 5% panel level; * significant at 10% level. Note: the Student T is6.3: Appendix Results of the data regression where

ROAE ROAE 2008 2008 -53.294 -53.294 (-1.178) (-1.178) -1.180 -1.180 (-0.148) (-0.148) -0.685 -0.685 (-0.069) (-0.069) -4.154 -4.154 (-0.595) (-0.595) 13.434 13.434 (1.255) (1.255) -0.022 -0.022 (-0.008) (-0.008) Not included Not included 0.045 0.045 -0.043 -0.043 0.515 0.515 60 60

ROAE ROAE 2009-11 2009-11 7.746 7.746 (2.083**) (2.083**) -4.764 -4.764 (-1.296) (-1.296) -1.813 -1.813 (-0.387) (-0.387) 0.148 0.148 (0.041) (0.041) -2.411 -2.411 (-2.496**) (-2.496**) 3.559 3.559 (2.366**) (2.366**) 0.061 0.061 0.032 0.032 2.155* 2.155* 173 173

the endogenous variable is the

Appendix 6.3: Results of the data regression where the endogenous variable is the *** significant at 1% level; ** significant at 5% level;panel * significant at 10% level. loans to assets ratio

loans to assets ratio Loansi, t, c = αi + β1i LHCoopi,t + β2i IHCoopi,t + β3i MHCoopi,t + β4c Ratec,t + ∑ φt Crisist + ɛi,t Loans + β1i LHCoop β2i IHCoop + β4c Rate φt cooperative Crisist + ɛi,t t, c = αiwhere i,t + β3 i MHCoopi,tbanks, c,t + ∑ ∀ i = i,1,..,64, LHCoopi,tis+ the least hybrid cooperative IHCoop the ∀ i = 1,..,64, where LHCoop is the least hybrid cooperative banks, IHCoop the cooperative

27 27

82 81 JEOD - Vol.3, Issue 1 (2014)


(-2.440**)

R² R² R² 0.061 Adjusted Adjusted R²

Not included

0.046 0.046

0.034 0.034

0.045 0.045

0.061 0.061

0.036 0.036

0.020 0.020

-0.043 -0.043

0.032 0.032

4.646*** 4.646***

2.391** 2.391**

0.515 0.515

2.155* 2.155*

Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

0.032 FF

of observations 2.155* Number Number of observations

581

348

60

581 348 60 Note: the Student T is in brackets. Note: the Student T islevel; in brackets. 173 *** significant at 1% ** significant at 5% level; * significant at 10% level. *** significant at 1% level; significant 5% level; * where significant 10% level. variable is the loans to assets Appendix 6.3: Results of the**data panelatregression theatendogenous

173 173

Appendix 6.3: Results of the data panel regression where the endogenous variable is the paneltoregression where the endogenous variable is the ratioAppendix 6.3: Results of the dataloans assets ratio us variable is the loans to assets ratio

Loansi, t, c = αi + β1i LHCoopi,t + β2i IHCoopi,t + β3i MHCoopi,t + β4c Ratec,t + ∑ φt Crisist + ɛi,t i, t, c = αi + β1i LHCoopi,t + β2i IHCoopi,t + β3i MHCoopi,t + β4c Ratec,t + ∑ φt Crisist + ɛi,t ∀Loans i = 1,..,64, where LHCoop is the least hybrid cooperative banks, IHCoop the cooperative + ∑ φt Crisist + ɛi,t ∀ i = 1,..,64, where whereLHCoop LHCoopis the is the hybrid cooperative banks, the IHCoop the cooperative leastleast hybrid cooperative banks, IHCoop cooperative banks op the cooperative that have an intermediate degree of hybridization, MHCoop the most hybrid cooperative banks and JSB banks that have an intermediate degree of hybridization, MHCoop the most hybrid

cooperative the joint-stock banks.banks and JSB the joint-stock banks. banks that have an intermediate degree of hybridization, MHCoop the most hybrid

cooperative banks and JSB the joint-stock banks. Results the regression data panelusing regression using the full sample Table Results6.3fora:the data for panel the full sample 27 6.3 a: Table

Table 6.3 a: Results for the data panel regression using the full sample Full sample Loans to assets Loans to assets Loans to assets 2002-11 2002-7 2008 Full sample Loans to41.919 assets Loans to56.075 assets Loans to-16.190 assets C (9.566***) (10.947***) (-0.291) 2002-11 2002-7 2008 11.491 12.854 11.272 41.919 56.075 -16.190 LHCOOP C (4.056***) (1.576) (1.111) (9.566***) (10.947***) (-0.291) 5.445 3.615 6.859 11.491 12.854 11.272 IHCOOP LHCOOP (1.735*) (0.400) (0.601) (4.056***) (1.576) (1.111) 0.733 0.445 -1.086 5.445 3.615 6.859 MHCOOP IHCOOP (0.276) (0.059) (-0.117) (1.735*) (0.400) (0.601) 2.813 -0.703 16.107 0.733 0.445 -1.086 Long-term MHCOOP interest rate (2.693***) (-0.668) (1.270) (0.276) (0.059) (-0.117) -0.112 2.813 -0.703 16.107 2008CRISIS Not included Long-term interest rate (-0.043) (2.693***) (-0.668) (1.270) -0.112 R² 0.044 0.008 0.058 2008CRISIS Not included (-0.043)

Loans to assets 2008-11 Loans to35.219 assets (6.660***) 2008-11 9.211 35.219 (1.974**) (6.660***) 9.011 9.211 (1.752*) (1.974**) 1.753 9.011 (0.388) (1.752*) 4.754 1.753 (3.602***) (0.388) -2.000 4.754 (-0.673) (3.602***) -2.000 0.081 (-0.673)

Loans to assets 2009-11 35.299 Loans to assets (6.757***) 2009-11 8.090 35.299 (1.535) (6.757***) 10.488 8.090 (1.809*) (1.535) 3.036 10.488 (0.585) (1.809*) 4.695 3.036 (3.613***) (0.585) 4.695 Not included (3.613***) 0.100 Not included

Adjusted R² R²

0.035 0.044

-0.003 0.008

-0.008 0.058

0.062 0.081

0.079 0.100

FAdjusted R²

5.476*** 0.035

0.746 -0.003

0.882 -0.008

4.176*** 0.062

4.809*** 0.079

Number of observations F

601 5.476***

360 0.746

62 0.882

241 4.176***

179 4.809***

601

360

62

241

179

Loans to assets 2009-11 Loans to assets 34.949 2009-11 (6.626***) 34.949 8.224 (6.626***) (1.552) 10.686 8.224 (1.552) (1.833*) 10.686 3.215 (1.833*) (0.615) 4.743 3.215 (3.627***) (0.615) 4.743 Not included (3.627***)

Note: the Student T is in brackets.

Number of observations

*** significant at 1% level; ** significant at 5% level; * significant at 10% level. Note: the Student T is in brackets. Note: the Student T is in brackets. *** significant 1% level; ** significant at 5%atlevel; * significant atat10% level. ***atsignificant at 1% level; ** significant 5% level; * significant 10% level.

Table 6.3 b: Results for the data panel regression using the full sample except Dexia

b:the Results the regression data panel using regression using the full sample except Dexia Table 6.3 b: Table Results6.3for datafor panel the full sample except Dexia Full sample except Dexia Full sample except Dexia C

Loans to assets 2002-11 Loans to assets 41.818 2002-11 (9.426***) 41.818 11.440 (9.426***) (4.010***) 5.416 11.440 (4.010***) (1.713*) 5.416 0.698 (1.713*) (0.260) 2.851 0.698 (2.701***) (0.260) -0.273 2.851 (-0.105) (2.701***) -0.273 0.044 (-0.105)

Loans to assets 2002-7 Loans to assets 56.780 2002-7 (10.903***) 56.780 12.699 (10.903***) (1.544) 3.435 12.699 (1.544) (0.377) 3.435 0.272 (0.377) (0.036) -0.831 0.272 (-0.778) (0.036) -0.831 (-0.778)

Loans to assets 2008 Loans to assets -18.215 2008 (-0.324) -18.215 11.396 (-0.324) (1.114) 7.124 11.396 (1.114) (0.619) 7.124 -0.892 (0.619) (-0.096) 16.529 -0.892 (1.289) (-0.096) 16.529 Not included (1.289)

0.009

Not included 0.060

Loans to assets 2008-11 Loans to assets 34.861 2008-11 (6.525***) 34.861 9.343 (6.525***) (1.991**) 9.199 9.343 (1.991**) (1.777*) 9.199 1.922 (1.777*) (0.423) 4.795 1.922 (3.614***) (0.423) -2.013 4.795 (-0.669) (3.614***) -2.013 0.083 (-0.669)

R² Adjusted R²

0.044 0.035

0.009 -0.003

0.060 -0.007

0.083 0.063

0.102 0.081

Adjusted R² F

0.035 5.381***

-0.003 0.759

-0.007 0.894

0.063 4.206***

0.081 4.852***

F Number of observations

5.381*** 591

0.759 354

0.894 61

4.206*** 237

4.852*** 176

61

237

176

C LHCOOP LHCOOP IHCOOP IHCOOP MHCOOP

Long-term MHCOOP interest rate 2008CRISIS Long-term interest rate 2008CRISIS R²

Note: the of Student T is in brackets. Number observations 591 354 *** significant at 1% level; ** significant at 5% level; * significant at 10% level. Note: the Student T is in brackets. *** significant at 1% level; ** significant at 5% level; * significant at 10% level.

Note: the Student T is in brackets.

*** significant at 1% level; ** significant at 5% level; * significant at 10% level.

83 82 JEOD - Vol.3, Issue 1 (2014)

Not included 0.102

28 28

27 27


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

Table 6.3 c: Results forc:the data panel the full sample except Rabobank Table 6.3 Results for theregression data panelusing regression using the full sample except Rabobank Full sample except Rabobank

Loans to assets 2002-11

Loans to assets 2002-7

Loans to assets 2008

Loans to assets 2008-11

Table 6.3 c: Results for the data panel regression using the full sample except Rabobank 41.714 56.133 -14.906 35.237 C (9.412***) (10.793***) (-0.266) (6.611***) Full sample except Loans to assets Loans to assets Loans to assets Loans to assets Rabobank 2002-11 2002-7 2008 2008-11 11.477 12.855 11.309 9.214 LHCOOP (4.024***) (1.563) (1.108) (1.962**) 41.714 56.133 -14.906 35.237 C 3.414 2.095 3.163 6.171 (9.412***) (10.793***) (-0.266) (6.611***) IHCOOP (0.946) (0.201) (0.242) (1.045) 11.477 12.855 11.309 9.214 LHCOOP (4.024***) (1.563) (1.108) (1.962**) 0.739 0.444 -1.097 1.751 MHCOOP (0.275) (0.059) (-0.118) (0.385) 3.414 2.095 3.163 6.171 IHCOOP (0.946) (0.201) (0.242) (1.045) 2.865 -0.716 15.814 4.739 Long-term interest rate (2.712***) (-0.669) (1.239) (3.570***) 0.739 0.444 -1.097 1.751 MHCOOP (0.275) (0.059) (-0.118) (0.385) -1.971 -0.206 Not included 2008CRISIS (-0.654) (-0.079) 2.865 -0.716 15.814 4.739 Long-term interest rate (2.712***) (-0.669) (1.239) (3.570***) R² 0.042 0.008 0.057 0.078 -0.206 -1.971 2008CRISIS Not included (-0.079) (-0.654) Adjusted R² 0.034 -0.003 -0.010 0.058 R² 0.042 0.008 0.057 0.078 F 5.224*** 0.719 0.847 3.922*** Adjusted R² 0.034 -0.003 -0.010 0.058 Number of observations 591 354 61 237 F 5.224*** 0.719 0.847 3.922***

Note: the Student T is in brackets. *** significant at 1% level; ** significant at 5% level; * significant at 10% level. Student T is in brackets.

Loans to assets 2009-11 35.321 (6.711***) Loans to assets 2009-11 8.094 (1.525) 35.321 7.984 (6.711***) (1.199) 8.094

(1.525) 3.035 (0.580) 7.984 (1.199) 4.690 (3.582***) 3.035

(0.580) Not included 4.690 (3.582***) 0.095 Not included 0.074 0.095 4.510*** 0.074 176 4.510***

Note: the observations 591 * significant354 *** significantNumber at 1%oflevel; ** significant at 5% level; at 10% level.

61 237 Table 6.3 d: Results for the data panel regression using the full sample except Dexia and Rabobank

176

Note: the Student T is in brackets. *** significant at 1% level; ** significant at 5% level; * significant at 10% level.

Full sample except

Loans to assets

Loans to assets

Loans to assets

Loans to assets

Loans to assets

41.609 Loans to assets (9.272***) 2002-11 11.427 (3.978***) 41.609

56.851 Loans to assets (10.748***) 2002-7 12.700 (1.531) 56.851

-16.924 Loans to assets (-0.299) 2008 11.432 (1.111) -16.924

34.877 Loans to assets (6.475***) 2008-11 9.347 (1.978**) 34.877

34.970 Loans to assets (6.580***) 2009-11 8.228 (1.542) 34.970

0.704 (3.978***) (0.260) 3.386 2.904 (0.932) (2.719***) 0.704 (0.260) -0.371 (-0.141) 2.904

0.270 (1.531) (0.035) 1.914 -0.848 (0.182) (-0.780) 0.270 (0.035)

-0.848 (-0.780) 0.008

-0.905 (1.111) (-0.096) 3.434 16.235 (0.260) (1.257) -0.905 (-0.096) Not included 16.235 (1.257) 0.059

1.920 (1.978**) (0.420) 6.360 4.789 (1.070) (3.582***) 1.920 (0.420) -1.983 (-0.650) 4.789

-0.003

Not included -0.010

(3.582***) 0.080 -1.983 (-0.650) 0.059

3.214 (1.542) (0.611) 8.182 4.737 (1.222) (3.595***) 3.214 (0.611) Not included 4.737 (3.595***) 0.098

Table d:the Results for theregression data panelusing regression using the full sample except and Rabobank Dexia6.3 and Rabobank 2002-11 2008 2008-11 2009-11 Table 6.3 d: Results for data panel the2002-7 full sample except Dexia and Dexia Rabobank C Full sample except Dexia and Rabobank LHCOOP

C IHCOOP LHCOOP MHCOOP IHCOOP Long-term interest rate MHCOOP 2008CRISIS Long-term interest rate R² 2008CRISIS Adjusted R²

(9.272***) 3.386 (0.932) 11.427

(2.719***) 0.042 -0.371 (-0.141) 0.034

(10.748***) 1.914 (0.182) 12.700

(-0.299) 3.434 (0.260) 11.432

(6.475***) 6.360 (1.070) 9.347

(6.580***) 8.182 (1.222) 8.228

Not included 0.076

0.042

0.008

0.059

0.080

0.098

Adjusted R²

0.034

-0.003

-0.010

0.059

0.076

5.139***

0.734

0.857

3.950***

4.549***

581

348

60

233

29 173

F Number of observations

29

Note: the Student T is in brackets. *** significant at 1% level; ** significant at 5% LEVEL; * significant at 10% level.

Note: the Student T is in brackets. *** significant at 1% level; ** significant at 5% LEVEL; * significant at 10% level. References

Angelini, P., Di Salvo, R. and Ferri, G. (1998). “Availability and Cost of Credit for Small Businesses: Customer Relationships and Credit Cooperatives”, Journal of Banking and Finance, pp. 925-954. http://dx.doi.org/10.1016/S0378-4266(98)00008-9 Akella, S.R. and Greenbaum, S.I. (1988). “Savings and Loans Ownership Structure and Expense Preference”, Journal of Banking and Finance, 12: 419-37. 84 83 http://dx.doi.org/10.1016/0378-4266(88)90007-6 JEOD - Vol.3, Issue 1 (2014)

Allen, F. and Gale, D. (2004). “Competition and Financial Stability”, Journal of Money, Credit and Banking, 36(3): 453-80. http://dx.doi.org/10.1353/mcb.2004.0038


Did the Extent of Hybridization better Enable Cooperative Banking Groups to Face the Financial Crisis? Lemzeri, Y.

References Akella, S.R. and Greenbaum, S.I. (1988). “Savings and Loans Ownership Structure and Expense Preference”, Journal of Banking and Finance, 12: 419-37. http://dx.doi.org/10.1016/0378-4266(88)90007-6 Allen, F. and Gale, D. (2004). “Competition and Financial Stability”, Journal of Money, Credit and Banking, 36(3): 453-80. http://dx.doi.org/10.1353/mcb.2004.0038 Altman, E.I. (1968). “Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy”, Journal of Finance, September, pp. 189-209. Amess, K. (2002). “Financial Institutions, the Theory of the Firm and Organizational Form”, The Service Industries Journal, 22(2): 129-148. http://dx.doi.org/10.1080/714005069 Angelini, P., Di Salvo, R. and Ferri, G. (1998). “Availability and Cost of Credit for Small Businesses: Customer Relationships and Credit Cooperatives”, Journal of Banking and Finance, pp. 925-954. http://dx.doi.org/10.1016/S0378-4266(98)00008-9 Ayadi, R., Llewellyn, D., Schmidt, R., Arbak, E. and De Groen, W.P. (2010). “Investing Diversity in the Banking Sector in Europe”, Centre for European Policy Studies, Brussels. Beltratti, A. and Stultz, R.M. (2011). “Why Did Some Banks Perform Better during the Credit Crisis? A Cross-Country Study of the Impact of Governance and Regulation”, Fisher College of Business, Working Paper No. 2009-03-012. Desrochers, M. and Fischer, K. (2005). “The Powers of Networks: Integration and Financial Cooperative Performance”, Annals of Public and Cooperative Economics, 76(3): 307-54. http:// dx.doi.org/10.1111/j.1370-4788.2005.00281.x Fonteyne, W. (2007). “Cooperative Banks in Europe: Policy Issues”, IMF Working Paper, July. Gurtner, E., Jaeger M. and Ory J.N. (2002). “Le statut de coopérative est-il source d’efficacité dans le secteur bancaire”, Revue d’Economie Financière, 67, Octobre, pp. 133-163. Gurtner, E., Jaeger M. and Ory J.N. (2006). “La banque à forme coopérative peut-elle soutenir durablement la compétition avec la banque SA? ”, Finance Contrôle Stratégie, 9(2) :121-157. Hansmann, H., and Kraakman, R. (2001). “The End of History for Corporate Law”, Harvard Discussion Paper, 280. Hesse, H. and Cihak (2007). “Cooperative Banks and Financial Stability”, IMF Working Paper WP/072. Laeven, L. and Levine, R. (2008). “Bank Governance, Regulation, and Risk Taking”, Journal of Financial Economics, 93(2): 259-275. http://dx.doi.org/10.1016/j.jfineco.2008.09.003 Levin, A. Lin, C.F. and Chu, C.S.J. (2002). “Unit Root Test in Panel Data: Asymptotic and Finite Sample Properties”, Journal of Econometrics, 108: 1-24. http://dx.doi.org/10.1016/S0304-4076(01)00098-7 Ory, J.N., Gurtner, E. and Jaeger, M. (2006b). “The Challenges of the Recent Changes in the French Cooperative Banking Groups”, Revue Internationale de l’Economie Sociale RECMA, October, pp. 43-59. Ory, J.N. and Lemzeri Y. (2012). “Efficiency and Hybridization in Cooperative Banking: the French case”, Annals of Public and Cooperative Economics, 83(2): 215-250. http://dx.doi.org/10.1111/j.14678292.2012.00462.x Ory, J.N, De Serres, A. and Jaeger M. (2012). “Comment résister à l’effet de normalisation: le défi des banques coopératives”, La Revue des Sciences de Gestion, 258 :69-82. Rasmusen, E. (1988). “Stock Banks and Mutual Banks, Journal of Law and Economics, 31 : 395-422”. http://dx.doi.org/10.1086/467162 Soulage, F. (2000). Les Fonds Propres des Coopératives, Analyse et Propositions. Groupe ESFIN-IDES. http://www.esfin-ides.com/esfin-anciensite/pages/publications/rapport030500.pdf 85 84 JEOD - Vol.3, Issue 1 (2014)


Part 2 Cooperative banking: assumptions and evidence


AT T R I B U T I O N 3 . 0

You are free to share and to remix, you must attribute the work

17 2014 June 2014 | Vol.3, Issue11(2014) (2014)86-107 89-109 Publication date: xx | Volume 3, Issue

AUTHOR SIMON CORNテ右 CREM UMR CNRS 6211 and CERMi, Universitテゥ de Rennes 1, France Facultテゥ des Sciences Economiques simon.cornee@univ-rennes1.fr

Soft Information and Default Prediction in Cooperative and Social Banks

ABSTRACT In this paper, to begin with, we define soft information as qualitative, subjective information produced by banks through the establishment of long-term lending relationships. We then highlight the importance of soft information for cooperative and social banks in the screening, pricing and monitoring of their borrowers as a result of their institutional features (governance, values, etc.) and the specificities of their clientele. We finally emphasise the value of qualitative (economic, social and/or environmental) factors stemming from the production of soft information in predicting credit default events.

KEY-WORDS RELATIONSHIP LENDING; SOFT INFORMATION; CREDIT RATING; COOPERATIVE AND SOCIAL BANKING

Acknowledgements The author gratefully thanks Yiorgos Alexopoulos, Silvio Goglio, and Panu Kalmi for their helpful comments and remarks.

JEL Classification: G21, L22, M21, P13 | DOI: http://dx.doi.org/10.5947/jeod.2014.005

89 86 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

1. Introduction In the aftermath of the financial crisis that broke out in 2007, The Economist (02/2010, the 11th) wrote: “these eggheads [quantitative analysts] are now in the dock, along with their probabilistic models. In America a congressional panel is investigating the models’ role in the crash. Wired, a publication that can hardly be accused of technophobia, has described default-probability models as ‘the formula that killed Wall Street’.” From an academic stance, the exclusive use of quantitative information in assessing borrower creditworthiness is also deemed to be one of the causes of the financial crisis (Diamond and Rajan, 2009)1. By using data on securitised subprime mortgages issued in the period 1997-2006, Rajan et al. (2008, 2010) highlight the increasing prevalence of hard information in setting interest rates. The authors also demonstrate that statistical default models solely equipped with this source of information do not yield satisfactory default predictions for borrowers for whom expert, subjective judgement using soft information is more valuable. This gradual retreat from the reliance on qualitative information results from the remarkable transformations undergone by the banking industry in the last few decades. Front offices (i.e. loan officers in local branches) have reduced in size and have been subject to intensified staff turnover, to the advantage of the back office and headquarters. Banks have, it would seem, exploited the opportunities offered by financial liberalisation, regulation and ICT progress to become large, consolidated financial institutions. As a consequence, old-fashioned, interpersonal lending relationships have been progressively replaced by more standardised and impersonal rapports (Ferri, 2010). Of course, not all financial institutions have followed this trend at the same pace. This is especially the case for cooperative banks, and even more so, that of a specific fraction of them - social banks - whose distinctive feature lies in their emphasis on sustainability (i.e. social and/or environmental) goals2. Because of their specific governance, regulation, values, and missions, cooperative banks continue to rely on soft information in order to assess their borrowers. Soft information summarises subjective, unquantifiable facts about a firm and its owner, which are idiosyncratic to individual long-term credit relationships. In this paper, we contend that these lending practices previously viewed as archaic should now be better considered by the sector and prudential regulation authorities, given the fallacy of quantitative models. Our objective consists of demonstrating the critical importance of soft knowledge when it comes to obtaining accurate default predictions for borrowers subject to credit rationing. This is typically the case for firms who are by nature informationally opaque on account of their size (e.g. SMEs) and/or their specific missions and values (e.g. social enterprises, cooperatives). Our main contribution lies in the aggregation of synthesised sets of banking literature, which are closely related and complementary, but seldom assembled. In particular, we make the connection between how soft information is produced and the way it could be used in credit rating models by cooperative banks. In doing so, this paper also contributes to the topical debate on the design of credit rating systems tailored to suit the peculiarities of the relational approach to banking. The rest of the paper is structured as follows. In Section 2, we deal with conceptual preliminaries involving the definition of soft information and the consequential effects of its properties on internal contracting issues. In Section 3, we document that credit cooperatives have behaved from scratch as veritable information machines, so as to be able to finance informationnally opaque borrowers. Notwithstanding the substantial changes they have experienced, cooperatives are still heavily reliant on the production of soft information 1

This analysis should of course be put into a broader context whereby financial intermediaries were not incited to conduct a thorough screening of credit applicants, since the loans originated by banks were then securitised and sold to financial markets.

2

In the rest of the article, we use interchangeably the terms “social” and “sustainable/sustainability”. 90 87 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

in the lending operations. We also argue that social banks follow a similar pattern, to an even greater extent, especially because they claim more asserted ethical values. In Section 4, we highlight the added merit of incorporating qualitative factors derived from soft information into credit default prediction models. We also show that these qualitative factors should not necessarily be economically-oriented but can also take the form of sustainability criteria. Section 5 concludes.

2. The problem of knowledge production in banks In this section, we make a general distinction between the two categories of knowledge produced by financial intermediaries, i.e. hard and soft information (2.1.). We then elucidate the internal contracting and organisational issues inherent to the use of soft information (2.2.). 2.1. Hard and soft information From a theoretical stance, banks’ raison d’être comes from their special capacity to overcome informational asymmetries in credit markets (Stiglitz and Weiss, 1981). The production of non-public knowledge on firms confers to banks an informational advantage over other financial operators (Leland and Pyle, 1977; Diamond, 1984; Bhattacharya and Thakor, 1993). This informational superiority enables banks to provide external debt funding to informationnally indefinite segments of the financial market, such as small- and medium-sized businesses (SMEs). Knowledge produced by banks stems from two sources: hard and soft information. Hard information is explicit knowledge reported through formal instruments such as audited financial statements, history of repayments, checking accounts, and other financial usage (Petersen, 2004). The collection of these quantitative data abides by standardised and third-party verifiable procedures. This implies that the content of information and its interpretation are not contingent on the agents in charge of its collection. What may then differ across banks is their storage capacity as well as their in-house computing and coding technologies, employed to synthesise the gathered information into decision-support indicators such as credit scores. In contrast, soft information refers to implicit (or idiosyncratic) knowledge that takes the form of unpublished, informal aspects of the firm’s management quality, inside conflicts, strategy, competencies, critical suppliers, underlying motives or customer dependencies, etc. (Uzzi and Lancaster, 2003). These qualitative inputs are especially valuable for banks, since their uniqueness enables them to tailor their credit conditions to the confidential specificities of their borrowers. This customisation provides banks with a competitive advantage, because it is difficult to be imitated by other market players. Whether a bank relies on soft or hard information depends on the extent to which its commercial transactions with borrowers are embedded in social ties. Social relationships can be represented as a continuum with arm’s length at one end, and embeddedness at the other (Uzzi, 1999; Uzzi and Gillespie, 1999; Uzzi and Lancaster, 2003). Arm’s length ties are characterised by meagre, sporadic, cool and impersonal transactions without any prolonged human and social contact between parties (Uzzi, 1999). Arm’s length ties typically refer to transaction technologies in which a bank exploits all hard, financial information available on credit applicants at the time of loan origination. The most frequent transactional techniques are financial statement lending, asset-based lending, and credit-scoring lending (Berger and Udell, 2002). Among these techniques, Small Business Credit Scoring (SBCS) is the most important and fastest-growing one. SBCS is typically used to evaluate under EUR 250,000 business loans by combining financial data and consumer data about 91 88 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

the owner (Mester, 1997; Akhavein et al., 2005; Berger and Frame, 2007). In sharp contrast to one-shot, transactional interactions, relationship lending enables a bank to accumulate qualitative information over time through repeated contacts with the entrepreneur, the firm’s stakeholders, and the local community. Moreover, we consider the quality of soft knowledge as varying in the degree to which lending relationships are imbued by nonmarket attachments between parties (Uzzi and Gillespie, 1999). A trustful and cooperative basis is essential for an easy flow of qualitative facts and informalities between lenders and borrowers (Uzzi and Lancester, 2003). Therefore, such soft knowledge transfer cannot occur properly if pure market mechanisms and narrow self-gain behaviours, as described by Jensen and Meckling (1995), are alone at work. Besides, knowledge transfer across firm boundaries, i.e. a borrowing firm and its bank, is strongly affected by the bank’s internal structure, which is generally shaped by environmental factors such as market structures, technological innovations, business conditions, as well as legal and regulatory aspects (Berger and Udell, 2002; Degryse and Ongena, 2008). As we shall see below, that for any bank regardless of its status (i.e. cooperative or not) and its social mission, there exists a strong correlation between a propensity to favour one lending technology over the other and its organisational architecture. 2.2. Soft information, internal contracting issues, and organisational architecture One property that largely differentiates hard from soft knowledge is transferability (Grant, 1996). This means that quantitative information is communicable with ease and at a low cost, whereas the transfer of qualitative information among agents proves to be costly and uncertain (Jensen and Mecking, 1995). To exemplify the concept of transferability in the context of banking, imagine two scenarios. In the first one, a loan officer puts into the bank’s information system hard facts extracted from audited financial statements. This information will be explicitly communicated across space, i.e., across hierarchical layers and physical distances. It will also flow easily across time, since its storage is facilitated by its quantitative nature. In the second situation, a loan officer who has spent a lot of time with a small-business owner may come to strongly believe that the latter is honest and hardworking - in other words the typical candidate for an unsecured “character loan”. Since the loan officer’s belief - which is a typical piece of intrinsic information - is not verifiable by anyone but themselves, it cannot be unambiguously transferred across space and across time. In this regard, Liberti and Mian (2009) report that a greater number of hierarchical layers or larger geographical distances between the agent who collects the information and the loan approving officer, leads to more reliance on objective, hard information at the expense of subjective, soft information in making credit-granting decisions. Thus, soft knowledge loses its informative power when it is transferred among agents. Another aspect associated with transferability is the recipient’s ability to analyse and compute the transmitted knowledge (Grant, 1996). With regards to the information format, hard information is presented in numeric format so as to facilitate its aggregation and comparison. In contrast, soft information, which comes via text format as well as by intuitions and judgements of loan officers, lacks objective benchmarks to render it comparable. Typically, the interpretation of honesty may differ from one loan officer to another. Moreover, qualitative information, even though hardened through a scale and an index, remains loan officer-dependent and idiosyncratic to one peculiar lending relationship (Petersen, 2004). From a broader perspective, the quality of soft knowledge, i.e. the accuracy of its interpretation, depends on the extent to which bank employees are embedded in local communities and whether they have established pre-existing, affinitive bonds with their customers. The sharing of similar cultural values and ethical traits homogenises – and stabilises – socio-demographic, behavioural and personal characteristics amongst local community 92 89 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

members (Katerinakis, 2012). The regularity of social networks and relations generates substantial economic outcomes, since it enables trading parties to infer markedly accurate predictions on the other’s future behaviour. From the bank’s viewpoint, producing accurate forecasts on its customers’ propensity to meet their commitments proves extremely valuable, especially for implicit, trust-based arrangements (i.e. not enforceable by third-parties) which rely on the prior collection of soft information. The difficulty in transferring and aggregating soft information entails inevitable internal contracting issues within the banking institution, particularly between the management and lending staff (Jensen and Meckling, 1995)3. When the credit eligibility process relies on intense lending relationships, loan officers – whose role is thus pivotal – typically benefit from more delegated authority. Entrusting lending staff with more discretion in credit-granting operations may potentially have undesirable effects in credit allocation. For example, overlending may arise from “social attachment” between loan officers and borrowers (Uzzi and Gillespie, 1999) or from loan officers’ willingness to manage larger budgets (Ozbas, 2005). In contrast, certain borrowing fractions may be rationed because loan officers exhibit preferences or/and stereotypes, which are incongruent with their organisations’ mission statement (Agier and Szafarz, 2013). This directly echoes the “group thinking” issue raised by Alexopoulos et al. (2013). The authors suggest that group thinking occurs when, in a context of deficient organisational control, a category of stakeholders (e.g. loan officers) rationalise all of their decisions autonomously from the organisation’s missions and objectives. This situation potentially brings about agency problems because the managerial discretion granted to the “thinking group” members may be detrimental to the organisation, since they may seek to pursue their own interests - rather than adopting creative and innovative practices beneficial to the organisation. For all these reasons, banks spend resources monitoring loan officers and the performance of their individual portfolios, when more authority is bestowed on field personnel (Udell, 1989). The magnitude of contracting problems is also likely to increase with the complexity and the size of the banking institution. In a theoretical model, Stein (2002) predicts that qualitative information production is only efficient in small or decentralised banks, which are characterised by few managerial layers. Loan officers will be more predisposed to collecting high-quality soft information on the condition that they have a sufficient authority on the bank’s fund allocation. Typically, contracting problems can be resolved if the loan officer is also the manager of the bank. By contrast, large, centralised banks with multiple layers of management and more hierarchical decision-making processes in which soft facts on borrowers are amassed and compiled lead to organisational diseconomies – low transferability of qualitative information resulting in prohibitive internal contracting costs. Large, hierarchical banks are therefore more inclined to adopt transaction-oriented technologies based on explicit and transmittable knowledge. Empirical evidence supports this theoretical view. For example, Cole et al. (2004) show that large banks, with over USD 1 billion in assets and numerous hierarchical layers, exhibit a higher propensity to base loan approvals on standard criteria obtained from financial statement, while small banks rely, to a greater extent, on information about borrowers’ character. In the same vein, Berger et al. (2005) find that large banks lend at a greater distance and interact more impersonally with their borrowers. In the past three decades, the banking industry has been characterised by a huge concentration, spurred by changes in the legal regulatory framework and by the advent of technological innovations. Larger bank size, greater distances with borrowers, as well as improved computing performance capacities are all factors explaining the adoption of transaction lending models, to the detriment of technologies based on soft information (Petersen and Rajan, 2002; Brevoort and Hannan, 2004). 3

Other contracting issues may arise, for instance between bank managers, on the one hand, and stockholders (or members), on the other (Berger and Udell, 2002; Alexopoulos et al., 2013). 93 90 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

In spite of this dramatic shift, recent evidence indicates that discretion in the lending process remains widespread and economically significant, especially in small banks (Cerqueiro et al., 2011; Puri et al., 2011). Moreover, loan approval decisions, based on loan officers’ discretionary judgment, do not necessarily lead to excess risk (Puri et al., 2011). Similarly, Gropp et al. (2012) document that loan officers may even use soft information too cautiously in their loan approval decisions. Interestingly, these findings would suggest that internal contracting problems may be less important than what conventional agency theory indicates. In spite of their discretionary latitude, loan officers do not seem to misuse soft information to adopt opportunistic strategies which would be detrimental to the bank.

3. The importance of soft information production in cooperative and social banks In this section, we explain why cooperative lending practices are heavily reliant on soft information produced via the establishment of long-term credit relationships (3.1. and 3.2.). We also show that the production of qualitative information is even more critical to social banks (3.3.). 3.1. Credit cooperatives as information machines In light of the advancements made in the previous section, the practices of the first credit cooperatives introduced in rural Germany in the nineteenth century appear surprisingly novel4. These financial institutions operated as veritable “information machines” given their great ability to both produce soft information on borrowers and exploit it efficiently. Guinnane (2001) captures the essential features of their successful functioning. In particular, the author provides compelling evidence that cooperatives were intentionally confined to limited geographical areas. By focusing on local, stable communities, they managed to access first-rate soft information concerning potential borrowers. This specific information was peer-produced within the community by congruent cooperative members who had an in-depth knowledge of each other’s habits, characters and competencies. Importantly, credit cooperatives were also able to exploit this high-quality soft knowledge produced by their members thanks to their decentralised organisational scheme. As documented by Guinnane (1997), these credit providers were not, legally speaking, cooperative branches but rather individual grass-root cooperatives with full autonomy in decision-making. Nonetheless, most of them took part in the formation of regional centrals and auditing associations. The former ensured liquidity on the condition that cooperatives were audited by the latter. Audits were not conducted in a pure top-down fashion but were rather a mix of formal control and customised counsel. Consequently, lending operations were carried out at the grass-root level. Given their (very) small size, cooperatives exhibited simple organisational structures with few hierarchical layers, thereby avoiding most internal contracting problems associated with the low transferability of soft knowledge. This informational superiority explains – at least partially5 – why these cooperatives were successful in screening and monitoring borrowers spurned by commercial banks, as well 4

We describe Raiffeisen’ s model in rather functionalist terms. However, we fully agree with Goglio and Leonardi (2012) and Kalmi (2012) on the fact that these credit cooperatives were primarily animated by non-instrumental goals and distinct ideological attitudes towards finance – compared to commercial banks.

5

While we focus on the sole production of information, other factors are evoked in the literature, such as the ability to enforce loan agreements due to proximity among cooperative members (Banerjee et al., 1994; Guinnane, 2001). For instance, deviant borrowers may suffer from social sanctions. 94 91 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

as in tailoring loans according to the specific needs of borrowers. The archetypical model of credit cooperatives, briefly outlined above, has undergone dramatic changes over the last century to form large banking groups (Ayadi et al., 2010). This integration process, fostered by market liberalisation and industry concentration, has resulted in central units taking more power, to the detriment of local, regional cooperatives. This phenomenon of centralisation in the decisional pyramid has de facto reduced the democratic power held by grass-root cooperative members (Di Salvo, 2002). As highlighted by this author, prudential authorities have favoured this institutional evolution by designating centralised bodies as their single interlocutor for regulatory aspects6. This critical evolution in governance structure is, among other things, a clarifying factor in the hybridisation process, observable in a number of cooperatives. Evidently, exceptions to this process would deserve to be mentioned. For instance, Katerinakis (2012), in his analysis of major Greek credit cooperatives, considers the latter as “polycentric self-organisations”. Hybridisation has led to various cooperatives abandoning the explicit goal of facilitating access to credit to the non-bankable, by evolving over time into full-service universal banks or by entering into activities that were not inherently entrenched, such as corporate and investment banking (Ayadi et al., 2010; Ory and Lemzeri, 2012). Nonetheless, the vast majority of cooperative banks remain firmly anchored to their business basics since they have preserved a unique ability to serve small borrowing firms characterised by informational opacity, such as SMEs. Retail activities remain their cash cow by providing cooperatives with a stable source of revenue7. In the context of France, Gianfaldoni and Richez-Battesti (2005) show that cooperative banks are by far the main external funders of SMEs and households. Labye et al. (2002) and Ayadi et al. (2010) generalise these findings to Europe by showing that cooperative banks remain major retail banking players in most EU countries. Cooperative banks still rely on soft, idiosyncratic information to evaluate their borrowers, but their relationship lending technologies have evolved over time. In the nineteenth century, cooperatives had recourse to peer monitoring to overcome asymmetric information problems such as adverse selection (Stiglitz, 1990; Banerjee et al., 1994). Members of rural, local communities were typically able to self-select themselves to form credit cooperatives, since they benefited from an intimate profile of each other via dayto-day interactions. This self-selection mechanism is no longer efficient when the cooperative grows and its membership is dispersed throughout urban and/or larger rural areas8. Cooperative banks have consequently switched to a different pattern in which they establish bilateral long-term credit relationships with borrowers. Interestingly, the reliance on lending relationships is still a distinctive feature of cooperative banking (Ferri, 2010).

6

This entails “bottom-up” constraints for local cooperatives (e.g. reporting to central units) as well as “top-down” constraints (e.g. risk models imposed by the regulator via central units).

7

Traditional retail activities operated by local banks act as a buffer, which in bad times dampens losses from market operations and corporate investment activities undertaken by central bodies (Ory and Lemzeri, 2012).

8

On a related note, the importance of peer-monitoring in explaining the current success of microfinance seems exaggerated. Other factors are at work (Morduch, 1999). For instance, dynamic incentives induced by long-term credit relationships may be a more powerful devise than peer-pressure mechanisms in enforcing microfinance contracts (Cornée and Masclet, 2013). 95 92 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

3.2. The reasons for relationship lending in cooperative banks Here, we explain why cooperative banks still rely on relationship lending while in the last decades the banking industry as a whole has massively adopted transactional technologies. Conversely, we do not intend to demonstrate that relationship lending is the preserve of cooperative banks9. In many instances, commercial financial institutions that follow the same organisational pattern (i.e. a decentralised decisionmaking) may exhibit quite a similar behaviour in their lending practices (e.g. De Young et al., 2004; Scott, 2004). This is especially the case for stakeholder-oriented banks, such as community banks and savings banks. As argued above, retail banking, and particularly SME financing, remains by far the business goodwill of cooperative banks. The latter still hold a noticeable expertise in serving informationally opaque borrowing fractions. This expertise stems mainly from the intensive use of qualitative information derived from the establishment of long-term credit relationships. Engaging in repeated interactions entails beneficial effects for both parties, allowing them to reach pareto-superior outcomes compared to what they could get in spot-market transactions. Repeated lender-borrower interactions enhance cooperative behaviour and mutual trust, which in turn facilitate the flow of soft information between parties (Uzzi and Lancaster, 2003). A sort of implicit arrangement is formed: borrowers are ready to deliver more private information on their situation to lenders provided that they believe they are guaranteed credit availability in the future. Lenders make the promise of ensuring further credit availability to borrowers because they have a good impression of the latter, thanks to the high-quality soft knowledge they have gathered (Sharpe, 1990; Rivaud-Danset, 1996). There exists a plethora of empirical work highlighting the benefits associated with relationship lending in terms of credit availability. Methodologically speaking, the strength of lending relationships is generally measured via three indicators: i) its duration, defined as the number of years the bank has offered loans, deposits or other financial services to the firm; ii) its scope, proxied by the quantity of financial services contracted by the firm; iii) its exclusivity, in terms of the bank being the firm’s sole debt provider. Regardless of the indicator, empirical evidence reports quite univocally that for the borrowing firm, stronger lending relationships result in a lower probability of being rationed (e.g. Angelini et al., 1998; Cole, 1998; Elsas and Krahnen, 1998; Uzzi, 1999; Machauer and Weber, 1998; Lehmann and Neuberger, 2001; Cole et al., 2004), or a lesser recourse to more expensive sources of corporate financing such as trade credit (e.g. Petersen and Rajan, 1994; Harhoff and Korting, 1998). This evidence shows that banks amplify credit availability directly through facilitated debt provision, but they can also do it indirectly by reducing collateral requirements. The repeated interactions in long-term credit relationships allow lenders to condition their credit terms on the past repayment behaviour of borrowers. By repeatedly repaying their debt rather than defaulting, borrowers can positively alter lenders’ beliefs about their creditworthiness (Fehr and Zehnder, 2006; Brown et al., 2009; Cornée et al., 2012). By strengthening their creditworthiness over time, borrowers build a reputation with lenders. Reputation can act as an intangible asset which may be pledged by small entrepreneurs to compensate their lack of real assets; thereby circumventing the problems associated with collateral requirements. Empirical evidence unambiguously supports this argumentation by documenting that stronger relationships are associated with reduced collateral requirements (e.g. Berger and Udell, 1995; Harhoff and Körting, 1998; Degryse and Van Cayseele, 2000; Machauer and Weber, 1998). 9

This justifies why the studies reviewed in this section are not specifically conducted on cooperative banks. There exists very little research comparing cooperative banks with their commercial counterparts, the studies being generally carried out on the banking industry as a whole. In addition, the dummies controlling for the bank status are unfortunately constructed in such a way that we cannot properly isolate the “cooperative bank” effect on credit conditions. 96 93 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

The information produced by repeated borrower-lender interactions may be regarded as hard, since it directly feeds the customer’s credit history. We cannot exclude that it is partially the case, if we consider information-sharing schemes between lenders (e.g. credit bureaus). Nonetheless, a closer look at the studies reveals that collateral requirements depend positively on the temporal length of the lending relationship, highlighting the fact that immaterial reputation assets are built over time. Furthermore, the less exclusive the relationship (i.e. the greater the number of banks the borrower trades with), the higher the collateral requirements. This indicates that borrowers’ creditworthiness assets are specific to each credit relationship and represent essentially soft, idiosyncratic knowledge amassed over time by the bank on its debtors (Guille, 1994). So far, we have insisted on the “win-win” consequences of relationship lending, however this technology may also bring about undesirable side effects. The principal effect is an informational hold-up for borrowers. By trading with only one bank, a borrower may, over the course of the relationship, become captive to their incumbent bank since they are unable to signal their quality to other banks - their debt provision relying mainly on soft, non-transmittable knowledge (Sharpe, 1990)10. Lenders may in turn take advantage of their bargaining power and extract rents from lending relationships by raising their interest rates (Boot, 2000; Von Thadden, 1995). Empirical evidence regarding this hold-up hypothesis is contradictory. For instance, D’Auria et al. (1999), Degryse and van Cayseele (2000), and Degryse and Ongena (2005) report that longer relationships results in a higher interest rate charged to borrowers. Berger and Udell (1995), Uzzi (1999), Bodenhorn (2003), as well as Berger et al. (2007) reach an opposite conclusion11. Moreover, many studies (e.g. Elsas and Krahnen, 1998; Machauer and Weber, 1998; Lehmann and Neuberger, 2001; Canovas and Solvano 2006) show no significant effect. Finally, the findings from Angelini et al. (1998) are of great interest with regards to this topic. The authors report a hold-up effect for borrowers trading with commercial banks. However, they do not observe such an effect for borrowers that are members of a credit cooperative. Altogether, these studies reveal that cooperative banks, like their commercial counterparts, guarantee credit availability and are less stringent for collateral requirements when long-term credit relationships are established. However, evidence (though limited) shows that cooperatives appear to be particular with regards to sharing the rent generated by credit relationships. Specifically, they do not charge higher interest rates even though their borrowers may be in a situation of informational capture. Another reason explaining cooperative banks’ reliance on relationship lending lies in the specificities of their decision-making processes. Cooperative banks still exhibit more decentralised and complex organisational architecture than their commercial counterparts, in spite of the dramatic consolidation and integration they have undergone (Ory and Lemzeri, 2012). Abdesselam et al. (2002) carry out a comparative analysis of the role played by loan officers in both cooperative and mainstream major French banks. They conclude that cooperatives grant their loan officers more discretion in the lending process. Loan approval decisions are more likely to be based on proximity, mutual trust and borrowers’ credit reputation. The temporal length of the lending relationship is also a key factor taken into consideration by loan officers. Furthermore, the management of credit lines, which crystallises the quality of the interactions, may also be more flexible.

10

To mitigate this “hold-up” problem, SMEs increasingly favour multiple sources of capital. By trading with several banks and by pitting one against the other, borrowers can reduce the cost of their debt. Nonetheless, multi-banking strategies affect the cost of capital in various ways. As explained in the rest of this section, the quality of debt supply (i.e. capital availability and reduced collateral requirement) deteriorates along with the number of financiers, thereby incurring indirect financing costs that will eventually be embedded in the interest rate.

11

On a related note, Ferri and Messori (2000) show that stronger relationships offer greater protection against the interest rate cycle. 97 94 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

In contrast, loan officers in commercial banks have less managerial latitude and adopt a more technical approach to lending and credit-granting decisions, which depend on the firm’s financial statements and on credit analysts’ sector expertise. Moreover, credit lines granted to borrowers are proportionally less important in terms of “volume”, and more stringently managed. Finally, cooperative banks rely more so on relationship lending, as a result of their ownership structure. The way equity is held and traded is still a major distinctive feature between cooperatives, which are not listed in stock markets, and commercial banks, which are often listed corporations (Hesse and Čihák, 2007; Ory et al., 2006). Subsequently, commercial banks are considerably more exposed to the pressure and “discipline” exerted by stock markets. The pursuit of shareholder value maximisation may also impact the lending strategy of commercial banks. Typically, they are incited to invest in assets that are currently profitable to meet the short-term profitability constraints dictated by shareholders12. In contrast, cooperative banks are notably less subject to market pressure and their goal is not profit maximisation – at least in the short run – but rather their members’ consumer surplus, through the provision of financial services at moderate prices to retail clients (Hesse and Čihák, 2007). Even though cooperatives do not face short-term market constraints, they need to attain a certain level of profitability in the medium or long term to be competitive in the industry. On the whole, cooperatives’ profitability constraints are not dictated by the short-term pressure exerted by stock markets but rather by the medium- or long-term product market (Gianfaldoni and Richez-Battesti, 2005). The medium- or long-term profitability constraint is more in line with the temporality of relationship lending. Financing young or de novo small borrowing firms is seldom profitable instantaneously. When a bank finances small businesses via relationship lending technology, it makes a substantial informational investment in the first interactions, which is then amortised over several periods, i.e. through repeated interactions with the businesses. In other words, lenders subsidise borrowers in early periods and get reimbursed for this subsidy thereafter (Petersen and Rajan, 1995). Such a long-term strategy may be feasible for cooperative banks, but rather incompatible with their commercial counterparts, which may be subject to short-term shareholder interests maximisation13. 3.3. Soft information in social banking In the previous point, we restricted our analysis to the case of cooperative banks relying massively on soft information to manage lending operations with informationally opaque SME borrowers. More generally, the collection of soft information proves to be essential when a bank explicitly claims that it pays attention to the non-economic (i.e. social, environmental, ethical) consequences of its financial activity. To exemplify this view, let us examine the case of social banks, which are burgeoning financial intermediaries falling within the broad scope of cooperative banking (Kalmi, 2012; Artis and Cornée, 2013)14. Social banks aim

12

Orléan (1999) provides an in-depth theoretical analysis explaining how the microstructural features of financial markets lead to short-term profit maximisation.

13

Similarly, a fiercely competitive credit market may be destructive to the formation of mutually beneficial relationships between lenders and borrowers. Lenders are reluctant to assist informationnally-opaque borrowers and consequently accept lower returns upfront if they fear that the future benefits associated with this early assistance may be reaped by competitors (Petersen and Rajan, 1995).

14

Admittedly, not all social banks operate under the legal status of “cooperative” even though it is the case for most major social banks. When social banks have a standard private-owned equity structure they are not listed, and most importantly, their shareholders’ power and profit are restricted through ad hoc mechanisms to favour the involvement of stakeholders (San-Jose et al., 2011). For instance, the number of voting shares at Alternative Bank Schweiz (Switzerland) and Triodos Bank (The Netherlands and Belgium) are limited (GABV, 2012). 98 95 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

to encourage a community of values by matching the two sides of financial intermediation: socially-minded investors (i.e. members/shareholders and savers) and socially-responsible borrowers (Cornée and Szafarz, 2013). Thus, social banks are special in the midst of cooperative banks, since their foundational principles go beyond the rigorous application of conventional cooperative principles, as they explicitly prioritise social aims over financial ones (Becchetti et al., 2011; San-Jose et al., 2011; Weber and Remer, 2011)15. In concrete terms, these foundational principles accommodate the selection of credit applicants that relies on both a financial basis and sustainability (i.e. social and/or environmental) criteria. In practice, social banks may be viewed as specialised financial intermediaries providing external debt funding to borrowers who meet this double-bottom line, i.e. organisations pertaining to the social economy sector, such as cooperatives, not-for-profit organisations, and community projects supported by the civil society (Borzaga and Defourny, 2001)16. To uphold this commitment to their funders, social banks assess the social dimension of projects in addition to conducting conventional economic analysis17. Evaluating the social aspects of an investment project involves compiling a hard, explicit knowledge component, including all the public information available on the project (e.g. activity sector), and a soft, idiosyncratic knowledge component. The latter cannot reasonably be built up by following a strict, automatic procedure, given the specificity of each project. Rather, it involves judgments on intangible characteristics such as borrowers’ social preferences, the ethicality and environmental-friendliness of the business activity, and responsibility towards stakeholders, etc. Interestingly, social screening is also subject to informational asymmetries, which can be attenuated over the course of the lending relationship. This may explain why social ratings of start-ups, which are particularly plagued by informational asymmetries, tend to be systematically lower than those of existing firms (Cornée and Szafarz, 2013). The gathering of soft knowledge on sustainability aspects has advantages other than the mere measurement of borrowers’ non-financial performance for accountability purposes. As we examine in Section 4, these sustainability criteria are likely to improve the forecast quality of borrowers’ credit default. At this stage, we explain why this is the case. Assessing the creditworthiness of social enterprises through the sole lens of economic viability may appear too restrictive. Since the ultimate goal of these organisations is the satisfaction of a social or a community need - and not profit maximisation per se - , the quality of ties between these organisations and their stakeholders (members, users, local community, public funders, etc.) is critical to the fulfilment of their mission. In other words, this “relational capital”, definable as an intangible, idiosyncratic asset composed of social relations and values embedded in the local community and reproduced through the interactions taking place within it, is vital to the success and survival of social enterprises in their environment (Gagliardi, 2009). By screening credit applicants through the collection of qualitative information on both financial and sustainability aspects on each of their borrowers, social banks manage to appraise quite accurately their relation capital and can in turn satisfactorily predict credit defaults of social enterprises (Artis and Cornée, 2013). Conversely, this may explain why mainstream

15

There is actually no clear-cut division between “social cooperative banks” and “non-social cooperative banks”. For instance, some small- or medium-sized credit cooperatives (e.g. the BBCs in Northern Italy) that do not explicitly self-identify as social banks could be considered as such.

16

For the sake of simplicity, we henceforth use the catchword of “social enterprise” to designate all of these organisations.

17

Social bank loan officers who are often at the crux of this double-screening process dedicate a substantial fraction of their workload to amassing qualitative facts on the social dimension of projects – compared to their counterparts from mainstream cooperative banks. In Cornée and Szafarz (2013), we approximate that loan officers dedicate about one third of their workload to social screening. Let us remark however that social assessment is not systematically conducted by loan officers. For example, in Banca Etica (Italy), social auditing is carried out by so-called “social auditors or experts”. 99 96 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

cooperative banks that do not engage in non-financial screening are sometimes at odds with financing the social economy sector (Gagliardi, 2009).

4. Soft information and credit rating In this section, we first depict the conventional view of credit rating (4.1.). We then highlight the value of qualitative economically-oriented factors, representative of the soft information produced by relational banks, in predicting credit default events (4.2.). Finally, we discuss whether and by what means sustainability criteria improve forecast quality of default models (4.3.). 4.1. The conventional view of credit rating Credit risk measurement has dramatically thrived over the past two decades, especially since the announcement of the Basel Accord on credit risk capital adequacy. Banks have modified existing internal credit risk systems or developed new systems to calculate the probability of default (PD) and, possibly, loss-given-default (LGD) on their credit assets (Altman, 2002). In this context, credit rating has become a widespread practice not only in capital markets, but also within banking institutions. Credit rating is used not only internally for screening borrowers, pricing loans, and managing credit risk thereafter (e.g. for loan-loss provisioning), but also externally for calibrating regulatory capital requirements. To correctly fulfil these functions, credit ratings should be a good predictor of default, as noted by Krahnen and Weber (2001) in their normative set of “generally accepted rating principles”. The construction of nearly all credit rating models follows the same pattern. A set of quantitative accounting ratios are combined and weighted to predict bankruptcy. These accounting ratios aim to capture the main aspects of a firm’s financial performance: capital structure (e.g. equity-to-assets ratio), profitability (e.g. ROA), and liquidity (e.g. current ratio). Market-based information (e.g. Tobin’s Q) has been used insofar as the original models have been primarily applied to large corporate firms. Since the seminal works by Beaver (1966) and Altman (1968), a substantial body of work has investigated whether and to what extent a combination of the aforementioned variables can predict corporate default events. Predictive models have generally been based on linear discriminant analysis, on logit and on probit regression analysis, or, more recently, on neural networks (Altman and Saunders, 1997). Throughout this extensive research, most academics and many sophisticated practitioners have systemically moved toward the elimination of nonquantitative factors in credit rating modelling. The survey carried out by Altman and Saunders (1997) on the developments of credit risk measurement since the 80’s is quite representative of the lack of interest for qualitative factors. The authors argue that the expert’s (i.e. the banker’s) subjective assessment, or the “4 Cs”18 methodology, tends to be outperformed by univariate accounting-based credit scoring systems. The authors (p. 1722) then conclude that “perhaps, not surprisingly, financial institutions themselves have increasingly moved away from subjective/expert systems over the past 20 years towards systems that are more objectively based”. This view is now widely accepted by the banking sector, and bankers’ expertise and subjectivity are disregarded by the Basel Committee on Banking Supervision (henceforth BCBS, 2000b, 18

The so-called four Cs encompass borrower character (reputation), capital (leverage), capacity (volatility of earnings) and collateral. This screening methodology confers a critical role to bankers’ expertise and subjective judgement. 100 97 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

p. 108)19. A good illustration of this is provided in BCBS (2006), which suggests that a regulator may rely on a third party to assess the probability of default on bank loans (Rajan et al., 2008, 2010). This means essentially that a bank is not expected to produce qualitative information because a remote third-party (e.g. rating agency) will only use public, explicit knowledge when it rates the bank’s borrowers (Diamond and Rajan, 2009). Thus, banks are in some way incited to not collect and use soft information in their lending activities. The BCBS favours quantitative default models. In this regard, the BCBS (2000b, p. 110) states that “all in all, multivariate accounting-based credit-scoring models have been shown to perform quite well”. At the same time, the BCBS points out shortcomings in these models that lack theoretical foundations and pivot on personal judgement when it comes to selecting the accounting data to be included in the statistical analysis. Moreover, accounting data are by nature backward-looking since they account for what happened in the past. Thus, they do not capture future firm prospects. To take into account a forwardlooking perspective, the BCBS suggests that financial institutions rely not on judgemental, subjective appraisals, but rather on market-price-based measures (e.g. the KMV model), or equivalents for non-listed companies (e.g. the Private Firm Model). Prices are supposed to give a correct estimate of a firm’s situation since they embody the synthesised views and forecasts of many investors who are constantly performing analysis on companies. Furthermore, the BCBS implicitly recognises the superiority of market-price-based measures over accounting-data-based measures, when it states that “they pick up more subtle and fast-moving changes in borrower conditions, such as those that are reflected in capital market data and values. In addition, accountingdata-based measures are often only tenuously linked to an underlying theoretical model (2000b, pp. 6-7)”. In all these developments, the regulator and, more generally speaking, the banking industry takes credit rating models applicable to large, conventional corporations as reference. These models obviously cannot be adopted in the same way for other agents such as SMEs or social enterprises. Furthermore, qualitative approaches to risk management such as bankers’ subjective judgements appear to be downgraded, benefiting quantitative approaches. The high degree of interest in market-price-based measures, expressed since the turn of the century, is one step further from the reliance on banker’s judgemental expertise, which is the only way of interpreting soft information20. 4.2. The value of qualitative factors in predicting default A number of empirical studies document that, depending on the bank, internal rating systems are characterised by a huge diversity, ranging from statistical methods to exclusive expert judgements (Elsas and Krahnen, 1998; Machauer and Weber, 1998; Treacy and Carey, 2000; Brunner et al., 2000). In this regard, pure quantitative approaches to credit risk management are incompatible with the lending practices of banks that are heavily reliant on the production of qualitative knowledge. As argued in Section 2, the main properties of soft information are its low transferability and capacity of aggregation for comparability purposes. Consequently,

19

The BCBS (2000b, p. 108) is quite clear on this point by stating that “the main advantage of this approach is also its limit: the analysis is subjective, so that it can take into account all those qualitative factors that are difficult to quantify. On the other hand, it is difficult to assess the creditworthiness of a firm whose balance-sheet ratios are discordant: a firm may have a poor profitability ratio but an above-average liquidity ratio, and in this case different experts may have different opinions on the same firms”.

20

Surprisingly enough, this evolution has occurred even though the BCBS itself has documented the variety of rating techniques employed by financial institutions (BCBS, 2000a). The BCBS (2000a, p. 4) states that “while there does not appear to be a single standard for the structure and operation of internal rating systems, the survey highlighted a few alternative approaches. These can be viewed as points on a continuum with, at one extreme, systems focussed on the judgement of expert personnel, and at the other, those based solely on statistical”. 101 98 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

the best interpretation of this type of knowledge is given subjectively by its producers (e.g. loan officers). In other words, the only way of exploiting soft information from a risk management perspective involves relying on bankers’ judgemental expertise. Furthermore, the financing of entire credit market segments may be hampered by inadequate risk models (Altman et al., 2010). This is particularly the case of SMEs and, to an even greater extent, of firms such as cooperatives and social enterprises. Some precursors of the accounting-based credit rating now recognise the value of qualitative factors for SME risk modelling. For instance, Altman and Sabado (2007, p. 335) “acknowledge that our analysis could still be improved using qualitative variables as predictors in the failure prediction model to better discriminate between small and medium enterprises”. To the best of our knowledge, the amount of empirical evidence supporting this insight is still limited. Table 1 takes stock of the few existing studies. All of these studies follow the same methodological pattern. They examine the benefits associated with the inclusion of qualitative factors in addition to the conventional accounting and financial factors in terms of predicting default events. The analysis is generally based on logit or probit regressions, and the main indicators of forecast quality used are the area under the ROC (receiver operating characteristic) curve, the percentage of observations correctly classified, and the Brier score21. As displayed in Table 1, the findings unambiguously show that mixed models combining both qualitative and quantitative factors significantly outperform models including only quantitative factors, thereby highlighting the value of subjective factors derived from soft information. The results hold true regardless of the type of banking institution, whether commercial, as in Lehmann (2003) and Grunert et al. (2005) or cooperative, as in Cornée (2013)22. It is worth mentioning that Altman et al. (2010) have a rather rigid definition of qualitative information, which they consider as non-accounting and non-subjective data such as the activity sector, the age of the firm, etc. In spite of this limited definition, qualitative factors still remain valuable in terms of forecast quality. Table 1.TThe value of qualitative factors in credit risk management ABLE 1. THE VALUE OF QUALITATIVE FACTORS IN CREDIT RISK MANAGEMENT Study

a

Sample

Qualitative factors

Results

Lehmann (2003)

20, 000 SMEs from a German commercial bank. 2/3 of firms have a turnover up to EUR 5 million

1) Factor synthesising subjective assessments of future financial prospects, market position, and quality of lending relationship. 2) “Behavioural” factor derived from an analysis of the checking account

FIN: AUC = 0.72 FINQUAL: AUC = 0.81

Grunert et al. (2005)

409 medium-sized firms from a German commercial bank. Firms’ turnover is EUR25 million to EUR250 million

1) Subjective assessment of management quality. 2) Subjective assessment of market position

FIN: OCCd = 0.89, BSe = 0.13 FINQUAL : OCC = 0.92, BS = 0.10

Altman et al. (2010)

5.8 million SMEs from UK Companies House. Firms’ turnover goes up to £22.8 million

1) County court judgements (y/n), 2) audited accounts (y/n), 3) subsidiary (y/n), cash-flow statement (y/n), 4) late filing days, 5) audit report judgement, 6) age of the firm, 7) sector.

FINa : AUCc = 0.67 (Model 1) and 0.71 (Model 2) FINQUALb: AUC = 0.76 (Model 1) and 0.75(Model 2)

Cornée (2013)

389 very small firms from a French financial cooperative. Firms’ average turnover is EUR540,000

1) Subjective assessment of management quality. 2) Subjective assessment of project quality

FIN: AUC = 0.68, BS = 0.17 FINQUAL: AUC = 0.73, BS = 0.16

b

c

FIN: only modelincluding only including quantitative financialfactors, factors, b FINQUAL: model including both quantitative and qualitative factors, FIN: model quantitative financial FINQUAL: model including both quantitative and qualitative factors, AUC: area under the curve, d OCC: % of observations correctly classified, e BS: Brier score. The differences between indicators c AUC:(OCC, area under the curve, d OCC: % of observations correctly classified, e BS: Brier score.The differences between indicators AUC, or BS) are always significant, mostly at the 1% threshold

a

(OCC, AUC, or BS) are always significant, mostly at the 1% threshold

21

22

How can the value added by qualitative factors be explained? The best way of assessing informationally opaque borrowers consists of gathering soft information over the course of longterm relationships. As argued in Section 3, financial statements are markedly insufficient in For further details on these various indicators, refer to Brier (1950), Güttler (2005), Behr and Güttler (2007), Krämer and bridging the informational gap between banks and SMEs or social enterprises, since they lack the Güttler (2008). relevant information to accurately assess a small business’ creditworthiness. Subsequently, if a riskthis model does not include that factors deriving asfrom softbank. information, it automatically leaves Morecredit precisely, is a financial cooperative self-identifies a social aside a substantial fraction of knowledge, thereby losing part of its predictive power. 102 99 default 4.3. The value of sustainability criteria in predicting JEOD - Vol.3, Issue 1 (2014)

As mentioned previously in Subsection 3.3., social banks condition their loan approval


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

How can the value added by qualitative factors be explained? The best way of assessing informationally opaque borrowers consists of gathering soft information over the course of long-term relationships. As argued in Section 3, financial statements are markedly insufficient in bridging the informational gap between banks and SMEs or social enterprises, since they lack the relevant information to accurately assess a small business’ creditworthiness. Subsequently, if a credit risk model does not include factors deriving from soft information, it automatically leaves aside a substantial fraction of knowledge, thereby losing part of its predictive power. 4.3. The value of sustainability criteria in predicting default As mentioned previously in Subsection 3.3., social banks condition their loan approval decisions on the satisfaction of sustainability criteria. Therefore, the relevant question from a risk management perspective is whether these sustainability criteria may be good predictors of default events – in addition to qualitative and quantitative economically-oriented factors. In other words, should the use of extra-economic ratings by social banks be restricted to not only the screening stage – as is already the case – but also throughout the whole credit risk management process? In Cornée and Szafarz (2013), we scrutinise the behaviour of a French social banking institution and provide theoretical and empirical evidence of the significantly negative impact of social ratings on default probability. Moreover, we report that this effect exhibits a similar magnitude as that of financial rating. This range is quite substantial since an additional unit of social rating leads to around a 10% decrease in default probability. These findings can be rationalised as follows. Firstly, the social rating that is determined in-house by the social bank can be viewed as a measurement of the degree of proximity between its own identity and values, on the one hand, and those of the borrowers, on the other. Secondly, the social bank signals their “privileged status” to the borrowers who share its social values by charging them a lower interest rate, ceteris paribus. Lastly, these privileged borrowers respond favourably to this signal by making safer investments than the rest of the clientele with similar ex ante creditworthiness, thereby reducing their probability of defaulting. Weber et al. (2010) provide evidence confirming our findings. Their study, based on a sample of 180 German SMEs, shows that incorporating sustainability criteria into risk management is advantageous. The impact is considerable, since the percentage of correctly classified observations increases from 78.9 per cent to 86.6 per cent with the inclusion of social and environmental factors. The authors also argue that sustainability may improve the creditworthiness of firms through the avoidance of certain economic or financial risks. Integrating extra-economic factors entails additional costs and impacts banks’ profitability. Thus, a more comprehensive cost-benefit evaluation should be carried out to effectuate the net advantage of using sustainability criteria in credit risk management. In Cornée and Szafarz (2013), we partially address this issue by conducting a simple cost-benefit analysis to assess the net benefit of the “reciprocity” policy implemented by social banks. This policy seems to be costly for the bank, and the benefits associated with the decrease in credit default do not offset the extra-costs incurred by social screening. The assessment of social rating appears to be particularly expensive in our case. This may be explained by the combination of two factors: i) social rating measurements imply that a considerable soft-informational investment is made at the outset of a lending relationship and then amortised in the subsequent interactions; ii) 50 per cent of our sample consists of start-ups. If the proportion of existing firms in a relationship with the bank had been higher, the result of the cost-benefit analysis would have been different, since the marginal cost of assessing social rating is lower for this type of firm. 103 100 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

5. Conclusion In this paper, we have attempted to show that banks’ risk management procedures, especially credit rating systems, should be congruent with the peculiarities of their screening, pricing, and monitoring practices. Our analysis involves three steps. First, we elaborate on the notion of soft information and highlight the internal contractual issues inherent to its use by banking organisations. We then provide compelling evidence that, regardless of the dramatic evolutions undergone by the banking industry in recent decades, soft information is still critical for cooperative banks and, even more so, for social banks in their lending operations. Finally, we attest to the fact that credit rating significantly benefits from the inclusion of qualitative factors – not only economically-oriented factors but also sustainability-oriented ones. In our view, allowing the incorporation of soft components into credit ratings may be a legitimate means of fostering diversity in the banking sector. In contrast, restricting credit rating models to hard, explicit information alone may incite financial intermediaries not to screen credit applicants on the basis of first-rate qualitative knowledge collected through repeated interactions. Yet, it appears that a relational approach to financing is essential in tackling the problem of credit rationing. The emergence of the Raiffeisen cooperatives in the late nineteenth century provides a good illustration of this approach, and so does the nascent phenomenon of social banking nowadays (Kalmi, 2012). In comparison, “pure” quantitative default models, disconnected from many facets of socio-economic reality, appear inconclusive when it comes to facilitating access to credit in a desirable manner. As suggested by the Economist (02/2010, the 11th), let us hope that, “[the] changes [in the banking industry resulting from the financial crisis] point toward greater use of judgement and less reliance on numbers in the future.”

References Abdesselam, R., Cieply, S., Le Pape, N. (2002). Les facteurs de différentiation des banques mutualistes et AFB en matière de financement des PME-PMI. Revue d’Economie Financière, 67: 121-131. http:// dx.doi.org/10.3406/ecofi.2002.3577 Agier, I., Szafarz, A. (2013). Microfinance and gender: Is there a glass ceiling on loan size?. World Development, 42: 165-181. http://dx.doi.org/10.1016/j.worlddev.2012.06.016 Akhavein, J., Frame, W.S., White, L.J. (2005). The diffusion of financial innovation: An examination of the adoption of small business credit scoring by large banking organizations. Journal of Business, 78(2): 577-596. http://dx.doi.org/10.1086/427639 Alexopoulos, Y., Catturani, I., Goglio, S. (2013). In search for a model of governance in cooperative banking. Paper presented at the 2013 EMES Conference, Liege. Altman, E.I. (1968). Financial ratios, discriminant analysis and the prediction of corporate bankruptcy. Journal of Finance, 23(4): 589-609. http://dx.doi.org/10.1111/j.1540-6261.1968.tb00843.x Altman, E.I. (2002). Revisiting credit scoring models in a Basel 2 environment. Working Paper, SSRN. Altman, E.I., Sabado, G. (2007). Modeling credit risk for SMEs: Evidence from the US market. Abacus, 43(3): 332-357. http://dx.doi.org/10.1111/j.1467-6281.2007.00234.x Altman, E.I., Saunders, A. (1997). Credit risk measurement: Developments over the last 20 years. Journal of Banking & Finance, 21(11-12): 1721-1742. http://dx.doi.org/10.1016/S0378-4266(97)00036-8 Altman, E.I., Sabado, G. and Wilson, N. (2010). The Value of Non-Financial Information in small and medium-sized enterprise risk management. The Journal of Credit Risk, 6(2): 1-33. 104 101 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

Angelini, P., Di Salvo, R., Ferri, G. (1998). Availability and cost of credit for small businesses: Customer relationships and credit cooperatives. Journal of Banking & Finance, 22(6-8)  : 925-954. http:// dx.doi.org/10.1016/S0378-4266(98)00008-9 Artis, A., Cornée, S. (2013). Transformation informationnelle, certification et intermédiation financière: le cas de la banque solidaire. CREM Working Paper, No. 210326, Université de Rennes 1. Ayadi, R., Llewellyn, D.T., Schmidt, R.H., Arbak, E., De Groen, W.P. (Eds.) (2010). Investigating diversity in the banking sector in Europe: Key developments, performance and role of cooperative banks. Brussels: Centre for European Policy Studies. Banerjee, A.V., Besley, T., Guinnane, T.W. (1994). The neighbor’s keeper: The design of a credit cooperative with theory and a test. Quarterly Journal of Economics, 109(2): 491-515. http://dx.doi. org/10.2307/2118471 BCBS (Basel Committee on Banking Supervision) (2000a). Range of practice in banks’ internal rating systems. Discussion Paper (January), Bank for International Settlements. BCBS (Basel Committee on Banking Supervision) (2000b). Credit ratings and complementary sources of credit quality information. Working Paper No3 (August), Bank for International Settlements. BCBS (Basel Committee on Banking Supervision) (2006). International convergence of capital measurement and capital standards. A revised framework comprehensive version. Bank for International Settlements. Beaver, W.H. (1966). Financial ratios as predictors of failure. Journal of Accounting Research, 4: 7-11. http://dx.doi.org/10.2307/2490171 Becchetti, L., Garcia, M., Trovato, G. (2011). Credit rationing and credit view: Empirical evidence from an ethical bank in Italy. Journal of Money, Credit and Banking, 43(6): 1217-1245. http://dx.doi. org/10.1111/j.1538-4616.2011.00423.x Behr, P.A., Güttler, A. (2007). Credit risk assessment and relationship lending: An empirical analysis of German small and medium-size enterprises. Journal of Small Business Management, 45(2): 194213. Berger, A.N., Frame, W.S. (2007). Small business credit scoring and credit availability. Journal of Small Business Management, 45(1): 5-22. http://dx.doi.org/10.1111/j.1540-627X.2007.00195.x Berger, A.N, Miller, N.H., Petersen, M.A., Rajan, R.G., Stein, J.C. (2005). Does function allow organizational form? Evidence from the lending practices of large and small banks. Journal of Financial Economics, 76(2): 237-269. http://dx.doi.org/10.1016/j.jfineco.2004.06.003 Berger, A.N., Rosen, R.J., Udell, G.F. (2007). Does market size structure affect competition: The case of small business lending. Journal of Banking & Finance, 31(1): 11-33. http://dx.doi.org/10.1016/j. jbankfin.2005.10.010 Berger, A.N., Udell, G.F. (1995). Relationship lending and lines of credit in small firm finance. Journal of Business, 68(3): 351-381. http://dx.doi.org/10.1086/296668 Berger, A.N., Udell, G.F. (2002). Small Business Credit Availability and Relationship Lending: the Importance of Bank Organizational Structure. Economic Journal, 112(477): 34-53. http://dx.doi. org/10.1111/1468-0297.00682 Bhattacharya, S., Thakor, A.V. (1993). Contemporary banking theory. Journal of Financial Intermediation, 3(1): 2-50. http://dx.doi.org/10.1006/jfin.1993.1001 Bodenhorn, H. (2003). Short term loans and long term relationships: Relationship lending in early America. Journal of Money, Credit and Banking, 35(4): 485-505. http://dx.doi.org/10.1353/mcb.2003.0025 Boot, A.W.A. (2000). Relationship banking: What do we know?. Journal of Financial Intermediation 9(1), pp. 7-25. http://dx.doi.org/10.1006/jfin.2000.0282. Borzaga, C., Defourny, J. (Eds.) (2001). The emergence of social enterprise. London - New York: Routledge. 105 102 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

Brevoort, K.P., Hannan, T.H. (2004). Commercial lending and distance: Evidence from community reinvestment act data. Journal of Money, Credit and Banking, 38(8): 1991-2012. http://dx.doi. org/10.1353/mcb.2007.0000 Brier, G.W. (1950). Verification of forecasts expressed in terms of probability. Monthly Weather Review, 78: 1-3. http://dx.doi.org/10.1175/1520-0493(1950)078<0001:VOFEIT>2.0.CO;2 Brown, M., Fehr, E., Zehnder, C. (2009). Reputation: A Microfoundation of Contract Enforcement and Price Rigidity. Economic Journal, 111(536): 333-353. Brunner, A., Krahnen, J.P., Weber, M. (2000). Information production in credit relationships: On the role of internal ratings in commercial banking. Working Paper n. 2000/10. Centre for Financial Studies. Canovas, G., Solvano, P. (2006). Banking relationships: Effect on debt terms for small Spanish firms. Journal of Small Business Management, 44(3): 315-333. http://dx.doi.org/10.1111/j.1540-627X.2006.00174.x Cerqueiro, G., Degryse, H., Ongena, S. (2011). Rules versus discretion in loan rate setting. Journal of Financial Intermediation, 20(4): 503-529. Cole, R.A. (1998). The importance of relationships to the availability of credit. Journal of Banking & Finance, 22(6-8): 959-977. http://dx.doi.org/10.1016/S0378-4266(98)00007-7 Cole, R.A., Goldberg, L.G., White, J.W. (2004). Cookie cutter vs. character: The micro structure of small business lending by large and small banks. Journal of Financial and Quantitative Analysis, 39(2): 227-250. http://dx.doi.org/10.1017/S0022109000003057 Cornée, S. (2013). The relevance of soft information for predicting small business credit default: Evidence from a social bank. WP-CREM No 201226, Université de Rennes 1. Cornée, S., Masclet, D. (2013). Long-term relationships, group lending and peer sanctioning in microfinance: New experimental evidence. CEB Working Paper No 13/026. Cornée, S., Masclet, D., Thenet, G. (2012). Credit relationships: Evidence from experiments with real bankers. Journal of Money, Credit and Banking, 44(5): 957-980. http://dx.doi.org/10.1111/j.15384616.2012.00517.x Cornée S., Szafarz, A. (2013). Vive la différence: Social banks and reciprocity in the credit market. Journal of Business Ethics. http://dx.doi.org/10.1007/s10551-013-1922-9 D’Auria, C., Foglia, A., Reedtz, P. (1999). Bank interest rates and credit relationships in Italy. Journal of Banking & Finance, 23(7): 1067-1093. http://dx.doi.org/10.1016/S0378-4266(98)00131-9 Degryse, H., van Cayseele, P. (2000). Relationship lending within a bank-based system: Evidence from European small business data. Journal of Financial Intermediation, 9(1): 90-109. http://dx.doi. org/10.1006/jfin.1999.0278 Degryse, H., Ongena, S. (2005). Distance, lending relationships, and competition. Journal of Finance, 60(1): pp. 231-266. http://dx.doi.org/10.1111/j.1540-6261.2005.00729.x Degryse, H., Ongena, S. (2008). Competition and regulation in the banking sector: A review of the empirical evidence on the sources of bank rents. In Thakor, A.V., Boot, A.W.A. (Eds.). Handbook of financial intermediation and banking. Amsterdam: Elsevier. pp. 483-554. http://dx.doi.org/10.1016/B978044451558-2.50023-4 DeYoung, R., Hunter, W.C., Udell, G.F. (2004). The past, present and probable future of community banks. Journal of Financial Service Research, 25(2-3): 85-133. http://dx.doi.org/10.1023/ B:FINA.0000020656.65653.79 Diamond, D.W. (1984). Financial intermediation and delegated monitoring. Review of Economic Studies, 51(3): 393-414. http://dx.doi.org/10.2307/2297430 Diamond, D.W., Rajan, R. (2009). The credit crisis: Conjectures about causes and remedies. Working Paper N. 14739, NBER. 106 103 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

Di Salvo., R. (2002). La gouvernance des systèmes bancaires mutualistes et coopératifs en Europe. Revue d’Economie Financière, 67: 65-177. Economist (The) (2010). Number-crunchers crunched. The uses and abuses of mathematical models. A special report on financial crisis, February 11th. Elsas, R., Krahnen, J.P. (1998). Is relationship lending special? Evidence from credit-file data in Germany?. Journal of Banking & Finance 22(10-11): 1283-1316. http://dx.doi.org/10.1016/S0378-4266(98)00063-6 Fehr, E., Zehnder, C. (2006). Reputation and credit market formation. FINRISK Working Paper, University of Zurich. Ferri, G. (2010). Foreword. In Ayadi, R., Llewellyn, D.T., Schmidt, R.H., Arbak E., De Groen W.P. (Eds.). Investigating diversity in the banking sector in Europe: Key developments, performance and role of cooperative banks. Brussels: Centre for European Policy Studies. Ferri, G., Messori, M. (2000). Bank-firm relationships and allocative efficiency in Northeastern and Central Italy and in the South. Journal of Banking & Finance, 24(6) : 1067-2095. http://dx.doi. org/10.1016/S0378-4266(99)00118-1 GABV (Global Alliance for Banking on Value) (2012). Strong, Straightforward and Sustainable Banking. Financial Capital and Impact Metrics of Values Based Banking. Full report. Available online at : http://www.gabv.org/wp-content/uploads/Full-Report-GABV-v9d.pdf [Accessed on 15 April 2014] Gagliardi, F. (2009). Financial development and the growth of cooperative firms. Small Business Economics, 32(4): 439-464. http://dx.doi.org/10.1007/s11187-007-9080-z Gianfaldoni, P., Richez-Battesti, N. (2005). Les banques coopératives en France: Entre spécificités et banalisation. Rapport pour la DIIESE. Goglio, S., Leonardi, A. (2012). The motivation of economic behavior: The case of credit cooperative. Journal of Entrepreneurial and Organizational Diversity 1(1): 65-84. Grant, R.M. (1996). Toward a knowledge-based theory of the firm. Strategic Management Journal 17: 109-122. Gropp, R., Gruendl, C., Guettler, A. (2012). Does discretion in lending increase bank risk? Borrower self-selection and loan officer capture effects. European Banking Center Discussion Paper No. 2012-010. Grunert, J., Norden, L., Weber, M. (2005). The role of non-financial factors in internal credit ratings. Journal of Banking & Finance, 29(2): 509-531. http://dx.doi.org/10.1016/j.jbankfin.2004.05.017 Guille, M. (1994). Savoir bancaire spécifique, marché du crédit et intermédiation financière. Economie Appliquée 67 : 49-77. Guinnane, T.W. (1997). Regional organizations in the German cooperative banking system in the late 19th Century. Research in Economics, 51(3): 251-274. http://dx.doi.org/10.1006/reec.1997.0042 Guinnane, T.W. (2001). Cooperatives as information machines: German rural credit cooperatives. Journal of Economic History, 61(2): 366-389. http://dx.doi.org/10.1017/S0022050701028042 Güttler, A. (2005). Using a bootstrap approach to rate the raters. Financial Markets and Portfolio Management, 19(3): 277-295. http://dx.doi.org/10.1007/s11408-005-4693-1 Harhoff, D., Körting, T. (1998). Lending relationships in Germany. Empirical evidence from survey data. Journal of Banking & Finance, 22(10-11): 1317-1353. http://dx.doi.org/10.1016/S03784266(98)00061-2 Hesse, H. and Čihák, M. (2007). Cooperative banks and financial stability. IMF Working Paper, No. 07/02. Jensen, M.C., Meckling, W.H. (1995). Specific and general and organizational structure. Journal of Applied Corporate Finance, 8(2): 4-18. http://dx.doi.org/10.1111/j.1745-6622.1995.tb00283.x Kalmi, P. (2012). Cooperative banking. In Toporowski, J., Michell, J. (Eds.). Handbook of critical issues in finance. Edward Elgar Publishing, pp. 56-65. http://dx.doi.org/10.4337/9781849805957.00015 107 104 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

Katerinakis, T.A. (2012). Homo Dictyous transforms Homo Oeconomicus: The social network culture of Hellenic cooperative banking in Crete. In Goglio, S., Alexopoulos, Y. (Eds.). Financial cooperatives and local development. London: Routledge, pp. 141-162. Krahnen, J.P. and Weber, M. (2001). Generally Accepted Rating Principles: A Primer. Journal of Banking and Finance 25(1): 3-23. http://dx.doi.org/10.1016/S0378-4266(00)00115-1 Krämer, W., Güttler, A. (2008). On comparing the accuracy of default predictions in the rating industry. Empirical Economics, 34(2): 343-356. http://dx.doi.org/10.1007/s00181-007-0125-6 Labye, A., Lagoutte, C., Renversez, F. (2002). Banques mutualistes et Systèmes financiers: une analyse comparative. Allemagne, Grande-Bretagne, France. Revue d’Economie Financière, 67  : 85-109. http://dx.doi.org/10.3406/ecofi.2002.3575 Lehmann, B. (2003). Is it worth the while? The relevance of qualitative information in credit rating. Working Paper, University of Konstanz. Lehmann, E., Neuberger, D. (2001). Do lending relationships matter, evidence from bank survey data in Germany. Journal of Economic Behavior and Organization, 45(4): 339-359. http://dx.doi. org/10.1016/S0167-2681(01)00151-2 Leland, H.E., Pyle, D.H. (1977). Informational asymmetries, financial structure, and financial intermediation. Journal of Finance, 32(2): 371-387. http://dx.doi.org/10.2307/2326770 Liberti, J.M., Mian, A.R. (2009). Estimating the effect of hierarchies on information use. Review of Financial Studies, 22(10): 4057-4090. http://dx.doi.org/10.1093/rfs/hhn118 Machauer, A., Weber, M. (1998). Bank behavior based on internal credit ratings of borrowers. Journal of Banking & Finance, 22(10-11): 1355-1383. http://dx.doi.org/10.1016/S0378-4266(98)00064-8 Mester, L.J. (1997). What’s the point of credit scoring? Federal Reserve Bank of Philadelphia Business Review September/October, pp. 3-16. Morduch, J. (1999). The microfinance promise. Journal of Economic Literature, 37(4): 1569-1614. http:// dx.doi.org/10.1257/jel.37.4.1569 Orléan, A. (1999). Le pouvoir de la finance. Paris: Odile Jacob. Ory, J.-N., Lemzeri, Y. (2012). Efficiency and hybridization in cooperative banking: The French case. Annals of Public and Cooperative Economics, 83(2): 215-250. http://dx.doi.org/10.1111/j.14678292.2012.00462.x Ory, J-N., Jaeger, M., Gurtner, E. (2006). La banque à forme coopérative peut-elle soutenir durablement la compétition avec la banque SA?. Finance Contrôle Stratégie, 9(2): 121-157. Ozbas, O. (2005). Integration, organizational processes and allocation of resources. Journal of Financial Economics, 75(1): 201-242. http://dx.doi.org/10.1016/j.jfineco.2004.01.003 Petersen, M.A. (2004). Information: Hard and soft. Working Paper, NBER. Petersen, M.A., Rajan, R. (1994). The benefits of lending relationships: Evidence from small business data. Journal of Finance, 49(1): 3-37. http://dx.doi.org/10.1111/j.1540-6261.1994.tb04418.x Petersen, M.A., Rajan, R. (1995). The effect of credit market competition on lending relationships. Quarterly Journal of Economics, 110(2): 407-443. http://dx.doi.org/10.2307/2118445 Petersen, M.A., Rajan, R. (2002). Does distance still matter? The information revolution in small business lending. Journal of Finance, 57(6): 2533-2570. http://dx.doi.org/10.1111/1540-6261.00505 Puri, M., Rocholl, J., Steffen, S. (2011). Rules versus discretion in bank lending decisions. Unpublished working paper. Rajan, U., Seru, A., Vig, V. (2008). The failure of models that predict failure: Distance, incentives and defaults. Working Paper, SSRN.

108 105 JEOD - Vol.3, Issue 1 (2014)


Soft Information and Default Prediction in Cooperative and Social Banks Cornée, S.

Rajan, U., Seru, A., Vig, V. (2010). Statistical default models of incentives. American Economic Review, 100(2): 1-5. http://dx.doi.org/10.1257/aer.100.2.506. Rivaud-Danset, D. (1996). Les contrats de crédit dans une relation de long terme. De la main invisible à la poignée de main. Revue Economique, 27(4) : 937-962. San-Jose, L., Retolaza, J.L., Gutierrez, J. (2011). Are ethical banks different? A comparative analysis using the radical affinity index. Journal of Business Ethics, 100(1): 151-173. http://dx.doi.org/10.1007/ s10551-011-0774-4 Scott, J.A. (2004). Small businesses and the value of community banks. Journal of Financial Service Research, 25(2-3): 207-230. http://dx.doi.org/10.1023/B:FINA.0000020661.30763.fe Sharpe, S.A. (1990). Asymmetric information, bank lending, and implicit contracts: A stylized model of customer relationships. Journal of Finance, 45(4): 1069-1087. Stein, J.C. (2002). Information production and capital allocation: Decentralized versus hierarchical firms. Journal of Finance, 57(5): 891-1921. http://dx.doi.org/10.1111/0022-1082.00483 Stiglitz, J. (1990). Peer monitoring and credit markets. World Bank Review, 4(3): 351-366. Stiglitz, J., Weiss, A. (1981). Credit rationing in markets with imperfect information. American Economic Review, 71(3): 393-410. http://dx.doi.org/10.1093/wber/4.3.351 Treacy, W.F., Carey, M. (2000). Credit rating systems at large US banks. Journal of Banking & Finance, 24(1-2): 167-201. http://dx.doi.org/10.1016/S0378-4266(99)00056-4 Udell, G.F. (1989). Loan quality, commercial loan review and loan officer contracting. Journal of Banking & Finance, 13(3): 367-382. http://dx.doi.org/10.1016/0378-4266(89)90048-4 Uzzi, B. (1999). Embeddedness in the making of financial capital: How social relations and networks benefit firms seeking financing. American Sociological Review, 64(4): 481-505. http://dx.doi. org/10.2307/2657252 Uzzi, B., Gillespie, J. (1999). What small firms get access to capital and at what costs: Notes on the role of social capital and banking networks. In Brinton, J.L., Williams, A., Rhine, S. (Eds.). Business access to capital and credit. Washington DC: Federal Reserve Bank Press, pp. 413-442. Uzzi, B., Lancaster, R. (2003). Relational embeddedness and learning: The case of bank loan managers and their clients. Management Science, 49(4): 383-399. http://dx.doi.org/10.1287/mnsc.49.4.383.14427 Von Thadden, E. (1995). Long-term contracts, short-term investments and monitoring. Review of Economic Studies 62(4), pp. 557-575. http://dx.doi.org/10.2307/2298077 Weber, O., Remer, S. (Eds.) (2011). Social banks and the future of sustainable finance. London, New York: Routledge. Weber, O., Scholz, R.W., Michalik, G. (2010). Incorporating sustainability criteria into credit risk management. Business Strategy and the Environment, 19(1): 39-50.

109 106 JEOD - Vol.3, Issue 1 (2014)


AT T R I B U T I O N 3 . 0

You are free to share and to remix, you must attribute the work

17 2014 June 2014 | Vol.3, 1 (2014) 111-129 Publication date: xx | Vol.3, Issue 1Issue (2014) 108-127

AUTHOR MATTEO ALESSI Federcasse and University of Rome Tor Vergata malessi@federcasse.bcc.it

Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence

STEFANO DI COLLI Federcasse and John Cabot University sdicolli@federcasse.bcc.it JUAN SERGIO LOPEZ Federcasse, Rome jlopez@federcasse.bcc.it

ABSTRACT A panel of Italian banks for the period 2006-2012 is used in this paper to examine LLP main determinants. Our analysis also focuses on the determinants of the sub-components of LLP, i.e. provisions associated to Bad Loans and Impaired Loans and Bad Loans and Impaired Loans Coverage Ratio. A specific analysis for cooperative credit banks is provided. We find that Loan Loss provisioning for Italian banks seems to be driven principally by non-discretionary behavior. Economic fluctuations, according to our results, do not play a significant role, nor do signaling and income smoothing. Provisioning strategies for cooperative credit banks also seem to be affected by collateralized loans.

KEY-WORDS LOAN LOSS PROVISIONS; BANK LENDING; FINANCIAL SYSTEM CYCLICALITY

Acknowledgements The authors are grateful to Yiorgos Alexopoulos, Carlo Borzaga, Silvio Goglio, Panu Kalmi and the other participants of the Fourth Euricse International Workshop on â&#x20AC;&#x153;Cooperative Finance and Sustainable Developmentâ&#x20AC;? (Trento, Italy, June 2013) for their useful comments and suggestions. The views expressed in this paper are personal and not necessarily reflect those of Federcasse.

JEL Classification: G21, G28 | DOI: http://dx.doi.org/10.5947/jeod.2014.006

111 108 JEOD - Vol.3, Issue 1 (2014)


Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence Alessi M.; Di Colli S.; Lopez J.S.

1. Introduction 1. Introduction During the past five years, the Italian economy has fallen into one of the deepest recessions of the post-war period. The Italian banking system was obviously affected by this crisis. Bad loans have started to pile up. During pasthave fiveayears, theeffect: Italianthey economy fallen and into increase one of the deepest recessions of Increasing badthe loans double reduce has revenues provisions (which further the post-war period. The Italian banking system was obviously affected by this crisis. Bad loans reduce revenues). As up. pointed out bybad Balla et al. (2012) “loaneffect: loss provisions have a significant effect on have started to pile Increasing loans have a double they reduce revenues and increase earnings and (which regulatory capital. Because loan loss provisions areout at the bank managers, provisions further reduce revenues). As pointed bydiscretion Balla etofal. (2012) “loanthere lossis provisions a significant effectmore on earnings andnecessary regulatory loan income”. loss provisions the potentialhave for banks to provision or less than as capital. a way toBecause smooth their are In at the discretion of bank managers, there is the potential for banks to provision more or lessdue thanto principle, loan loss provisions (LLP) must be used to cover expected losses; however, necessary as a way to smooth their income”. the discretionary behavior of bank managers, they can become an important tool to pursue goals that In principle, loan loss provisions (LLP) must be used to cover expected losses; however, due are different from a fair representation of the expected evolution a bank’s loantool losses. In a situation to the discretionary behavior of bank managers, they can becomeofan important to pursue goals characterized by an ample fluctuation of the business cycle, provisioning policy can be used to stabilize that are different from a fair representation of the expected evolution of a bank’s loan losses. In a earnings dividends. by an ample fluctuation of the business cycle, provisioning policy can be situationand characterized used to example, stabilize earnings and the dividends. For just recently Bank of Italy has put pressure on the banking industry to correctly For example, just recently Bank of provide Italy has the banking industry datatoon evaluate the viability of loans and totheadequately forput the pressure increasingoncredit risk1. Indeed, correctly evaluate the viability of loans and to adequately provide for the increasing credit risk1. credit provisioning showprovisioning a non-homogeneous picture. The coverage ratio (the of loan loss(the reserves Indeed, data on credit show a non-homogeneous picture. Theratio coverage ratio ratioto total badloss loans - Table to 1)total appears types of different banks. between types of banks. of loan reserves badquite loansdifferent - Table between 1) appears quite TABLE 1. LOAN QUALITY : RATIO OF PERFORMING LOANS AND NON -PERFORMING LOANS TO TOTAL LENDING AND 2012 Table 1. Loan quality: ratio of performing loans and non-performing loans to total lending and coverage ratios - December COVERAGE RATIOS - DECEMBER 2012 Top 5 groups

Customer loans

Large banks

Small banks

Minor banks

Financial companies

Total banking industry

Compos. (%)

Cover Ratio

Compos. (%)

Cover Ratio

Compos (%)

Cover Ratio

Compos. (%)

Cover Ratio

Compos. (%)

Cover Ratio

Compos. (%)

Cover Ratio

100

6,3

100

4,7

100

5,9

100

4,1

100

6,6

100

5,7

of which - performing

86,0

0,6

88,5

0,5

85,6

0,6

86,2

0,5

86,2

1,3

86,6

0,6

- non performing

14,0

41,1

11,5

36,7

14,4

37,8

13,8

27,2

13,8

40,2

13,4

38,8

- bad debts

7,7

56,1

6,1

52,2

7,4

56,0

6,1

46,1

8,1

55,1

7,2

54,6

- substandard

4,1

25,2

3,7

23,1

4,7

22,7

5,8

14,1

3,8

22,2

4,2

23,2

- restructured

1,2

24,0

0,6

17,0

0,5

15,7

0,4

16,1

0,2

10,0

1,0

22,4

- past due

1,0

10,8

1,1

7,5

1,9

10,1

1,6

4,1

1,7

13,4

1,1

9,4

Memorandum item: 1.344.548 487.923 customer loans Source: Bank of Italy: “Financial Stability Report n. 5”

137.323

186.948

71.286

2.218.028

Source: Bank of Italy: “Financial Stability Report n. 5”

Prior theoretical and empirical research suggests three central reasons to explain managerial discretionary behavior: income smoothing, signaling and capital regulation. These motives, together Prior theoretical and empirical research suggests three central reasons to explain managerial with non-discretionary components and economic fluctuation, contribute to explaining provisioning discretionary behavior: income smoothing, signaling and capital regulation. These motives, together with policy. non-discretionary components and economic fluctuation, contribute provisioning policy. A further aspect that may affect provisioning policy is to theexplaining transactional or relationship approach to clients. As pointed out by Dewenter and Hess (2003), these two approaches may yieldto A further aspect that may affect provisioning policy is the transactional or relationship approach different outcomes when banks evaluate doubtful loans: relationship banks may have better clients. As pointed out by Dewenter and Hess (2003), these two approaches may yield different outcomes information on customers than transactional banks and therefore less risky loans (or higher recovery rates); on the other hand relationship banks may have a stronger incentive to “evergreen” loans 1

                                                                                                                        1   “Il economicoimpone imponeallealle banche rischi creditizi elevati, da fronteggiare con patrimoniali. riserve patrimoniali. Bancasta “Il ciclo ciclo economico banche rischi creditizi elevati, da fronteggiare con riserve La BancaLad’Italia

d’Italia sta conducendo verifiche sull’adeguatezza di un valore effettuate dagruppi un ampio numero gruppi conducendo verifiche sull’adeguatezza delle rettifiche di delle valorerettifiche effettuate da ampio numero di bancari grandi di e medi. Ove necessarie, sono richieste correttive”. Speech given by F.azioni Panettacorrettive”. Vice General Manager of Bank of Italy, Perugia, 23rd bancari grandi e medi. azioni Ove necessarie, sono richieste Speech given by F. Panetta Vice March General rd 2013. of Bank of Italy, Perugia, March 23 2013.   Manager 2     112 109

JEOD - Vol.3, Issue 1 (2014)

barbara.franch

Commenta [1 X  FEDE:   Per  il  formato  io   pubblicati  sul  sito carattere  normal che  la  editor  abb numero  però  ved sono  paper  con  u facciamo?    


Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence Alessi M.; Di Colli S.; Lopez J.S.

when banks evaluate doubtful loans: relationship banks may have better information on customers than transactional banks and therefore less risky loans (or higher recovery rates); on the other hand relationship banks may have a stronger incentive to “evergreen” loans compared to transactional banks. In both cases a relationship bank would show a lower LLP even though in the first case it is a correct evaluation of the expected loss, while in the second case it is a managerial discretionary behavior. Based on a panel of more than 400 Italian banks for the period 2006-2012, we examine LLP main determinants. Our analysis also focuses on the determinants of the sub-components of LLP, i.e. provisions associated to Bad Loans and Impaired Loans. Along with the standard explanatory variables commonly used in empirical literature, we also analyze the effect of guaranteed loans on the non-discretionary component. A bank with a higher stock of collateralized loans can, in principle, reduce expected future losses and consequently affect LLP decision-making process. In this paper we also try to model two particular indicators: Bad Loans and Impaired Loans Coverage Ratios. We extend our analysis to these two indicators due to their crucial significance on bank’s report activity and strategic planning. The main findings of this analysis are as follows. First, Loan Loss provisioning for Italian banks seems to be driven principally by non-discretionary behavior. Expectations about future losses and credit risk perception appear to be relevant components for determining LLP. According to our results, neither the economic fluctuations nor signaling or income smoothing hypothesis play a significant role. The non-discretionary component is also significant for cooperative banks. Provisioning strategies for this particular category of banks seem to be affected by collateralized loans as well. We find a negative relationship between provisions and total guaranteed loans. This factor could explain the lower level of hini 5/5/14 17:31 1]:   provisions and Coverage Ratios experienced during the past years by Cooperative Credit Banks (CCBs) with respect to other categories of banks2.  avevo  guardato  i  paper  già   o,  dove  per  le  tabelle  avevi  usato  il   The paper is organized as follows. Section 2 reviews the literature on bank’s Loan Loss Provisions le,  non  maiuscoletto…  quindi  sia  io   biamo  seguito  quello  stile.  determinants. In  questo   Section 3 reports the empirical methodology. Section 4 presents data and empirical estimates do  che  usi  il  maiuscoletto,  quindi  ci   forCosa   the Italian banking system and for the sub sample of cooperative credit banks. Concluding remarks are uno  stile  e  paper  in  un  altro.   presented in the final section.

2. Literature review A recent strand of literature has focused on how provisions contribute to the procyclicality of financial systems by being lower when output and credit are expanding and higher in periods of contraction. Bank lending behavior is generally affected by a strong relationship with the economic cycle. There is a large body of literature which provides evidence in favor of this phenomenon. Asea and Blomberg (1998), using U.S. data from 1977 to 1993, show that bank lending evolves cyclically, affecting aggregate economic activity. The same conclusions are reached by Gambacorta and Mistrulli (2004), Gambacorta (2005), Di Giulio (2009), Di Colli and Girardi (2010) for the Italian banking system. Bikker (2004), considering a panel dataset with 26 OECD countries over the period 1979-1999, finds that lending is strongly dependent on demand factors, measured by cyclical variables such as real GDP growth, inflation, unemployment and real money supply. In addition, Peek et al. (2003) and Lown and Morgan (2006) clearly identify the effects of loan supply on fluctuations in credit and GDP which supports the existence of the bank lending 2

The higher level of collateralized loans may be due to the fact that CCB clients are mainly small and micro enterprises, typically riskier borrowers than large enterprises, households and the public sector. 113 110 JEOD - Vol.3, Issue 1 (2014)


Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence Alessi M.; Di Colli S.; Lopez J.S.

channel. These studies are in line with empirical findings related to the 1990-1992 “credit crunch” in the United States. Bernanke and Lown (1991) find a positive correlation between loan growth and changes in bank capital during 1990-1991 while Hancock and Wilcox (1998) and Peek and Rosengren (1995) detect a positive effect of bank capital requirement on credit growth during the same period. Brinkmann and Horvitz (1995) also find a positive effect on loan growth, but only for large banks. Wagster (1999) shows that stricter supervision, which occurred during the period 1990-1992 in Canada, UK and the USA, implies that less credits were extended to lower-risk investments such as government bonds. Focusing on the impact of monetary policy, the point is that banks could react to changes in monetary policy in a different way. As a consequence, changes in the money market rate affect the cost of funding but this has a limited effect on lending when banks can easily raise non-deposit funding or when banks own a buffer of liquid assets. Kashyap and Stein (1995) findings are consistent with the bank lending channel view showing that loan growth of large banks and small banks respond differently to a monetary policy shock. Other studies in this field demonstrate that the impact of the bank lending channel in the US banking system is also greater for banks with less liquid assets and less capital (Kashyap and Stein, 2000; Kishan and Opiela, 2000). The bank lending view is also relevant for European banks (Altunbas et al., 2002; Ehrmann et al., 2003) but with less conclusive results. Theoretical investigations (Chami and Cosimano, 2001; Zicchino, 2005; Furfine, 2001) also emphasize the role of macroeconomic conditions and changes in banking regulation to explain the impact of capital requirements on bank lending. Fluctuations in bank lending over the business cycle could also be explained by misevaluation of credit risk phenomena. In phases of economic boom, banks are inclined to take on greater risks. By contrast, banks are excessively pessimistic during cyclical downturns if they overstate credit risk. Disaster myopia (Guttentag and Herring, 1984, 1986), herd behavior (Rajan, 1994) and the institutional memory hypothesis (Berger and Udell, 2002) account for misevaluation of credit risk. Caporale et al. (2013) demonstrate that a bad loans surplus (more bad loans than could be explained by macroeconomic and financial determinants) occurred during the 2008-2009 and 2011-2012 recessions in Italy. They also determine that bad loans surplus during the crisis is partially due to the lending policy adopted during economic good times. Another issue in analyzing the relationship between loan behavior, economic cycle and the problem of misevaluation of credit risk is backward-looking provisioning systems. Beaver and Engel (1996) identify a non-discretionary component in loan loss provisions related to contemporaneous loans, while Laeven and Majnoni (2003) and Bikker and Metzemakers (2005) show that provisioning behavior, in particular the ratio of loan loss provisions to total loans, is related to the business cycle. Additional empirical evidence can be found for France (Clerc et al., 2001), Austria (Arpa et al., 2001), Spain (Fernandez de Lis et al., 2001) and the United Kingdom (Pain, 2003). A time-lag can notably be stressed between riskier loans which are granted during the peak of the business cycle (Keeton, 1999; Jiménez and Saurina, 2006) and loan loss provisions which are built up only during the next downturn according to backward-looking rules (Caporale et al., 2013). In particular, Jordan et al. (2002) emphasize that the cyclicality of loan loss provisions is also reflected in bank capital. Loan loss provisions could also be affected by discretionary components. In this way, income smoothing is a specific way to manage earnings in firms and in banks. Income smoothing could be defined as a practice aiming at the reduction of variability of net profit over time. In other words, managers will increase (decrease) loan-loss provisions when earnings are high (low) in order to stabilize net-profit. In the banking field, bank managers might have significant incentives to adopt income smoothing procedures: adjusting a bank’s current performance to a firm-specific mean (Collins et al., 1995) or to the average performance of other benchmark-banks (Kanagaretnam et al., 2005), allowing managers to grant a steady flow of dividends to bank stockholders, improving the risk perception that regulators have about the bank, maintaining the 114 111 JEOD - Vol.3, Issue 1 (2014)


Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence Alessi M.; Di Colli S.; Lopez J.S.

stability of the bank’s stock price by reducing earnings volatility. Other motivations behind adopting an income smoothing approach are to exploit the signaling power of a stable income (Ronen and Sadan, 1981) and reducing the perceived bankruptcy probability of the firm (Trueman and Titman, 1988). Managerial self-interest incentives could also lead to income smoothing, as well as stabilizing managers’ compensation over time, and minimizing the probability of being fired (Fudenberg and Tirole, 1995). Furthermore, from the supervisory authority’s point of view, regulators are interested in reducing banks’ pro-cyclical behavior. In other words, banks are asked to increase loan loss reserves during good times, and to draw resources from these reserves when the economy slows down. Finally, transactional and relationship banks may behave differently when facing the decision to make provisions for bad loans. Due to the strict relationship with their clients, relationship banks may have a stronger incentive to renegotiate or roll-over doubtful loans compared to transactional banks. Remaining in the field of banking, there is a vast literature regarding the use of loan-loss provisions for income smoothing purposes. Greenwalt and Sinkey (1988) find that regional banks are more likely to be involved in income smoothing behavior, while Ma (1988) shows that U.S. commercial banks used loan-loss provisions to smooth earnings, finding no relationship between loan portfolio quality and loan-loss provisions. Collins et al. (1995) also find a positive relationship between earnings management and LLPs, thus supporting the notion that banks smooth income over time to a firm-specific mean. Bhat (1996) demonstrates that banks are more likely to be involved in income smoothing practices if they are small and in poor financial condition. More recently, Anandarajan et al. (2007) show that Australian commercial banks are engaged in earnings management practices, especially if they are publicly traded. Fonseca and González (2008), considering a cross-country dataset, find that the incentive to smooth earnings is positively related with developed and market-oriented financial systems but negatively related with banking systems characterized by higher levels of accounting disclosure, the existence of a supervisory framework, and by stricter restrictions on banking activities. Dewenter and Hess (2003) find that transactional and relationship banks differ in their loan loss provisioning and write off due to different incentives. Finally, Curcio and Hasan (2013) explicitly examine the impact of loan loss provisions on bank lending. Shrieves and Dahl (2002) - analyzing the utilization of the discretionary accounting practice of Japanese banks during 1989-1996 - find a negative and significant relationship between loan loss provisions and year-on-year change in total loans. This result is consistent with the hypothesis that loan loss provisions influence credit cycles. However, to explicitly test the impact of loan loss provisions on the fluctuations of bank lending, the discretionary component and the non-discretionary component need to be distinguished. Indeed, the cyclical behavior of non-discretionary provisions should reinforce the cyclical nature of bank lending. On the contrary, the discretionary component, through the income smoothing behavior, may reduce the procyclicality of bank lending.

3. Methodology The key objective of this analysis is to investigate the determinants of LLP and Coverage Ratios for the Italian banking system. Regarding LLPs, both empirical and theoretical literature suggest three main classes of factors which may explain loan loss provisioning: non-discretionary behavior, discretionary behavior and economic cycle. The non-discretionary component is related to expected losses and credit risk of a bank’s portfolio. This factor, together with economic fluctuation, could be strongly cyclical. In order

115 112 JEOD - Vol.3, Issue 1 (2014)


reinforce the cyclical of bank lending. the thebehavior inguished. Indeed, thenature behavior ofOnnondiscretionary provisions component need to becyclical distinguished. Indeed, thecontrary, cyclical of non-should reinforce the cyclical nature of bank lending. On the contrary, the discretionary should reinforce the cyclical nature of lending. Onthrough the contrary, the smoothing behavior, may reduce the procyclicality of gh the income smoothing behavior, procyclicality of bank cyclical nature ofprovisions bank lending. Onmay the reduce contrary, discretionary component, the income provisions should reinforce the cyclical nature the of the bank lending. On the contrary, the discretionary component, through the income smoothing behavior, may reduce the procyclicality of smoothing behavior, may reduce the procyclicality of bank lending. component, through the income smoothing behavior, may reduce the procyclicality of bank lending.

3. Methodology Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence 3. Methodology Alessi M.; Di Colli S.; Lopez J.S. ogy s analysis is to investigate the determinants of LLP and Coverage The key objective of this analysis is to investigate the determinants of LLP and Coverage Thethe key objective ofofempirical this analysis istheoretical to investigate the determinants ofbanking LLP andsystem. Coverage Regarding LLPs, both literature investigate LLP andand Coverage Ratios for the Regarding LLPs, both empirical and theoretical literature yosystem. objective of thisdeterminants analysis is to investigate the determinants of LLP andItalian Coverage Ratios for theempirical Italian loan banking system. Regarding LLPs, both empirical and theoretical literature actors which may explain loss provisioning: non-discretionary rding LLPs, both and theoretical literature suggest three main classes of factors which may explain loan loss provisioning: non-discretionary e Italian banking system. Regarding LLPs, both empirical and theoretical literature three mainThe classes of factors which may explain loantolossdiscretionary provisioning: non-discretionary rmain andsuggest economic non-discretionary is related may explain loss provisioning: behavior, behavior and economic cycle. The non-discretionary component is related to classesloan ofcycle. factors which may non-discretionary explain component loan loss provisioning: non-discretionary behavior, discretionary behavior and economic cycle. The non-discretionary component is related to of aThe bank’s portfolio. This factor, with economic cskcycle. non-discretionary component istogether related to expected losses andwe credit risk of bank’s portfolio. together with economic cretionary behavior and economic cycle. The non-discretionary component is related toestimate to capture the effect of non-discretionary behaviors, fivea different models, This each factor, one with a expected losses and the credit riskofwith of a economic bank’s portfolio. This factor, together with economic yclical. In order to capture effect non-discretionary behaviors, k’s portfolio. This factor, together fluctuation, could be strongly cyclical. In order to capture the effect of non-discretionary behaviors, ses and credit risk of a bank’s portfolio. This factor, together with economic fluctuation, be cyclical. In order capture effect of non-discretionary different dependent variable. The first one is the ratio of Non-Performing Loans over Total Loansvariable. at the The first one is the els, one the withcould a different dependent variable. Thetofirst one isestimate the er toeach capture effect of strongly non-discretionary behaviors, wethe five different models,behaviors, each one with a different dependent ould be strongly cyclical. In order to capture the effect of non-discretionary behaviors, we estimate five different models, each one with a different dependent variable. The first one is the overdifferent Loans at end the end of period analysis with a Total different dependent variable. The first is the).. Our ratio ofThe Non-Performing over Total at the end ofofthe period t ( but ). Our analysis Our analysis aims to estimate only theLoans determinants total LLP, of thethe period tt ( onedependent ive models, each one with a different variable. first one is theLoansnot Non-Performing over Loans at the end ofon theestimate period tnot ( only the).determinants Our analysisof total LLP, but also the sub component (provisions on eterminants of total LLP, but also the sub component (provisions ans atratio the of end of the period t (Loans ).Total Our analysis aims to Performing Loans over Total Loans at the end of the period t ( ). Our analysis also thecomponent sub component (provisions on Bad Loans and Impaired Loans). aims to estimate not only the determinants ofTotal total LLP,Bad but subImpaired component (provisions on So we also use the ratio of Bad s). So we also the ratio ofof Bad over ( also ) the(provisions total but also the sub (provisions on Loans and ) ate notLLP, only theuse determinants totalLoans LLP, but also theLoans sub component on Loans). So we also use the ratio of Bad Loans over Total Loans ( Bad Loans and Impaired Loans). So we also use the ratio of Bad Loans over Total Loans ( and Impaired Loans over Total Loans asdependent dependent variables. Loans over Total Loans Loans ( ratio ) Loans). asofdependent use the BadSo Loans overuse Total ( Bad) Loansand Loans( over )Total Loans ( )) as variables. nd Impaired wevariables. also the Loans ratio of overImpaired Total Loans andweImpaired Loans over Loans ( ) asa dependent variables. cators, consider their )Total first differences in forward we looking sLoans dependent variables. Together withtheir these indicators, we also consider their first differences in a forward looking overalso Total Loans (Together as dependent variables. with these indicators, also differences Together with these indicators, we also consider their firstconsider differences in first a forward lookingin a forward looking prospective oerconsider their first differences a forward ; inconsider ).prospective ; ). In with these indicators, we also theirlooking first differences inIna forward( looking prospective ( ( ).. In addition, In we include). the of of loans to total ;assets ). The coefficients associated with; we include ). ; In addition, The ratio coefficients associated with ( ). theIn ratio of loans to total assets ( addition, we the ratio offactors loans could to total assets (these as ). The with to be positive, as these factors could be considered as einclude expected be include positive, astotal theseassets be considered tal assets ( to ). loans The associated with three setscoefficients of variables are expected . The coefficients associated withassociated these three sets of variables are expected to be to the ratio of loanscoefficients to total ( ). The coefficients associated with these three sets of variables are expected to be positive, as these factors could be considered osses and overall default risk. to Another feature that could be of positive, as are these factors could be important considered as factors proxies expected potential losses and as overall default risk. Another important feature that could ets variables expected be positive, ascould these could of beas considered asexpected positive, as these factors be considered proxies of potential losses and overall default risk. proxies of expected potential losses andthat overall default risk. Another that could erall losses) is represented guarantees on loans. with ainfluence higher default risk. Another important feature could creditimportant risk could (and feature future losses) is represented by guarantees on loans. Banks with a higher pected potential losses by and overall default risk.Banks Another important feature that influence credit risk (and future losses) is represented by guarantees on loans. Banks with a higher Another important feature that could influence credit risk (and losses) is represented by guarantees editexpected totallylosses) or partially recover the value of theon investment presented bytoguarantees on is loans. Banks with a higher level of collateralized assets are future expected to totally or partially recover the value of the investment risk (and future represented by guarantees loans. Banks with a higher level ofbank collateralized assets are of expected toprovisioning totally or partially recover the value of the investment g to this, a should exhibit a lower level of if there totally or partially recover the value the investment in case of insolvency. According to this, a bank should exhibit a lower levelthe of provisioning if there teralized assets are expected to Banks totally or partially recover theof value of the investment on loans. with a higher level collateralized assets are expected to totally or partially recover in case of insolvency. According to this, a if bank should exhibit a lower level of provisioning if there nk should exhibit a5  lower level of provisioning there olvency. According to this, a bank should exhibit a lower level of provisioning if there 5  

value of the investment in case 5   of  insolvency. According to this, a bank should exhibit a lower level of 5   provisioning if there is a high percentageofofcollateralized collateralized loans in its credit portfolio. As indicators, we is a high percentage its credit As portfolio. As indicators, is a high percentage of collateralized loans in itsloans creditin portfolio. indicators, we choosewe twochoose two choose twopercentage different variables. The first one is ratio the of totally guaranteed loans totototal different variables. The first one isof the ratio ofportfolio. totally guaranteed loans total loans ).. The is a high of collateralized inratio its credit Asto indicators, weloans choose different variables. The first one isloans the totally guaranteed loans total ). (The two 3 loans ( 3( second one is the ratio of guaranteed loans to total loans ) . is a high percentage of collateralized loans credit portfolio. Asloans indicators, we)loans second one is ratio ofin guaranteed to total loans ( .choose The second one isloans the ratio guaranteed loans tototally total different The first one isitsthe ratio of guaranteed to two total loans ( ). The is aofhigh percentage of collateralized in the its credit portfolio. Asloans indicators, we choose two ge collateralized loans in itsvariables. credit Asof indicators, wecredit choose two second ofthe LLP, i.e. the discretionary one, related tomanagement different management is a high percentage of portfolio. collateralized loans in component its portfolio. As indicators, we choose 3 related different variables. The first one istotally theThe ratio ofAstotally guaranteed loans totwo total loans ()is ). The The second component ofloans LLP,to i.e. discretionary one, toistwo different percentage of collateralized loans in its credit portfolio. indicators, we choose second one is the ratio of guaranteed to total loans ( . different variables. The first one is the ratio of guaranteed loans total loans ( ). The The second component of LLP, i.e. the discretionary one, is related to management The first one isdifferent the ratiovariables. of totally guaranteed loans toratio total ( to ).“income The objectives. smoothing theory”, banks tend to(increase) decrease objectives. (increase) LLP The first one is loans the ofloans totally guaranteed to total loans ( different The one isguaranteed the ratio of guaranteed toAccording loans ( thesmoothing )3. loans According the “income theory”, banks to ).decrease LLP ariables. Theissecond first isof the ratio ofobjectives. totally guaranteed loans to total second one the one ratio loans towhen total (tototal )3i.e. . loans 3 ). The atio of guaranteed loans to total loans (of guaranteed )3. loans The second component of LLP, the(banks discretionary one, is tend related to different management earnings are expected to be low (high). Following this approach, the sign associated to second one is the ratio loans to total loans ( ) . 3 According to the “income smoothing theory”, tend to decrease (increase) LLP when earnings are The second component of (LLP, i.e.)expected the discretionary one, is related to different when earnings are tois be low Following this management approach, the sign associated to is the The ratiosecond ofLLP, guaranteed loans toLLP, total i.e. loans .to different component of the one, related to(high). different management omponent of i.e. thesecond discretionary one, isdiscretionary related management objectives. According to the “income smoothing theory”, banks tend to decrease (increase) LLP earnings could be positive or negative. If banks use provisions to smooth earnings, there should be a The component of LLP, i.e. the discretionary one, is related to different management objectives. to low the “income smoothing theory”, banks tend toprovisions decrease (increase) LLP earnings could one, be Following positive or negative. If decrease banks use(increase) to smooth earnings, there should be a second component ofexpected LLP, i.e.“income the discretionary is decrease related to different management to be (high). this approach, the sign associated to earnings could be positive or the objectives. According toAccording the smoothing banks tend to LLP ing to the “income smoothing theory”, banks tendtheory”, to (increase) LLPhand, positive relationship. On the other negative sign should indicate pro-cyclicality. We use objectives. According toare the “income smoothing theory”, banks tendasign to decrease (increase) LLP when earnings expected to be low (high). Following this approach, the sign associated to when earnings are expected to be low (high). Following this approach, the sign associated to positive relationship. On the other hand, a negative should indicate pro-cyclicality. We use the According the “income banks tend to decrease (increase) LLP when earnings are expected to be lowtheory”, (high). Following this approach, the sign associated toa positive expected toto be low (high).smoothing Following this thesmooth sign associated to there ratio earnings before taxes and loanthe lossthere provision over ) as other abe variable negative. If banks provisions to earnings, should be the when earnings are expected toapproach, beofbefore low (high). Following approach, sign associated to( assets earnings could positive oruse negative. If banks use provisions tothis smooth earnings, berelationship. a total ratio of earnings interest, taxes and loan loss provision over total assets )there as a(On variable could be positive or negative. Ifinterest, banks use provisions to smooth earnings, should a ings areorexpected toIfearnings be low (high). Following this approach, the sign associated to earnings could be positive orbe negative. If banks use provisions to smooth earnings, there should be a should positive negative. banks use provisions to smooth earnings, there should be a to test the income smoothing hypothesis. earnings could be positive or negative. If banks use provisions to smooth earnings, there should be a positive relationship. On the other hand, a negative sign should indicate pro-cyclicality. We use the to test the income smoothing hypothesis. hand, a negative sign should indicate pro-cyclicality. We use the ratio of earnings before interest, taxes and uld be positive or negative. If banks use provisions to smooth earnings, there should be a positive relationship. On the other hand, a negative sign should indicate pro-cyclicality. We use the positive relationship. On the other hand, a negative sign should indicate pro-cyclicality. We use the p. On the otherpositive hand, a relationship. negative signOn should indicate pro-cyclicality. Weshould use the Loan provisions could alsoforbe used( formanagement” “capital management” purposes. with a the other hand, aLoss negative sign indicate pro-cyclicality. usepurposes. the of earnings interest, taxes and loan loss provision ) as aWe variable Loan Loss provisions could also beover used “capital Banks withBanks a ationship. Onratio the before other hand, abefore negative sign should indicate We use the ratio interest, of earnings interest, taxes and loan loss provision over total assets (total )assets asbuild athe variable efore taxes and loan provision over total assets ( pro-cyclicality. )can as aause variable ratio ofloss earnings before interest, taxes and loan loss provision over total assetsbuffer. (hypothesis. ) asmeasure a variable variable to test income smoothing loan loss provision over total assets lower level of capital provisions to up a greater reserve To the effect ratio of earnings before interest, taxes and loan loss provision over total assets ( ) as a variable to testsmoothing thetaxes income smoothing hypothesis. lower of capital provisions up a greater reserve buffer. To measure the effect nings interest, and loan losslevel provision over can totaluse assets ( ) to asbuild a variable to testbefore the hypothesis. income hypothesis. moothing tothetest the income smoothing hypothesis. of regulatory managing regulatory capital wethe compute theof deviation ofCapital the Total Capital Ratio with respect to to test income smoothing hypothesis. Loan Loss provisions could also be used for “capital management” purposes. Banks with a lower level of Loan Loss provisions could also be used for “capital management” purposes. Banks with a of managing capital we compute deviation the Total Ratio with respect to ncomeLoan smoothing hypothesis. Lossalso provisions could also bemanagement” used forcent, “capital management” purposes. Banks a value of this indicator indicates a wellrovisions could be used forprovisions “capital purposes. Banks with a ( “capital 8 divided per divided by 8 ( per cent ).To Awith higher Loan Loss could alsotocould be used foracent “capital management” purposes. Banks withpurposes. a indicatesBanks Loan Loss provisions also be used for management” with a lower level of capital can use provisions build up greater reserve buffer. measure the effect 8 per cent, by 8 per ). A higher value of this indicator a wellLoss could also beprovisions for “capital management” purposes. Banks with a To lower level of capital can use to build upbuffer. a greater reserve buffer. To buffer. measure themeasure effect the effect of managing regulatory capital can use provisions to build up measure a greater reserve tal canprovisions use provisions to build upused a greater To the effect capitalized lower level oflevel capital can usereserve provisions to build up a greater buffer. Toreserve measure the effect of managing regulatory capital we compute the deviation of the reserve Total Capital Ratio with to To measure the effect capitalized bank. lower of capital can use bank. provisions build upeffect awith greater buffer. of managing capital use provisions towe build up athe greater reserve buffer. To to measure the of regulatory capital compute the deviation of the Total Capital Ratio respect to respect atory capitalcan we compute the deviation of Total Capital Ratio with respect to Banks can also use LLP to signal financial strength. We one-year-ahead capital we compute the deviation of the Total Capital Ratio with respect to 8use pertothe cent, divided by percentage 8 per of managing regulatory capital we compute the deviation of the Total Capital Ratio with respect 8 per cent, divided by 8 per cent ( ). A higher value of this indicator indicates a wellBanks can also use LLP to signal financial strength. We use the one-year-ahead percentage g regulatory capital we compute the deviation of the Total Capital Ratio with respect to of managing regulatory capital we compute the deviation of the Total Capital Ratio with respect to 8 per cent, divided by 8 per cent ( ). A higher value of this indicator indicates a welld by 8 per cent ( cent,). divided A higherbyvalue this ( indicator a well8 per 8 perofvalue cent ).indicator Aindicates higher indicates value of this indicator indicates a wellcapitalized bank. . A higher of this a well-capitalized bank. cent ,capitalized divided bybank. 8 per 8cent ( cent,). divided A higher by 8 perthis indicator indicates a well- value of this indicator indicates a wellpercan cent ( “signaling ). A higher change tostrength. test ( ). We expect a positive capitalized bank. Banks also use LLP signal financial We usehypothesis” the percentage change of to toofone-year-ahead test “signaling hypothesis” ( one-year-ahead ). We expect a positive bank. cansignal also use LLP to signal financial strength. We use the one-year-ahead percentage so useBanks LLP to financial strength. We use the percentage Banks can also use LLP to signal financial strength. use the one-year-ahead percentage change capitalized bank. and ouruse setthe ofWe dependent variables. Banks can also use LLPcorrelation to signal between financial strength. We one-year-ahead percentage correlation between and our set of dependent variables. s can also use LLP to signal financial strength. We use the one-year-ahead percentage Finally, the business cycle could affect borrower’s ability to repay loans. Several empirical Banks also use LLP to (cycle signal financial strength. We use the percentage to “signaling ( could We expect a one-year-ahead positive correlation between of “signaling to test testcan “signaling hypothesis” We expect positive Finally, the business affect ability to arepay loans. Several empirical change of change to of test hypothesis” (hypothesis” ). and Weborrower’s expect). a). positive test “signaling hypothesis” ( to test ). found We expect a positive studies have a negative significant correlation between provisioning and (real) GDP change of “signaling hypothesis” ( ). We expect a positive correlation between ourfound set ofvariables. studies have adependent negative). variables. and significant correlation between provisioning and (real) GDP to test “signaling hypothesis” (setofand We expect a positive between and our set dependent and our of dependent ncorrelation and our set of dependent variables. growth. Asofavariables. consequence of this, to therepay percentage GDP growth isinincluded in our analysis. correlation between and our setaffect dependent variables. Finally, the business cycle could borrower’s ability loans. Several empirical growth. As a consequence of this, the percentage GDP growth is included our analysis. between and our setborrower’s ofofdependent change tovariables. test “signaling hypothesis” ( ability ). We expect a positive Finally, business cycle could affect borrower’s ability to borrower’s repay loans. Severaltoempirical business cycle the could affect ability to repay loans. Several empirical Finally, business cycle could affect repay loans. Several Finally, the business cycle could affect borrower’s ability to repay loans. Several have found athe negative andability significant correlation between provisioning and (real)empirical GDP empirical studies have business cycle affect borrower’s to repay loans. Several empirical studies havestudies found a could negative and significant correlation between provisioning and (real) GDP dly, a the negative and significant correlation between provisioning and (real) GDP correlation between and our set of dependent variables. 3.1. Model specification studies have a negative and significant correlation between provisioning and (real) As consequence of this, the percentage GDP growth included in our analysis. 3.1. Model specification found afound negative and correlation between provisioning and (real) GDPGDP growth. As a consequence ve foundofAs athis, negative anda significant correlation between provisioning and (real) GDP growth. agrowth. consequence of this, percentage GDP growth is included inisour analysis. quence the percentage GDPthe growth issignificant included in our analysis. the cycle could to repay loans. Several empirical growth. As aFinally, consequence ofbusiness this, the percentage GDPaffect growthborrower’s is included in ability our analysis. a consequence of this, the percentage GDP growth is included in our analysis. of this, the percentage isaincluded in our analysis. We specify set of in equations order tothe estimate the determinants of several studies have found aGDP negative significant correlation between provisioning (real)endogenous GDP 3.1. Model specification We specify agrowth set ofand equations order toin estimate determinants of severaland endogenous 3.1. Model specification ation variables related to provisioning. Equations (1) and (2) model the relationship between 3.1. Model specification related toofprovisioning. Equations (1) andgrowth (2) model the relationship total LLP total LLP growth. As variables a consequence this, the percentage GDP is included in ourbetween analysis. specification and the variables: specify a specification set equations in explanatory order to estimate the determinants of several endogenous andofin the explanatory variables: specify We ain3.1. set ofModel equations order to estimate determinants of several endogenous a set ofWe equations order to estimate the determinants ofthe several endogenous We specify a set of equations in order to estimate thethe determinants of several total endogenous variables related to provisioning. Equations (1) and (2) model relationship LLP a set of Equations equations in order to model estimate of relationship several endogenous variables related to 3.1. provisioning. Equations (1)the anddeterminants (2) model the between totalbetween LLP opecify provisioning. (1) andtospecification (2) the relationship between total LLP Model variables related provisioning. Equations (1) and (2) model the relationship between total LLP (1) and the explanatory variables: 5    

(1) provisioning. Equations (1) and (2) model the relationship between total LLP and thetoexplanatory variables: yelated variables: and the explanatory variables: planatory variables: We specify a set of equations in order to estimate the determinants of several endogenous variables (2) We specify a set of equations in order to estimate the determinants (1) of several endogenous (2) (1) (1) relationship between related to provisioning. Equations (1)Equations and (2) model total(1) LLP between and the explanatory variables related to provisioning. (1) the and (2) model the relationship total LLP (1) ourisanalysis is interested in the of the collateralized loans on provisioning variables: (2) Sincevariables: our Since analysis interested in the impact of impact the collateralized loans on provisioning and the explanatory (2) are estimated (2) behavior, two equations with the two different variables separately. (2) behavior, two equations with the two different variables (2) are estimated separately. To obtain a more comprehensive assessment, it be canuseful also be Since our analysis isTo thethe impact of the collateralized loans provisioning obtain ainmore comprehensive assessment, it canonalso to useful model to themodel sub the sub our analysis isimpact interested ininterested the impact of collateralized loans on provisioning nalysis Since is interested in the of the collateralized loans on provisioning Since our analysis is interested in the impact of the collateralized loans on provisioning ofvariables loan loans loss provisions, i.e. Loan Loss Provisions on Bad ( on ) (1) and on behavior, two equations with the two are estimated our with analysis is interested inthethe impact ofcomponents thedifferent collateralized on components of loan loss provisions, i.e. provisioning Loan separately. Loss Provisions on Bad Loans ( Loans ) and behavior, two equations with two different variables are estimated separately. tions the two different are estimated behavior, twovariables equations with the twoseparately. different variables are estimated separately. To obtain a more comprehensive assessment, it can also be useful to model the sub wo equations with the two different variables are estimated separately. Impaired Loans ( ) and their relative determinants: To obtain a more comprehensive assessment, it can also be useful to model the sub more comprehensive it can also (be useful totheir model sub Loans )assessment, and relative To assessment, obtain a Impaired more comprehensive itthe candeterminants: also be useful to model the sub (2) btain a more it can also be Loss useful toBad model subLoans components ofprovisions, loanassessment, loss i.e. provisions, i.e. Loan Provisions onthe(Bad components of comprehensive loani.e.loss Loan Loss Provisions on Loans ) and( on ) and on an loss provisions, Loan Loss Provisions on Bad Loans ( ) and on on components of loan loss provisions, i.e. Loan Loss Provisions Bad Loans ( ) and on sImpaired of loan Loans loss provisions, i.e. Loan Loss Provisions on Bad Loans ( ) and on (3) Impaired Loans ( ) and their relative determinants: (3) ( ) and their relative determinants: ) and theirImpaired relative determinants: Loans ( our) and their relative determinants: Since analysis is interested in the ofimpact of the collateralized loans on behavior, provisioning oans ( ) and their relative determinants: Since our analysis is interested in the impact the collateralized loans on provisioning two behavior, with two equations with the two different variables are estimated separately. (4) (3) equations the two different variables are estimated (4) (3) separately. (3) (3) to model the sub To obtain a more comprehensive assessment,(3)it can also be useful To obtain a more comprehensive assessment, it can also be useful to model the sub components of loan theloan ratioloss of provisions loanProvisions loss provisions on Loans Total Assets. (4) is the ratioisi.e. of on (4) BadonLoans over Total Assets. components of where loan losswhere provisions, Loan BadBad Loans ( over )We and on We (4)Loss (4) variables consider in the non-discretionary behavior part only the explanatory associated with the (4)the explanatory variables associated with the consider in the non-discretionary behavior part only Impaired Loans ( ) and their relative determinants: related provisions sub components, i.e. and in equations (3) and (4); and where is ratio of loan provisions on over Bad Loans over Total We (3) and (4); related provisions subloss components, i.e. and equations and where is loss the ratio of the loan loss on Bad Loans Total Assets. WeinAssets. is the ratio of loan provisions on Badprovisions Loans over Total Assets. We where is the ratio of loanin loss provisions on(6). Bad variables Loans over Total Assets. equations and 116 113 consider inofthe non-discretionary behavior part only the(5) explanatory associated with theWe in equations (5) and (6). e is the ratio loan loss provisions on Bad Loans over Total Assets. We consider in the non-discretionary behavior part only the explanatory variables associated with the n-discretionary consider behavior in part the explanatory the theonly non-discretionary with                      behavior          variables            and        associated                    only              the    JEOD          with  explanatory      -   Vol.3, Issue 1variables (2014) (3) associated relatedsub provisions sub in(3) equations and andthe          components,    only  i.e.                    the          3        in      explanatory      i.e.                                                        part        in the components, non-discretionary behavior variables associated with the (3) related provisions components, and equations and (4);guaranteed, and (4); sub i.e. provisions and part equations (3) and (4); and 3sub components,   T he first variable considers only loans that are totally i.e. that are for the total related i.e. and in equations (3) and (4); and he only loans (3) that are totally guaranteed, that are collateralizedcollateralized for the total amount of theamount of the equations (5)first andvariable (6). considers visions in equations and (4);loans andarei.e. equations in (5)i.e. and (6).  Tand tions (5) sub andincomponents, (6). exposition, while the second one contains that partially guaranteed as well.   (6). while the second one contains loans that are partially guaranteed as well.                                                          in          equations                  exposition,                    (5)                and


e variables related to provisioning. Equations (1) and (2) model the relationship between total LLP We specify a set of equations in order to estimate th We specify a set of equations in order to estimate the determinants of several endogenous the explanatory variables: f andWe variables related to provisioning. Equations (1) and (2) mo Wespecify specifya aset set of equations in order to estimate the determinants of several endogenous of equations in order to estimate the determinants of several endogenous (1) variables related to provisioning. Equations (1) and (2) model the relationship between total LLP and the explanatory variables: variables ariables relatedtotoprovisioning. provisioning. Equations (1) and (2) model the relationship between total LLP riables related Equations (1) and (2) model the relationship between total LLP and the explanatory variables: (1) nd theexplanatory explanatoryvariables: variables: dndthe

(2)

Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence Alessi M.; Di Colli S.; Lopez J.S. (2)

(1) (1) (1) Since our analysis is interested in the impact o e Since our analysis is interested in the impact of the collateralized loans on provisioning behavior, two with the two different variables (2)equations Since our analysis is interested in the impact of the collateralized loans on provisioning e (2) (2) behavior, two equations with the two different variables are estimated separately. To obtain a more comprehensive assessment, y behavior, two equations with the two different variables are estimated separately. Since our analysis is interested in the impact of th To our obtain a assessment, more comprehensive it can also be useful toof model sub i.e. Loan Loss P analysis is interested impact of the tocollateralized on provisioning To obtain a moreSince comprehensive it in canthealso beassessment, useful model the loans subbehavior, o components loanwith lossthe provisions, two equations the two different variables are e Sinceour ouranalysis analysis is interested in the impact of the collateralized loans on provisioning Since is interested in the impact of the collateralized loans on provisioning behavior, two equations with Loss the two different variables are estimated separately. c components of loan loss provisions, i.e. Loan Provisions on Bad Loans and on Impaired Loans and their loss provisions, i.e. Loan Provisions on Bad Loans ( Provisions ) and onon Bad components of loan lossLoss provisions, i.e. Loan Loss ) and andtheir on relative To Loans obtain a( more comprehensive assessment, it c Impaired Loans ( determinants: behavior, ehavior, twoequations equations with the two different variables are estimated separately. havior, two with the two different variables are estimated separately. To obtain a more comprehensive assessment, it can also be useful to model the sub , To obtain obtain more comprehensive assessment, it can also be useful to model the sub To comprehensive assessment, it can also be useful to model the sub Impaired Loansa a( more ) and their relative determinants: components of loan loss provisions, i.e. Loan Loss Provi relative determinants: Impaired Loans ) and their determinants: e components of loan( loss provisions, i.e. relative Loan Loss Provisions on Bad Loans ( ) and on omponents omponentsofofloan loanloss lossprovisions, provisions,i.e. i.e.Loan LoanLoss LossProvisions ProvisionsononBad BadLoans Loans( ( andonon Impaired Loans ( ) )and ) and their relative determinants: smponents Impaired Loans ( ) and their relative determinants: (3) n mpairedLoans Loans( ( andtheir theirrelative relativedeterminants: determinants: mpaired ) )and (3) ) (3) (4) (3) (3) g where is the ratio (4)of loan loss provisio (4) non-discretionary behavior part only the n consider in the where is the ratio of loan loss provisions on Bad Loans over Total Assets.(4) We h consider in the non-discretionary behavior part only the explanatory variables associated with(4) related provisions sub components, i.e. and the where isconsider the ratioinof loan loss provisions o where isratio the of ratio ofloss loan loss provisions on over Bad Loans overin Total Assets. We s where is the loan loss provisions on Bad Loans over Total Assets. We the where is the ratio of loan provisions on Bad Loans Total Assets. We equations (5) and (6). related provisions sub components, i.e. and in equations (3) and (4); andconsider in the non-discretionary behavior part only the ex where is theratio ratio loannon-discretionary lossprovisions provisions Bad Loans over Totalthe Assets. We associated is the ofofthe loan loss ononBad over Total Assets. We d where consider in behavior only explanatory consider in the non-discretionary behavior partLoans only part the explanatory variables the  variables  with                      the  with                related    associated                              provisions                    with                       the non-discretionary behavior part only the explanatory variables associated in equations (5) and (6). related provisions sub components,sub i.e. and 3 onsider onsiderininthe thenon-discretionary non-discretionary behavior part only the explanatory variables associated with the behavior part only the explanatory variables associated with the rnsider   T he first variable considers only loans sub components, i.e. and (3) equations and (4); (3) andand (4); sub components, i.e. and in equations in andthat are totally guaranteed,                                                                            related  components,   related                          provisions            provisions   equations (5) and (6). and in equations (3) and (4); and in equations (5) and (6). i.e elated provisions sub components, i.e. and in equations (3) and (4); and lated provisions sub components, i.e. and in equations (3) and (4); and t 3  The first variable considers only loans exposition, while the second one contains loans that are partially gu in equations (5) guaranteed, and that are totally that are collateralized for the total amount of the in equations (5)(6).andi.e.(6).                                                                                                                         6   inequations equations e exposition,inwhile the second guaranteed as well.            (5)      (5)      one  and      and      contains    (6).      (6).                  loans                    that              are            partially           3

 The   first variable considers only loans that are totally guaranteed, i.e. th            he                        first                            only              loans      6            that             are totally guaranteed, i.e. that are collateralized for the                            considers                                                                                                                    3          T                                  variable total amount of the 3 exposition, while second contains   T he first variable considers only loans that are totally guaranteed, i.e. that are collateralized forthethe totalone amount ofloans the that are partially guarant The The firstvariable variableconsiders considers only loans that are totally guaranteed, i.e. that are collateralized for the total amount of the he only loans that are totally guaranteed, i.e. that are collateralized for the total amount of the   first (5) exposition, while the second one contains loans that are partially guaranteed as well.   6   (5) (5) xposition,while whilethe thesecond second onecontains contains loansthat thatare arepartially partially guaranteed aswell.   well.   position, one loans guaranteed (5) exposition, while the second one containsasloans that 6   are partially guaranteed as well.   (5)   6  6     6  

 

(5) (5)(6)

(6)

(6)

(6) (6)

Where isis the ratio ofloss loss on Impaired Loans over Total Assets. the ratio of provisions on Impaired Loans over Total Assets. isprovisions the ratio ofprovisions loss provisions onon Impaired Loans over Total Assets. Where is the ratio of loss provisions Where the ratio of loss provisions Impaired Loans over Total Assets. Where Where is Where the ratio ofis loss on Impaired Loans over Total Assets. (6) (6) Total Assets. Where is the ratio of loss provisions on Impaired Loans over We are also interested in the determinants of the coverage ratio of Bad Loans We are also interested in the determinants of the coverage ratio of Bad Loans ( and ) and We are also interested in the determinants of the coverage ratio of Bad Loans We are also interested in interested the determinants of determinants the coverage ratio of Bad Loans ( ratio ) of andBad) (Loans We are also also interested in the the determinants We are in of the coverage (and Impaired ) and We are also interested in the determinants of the coverage ratio of Bad Loans ( ) and Impaired Loans ( ). These two indicators are an important source of information for the bank’s Impaired Loans ( ). These two indicators are an important source of information for the bank’s . These two indicators are an important source of information for the bank’s reporting activity. Loans Impaired Loans ( ). These two indicators are an important source of information for the bank’s Where WhereImpaired isisthe the ratio ratioof(of lossprovisions provisions on on Impaired Impaired Loans Loansare over over Total Assets. Assets. source of information for the bank’s Impaired Loans ( loss These two indicators are Loans ).).toThese two indicators anTotal important reporting activity. We decide model these alternative endogenous variables as follows: reporting activity. We decide to model these alternative endogenous variables as follows: Impaired (to ). alternative Thesealternative twothe indicators are an important source of information for the bank’s reporting activity. WeLoans these endogenous variables asfollows: follows: We decide todecide model these endogenous variables We Weare are also also interested interested ininthe themodel determinants determinants ofofthe coverage coverage ratio ratio ofofBad Badas Loans Loans (( ) )and and reporting activity. activity. We We decide decide to to model model these these alternative alternative endogenous variables as follows: reporting reporting activity. We decide to model these alternative endogenous variables as follows:

Impaired ImpairedLoans Loans( ( ).).These Thesetwo twoindicators indicatorsare areananimportant importantsource sourceofofinformation informationfor forthe thebank’s bank’s (7) reporting reportingactivity. activity.We Wedecide decidetotomodel modelthese thesealternative alternativeendogenous endogenousvariables variablesasasfollows: follows:

(7)

(7)

(7) (7)(8)

(8)

(8)

(7) (7)

(8)

in equations, theAsLLP equations, thedependent lagged variable is included as an explanatory variablevariable in the LLP equations, thedependent lagged dependent variable included asvariable an explanatory As in theAs LLP the lagged variable is included as an isexplanatory (8) (8) (8) we in the regression. Ofdue course, due to the construction of the Coverage Ratios, decided inOfthecourse, regression. course, dueparticular to the particular construction of the Coverage Ratios, we decided in the regression. toOfthe particular construction of the Coverage Ratios, we decided As in respectively the LLP equations, the lagged dependent not to include and in the equations. As the LLP equations, lagged dependent variable is included an explanatory variable not to include respectively in the equations. As in in the LLP equations, thethe lagged dependent variable is included as anas variable in the not to include respectively and inand the equations. As in the LLP equations, the lagged variable is included asexplanatory an explanatory variable As Asininthe the LLP LLP equations, equations, the theOf lagged lagged dependent dependent variable variable isisdependent included includedconstruction asasan anexplanatory explanatory variable variable in the regression. Of course, due to the particular in the regression. course, due to the particular of the Coverage Ratios, we decided regression. Of course, due to thedue particular construction of the Coverage Ratios, we decided not include in the regression. course, to the particular construction of the Coverage Ratios, wetodecided ininthe theregression. regression. Of course, course, due duetoOf tothe the particular particular construction construction ofofthe theCoverage Coverage Ratios, Ratios, we wedecided decided (9) not to toOfinclude include respectively and in the equations. (9) not respectively and in (9) not nottotoinclude include respectively respectively and and ininthe the equations. and theequations. equations. in the equations. respectively not to include respectively and (9) (9)(10)

(10)

(10)

Therefore, tenequations sets ten of equations are estimated for two for samples: the full containing sample, containing all Therefore, sets of equations the full sample, all Therefore, ten sets of are estimated forare twoestimated samples: thetwo full samples: sample, all containing (10) (10) banks and the restricted sample, focusing only on cooperative credit banks. banks and the restricted sample, focusing only on cooperative credit banks. banks and the restricted sample, focusing only on cooperative credit banks.

(9) (9)

(10) (10)

Therefore, ten ten sets sets ofof equations equations are estimated estimated for for two two samples: samples: the thefull full sample, sample, containing containingall all 4. Estimation of loan loss provisioning and coverage ratios determinants 4. Estimation of are loan loss provisioning and determinants coverage ratios determinants 4.Therefore, Estimation ofTherefore, loan loss provisioning coverage ratios ten setsonly ofand equations are estimated Therefore, ten sets of equations are estimated for two samples: the full sample, containing all banks banksand andthe therestricted restricted sample, sample, focusing focusing only on on cooperative cooperative credit credit banks. banks. Therefore, ten of equations are only estimated for two samples: the full containing sample, containing all Therefore, ten setstosets ofinvestigate equations aredeterminants estimated for two samples: the full Coverage sample, all banks and banks and the restricted sample, focusing on cooperative credit the restricted sample, focusing only on cooperative Intoand order investigate the determinants of the loan loss provision and banks. Ratios Ratios Intoorder the ofprovision the loan lossCoverage provision and Coverage In banks order investigate the determinants of the loan loss and Ratios banks and the restricted sample, focusing only on cooperative credit banks. 4.4.Estimation Estimation loan loan loss loss provisioning provisioning and and coverage coverage ratios ratios determinants restricted sample, focusing only on cooperative credit banks. behavior in thebanking Italian banking system, we use an approach similar to the one used by used Bouvatier and behavior in thesystem, Italian banking wedeterminants use an approach similar toBouvatier the one and behaviorthe inofof the Italian we use ansystem, approach similar to the one used by andby Bouvatier Lepetit (2008) and Packer and Zhu (2012). Lepetit (2008) and Packer and Zhu (2012). 4. Estimation of loan loss provisioning and coverage ratios determinants Lepetit (2008) and Packer and Zhu (2012). 4. Estimation of loan loss provisioning and coverage 4. toEstimation oftheloan loss provisioning coverage ratios determinants InInorder order toinvestigate investigatethe determinants determinants ofofthe theloan loanand loss lossprovision provisionand and Coverage Coverage Ratios Ratios behavior behavior in in the the Italian Italian banking banking system, system, we we use use an an approach approach similar similar to to the the one one used used by by Bouvatier Bouvatier and and 4.1. Data and descriptive analysis 4.1. Data and descriptive analysis 4.1. Data and descriptive analysis In order order to (2012). investigate the determinants determinants of of the the loan loan loss loss provision provision and and Coverage Coverage Ratios Ratios In to investigate the Lepetit Lepetit(2008) (2008)and andPacker Packer and and Zhu Zhu (2012). In order to investigate the determinants of the loan loss to provision and Coverage Ratios behavior in the Italian banking system, we use an approach similar the one used by Bouvatier and behavior in the Italian banking system, we use an approach similar to the one used by Bouvatier and The consists sample consists of an unbalanced ofpanel Italian banks’ balance sheets and income sample consists ofsystem, an unbalanced ofbalance Italian similar banks’ and balance sheets andbyincome The4. sample ofloan an unbalanced panel of panel Italian banks’ sheets income behavior inThe theofand Italian banking we use an approach to the one used Bouvatier and Estimation loss provisioning and coverage ratios determinants Lepetit (2008) Packer and Zhu (2012). 4.1. 4.1.statements Data Dataand and descriptive descriptive analysis analysis statements from 2006 to 2012. Data are provided by the Italian Banking Association (ABI) balance statements from 2006 to 2012. Data are provided by the Italian Banking Association (ABI) balance Lepetit (2008) and Data Packer and Zhuby(2012). from 2006 to 2012. are provided the Italian Banking Association (ABI) balance Lepetit (2008) and Packer Zhu (2012). sheets database. We preferred toand exclude balance sheets prior to prior 2006 due to in changes in accounting sheets database. We preferred to sheets exclude balance sheets to 2006 due to changes in accounting sheets database. We preferred to exclude balance prior to 2006 due to changes accounting The Thesample sample consists consists of ofdescriptive anan unbalanced unbalanced panel panel ofof Italian Italian banks’ banks’ balance balance sheets sheets and and income income standards implemented that year. In order to focus our attention only on commercial banks, we do we standards implemented that year. In order to focus our attention only on commercial do 4.1. Data and analysis standards implemented that year. In order to focus our attention only on commercial banks, we do In order to descriptive investigate the determinants of the loan loss provision and Coveragebanks, Ratios behavior in the 4.1.2006 Data and analysis statements statements from from 2006 toto2012. 2012. Data Data are are provided provided by bythe the Italian Italian Banking Banking Association (ABI) (ABI) balance balance 4.1.consider Data and descriptive analysis not any other categories other than those (investment trust corporations, consumer not consider any other categories other than those (investment and trust corporations, consumer not consider any other categories other than those (investment andAssociation trustand corporations, consumer Italian banking system, we use an approach similar to the one used by Bouvatier and Lepetit (2008) and sheets sheets database. database. We We preferred preferred exclude exclude balance balance sheets sheets prior prior to to2006 2006 due due to to changes changes in inaccounting accounting credit and finance companies). We also delete banks with less than four consecutive time series credit andtoto finance We also delete banks with less than four time series credit and finance companies). We companies). also delete banks with less than four consecutive timeconsecutive series The sample consists of an unbalanced panel of Italian banks’ balance sheets and income The sample consists of an unbalanced panel of Italian banks’ balance sheets and income standards standards implemented implemented that that year. year. In In order order to to focus focus our our attention attention only only on on commercial commercial banks, banks, we we do do observations, in explore order toconsists explore inofway a robust way the way phenomena from not only a cross-sectional, but observations, in order explore in aphenomena robust the phenomena from not onlybalance abut cross-sectional, but Packer and Zhu (2012). observations, in The order to in a robust the from notof only a cross-sectional, sample an unbalanced panel Italian banks’ sheets and balance income statements from 2006 tothan 2012. Data are provided by the the Italian Banking Association (ABI) not notalso consider consider any anyaother other categories other other than those those (investment (investment and and trust corporations, corporations, consumer consumer also dynamic point of view. Outliers were excluded bytrust eliminating theobservations bank’s for (ABI) statements 2006 to 2012. Data are provided by the Italian Banking Association alsocategories afrom dynamic point of view. Outliers were excluded by eliminating the observations bank’s observations for balance a dynamic point of view. Outliers were excluded by eliminating bank’s for statements from 2006 to 2012. Data are provided by the Italian Banking Association (ABI) balance credit credit and finance finance companies). companies). We We also also delete delete banks with with less less than than four consecutive consecutive time time series series sheets database. We preferred to exclude balance sheets prior tosample, 2006 dueyear to by changes inby accounting that year. Table shows the number of banks in four the final sample, divided by andyear by and year.2the Table 2preferred shows thebanks number of banks present in the finalto divided sheets database. We to exclude balance sheets prior 2006 due to changes in accounting thatand year. Table 2that shows number of banks present inpresent the final sample, divided year and by sheets database. We preferred to exclude balance sheets prior toby2006 due to changes in accounting observations, observations, ininorder order to toexplore explore in inaarobust robust way way the thephenomena phenomena from from not not only only afinal across-sectional, cross-sectional, but butacovers (shareholders/cooperative credit banks). However, the sample covers significant category (shareholders/cooperative credit banks). However, the final sample a significant standards implemented that year. In order to focus our attention only on commercial banks, we do category category (shareholders/cooperative credit banks). However, the final sample covers a significant standards implemented that year. In order to focus our attention only on commercial banks, we standards implemented that year. order to focus our attention on commercial we do do also alsopart aadynamic dynamic point point ofof view. Outliers Outliers were were excluded excluded by by eliminating eliminating the bank’s bank’s observations for average, part of the Italian banking system. CCBsIn represent a significant part ofobservations our sample. On 80 banks, part ofview. the Italian banking system. CCBs represent athe significant partOn ofonly our for sample. On average, 80consumer of the Italian banking system. CCBs represent a significant part of our sample. average, 80 not consider any other categories other than those (investment and trust corporations, not consider any other categories other than those (investment and trust corporations, consumer that thatper year. year. Table Table 2 2 shows shows the the number number of of banks banks present present in in the the final final sample, sample, divided divided by by year year and and by by per cent of banks in our dataset are cooperative banks. per cent of banks in our dataset are cooperative banks. not consider any other categories other than those (investment and trust corporations, consumer cent ofcredit banks and in ourfinance dataset are cooperative banks. companies). We alsothe delete banks with less than four four consecutive consecutive time time series credit and companies). We also delete less than category category(shareholders/cooperative (shareholders/cooperative credit creditbanks). banks).However, However, the final finalbanks sample samplewith covers covers aasignificant significant credit and finance finance companies). We also delete banks with less than not fouronly consecutive time series series observations, in order to explore in a robust way the phenomena from a cross-sectional, but ABLEbanking 2. ITALIAN BANKING SYSTEM SAMPLE T ABLE system. 2.system. Iin TALIAN BANKING SYSTEMain SAMPLE part partTofABLE ofthe the2.Italian Italian banking CCBs CCBs represent represent asignificant significant part partofof our ourphenomena sample. sample.On Onaverage, average, 80 80only a cross-sectional, but observations, order to explore a robust way the from not IT TALIAN BANKING SYSTEM SAMPLE observations, in order to explore in a robust way the phenomena from not only a cross-sectional, but alsoinin dynamic point of view. view. Outliers were were excluded by eliminating eliminating theFull bank’s observations for per percent centofofbanks banks our dataset datasetare are cooperative cooperative banks. banks. also aaour dynamic point of Outliers excluded by the bank’s Year CCB Other banks Full sample Year CCB Other banks sample observations for Other banks Full sample 117 114 also a dynamic of view. Outliers were excluded by eliminating the bank’s observations 133 496 133 in the final 496 that year. year. Table point 2 shows shows the363 number of banks present sample, divided by year year and and for by 133banks 496 sample, that Table 2 the number of present in the final divided by by JEOD Vol.3, Issue 1 (2014) 140 458 318 140 in the final 458 that year. Table 2 shows the number of banks present sample, divided by year and by 140 458 category (shareholders/cooperative credit banks). However, the final sample covers a significant category (shareholders/cooperative credit banks). the final covers a significant 7   However, category credit thepart finalof sample sample covers significant 7   CCB CCB Other Other banks banks7   banks). Full Fullsample sample part of of   the the(shareholders/cooperative Italian banking system. system. CCBs represent However, significant our sample. sample. Onaaverage, average, 80  part Italian banking CCBs represent aa significant part of our On 80

Year CCB 2006 363 2006 2006 363 2007BANKING TT ABLE ABLE 2.2.ITALIAN ITALIAN BANKING SYSTEM SAMPLE SAMPLE318 2007SYSTEM 318 2007 Year Year


Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence Alessi M.; Di Colli S.; Lopez J.S.

4.1. Data and descriptive analysis The sample consists of an unbalanced panel of Italian banks’ balance sheets and income statements from 2006 to 2012. Data are provided by the Italian Banking Association (ABI) balance sheets database. We preferred to exclude balance sheets prior to 2006 due to changes in accounting standards implemented that year. In order to focus our attention only on commercial banks, we do not consider any other categories other than those (investment and trust corporations, consumer credit and finance companies). We also delete banks with less than four consecutive time series observations, in order to explore in a robust way the phenomena from not only a cross-sectional, but also a dynamic point of view. Outliers were excluded by eliminating the bank’s observations for that year. Table 2 shows the number of banks present in the final sample, divided by year and by category (shareholders/cooperative credit banks). However, the final sample covers a significant part of the Italian banking system. CCBs represent a significant part of our sample. On average, 80 per cent of banks in our dataset are cooperative banks. Table 2. Italian banking system sample TABLE 2. ITALIAN BANKING SYSTEM SAMPLE CCB SAMPLE TYear ABLE 2. ITALIAN BANKING SYSTEM

Other banks 2006 363 133 Year CCB Other140 banks 318 2007 2006 363 133 367 141 2008 318 140 2007 367 141 2009 367 141 2008 360 132 2010 367 141 360 122 2009 2011 360 132 2010 2012 325 115 360 122 2011 Total Obs. 2012 325 115 Total Obs. Source: Italian Banking Association Balance sheets and Financial Statement database Source: Italian Banking Association Balance sheets and Financial Statement database

Full sample 496 Full458 sample 496 508 458 508 508 492 508 482 492 440 482 3386 440 3386

Source: Italian Banking Association Balance sheets and Financial Statement database

Descriptive statistics for the key variables are presented in Table 3. Descriptive statistics for the key variables are presented in Table 3. Descriptive statistics for the key variables are presented in Table 3. TABLE 3. SUMMARY STATISTICS OF MAIN VARIABLES

Table 3. Summary statistics of main variables Full Sample TABLE 3. SUMMARY STATISTICS OF MAIN VARIABLES

Standard 95% Confidence Mean Mean Full Sample Error Interval Standard 95% Confidence LLPtot 0.01955 0.00032 0.01891 0.02019 0.01776 Mean Mean Error 0.01613 0.00028 0.01557Interval0.01669 0.01473 LLPtot 0.01955 0.00032 0.01891 0.02019 0.01776 0.00304 0.00006 0.00291 0.00316 0.00270 0.01613 0.00028 0.01557 0.01669 0.01473 0.51899 0.00279 0.51350 0.52448 0.50261 0.00304 0.00006 0.00291 0.00316 0.00270 0.11764 0.00162 0.11445 0.12084 0.09193 0.51899 0.00279 0.51350 0.52448 0.50261 NPLtot 0.10029 0.00114 0.09804 0.10254 0.10339 0.11764 0.00162 0.11445 0.12084 0.09193 BL 0.04653 0.00072 0.04511 0.04795 0.04592 NPLtot 0.10029 0.00114 0.09804 0.10254 0.10339 IL 0.04203 0.00054 0.04095 0.04311 0.04618 BL 0.04653 0.00072 0.04511 0.04795 0.04592 GUA1 0.72745 0.00268 0.72219 0.73271 0.76076 IL 0.04203 0.00054 0.04095 0.04311 0.04618 GUA2 0.76373 0.00262 0.75859 0.76887 0.79555 GUA1 0.72745 0.00268 0.72219 0.73271 0.76076 LOAN 0.69049 0.00225 0.68606 0.69491 0.67841 GUA2 0.76373 0.00262 0.75859 0.76887 0.79555 CAP -0.00410 0.00321 -0.03102 0.00922 3.15813 LOAN 0.69049 0.00225 0.68606 0.69491 0.67841 ER 0.00647 0.00014 0.00618 0.00676 0.00682 CAP -0.00410 0.00321 -0.03102 0.00922 3.15813 SIGN -0.36025 0.13904 -0.63288 -0.08761 -0.36103 ER 0.00647 0.00014 0.00618 0.00676 0.00682 SIGN -0.63288 -0.36103 Note: Mean and -0.36025 standard errors 0.13904 are calculated over the sample-0.08761 period 2006-2012

Note: Mean and standard errors are calculated over the sample period 2006-2012

Total 1.95period per cent of the Note: Mean and Loan standardLoss errorsProvisions are calculatedaverages over the sample 2006-2012

CCB Standard Error CCB Standard 0.00033 Error 0.00029 0.00033 0.00007 0.00029 0.00347 0.00007 0.00165 0.00347 0.00131 0.00165 0.00083 0.00131 0.00065 0.00083 0.00279 0.00065 0.00272 0.00279 0.00246 0.00272 1.33322 0.00246 0.00014 1.33322 0.15825 0.00014 0.15825

95% Confidence Interval 95% Confidence 0.01710 0.01841 0.01414Interval0.01531 0.01710 0.01841 0.00256 0.00284 0.01414 0.01531 0.49580 0.50943 0.00256 0.00284 0.08869 0.09518 0.49580 0.50943 0.10081 0.10597 0.08869 0.09518 0.04427 0.04756 0.10081 0.10597 0.04489 0.04747 0.04427 0.04756 0.75528 0.76625 0.04489 0.04747 0.79020 0.80090 0.75528 0.76625 0.67358 0.68324 0.79020 0.80090 0.53320 5.75250 0.67358 0.68324 0.00653 0.00710 0.53320 5.75250 -0.67139 -0.05066 0.00653 0.00710 -0.67139 -0.05066

Total Assets, while LLP on Bad Loans (BL) and on Impaired Loans (IL) are respectively 1.61 per cent and 0.30 per cent of the Total Total Loan Loss Provisions averages 1.95 per cent of the Total Assets, while LLP on Bad Assets. On average, the coverage ratio of BL is 51.89 per cent, while that of IL is 11.76 per cent. Loans and on Provisions Impaired Loans (IL) areper respectively 1.61 perAssets, cent and 0.30 peron cent the Total Total(BL) Loan Loss averages 1.95 cent of the while Badof Loans (BL) CCBs have lower coverage ratios, respectively 50.26 perTotal cent and 9.19 perLLP cent. The Bad Loans Assets. On average, the coverage ratio of BL is 51.89 per cent, while that of IL is 11.76 per cent. ratio per (IL) cent are for respectively the whole sample and 4.59 per0.30 centper if we to and onaverages Impaired4.6 Loans 1.61 per cent and centrestrict of the our Totalattention Assets. On CCBs have lower coverage ratios, respectively 50.26 per cent and 9.19 per cent. The Bad Loans CCBs, the IL ratio is lower (4.20%). ratio averages 4.6 per cent for the whole sample and 4.59 per cent if we restrict our attention to Loans represent the bank’s main assets (69.04% of Total Assets). A substantial part of bank CCBs, the IL ratio is lower (4.20%). Loans is guaranteed. Approximately 76.37 per115 cent of Loans are guaranteed, 72.74 per cent of 118 Loans represent the bank’s main assets (69.04% of Total Assets). A substantial part of bank which are totally guaranteed. These percentages JEOD - Vol.3,increase Issue 1 (2014)in the case of CCBs. Cooperatives banks Loans is guaranteed. Approximately 76.37 per cent of Loans are guaranteed, 72.74 per cent of tend to collateralize their assets more and subsequently reduce credit risk. which are totally guaranteed. These percentages increase in the case of CCBs. Cooperatives banks


IL 0.04203 0.00054 0.04095 0.04311 0.04618 0.00065 0.04489 GUA1 0.72745 0.00268 0.72219 0.73271 0.76076 0.00279 0.75528 GUA2 0.76373 0.00262 0.75859 0.76887 0.79555 0.00272 0.79020 LOAN 0.69049 0.00225 0.68606 0.69491 0.67841 0.00246 0.67358 CAP -0.00410 0.00321 -0.03102 0.00922 3.15813 1.33322 0.53320 Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence ER 0.00647 0.00014 0.00618 0.00676 0.00682 0.00014 0.00653 Alessi M.; Di Colli S.; Lopez J.S. SIGN -0.36025 0.13904 -0.63288 -0.08761 -0.36103 0.15825 -0.67139 Note: Mean and standard errors are calculated over the sample period 2006-2012

0.04747 0.76625 0.80090 0.68324 5.75250 0.00710 -0.05066

Total Loan Loss Provisions averages 1.95 per cent of the Total Assets, while LLP on Bad average, the coverage ratio of BL is 51.89 cent, while that IL cent is 11.76 per cent. CCBs lower Loans (BL) and on Impaired Loans (IL) per are respectively 1.61ofper and 0.30 per cent of have the Total Assets. On average, the coverage ratio BL9.19 is 51.89 per cent, while thatratio of ILaverages is 11.764.6per coverage ratios, respectively 50.26 per centofand per cent. The Bad Loans percent. cent CCBs have lower coverage ratios, respectively 50.26 per cent and 9.19 per cent. The Bad Loans for the whole sample and 4.59 per cent if we restrict our attention to CCBs, the IL ratio is lower (4.20%). ratio averages 4.6 per cent for the whole sample and 4.59 per cent if we restrict our attention to Loans represent the bank’s main assets (69.04% of Total Assets). A substantial part of bank Loans is CCBs, the IL ratio is lower (4.20%). guaranteed. centassets of Loans are guaranteed, 72.74 per cent of which LoansApproximately represent the 76.37 bank’sper main (69.04% of Total Assets). A substantial partare of totally bank guaranteed. These percentages increase in76.37 the case CCBs. banks tend 72.74 to collateralize Loans is guaranteed. Approximately perofcent of Cooperatives Loans are guaranteed, per centtheir of whichmore are and totally guaranteed. Thesecredit percentages increase in the case of CCBs. Cooperatives banks assets subsequently reduce risk. tend to collateralize their assets more and subsequently reduce credit risk.

4.2. Empirical results 4.2. Empirical results TheThe empirical analysis is based on the of generalized method of moments (GMM) using empirical analysis is based on estimation the estimation of generalized method of moments (GMM) first differences (see Arellano and Bond, 1991) and orthogonal deviations (Arellano and Bover, 1995) using first differences (see Arellano and Bond 1991) and orthogonal deviations (Arellano and Bover 1995) regressions. are in difference to control for unobservable bank’s effects. regressions. VariablesVariables are in difference to control for unobservable bank’s specific effects.specific Estimations are Estimations are performed in order to obtain robust standard errors. Results for equations 1-6 are performed in order to obtain robust standard errors. Results for equations 1-6 are reported in Tables 4 and 5. reported in Tables 4 and 5. Table 4. Estimation of LLP determinant - full sample - Arellano - Bond TABLE 4. ESTIMATION OF LLP DETERMINANT - FULL SAMPLE - ARELLANO - BOND

Constant LLPtot (-1)

(1) -0.01326 (-4.18) 0.43177 (5.55)

 

LLPtot

Dependent variable (2) -0.01308 -(4.14) 0.43186 (5.54)

(3) -0.01153 (-6.42)

8  

0.21649 (1.38)

(-1)

(4) -0.01162 (-6.48)

BL

0.10702 (9.71)

0.10707 (9.72)

IL 0.00501 (1.03)

GUA1 GUA2 LOAN CAP ER SIGN

-0.00030 (-0.45) 0.02258 (5.74) -0.00000 (-0.68) -0.23474 (-5.77) 0.00000 (-0.03) 0.00161

(0.58)

0.00505 (1.04)

-0.00053 (-0.86) 0.02258 (5.77) -0.00000 (-0.68) -0.23478 (-5.76) 0.00000 (-0.03) 0.00160

(0.57)

0.26916 (7.43)

0.26857 (7.43)

-0.00400 (-0.48)

-0.00406 (-0.48)

-0.00114 (-2.01) 0.01886 (9.60) -0.00000 (-0.06) -0.06059 (-1.99) 0.00000 (0.96) -0.00341

(-1.49)

(6) -0.00078 (-0.88)

0.16758 (1.04)

0.16769 (1.04)

0.04877 (9.35)

0.04875 (9.37)

-0.00162 (-0.60) -0.00013 (-0.35)

-0.00162 (-0.60)

0.21712 (1.39)

(-1) NPLtot

(5) -0.0008 (-0.91)

-0.00097 (-2.03) 0.01888 (9.62) -0.00000 (-0.07) -0.06048 (-1.99) 0.00000 (0.99) -0.00340

(-1.49)

0.00240 (2.16) 0.00000 (6.01) -0.06219 (-5.59) -0.00000 (-1.37) -0.00012

(-0.11)

-0.00016 (-0.46) 0.00240 (2.15) 0.00000 (5.99) -0.06226 (-5.59) -0.00000 (-1.36) -0.00011

(-0.11)

Note: t – statistics in brackets. Arellano and Bond GMM two-step estimation. Lagged explanatory variables have been used as Note: t – statistics in brackets. Arellano and Bond GMM two-step estimation. Lagged explanatory variables have been used as instruments for differenced equations estimations

instruments for differenced equations estimations

Concerning the estimation results for equations (1) and (2), we find that total LLP are Concerning the estimation results forofequations (2), we find that total LLP significantly significantly correlated with the stock NPL. As(1)weand expected, the coefficients areare positive, and indicate that the cyclicality of Non Performing Loans influences provisioning via backward induction. 119 116 TABLE 5. ESTIMATION OF LLP DETERMINANTS -JEOD FULL SAMPLE ARELLANO - BOVER - Vol.3, Issue 1-(2014) Dependent variable


ER

-0.23474 -0.23478 -0.06059 -0.06048 -0.06219 (-5.77) (-5.76) (-1.99) (-1.99) (-5.59) 0.00000 0.00000 0.00000 0.00000 -0.00000 (-0.03) (-0.03) (0.96) (0.99) (-1.37) 0.00161 0.00160 -0.00341 -0.00340 -0.00012 Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence

SIGN

(0.58)

S.; Lopez J.S. (0.57) Alessi M.; Di Colli (-1.49)

(-1.49)

(-0.11)

-0.06226 (-5.59) -0.00000 (-1.36) -0.00011

(-0.11)

Note: t – statistics in brackets. Arellano and Bond GMM two-step estimation. Lagged explanatory variables have been used as instruments for differenced equations estimations

Concerning the estimation results for equations (1) and (2), we find that total LLP are correlated withcorrelated the stock with of NPL. As weofexpected, are coefficients positive, andare indicate thatand the significantly the stock NPL. Asthe wecoefficients expected, the positive, indicate of that thePerforming cyclicalityLoans of Non Performing Loansviainfluences provisioning via backward cyclicality Non influences provisioning backward induction. induction. Table 5. Estimation of LLP determinants - full sample - Arellano - Bover TABLE 5. ESTIMATION OF LLP DETERMINANTS - FULL SAMPLE - ARELLANO - BOVER Dependent variable

LLPtot Constant LLPtot (-1)

(1) -0.01736 (-5.82) 0.74947 (10.77)

(2) -0.01726 (-5.83) .074960 (10.80)

(-1)

(3) -0.01369 (-8.34)

(4) -0.01389 (-8.56)

0.44907 (3.62)

0.45019 (3.62)

(-1) NPLtot BL

0.08212 (7.32)

0.08207 (7.35)

IL 0.00220 (0.37)

GUA1

 

GUA2

-0.00032 (-0.36)

0.00219 (0.37)

0.22182 (6.73)

0.22102 (6.73)

-0.00885 (-0.99)

-0.00903 (-1.01)

-0.00112 9   (-1.53)

(5) 0.00091 (0.33)

(6) -0.00163 (-1.81)

0.41470 (5.08)

0.48788 (8.83)

0.04904 (8.89)

0.04919 (8.98)

-0.00317 (-1.16) -0.00032 (-0.77)

-0.00326 (-1.17)

-0.00042 -0.00083 -0.00020 (-0.52) (-1.33) (-0.47) LOAN 0.02437 0.02437 0.02076 0.02080 0.00248 0.00256 GUA1 -0.00032 -0.00112 -0.00032 (6.11) (6.12) (10.59) (10.64) (-0.77) (2.19) (2.30) (-0.36) (-1.53) CAPGUA2 -9.22e-06 -9.22e-06 -5.25e-06 -5.24e-06 3.69e-06 3.72e-06 -0.00042 -0.00083 -0.00020 GUA1 (-1.93) (-1.94) (-1.45) (5.20)(-0.47) (4.13) (-0.52) (-1.33) (-1.45) GUA1 0.02437 0.02437 0.02076 0.00256 ER LOAN -0.21653 -0.21702 -0.09206 0.02080 -0.092120.00248 -0.06542 -0.05916 GUA1 -0.00032 -0.00112 -0.00032 GUA2 (6.11) (6.12) (10.59)(-2.58) (10.64) (-2.57) (2.19) (-4.56) (-4.56) (-4.73)(2.30) (-4.50) GUA2 (-0.36) -9.22e-06 (-1.53) -5.25e-06 CAP -9.22e-06 -5.24e-06(-0.77) 3.72e-06 1.31e-06 1.31e-06 4.36e-06 4.63e-063.69e-06 3.83e-06 -3.31e-06 GUA2 SIGN -0.00042 -0.00083 -0.00020 LOAN (-1.93) (-1.94) (-1.45) (-1.45) (5.20) (4.13) LOAN (0.12) (-0.52) -0.21702 (0.12) (0.64) (-1.13) (-0.91) (-1.33) ER -0.21653 -0.09206 -0.09212 (0.69) -0.06542(-0.47) -0.05916 LOAN 0.02437 -0.00395 0.02437 0.02076 0.02080 0.00248 0.00256 -0.00394 -0.00650 -0.00651 -0.00038 -0.00063 CAP (-4.56) (-4.56) (-2.58) (-2.57) (-4.73) (-4.50) CAP (6.11) (10.59) (10.64) SIGN 1.31e-06 1.31e-06 4.36e-06 4.63e-06(2.19) -3.31e-06 (-1.32) (6.12) (-1.32) (-3.03) (-3.03) 3.83e-06(2.30)(-0.35) (-0.58) CAP -9.22e-06 -5.25e-06 -5.24e-06 3.69e-06 ER variables (0.12)-9.22e-06 (0.12) (0.64) (0.69) (-1.13)3.72e-06 (-0.91) Note: t – statistics in brackets. Arellano and Bover two-step estimation with orthogonal deviations. Lagged explanatory ER (-1.93) (-1.94) (-1.45) (-1.45) (5.20) (4.13) -0.00395 -0.00394 -0.00650 -0.00651 -0.00038 -0.00063 have tbeen used as instruments for differenced equations estimations ER Note: -0.21653 -0.21702 -0.09206 -0.09212 with -0.06542 -0.05916Lagged – statistics in brackets. Arellano and Bover two-step estimation orthogonal deviations. SIGN variables (-1.32) (-1.32) (-3.03) (-3.03) (-0.35) (-0.58)explanatory

-0.00032 -0.00032 (-0.36) (-0.36)

-0( 0. ( -9.

0.02437 0.02437 (6.11) (6.11) -9.22e-06 -9.22e-06 (-1.93) (-1.93) ( -0.21653 -0.21653 -0(-4.56) (-4.56) ( 1.31e-06 SIGN 1.31e-06 1.1 (-4.56) (-4.56) (-2.58) (-2.57) (-4.73) (-4.50) (0.12) Note: t – statistics in brackets. for Arellano and Bover two-step estimation with orthogonal deviations. Lagged explanatory variables instruments differenced equations (0.12) ( SIGNhave been used as 1.31e-06 1.31e-06 4.36e-06 estimations 4.63e-06 3.83e-06 -3.31e-06 -0.00395 have been used(0.12) as instruments for (0.12) differenced equations estimations -0.00395 -0(0.64) (-1.13) (-0.91) The coefficient associated to Loan to Assets(0.69) ratio ( ) is positive and significant at the 1 (-1.32) (-1.32) ( -0.00395 -0.00394 -0.00650 -0.00651 -0.00038 -0.00063 Note:t trisk statistics brackets.Arellano Arellanoand and per cent The level. This finding suggests that Italian banks make higher provisions when credit is Note: – –statistics ininbrackets. (-1.32) (-1.32) (-3.03) (-3.03) (-0.35) (-0.58) coefficient associated to Loan to Assets ratio ( ) is positive and significant at the 1 have havebeen beenused usedasasinstruments instrumentsfor fordifferenced differencede Note: thigher, – statistics consistent in brackets. Arellano andboth Bover two-step estimation with orthogonal deviations. explanatory variablesstudies. Total LLP the standard accounting and previous per cent level. Thiswith finding suggests that Italian banks makeprinciples higherLagged provisions when credit risk is have been used as instruments for differenced equations estimations

The coefficient associated to Loan to Assets ratio is positive and significant at the 1 per cent level. This suggests that Italian management banks make higher provisions when credit risk isrelated higher, consistent seems notfinding to be affected by capital In fact, the coefficient higher, consistent with both the standard accountingpurposes. principles and previous studies. Total LLP to capital Thecoefficient coefficientassociated associatedtoto The with both the standard accounting principles and previous studies. Total LLP seems not to be affected by This finding sugge seems not be significant affected by capital management In fact, the coefficient related to capital adequacy is tonot and very close topurposes. zero. As regards the income smoothing hypothesis, per centlevel. level. per cent This finding sugges The coefficient associated to Loan to Assets ratio ( ) is positive and significant at the 1 adequacy is not and close tobanks zero. tend Asrelated regards income smoothing hypothesis, higher, consistent withboth boththe thes results appear to significant bepurposes. inconsistent. Italian toprovisions reduce loan loss when earnings management In very fact, thebanks coefficient tothecapital adequacy is not andconsistent very with higher, per capital cent level. This finding suggests that Italian make higher when creditprovisions risk is significant results appear to be inconsistent. Italian banks tend to reduce loan loss provisions when earningsbyseems nottotobe beaffected affectedby bycapital capita before taxes and loan loss provisions increase, confirming the cyclicality suggested the nonseems not higher, consistent with both the standard accounting principles and previous studies. Total LLP closebefore to zero. As and regards smoothing results appearsuggested to be inconsistent. banks taxes loan the lossincome provisions increase, hypothesis, confirming the cyclicality by the non- Italian adequacy notsignificant significantand andve v adequacy isisnot seemsdiscretionary not to be affected by capital management purposes.fluctuation In fact, the coefficient related to capital variables. Otherwise, economic and business cycles seem to not affect total discretionary variables. Otherwise, economic fluctuation and business cycles seem to not affect total tend to reduce loan loss provisions when earnings before taxes and loan loss provisions increase, confirming results appear to be inconsistent. results appear to be inconsistent. I adequacy is not significant andwith very the closesignaling to zero. As regards theCyclicality income smoothing hypothesis, provisioning, hypothesis. Total LLPtoappears be driven provisioning,together together with the signaling hypothesis. Cyclicality of TotalofLLP appears be driventobefore taxes andloan loanloss lossprovisi provis before taxes the cyclicality suggested by the non-discretionary variables. Otherwise, economic fluctuation and business results appear to be inconsistent. Italian banks tend to reduce loan loss provisions when earnings primarily specificmicroeconomic microeconomic factors, the macroeconomic situation does and primarilybybythe the bank’s bank’s specific factors, while while the macroeconomic situation does discretionary variables.Otherwise, Otherwise discretionary variables. before taxes and loan loss provisions increase, confirming the cyclicality suggested by the noncycles seem not totalrole. provisioning, together with the signaling hypothesis. Cyclicality of Total LLP together with the sig seem playaffect relevant not not seem tototoplay aarelevant role. provisioning, provisioning, together with the sig discretionary variables. Otherwise, economic fluctuation and business cycles seem to not affect total The results for equations (3) and (4) are quite similar, but differ in some key aspects. Bad primarily bythe thebank’s bank’sspecific specificm results equations (3) (4) are quite similar, differ indriven some the keymacroeconomic aspects. appearsThe to be driven primarily theand bank’s specific microeconomic factors, while primarilyBad by provisioning, together withfor the signalingbyhypothesis. Cyclicality of Total LLP but appears to be Loans ( ) and forward looking differences ( ), together with the Loans to Assets ratio not not seem to play a relevant role. seemratio to play a relevant role. primarily by (the bank’s specific factors, does Loans )not andseem forward looking differences ( the macroeconomic ), togethersituation with the Loans to Assets situation does to microeconomic play a relevant role. while ( ) have signs equal to the ones estimated for equations (1) and (2). The non-discretionary The results for equations (3 The results for equations (3) not seem to play a relevant role. ( The ) have signs to the ones estimated equations (1)some and (2).with Thea non-discretionary component seems to equal also be(3) affected byare the amount offor guarantees on in loans. Banks higher results for equations and (4) quite similar, but indiffer key aspects. Bad Loans Loans Loans( ( )and andforward forwardlooking looking The results for equations (3) and (4) are quite similar, but differ some key aspects. Bad ) component to also beloans affected by or thetotally) amount guarantees on loans. Banks with a higher percentageseems of collateralized (partially are of willing to set lower provisions on Bad havesigns signsequal equaltotothe the ) )have have looking , to together with Loans tolower Assets Loans ( and) forward and forward lookingdifferences differences ( with the the Loans to Assets ratio ratio ( ( Loans, due the fact that they are(partially less exposed credit are default and totothe (expected) percentage of tocollateralized loans or), together totally) willing sethigher provisions on Bad component seems to also be affe component to also be affec ( signs) equal have signs equal to the ones estimated for equations (1) (2). and (2). The non-discretionarycomponent ones estimated equations (1)interest, and Theand non-discretionary seems seems to collateralized recovery The contribution offorearnings taxes loan loss is not (expected) Loans, duetorate. tothe the that by they lessbefore tooncredit andprovisions the higher percentage loans percentage ofofcollateralized loans component seems to also befact affected the are amount ofexposed guarantees loans.default Banks with a to higher trivial. The associated to this variable is negative and significant for both Arellano-Bond Loans, due thefact factthat thatthey theya recovery rate.coefficient The contribution taxes and loan loss provisions isdue nottotothe Loans, percentage of collateralized loans (partiallyofor earnings totally) arebefore willinginterest, to set lower provisions on Bad and Arellano-Bover estimations. Also in this case, the variation of GDP is not significant, nor is the recovery rate. The contribution o recovery rate. The contribution of Loans, due toThe the coefficient fact that they are less exposed credit default and toand the significant higher (expected) trivial. associated to this to variable is negative for both Arellano-Bond Signaling variable. 120 117 trivial.The The coefficientassociated associatedto trivial. recovery The contribution of earningsAlso before interest, taxes and loan loss provisions is not and rate. Arellano-Bover estimations. in this case, the variation of GDP is not significant, norArellano-Bover is thecoefficientestimations. Regressions for equations (5) and (6) suggest that the1 (2014) non-discretionary behavior component and JEOD - Vol.3, Issue and Arellano-Bover estimations. AA trivial. The coefficient associated to this variable is negative and significant for both Arellano-Bond Signaling variable. is relevant for provisions associated to Impaired Loans. However, the forward looking indicator Signaling variable. Signaling variable. and Arellano-Bover estimations. Also in this case, the variation of GDP is not significant, nor is the ( Regressions ) doesfor not equations seem to affect variable, is also the case with collateralized (5) the anddependent (6) suggest thatasthe non-discretionary behavior component Regressions for equations (5


coefficient to Loan to Assets ratio ( ) isfact, positive significant at the seems not1to affected by capital manage seems not The to be affectedassociated by capital management purposes. In theand coefficient related tobecapital per cent level. This suggestsand thatvery Italian ba The coefficient associated tolevel. Loan This to Assets ratio ( ) is Italian positivebanks and significant at the 1 per cent finding suggests that make higher provisions when credit risk isis finding adequacy not significant close adequacy is not significant and very close to zero. As regards the income smoothing hypothesis, higher, consistent with both the standard accounti per cent level. This finding suggests that Italian banks make higher provisions when credit risk is higher, consistent with both the standard accounting principles and previous studies. results Total LLP appear to be inconsistent. Italian ba results appear to be accounting inconsistent. Italian banks tendstudies. to In reduce loan loss provisions when earnings seems not betaxes affected capital pu higher, consistent with both thenot standard principles and previous Totalthe LLP seems to be affected by capital management purposes. fact, coefficient related totocapital before andby loan loss management provisions incr adequacy is not significant and very close to zero. beforeby taxes and loan loss provisions increase, confirming the cyclicality suggested by the nonseems not to be affected capital management purposes. In fact, the coefficient related to capital adequacy is not significant and very to zero. As inregards the income smoothing discretionary variables. Otherwise, economi Loan Loss Provisioning andclose Relationship Banking Italy: Practices and Empirical Evidence hypothesis, results appear to be inconsistent. banks tend adequacy is not significant and appear very close to inconsistent. zero. As regards the income smoothing hypothesis, discretionary variables. Otherwise, economic business cycles seem to not affect total withItalian Alessi M.;fluctuation Di Colli S.; reduce Lopezand J.S. results to be Italian banks tend to loan loss provisions when earnings provisioning, together the signaling hy before taxes and loan loss provisions increase, co results appear to beprovisioning, inconsistent. Italian tend to reduce loan loss provisions when earnings before taxes andbanks loan lossthe provisions increase, confirming the cyclicality suggested by the nontogether with signaling hypothesis. Cyclicality of Total LLP appears to be driven primarily by the bank’s specific microecon discretionary variables. fluctua before taxes and loan loss provisions variables. confirming economic the cyclicality suggested by the nondiscretionary Otherwise, fluctuation and business seem to not not affect total seem to playOtherwise, a relevant economic role. primarily by the increase, bank’s specific microeconomic factors, while cycles the macroeconomic situation does provisioning, together with the signaling hypothesis discretionary variables. Otherwise, economic fluctuation and business cycles seem to not affect total provisioning, with the signaling hypothesis. Cyclicality of Total LLP appears to be driven The results for equations (3) and (4) notwith seem play atogether relevant role. primarily by the bank’s specific microeconomic fa provisioning, together thetosignaling Cyclicality of Total LLP appears to be primarily by thehypothesis. bank’s specific microeconomic factors, while thedriven macroeconomic situation Loans does ( ) and forward looking differenc Theseem results for equations (3) andthe(4) are quite similar, differ in key notsome seem to playaspects. a relevantBad role. primarily by the bank’s specific microeconomic factors, while macroeconomic situationbut does not toby play a relevant role. ( )collateralized have signs equal to the ones esti also be affected the amount of guarantees on loans. Banks with a higher percentage of The results for equations (3) and (4) are quit not seem to play a relevant Loans (role.The) and forward looking differences ( ), together with the Loans to Assets ratio results for equations (3) and (4) are quite similar, but differ in some key aspects. Bad seems to also component be affected by th (partially are willing set Bad Loans, to the fact they are less Loans )that and forward looking differences ( The results loans for and (4) forward are quite similar, but lower differ( provisions in some aspects. Bad and ( equations ) have signs equal to the to ones estimated for key equations (1) (2). The non-discretionary Loans ((3) or )totally) and looking differences ),on together with thedue Loans to(Assets ratio percentage of collateralized loans (partially ( ) have signs equal to the ones estimated Loans ( ) and exposed forward looking differences ( ), together with the Loans to Assets ratio ( to credit ) have signs equal the higher onesby estimated for equations (1) andThe (2). The non-discretionary default and the (expected) recovery rate. of earnings before Loans, due a to higher the fact that they are less foe component seems to also betotoaffected the amount of guarantees oncontribution loans. Banks with component seems to also be affected by the amoun ( ) have signs equal to the ones estimated for equations (1) and (2). The non-discretionary component seems to also beloans affected amount of guarantees on loans. with a higherrate.on recovery The contribution of earning percentage ofand collateralized (partially or totally) are to setBanks lower provisions Bad loan provisions isby notthetrivial. The coefficient associated to this variable is negative of collateralized loans (partially or component seemsinterest, to alsopercentage betaxes affected theloss amount of guarantees Banks withwilling a higher ofbycollateralized loans (partiallyonorloans. totally) are willing to set lower percentage provisions on Bad trivial. The(expected) coefficient associated to thistota var Loans, due to the fact that they are less exposed to credit default and to the higher Loans, due to the fact that they exposed percentage of collateralized loans (partially or totally) are willing to set lower provisions on Bad and significant fortoboth Arellano-Bond Also this case, the variation of are lessAlso Loans, due the fact that they are and less Arellano-Bover exposed to creditestimations. default and to thein higher (expected) and Arellano-Bover estimations. in thi recovery rate. contribution rate.are The contribution of earnings before interest, taxes andloss loan lossSignaling provisions is not of earnings before Loans, due to the recovery fact that they less exposed to credit default and to the highertaxes (expected) rate. The contribution of earnings before interest, and loan provisions is The notvariable. GDP isrecovery not significant, nor is the Signaling variable. trivial. The coefficient associated to this variable is recovery rate. The trivial. contribution of earnings before interest, and loan loss provisions not for both The The coefficient associated to thisvariable variable is negative andis significant for both Arellano-Bond trivial. coefficient associated to taxes this is negative and significant Arellano-Bond Regressions for equations (5) and (6) estimations. in this case, andAlso (6) suggest thatfor the non-discretionary component is relevant trivial. The coefficient associated to for thisequations variable is (5) negative andin significant both Arellano-Bond and Arellano-Bover estimations. Also this the variation of GDP is notbehavior significant, is the nor andRegressions Arellano-Bover estimations. in thiscase, case, the variation of GDP isand notArellano-Bover significant, is the Also isnor relevant for provisions associated to Imth Signaling variable. and Arellano-Bover estimations. Also in this case, the variation of GDP is not significant, nor is the Signaling variable. to Impaired Loans. However, the forward looking indicator ( does not notseem to affect the de for provisions associated ) does Signaling variable. Regressions for equations (5) and (6) suggest Signaling variable. Regressions for equations (5) (5) and (6) suggest thatthat the non-discretionary behavior component loans. Provisions appear counter c Regressions for equations and (6) suggest the non-discretionary behavior component tois affect dependent astoisImpaired also the Loans. case with collateralized loans. Provisions appear toassociated be to tobe Impaired relevant for provisions L Regressionsseem for equations (5)the and suggestvariable, that the non-discretionary behavior component relevant for (6) provisions associated However, the forwardislooking indicator coefficient associated to the earnings. The r is relevant for provisions associated to Impaired Loans. However, the forward looking indicator ( ) does not seem to affect the dependent is relevant for provisions to Impaired Loans. However, the forward looking indicator counter( associated cyclical, as indicated by the negative and significant coefficient associated to the earnings. The risk ) does not seem to affect the dependent variable, as is also the case with collateralized lower (if compared to the coefficients obtain loans. appear to be counter cyclical, ( ) does not to affect the not dependent variable, asthe is also the case collateralized ( seem ) does seem to to be affect dependent variable, as the is also the case with collateralized loans. Provisions appear counter cyclical, as with indicated by negative andProvisions significant component, given by the loans to assets ratio, has a lower (if compared to the coefficients obtained in the Table 6 to contains the estimation associated the earnings. The riskrelate com loans. Provisions appear toProvisions be counter cyclical, indicated byrisk thecomponent, negative significant loans.coefficient appear toasearnings. be counter cyclical, as and indicated thetocoefficient negative and significant associated to the The given by theby loans assets ratio, has a respective order 1 autoregressive componen lower (if compared to the coefficients obtained in th other equations) but significant effect. coefficient associated to the earnings. The risk component, given by the loans to assets ratio, has a lower (if compared to in the other equations) coefficient associated tothe thecoefficients earnings.obtained The risk component, givenbutbysignificant the loanseffect. toofassets ratio, the hascoefficient a Bad Loans, associated t lower (if compared to the coefficients obtained in the other equations) but significant effect. Table 6 contains the estimation related to the and the respective Table contains the estimation related to the Coverage Ratios. For Table 6 contains the estimation related to the Ratios. Forboth both andbehavior the lower (if6compared to the coefficients obtained in Coverage the other equations) but significant effect. variable ( ) has a positive bu order 1 autoregressive component has a Table 6 contains therespective estimationorder related to the Coverage Ratios. For andeffect. As theregardsrespective 1component autoregressive component hasboth aeffect. significant the coverage ratio order 1Table autoregressive has a significant As regards the coverage ratio of Bad Loans, the 6component contains theaestimation related to theisCoverage Ratios. For of both and the Bad coefficient associated to respective order 1 autoregressive has significant effect. the ratio TABLE 6.the ESTIMATION OF COVERAGE RATIOS DETE of Bad Loans, the coefficient associated toAs regards not coverage significant. In this case, theLoans, signaling is not significant. In this case, the signaling behavior variable hasa positive but relativ associated to behavior variable ( ) has respective order 1 autoregressive component has a significant effect. As regards the coverage ratio of Bad Loans, thecoefficient coefficient associated to is not significant. In this case, the signaling behavior variable ( ) has a positive but relative small impact. (Arellano – Bond) behavior variable a( of a positive relative small impact. to Bad) has Loans, the but coefficient associated is not significant. In this case, the signaling positive but relative small impact. TABLE 6. ESTIMATION ATIOS DETERMINANTS – FULL SAMPLE behavior variable ( OF COVERAGE ) has a Rpositive but relative small impact. TABLE 6. ESTIMATION OF COVERAGE RATIOS DETERMINANTS – FULL SAMPLE

TABLE 6. ESTIMATION OF C OVERAGE RATIOS DETERMINANT (7) (8) (9)

(Arellano – Bond) (Arellano – Bover) Table 6. (Arellano Estimation of Coverage Ratios determinants – full sample – Bond) OF COVERAGE RATIOS DETERMINANTS (Arellano – Bover) TABLE 6. ESTIMATION – FULL SAMPLE

Constant

 

(7) 0.00428 (0.05)

(8) Constant -0.01385 (-0.16)

Constant   (-1)

 

(7) (9) 0.00428 0.14584 (0.05) (3.72)

(7) 0.00428 (0.05)

(8) (9) (10) -0.01385 0.14584 (Arellano – (7) Bond) 0.14011 0.15284 (-0.16) (3.72) (3.57) (1.86)

(8) 10   -0.01385 (-0.16)

0.95617 (9.44)

(9) 0.14584 (3.72)

(10) (8) 0.14011 0.07729 (3.57) (0.70)

0.96328 (9.36)

(7) (9) 0.15284 0.13618 (1.86) (3.73)

Constant

(7) (8) (9) (10) Constant 0.00428 (10) 0.07729 (Arellano 0.13618– Bover)   0.13039 (0.05) 0.13039 (0.70) (3.73) (3.54) (3.54)

10  (10) 0.14011 (3.57)

(7) 0.15284 (1.86)

(8)   0.07729 (0.70)

10  

0.72809 (9.12)

0.73766 (8.86)

(9) 0.13618 (3.73)

0.00428 -0.01385 (Arellano – Bond) (0.05) (-0.16) (8) -0.01385 (-0.16)

(10) 0.13039 (3.54)

0.59376 0.60108 0.57099 0.57295 (4.73) (4.75) (10.68) (10.68) -0.03287 -0.01885 -0.02911 -0.00042 -0.00083 (-0.66) -0.00020 (-1.82) (-0.77) (-1.79) 0.00042 -0.00083 -0.00020 (-0.52) (-1.33) (-0.47) GUA2 (-1.33) 0.00217(-0.47) -0.02535 -0.00018 -0.02131 (-0.52) 0.02437 0.02076 0.02080 0.00248 0.00256 .02437 0.02076 0.02080 0.00248 (0.09) 0.00256 (-1.40) (-0.01) (-1.31) (6.12) (10.59) (10.64) (2.19) (2.30) (6.12) (10.59) (10.64) (2.19) (2.30) LOAN 0.02361 0.02453 -0.09430 -0.09401 -0.02176 -0.02138 -0.08262 -0.08187 -9.22e-06 -5.25e-06 -5.24e-06 3.69e-06 3.72e-06 .22e-06 -5.25e-06 -5.24e-06 3.69e-06 3.72e-06 (0.29) (-2.03) (-0.25) (-0.25) (-1.89) (-1.87) (-1.94) (-1.45) (-1.45) (5.20) (0.30) (4.13) (4.13) (-2.04) (-1.94) (-1.45) (-1.45) (5.20) CAP 0.00017 -0.00017 -0.00015 0.00018 0.00017 -0.00013 -0.00013 -0.21702 -0.09206 -0.09212 -0.06542 -0.05916-0.00015 0.21702 -0.09206 -0.09212 -0.06542 -0.05916 (-4.56) (-2.58) (-2.57) (-4.73) (1.08) (-4.50) (-4.50) (-0.84) (1.07)(-4.73) (-0.85) (1.08) (1.06) (-0.84) (-0.85) (-4.56) (-2.58) (-2.57) 1.31e-06 4.36e-06 ER 4.63e-06 4.63e-06 -0.27238 3.83e-06 -0.23696 -3.31e-06-1.2080 -1.1901 0.76295 0.83173 -1.09208 -1.07464 .31e-06 4.36e-06 3.83e-06 -3.31e-06 (0.12) (0.64) (0.69) (-1.13) (-0.32) (-0.91) (-0.91) (-3.61) (0.12) (0.64) (0.69) (-0.38)(-1.13) (-3.52) (1.11) (1.20) (-3.55) (-3.48) -0.00394 -0.00650 -0.00651 -0.00038 -0.00063 0.00394 -0.00650 -0.00038 -0.00063 SIGN -0.00651 0.00068 0.00069 -0.00012 -0.00012 0.00055 0.00057 -0.00014 -0.00013 (-1.32) (-3.03) (-3.03) (-0.35) (-0.58) (-1.32) (-3.03) (-3.03) (2.33)(-0.35) (2.39) (-0.58) (-0.99) (-0.95) (1.78) (1.85) (-1.12) (-1.11) dBover Bovertwo-step two-stepestimation estimationwith withorthogonal orthogonaldeviations. deviations. Laggedexplanatory explanatory variables Lagged 0.07382 0.07266 variables 0.07571 0.07660 -0.01309 -0.02529 0.06536 0.06476 dequations equationsestimations estimations (0.86) (0.84) (1.74) (1.75) (-0.17) (-0.33) (1.72) (1.70) Note: t – statistics in brackets. Arellano - Bond and Arellano - Bover GMM two-step estimation. Lagged explanatory variables have oLoan LoantotoAssets Assetsratio ratio ( tused positive and significant atthe the1and 1 estimations (been ) )as isisinstruments positive and Note: – statistics in brackets. Arellano -at Bond Arellano - Bover GMM two-step estimation. Lagged explanatory variables forsignificant differenced equations (-1)

-0.00112 -0.00112 (-1.53) GUA1 (-1.53)

-0.00032 -0.00032 -0.01615 (-0.77) (-0.77)

ests thatItalian Italianbanks banksmake makehigher higherprovisions provisionswhen whencredit creditrisk riskisis sts that have been used as instruments for differenced equations estimations standard accounting principles and previous studies. Total LLP standard accounting principles and Arellano-Bond previous studies. and TotalArellano-Bover LLP Both estimation techniques confirm that Impaired Loans al managementpurposes. purposes.InInfact, fact,the thecoefficient coefficientrelated relatedtotocapital capital l management coverage ratio tends to be counter cyclical with respect to earnings. very close to zero. As regards the income smoothing hypothesis, ery close to zero. As regards income smoothing hypothesis, Boththe Arellano-Bond andwhen Arellano-Bover estimation techniques confirm that Impaired Loans coverage Italian banks tend to reduce loan loss provisions earnings Italian banks tend to reduce loan loss provisions when earnings ratio tends be FOR counter cyclical with respect earnings. – FULL SAMPLE – EQUATION 1-6 sionsincrease, increase,confirming confirming the cyclicality suggested bythe the nonTABLE 7Acyclicality . to TEST AUTOCORRELATION OF FIRSTto DIFFERENCE ions the suggested by none,economic economicfluctuation fluctuationand andbusiness businesscycles cyclesseem seemtotonot notaffect affecttotal total (Arellano – Bond) (Arellano – Bover) gnalinghypothesis. hypothesis.Cyclicality CyclicalityofofTotal TotalLLP LLP appears to be driven gnaling LLPtotappears to be driven LLPtot microeconomic factors, while the macroeconomic situation does microeconomic factors, while the macroeconomic does (1) (2) situation (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) Order 1

-3.4399

-3.4414

-0.9656

-0.9637

-1.8576

-1.8592

-4.9799

-4.983

-3.0183

-3.0123

-4.465

-5.4687

3) and(4) (4)are arequite quitesimilar, similar, butdiffer differ some keyaspects. aspects.Bad Bad 0.0006 0.0006 0.3342 0.3352 0.0632 0.0630 0.0000 0.0000 0.0025 0.0026 0.0000 0.0000 ) and but ininsome key (p-value) gdifferences differences( ( togetherwith withthe theLoans LoanstotoAssets Assetsratio ratio ),),together -0.4665 -0.4880 0.5623 0.5540 -1.274 -1.2681 -1.1016 -1.1072 -0.9961 -1.0023 -0.7568 -0.6010 eones onesestimated estimatedfor forequations equations and (2).The The non-discretionary Order (1) 2(1)and (2). non-discretionary 0.6408 0.6255 0.5739a higher 0.5795 0.2027 0.2048 0.2706 0.2682 0.3192 0.3162 0.4492 0.5478 (p-value) on ected by the amount of guarantees loans. Banks with cted by the amount of guarantees on loans. Banks with a higher Note: Arellano Bond test for zero autocorrelation on first differenced errors. H : no autocorrelation 0 s(partially (partiallyorortotally) totally)are arewilling willingtotoset setlower lowerprovisions provisionson onBad Bad areless lessexposed exposedtotocredit creditdefault defaultand andtotothe thehigher higher(expected) (expected) are of earningsbefore beforeinterest, interest,taxes taxes andloan loanloss lossprovisions provisions not Autocorrelation test results, contained in Tables 7a and 7b, confirm that autocorrelation in f earnings and isisnot 121 118issue. thisvariable variableisisnegative negative and significantcould forboth both Arellano-Bond firstand differences not be considered as a major otothis significant for Arellano-Bond JEOD - Vol.3, Issue 1 (2014) Also in this case, the variation of GDP is not significant, nor is the Also in this case, the variation of GDP is not significant, nor is the  

TABLE 7B. TEST FOR AUTOCORRELATION OF FIRST DIFFERENCE – FULL SAMPLE – EQUATION 7-10 5) and (6) suggest that the non-discretionary behavior component

(9) 0.14584 (3.72)

0.14584 (3.72)

(10) 0.1401 (3.57)

10  


CAP SIGN

0.00017 -0.00017 -0.00015 -0.00015 0.00018 0.00017 -0.00013 0.00068 0.00069 -0.00012 -0.00012 0.00055 0.00057 -0.00014 -0.00013 -0.00013 (1.07) (1.08) (-0.84) (-0.85) (1.08) (1.06) (-0.84) (-0.85) (2.33) (2.39) (-0.99) (-0.95) (1.78) (1.85) (-1.12) (-1.11) ER -0.27238 -0.23696 -1.2080 -1.1901 0.76295 0.83173 -1.09208 0.07382 0.07266 0.07571 0.07660 -0.01309 -0.02529 0.06536 -1.07464 0.06476 (-0.38) (-0.32) (-3.61) (-3.52) (1.11) (1.20) (-3.55) (-3.48) (0.86) (0.84) (1.74) (1.75) (-0.17) (-0.33) (1.72) (1.70) SIGN 0.00068 0.00069 -0.00012 -0.00012 0.00055 0.00057 -0.00014 -0.00013 Note: t – statistics in brackets. Arellano Bond and Arellano Bover GMM two-step estimation. Lagged explanatory variables have Provisioning(-0.99) and Relationship (-0.95) Banking in Italy: Practices Evidence (-1.12) (2.33) Loan Loss (2.39) (1.78) and Empirical (1.85) (-1.11) Alessi M.; Di Colli S.; Lopez J.S. been used as instruments for differenced equations estimations 0.07382 0.07266 0.07571 0.07660 -0.01309 -0.02529 0.06536 0.06476 (0.86) (0.84) (1.74) (1.75) (-0.17) (-0.33) (1.72) (1.70) Note: Both t – statistics in brackets. Arellano Bond and Arellano - Bover GMM two-step estimation. Lagged explanatory variables Loans have Arellano-Bond and- Arellano-Bover estimation techniques confirm that Impaired been used as instruments for differenced equations estimations

coverage ratio tends to be counter cyclical with respect to earnings.

Table 7a.Both Test for autocorrelation ofand first difference – full sample – Equationtechniques 1-6 Arellano-Bond Arellano-Bover estimation confirm that Impaired Loans TABLE 7A. TEST FOR AUTOCORRELATION OF FIRST DIFFERENCE – FULL SAMPLE – EQUATION 1-6

coverage ratio tends to be counter cyclical with respect to earnings. (Arellano – Bond)

(Arellano – Bover)

LLPtot TABLE 7A. TESTLLPtot FOR AUTOCORRELATION OF FIRST DIFFERENCE – FULL SAMPLE – EQUATION 1-6 Order 1 (p-value)

(1)

(2)

-3.4399 -3.4414 0.0006LLPtot 0.0006

(3)

(4)

(5)

(Arellano -0.9637 – Bond) -1.8576 -0.9656 0.3342 0.3352 0.0632

(6)

-1.8592 0.0630

(1)

(2)

(3)

(4)

-4.9799 -4.983 (Arellano -3.0183 – Bover) -3.0123 LLPtot0.0000 0.0000 0.0025 0.0026

(5)

-4.465 0.0000

(6)

-5.4687 0.0000

(1) (2) (3) (4) (5) (6) (1) (2) (3) (4) (5) (6) -3.4399 -0.4880 -3.4414 0.5623 -0.9656 -0.9637 Order 0.5540 -1.8576 -1.274 -1.8592 -1.2681 -4.9799 -1.1016 -4.983 -1.1072 -3.0183 -0.9961 -3.0123 -1.0023 -4.465 -0.7568-5.4687 -0.6010 Order 2 1 -0.4665 0.0006 0.6255 0.0006 0.5739 0.3342 0.5795 0.3352 0.0632 0.0630 (p-value) 0.6408 0.2027 0.2048 0.0000 0.2706 0.0000 0.2682 0.0025 0.3192 0.0026 0.3162 0.0000 0.4492 0.0000 0.5478 (p-value) Note: Arellano - Bond test for zero autocorrelation on first differenced errors. H0: no autocorrelation -0.4880 0.5540 on-1.274 -1.2681 errors. -1.1016 H -1.1072 -0.9961 -1.0023 -0.7568 -0.6010 Order 2 Note: Arellano --0.4665 Bond test for zero 0.5623 autocorrelation first differenced : no autocorrelation 0.6408 0.6255 0.5739 0.5795 0.2027 0.2048 0.2706 0 0.2682 0.3192 0.3162 0.4492 0.5478 (p-value) contained in Tables 7b, confirm that autocorrelation in Note: Autocorrelation Arellano - Bond test fortest zero results, autocorrelation on first differenced errors.7a H0:and no autocorrelation

Autocorrelation testnotresults, containedasina major Tables issue. 7a and 7b, confirm that autocorrelation in first first differences could be considered Autocorrelation results,ascontained in Tables 7a and 7b, confirm that autocorrelation in   differences could not be test considered a major issue. TABLE B. TEST FORcould AUTOCORRELATION OF FIRST – FULL SAMPLE – EQUATION 7-10 first 7differences not be considered as aDIFFERENCE major issue.   7b. Test for autocorrelation of first difference – full sample – Equation 7-10 Table (Arellano – Bond) (Arellano – Bover) TABLE 7B. TEST FOR AUTOCORRELATION OF FIRST DIFFERENCE – FULL SAMPLE – EQUATION 7-10

(7) (8) (9) (10) (7) (8) (Arellano – Bond) (Arellano – Bover)(9) Order 1 -5.2988 -5.299 -5.6179 -5.6081 -5.4275 -5.4566 -7.3192 (p-value) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 (7) (8) (9) (10) (7) (8) (9) Order 1 -5.2988 -5.299 -5.6179 -5.6081 -5.4275 -5.4566 -7.3192 Order 2 1.059 1.0588 1.5253 1.5531 0.9699 0.9898 1.4585 (p-value) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 (p-value) 0.2896 0.2897 0.1272 0.1204 0.3321 0.3223 0.1447 Order 2 1.059 1.5253 1.5531 errors. 0.9699 0.9898 1.4585 Note: Arellano - Bond test for zero1.0588 autocorrelation on first differenced H0: no autocorrelation (p-value) 0.2896 0.2897 0.1272 0.1204 0.3321 0.3223 0.1447 Note: Arellano - Bond test for zero autocorrelation on first differenced errors. H0: no autocorrelation

4.3. Focus on cooperative credit banks Note: Arellano - Bond test for zero autocorrelation on first differenced errors. H : no autocorrelation 0

(10) -7.2767 0.0000 (10) -7.2767 1.4815 0.0000 0.1385 1.4815 0.1385

4.3. Focus on cooperative credit banks In this section, we focus our attention on a restricted sample containing solely cooperative 4.3. Focus on cooperative credit banks are estimated using the same econometric technique applied in credit banks. the ten In thisAll section, weequations focus our attention on a restricted sample containing solely cooperative thecredit previous section. banks. All the ten equations are estimated using the same econometric technique applied in Regression results forour LLP (TotalonLLP on Badsample Loanscontaining and on Impaired Loans) are illustrated this section, we focus attention a restricted solely cooperative credit banks. theIn previous section. inAllTables and 9. results for LLPusing (Totalthe LLP oneconometric Bad Loans technique and on Impaired are illustrated the Regression ten8 equations are estimated same applied Loans) in the previous section. in Tables 8 and 9. for LLP (Total LLP on Bad Loans and on Impaired Loans) are illustrated in Tables Regression results TABLE 8. ESTIMATION OF LLP DETERMINANT – CCB SAMPLE - ARELLANO BOND 8Tand 9. ABLE 8. ESTIMATION OF LLP DETERMINANT – CCB SAMPLE - ARELLANO BOND Dependent variable

Dependent variable

 

 

11   11  

122 119 JEOD - Vol.3, Issue 1 (2014)


4.3. Focus on cooperative credit banks In this section,Loan weLossfocus our and attention a restricted sample containing Provisioning Relationshipon Banking in Italy: Practices and Empirical Evidence solely cooperative M.; Di Colli S.; Lopez J.S. same econometric technique applied in credit banks. All the ten equations are Alessi estimated using the the previous section. Regression results for LLP (Total LLP on Bad Loans and on Impaired Loans) are illustrated in Tables 8 and 9. Table 8. Estimation of LLP determinant – CCB sample - Arellano Bond TABLE 8. ESTIMATION OF LLP DETERMINANT – CCB SAMPLE - ARELLANO BOND Dependent variable

  Constant LLPtot (-1)

(1) 0.00056 (0.16) -0.04038 (-0.34)

LLPtot

(2) 0.00073 (0.22) -0.03954 (-0.34)

(-1)

11   (3)

-0.00716 (-3.38)

(4) -0.00738 (-3.47)

-0.12893 (-1.33)

-0.13517 (-1.38)

(-1) NPLtot

0.12256 (8.74)

BL

0.12264 (8.74)

IL 0.01250 (2.23)

GUA1

0.01256 (2.24)

0.00004 (0.06)

GUA2

0.30077 (14.66)

0.30084 (14.66)

0.00306 (0.40)

0.00306 (0.40)

-0.00146 (-2.01)

(5) -0.00028 (-0.26)

(6) -0.00016 (-0.14)

0.03842 (0.28)

0.03562 (0.26)

0.04039 (7.05)

0.04049 (7.05)

-0.00444 (-1.47) 0.00027 (0.66)

0.00171 (1.24)

-0.00023 -0.00106 0.00010 (-0.34) (-1.97) (0.26 LOAN 0.00826 0.00830 0.01569 0.01570 0.00171 0.00171 (2.12) (2.13) (6.20) (6.18) (1.24) (1.24) CAP 0.00000 0.00000 0.00000 0.00000 0.00000 0.00002 (0.30) (0.30) (1.78) (1.81) (4.95) (4.87) ER -0.23542 -0.23578 -0.04832 -0.04738 -0.08050 -0.08088 (-6.06) (-6.09) (-1.60) (-1.56) (-5.32) (-5.36) SIGN -0.00001 -0.00001 0.00000 0.0000 0.00000 -0.00000 (-1.20) (-1.20) (0.32) (0.34) (-1.47) (-1.49) 0.00824 0.00823 0.00057 0.00062 0.00033 0.00032 (2.91) (2.91) (0.31) (0.33) (0.26) (0.25) Note: t – statistics in brackets. Arellano and Bond GMM two-step estimation. Lagged explanatory variables have been used as instruments for differenced equations estimations Note: t – statistics in brackets. Arellano and Bond GMM two-step estimation. Lagged explanatory variables have been used as

  instruments for differenced equations estimations

Results for equations (1) and (2) are similar to the ones obtained for the Italian banking system, with the exception of Expectations on future NLP dynamic which, in this case seem to Results for equations (1) and (2) are similar to the ones obtained for the Italian banking system, with affect Total LLP for cooperative credit banks (the coefficients in both the regressions are positive the of Expectations future dynamic which, inwith this the caseearnings seem to suggests, affect Total for andexception significant). At the sameontime, theNLP coefficient associated as LLP in the cooperative (thethat coefficients in both theare regressions are positive andthe significant). At the same case of thecredit full banks sample, Total Provisions pro-cyclical and that income smoothing hypothesis is not verified for CCBs either. time, the coefficient associated with the earnings suggests, as in the case of the full sample, that Total Provisions are pro-cyclical and that the income smoothing hypothesis is not verified for CCBs either. TABLE 9. ESTIMATION OF LLP DETERMINANT – CCB SAMPLE- ARELLANO BOVER

Constant LLPtot (-1)

(1) -0.00882 (-2.49) 0.68214 (7.44)

LLPtot

(2) -0.00874 (-2.45) 0.68357 (7.46)

Dependent variable LLPBL (3) (4) -0.00796 -0.00837 (-3.66) (-3.81) 0.08012 (0.43)

(-1)

BL IL

 

0.07759 (4.53)

0.07720 (4.53)

0.28335 (8.03)

12   123 120 JEOD - Vol.3, Issue 1 (2014)

LLPIL

(6) -0.00110 (-0.91)

0.07973 (0.43)

(-1) NPLtot

(5) -0.00121 (-0.99)

0.28214 (7.99)

0.45907 (5.32)

0.45850 (5.29)

0.04392

0.04407


instruments for differenced equations estimations

 

Results for equations (1) and (2) are similar to the ones obtained for the Italian banking system, with the exception of Expectations on future NLP dynamic which, in this case seem to Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence M.; Di Colli S.; Lopez J.S. affect Total LLP for cooperative credit Alessi banks (the coefficients in both the regressions are positive and significant). At the same time, the coefficient associated with the earnings suggests, as in the case of the full sample, that Total Provisions are pro-cyclical and that the income smoothing hypothesis is not verified for CCBs either. Table 9. Estimation of LLP determinant – CCB sample- Arellano Bover TABLE 9. ESTIMATION OF LLP DETERMINANT – CCB SAMPLE- ARELLANO BOVER

(1) -0.00882 (-2.49) 0.68214 (7.44)

Constant LLPtot (-1)

LLPtot

(2) -0.00874 (-2.45) 0.68357 (7.46)

Dependent variable LLPBL (3) (4) -0.00796 -0.00837 (-3.66) (-3.81) 0.08012 (0.43)

(-1)

(5) -0.00121 (-0.99)

0.07759 (4.53)

BL

0.07720 (4.53)

IL

-0.00433 (-0.78)

 

GUA1

-0.00450 (-0.82)

-0.00109 (-1.01)

GUA2

0.28335 (8.03)

0.28214 (7.99)

12   -0.00254 (-0.24)

(6) -0.00110 (-0.91)

0.07973 (0.43)

(-1) NPLtot

LLPIL

-0.00275 (-0.25)

-0.0021 (-2.13)

0.45907 (5.32)

0.45850 (5.29)

0.04392 (7.65)

0.04407 (7.69)

-0.00507 (-1.61) 0.00032 (0.64)

-0.00495 (-1.58)

-0.00118 -0.00140 0.00018 (-1.21) (-1.80) (0.40) LOAN 0.01475 0.01482 0.01522 0.01517 0.00151 0.00150 (3.38) (3.40) (6.38) (6.36) (1.00) (0.99) CAP -7.37e-06 -7.45e-06 9.84e-07 9.20e-07 3.83e-06 3.83e-06 (-1.63) (-1.66) (0.23) (0.22) (5.53) (5.57) ER -0.29556 -0.29588 -0.04279 -0.04207 -0.07845 -0.07872 (-4.60) (-4.57) (-1.24) (-1.22) (-5.18) (-5.21) SIGN -7.21e-07 -3.98e-07 8.41e-07 1.22e-06 -2.70e-06 -2.76e-06 (-0.07) (-0.04) (0.08) (0.12) (-0.98) (-1.00) 0.00281 0.00286 -0.00110 -0.00111 0.00099 0.00010 (0.93) (0.94) (-0.49) (-0.50) (0.81) (0.82) Note: t – statistics in brackets. Arellano and Bover two-step estimation with orthogonal deviations. Lagged explanatory variables Note: – statistics brackets. for Arellano and Bover two-step estimation with orthogonal deviations. Lagged explanatory variables have tbeen used as in instruments differenced equations estimations

have been used as instruments for differenced equations estimations

Concerning Loan Loss Provisions on Bad Loans, we find that for cooperative credit banks they are negatively with (totally partially) guaranteed loans and positively associated Concerning Loancorrelated Loss Provisions on Badand Loans, we find that for cooperative credit banks they are to the ratio of Bad Loans over total Loans ( ) and to the Loan to Assets ratio, while there is no negatively andInpartially) guaranteed loans and associated to the significantcorrelated impact ofwith the (totally earnings. particular, the guaranteed loanspositively (totally and partially) haveratio a ) and to the Loan to Assets ratio, while there is no significant ofhigher Bad Loans over total Loans ( impact on CCB LLPs on bad loans than in the case of the full sample, confirming the fact impact of reason the earnings. In particular, the guaranteed loanslevel (totally andcollateralization. partially) have a Furthermore, higher impact that one for lower CCB provisioning is a higher of loan coefficients significant (and negative) in the estimation of equations andone (6).reason for onearnings CCB LLPs on bad are loans than in the case of the full sample, confirming the fact(5) that Finally, the equations for Bad Loans and Impaired Loans Coverage Ratio are estimated for the lower CCB provisioning is a higher level of loan collateralization. Furthermore, earnings coefficients are restricted sample (Tables 10, 11a and 11b). significant (and negative) in the estimation of equations (5) and (6). Finally, equations for Bad Loans andDETERMINANTS Impaired Loans Coverage Ratio are estimated for the restricted TABLE 10. Ethe STIMATION OF C OVERAGE RATIOS – CCB SAMPLE sample (Tables 10, 11a and 11b). (Arellano – Bond) (Arellano – Bover) Constant (-1)

(7) 0.15445 (1.47) 0.82060 (6.40)

(8) 0.13307 (1.23) 0.83115 (6.30)

(-1) GUA1 GUA2 LOAN CAP ER

-0.05621 (-1.67) -0.04966 (-0.44) 0.00016 (1.10) -0.01155 (-0.01)

-0.03593 (-1.14) -0.04798 (-0.43) 0.00016 (1.10) 0.06193 (0.06)

(9) 0.10447 (2.91)

(10) 0.09987 (2.83)

0.33895 (2.53) -0.03775 (-1.90)

0.34515 (2.53)

(7) 0.24425 (2.55) 0.63929 (7.27) -0.05458 (-1.64)

-0.03305 (-1.79) -0.00881 -0.00667 -0.05554 (-0.21) (-0.16) (-0.53) 124 121 -0.00011 -0.00011 0.00017 Issue 1 (2014) (1.15) (-1.26) JEOD - Vol.3, (-1.26) -1.70093 -1.67737 0.94656 (-4.10) -4.01) (1.07)

(8) 0.22740 (2.32) 0.64485 (7.24)

-0.03570 (-1.16) -0.05559 (-0.53) 0.00017 (1.16) 1.05549 (1.19)

(9) 0.08881 (2.12)

(10) 0.08789 (2.18)

0.53341 (7.11) -0.02995 (-1.33)

0.53502 (7.16)

-0.01831 (-0.39) -0.00014 (-1.17) -1.6389 (-3.81)

-0.02868 (-1.35) -0.01718 (-0.37) -0.00014 (-1.19) -1.62613 (-3.79)


they are negatively correlated with (totally and partially) guaranteed loans and positively associated to the ratio of Bad Loans over total Loans ( ) and to the Loan to Assets ratio, while there is no significant impact of the earnings. In particular, the guaranteed loans (totally and partially) have a higher impact on CCB LLPs on bad than in the case of the full sample, Loan Loss Provisioning andloans Relationship Banking in Italy: Practices and Empirical Evidence confirming the fact Alessi M.; Di Colli S.; Lopez J.S. that one reason for lower CCB provisioning is a higher level of loan collateralization. Furthermore, earnings coefficients are significant (and negative) in the estimation of equations (5) and (6). Finally, the equations for Bad Loans and Impaired Loans Coverage Ratio are estimated for the restricted sample (Tables 10, 11a and 11b). Table 10. Estimation of Coverage Ratios determinants – CCB sample TABLE 10. ESTIMATION OF COVERAGE RATIOS DETERMINANTS – CCB SAMPLE (Arellano – Bond) Constant (-1)

(7) 0.15445 (1.47) 0.82060 (6.40)

(8) 0.13307 (1.23) 0.83115 (6.30)

(-1) GUA1 GUA2

-0.05621 (-1.67)

(Arellano – Bover)

(9) 0.10447 (2.91)

(10) 0.09987 (2.83)

0.33895 (2.53) -0.03775 (-1.90)

0.34515 (2.53)

(7) 0.24425 (2.55) 0.63929 (7.27)

(8) 0.22740 (2.32) 0.64485 (7.24)

-0.05458 (-1.64)

(9) 0.08881 (2.12)

(10) 0.08789 (2.18)

0.53341 (7.11) -0.02995 (-1.33)

0.53502 (7.16)

-0.03593 -0.03305 -0.03570 -0.02868 (-1.14) (-1.79) (-1.16) (-1.35) LOAN -0.04966 -0.04798 -0.00881 -0.00667 -0.05554 -0.05559 -0.01831 -0.01718 (-0.44) (-0.43) (-0.21) (-0.16) (-0.53) (-0.53) (-0.39) (-0.37) CAP 0.00016 0.00016 -0.00011 -0.00011 0.00017 0.00017 -0.00014 -0.00014 (1.10) (1.10) (-1.26) (-1.26) (1.15) (1.16) (-1.17) (-1.19) ER -0.01155 0.06193 -1.70093 -1.67737 0.94656 1.05549 -1.6389 -1.62613 (-0.01) (0.06) (-4.10) -4.01) (1.07) (1.19) (-3.81) (-3.79) SIGN 0.00061 0.00062 -0.00003 0.00800 0.00051 0.00051 -0.00003 -0.00003 (1.51) (1.57) (-0.43) (0.17) (1.19) (1.24) (-0.28) (-0.25) 0.11299 0.10796 0.00738 0.00800 0.02209 0.01697 0.03490 0.03397 (1.47) (0.99) (0.16) (0.17) (0.25) (0.19) (0.80) (0.78) Note: t – statistics in brackets. Arellano - Bond and Arellano - Bover GMM two-step estimation. Lagged explanatory variables have been tused as instruments for differenced Note: – statistics in brackets. Arellano - equations Bond andestimations Arellano - Bover GMM two-step estimation. Lagged explanatory variables have

been used as instruments for differenced equations estimations

13   case of the full sample, ha in in thethecase has andsignificant significant autoregressive of As order 1 component. caseofofthe thefull full sample, sample, has aa strong strong and autoregressive of order 1in the   AsAs component. Well capitalized banks do not appe As in the case of the full sample, has a strong and significant autoregressive of order 1 component. Well banks capitalized banks do nottoappear have necessarily lowerLoans Bad Loans Coverage Well capitalized do not appear have to necessarily lower Bad Coverage Ratio.respect Withtorespect Ratio. the national case, the amo component. Welltocapitalized banksthedo not appear have necessarily lower BadWith Loans Coverage Ratio. With respect the national case, amount of totallytoguaranteed loans seems to negatively . This negative toRatio. the national case, the amount of totally guaranteed loans seems to negatively affect affect . This negative relationship appears With respect to the national case, the amount of totally guaranteed loans seems to negatively affect . This negative relationship appears much clearer for CCBs. As in the full sample, the Signaling variable is positively correlated with relationship appears much clearer for CCBs. As in the full sample, the Signaling variable is positively Signaling variable is negative positively relationship correlated withappears the dependent variable,for even if the As sizeinofthe thefull sample, theAs in the case of the f affect . This much clearer CCBs. coefficient is close to zero. coefficient is close to zero. correlated with the dependent variable, even if the size of the coefficient is close to zero. component. Well capitalize As in the case of the full sample, has a strong and significant autoregressive of order 1 Signaling variable is positively correlated with the dependent variable, even if the size of the Ratio. With respect to the n component. Well capitalized banks do not appear to have necessarily lower Bad Loans Coverage T ABLE 11 A . T EST FOR AUTOCORRELATION OF FIRST DIFF of to thezero. full sample, has a strong andSAMPLE significant autoregressive of order 1 coefficient is case close TABLEAs 11Ain . Tthe EST FOR AUTOCORRELATION OF FIRST DIFFERENCE – CCB – EQUATION 1-6 Ratio. With respect to the national amount of totally guaranteed negatively affect . This negative component. Well capitalized banks docase, not the appear to have necessarily lowerloans Bad seems Loans to Coverage (Arellano – Bond) (Arellano of – Bond) (Arellano – Bover) Table 11a. Testrespect for. This autocorrelation first difference – CCB sample –guaranteed Equation 1-6 Signaling variable is posit Ratio. With tonegative the national case, theappears amount of totally loans to affect relationship clearer for CCBs. As seems in –the fullnegatively sample, TABLE 11A. LLPtot TEST FOR AUTOCORRELATION OF FIRSTmuch DIFFERENCE – CCB SAMPLE EQUATION 1-6 theLLPtot LLPtot (1) (2) coefficient (3) (5)zero. is(4)close to Signaling is positively with dependent variable, even if the (1) . variable (3) relationship (4) correlated (5) (6) theclearer (1) (3)As in (4) full (5) size of (6) the affect This(2)negative appears much for(2)CCBs. the sample, Order 1 the 0.8553 0.84758 0.3869 0.4684 -1.1292 -1 (Arellano (Arellano Order 1 0.8553 is 0.84758 0.3869 0.4684 – Bond) -1.1292 -1.1181 -4.5866 -4.5986 -1.1303 -1.0975 -3.6585 – Bover) -3.6532 coefficient close to zero. (p-value) Signaling variable is positively correlated the dependent variable, if the 0.0003 size of0.0003 the0.3923 0.3967 0.6988 0.6395 0.2588 0 (p-value) 0.3923 0.3967 0.6988 0.6395 0.2588 with 0.2635 0.0000 0.0000 0.2583even0.2724 LLPtot LLPtot TABLE 11A. TEST FOR AUTOCOR coefficient is close to zero. Order 2 -0.208 -0.2101 0.6619 0.7037 -1.6805 -1 (2)AUTOCORRELATION (4) (5) (6) (1) 0.4123 (2) 0.46811-6(3)-0.9568 -.95513 (4) Order 2 -0.208 -0.2101 0.6619 (3)0.7037 -1.6805 -1.6848 -0.3996 -0.4183 TABLE 11A(1) . TEST FOR OF FIRST DIFFERENCE – CCB SAMPLE – EQUATION (p-value) 0.8350

(5)

0.8336

(6)

0.5080

0.4816

0.0929 (Arel 0

(p-value) 0.4816 0.4684 0.0929 -1.1292 0.0920 0.6894 0.6397-1.1303 0.3387 -1.0975 0.3395 Order 1 0.8350 0.8553 0.8336 0.847580.50800.3869 -1.1181 0.6757 -4.58660.6801 -4.5986 -3.6585 -3.6532 LLPtoton first differe Note: Arellano - Bond test for zero autocorrelation T ABLE 11A.0.3923 T-EST FOR AUTOCORRELATION OF DIFFERENCE –0.2635 CCB – EQUATION (Arellano – Bover) Note: Arellano Bond test for zero (Arellano autocorrelation onFIRST first differenced H0: noSAMPLE autocorrelation (p-value) 0.3967 0.6988 – Bond) 0.6395 0.2588 errors. 0.0000 0.0000 1-6 0.2583 0.2724 0.0003 0.0003

LLPtot

LLPtot

(Arellano – Bond)

(Arellano – Bover)

(1)

(2)

(3)

The(6)negative with is (1) (2) (3) (6) -1.6848 (1) (2) (3) Loans(4) Order 2The -0.208 0.6619 -1.6805 -0.4183 0.4123 (5) 0.4681 -0.9568 relationship -.95513 0.3923 negative relationship with(4)0.7037 (5) is confirmed for the-0.3996 Impaired Coverage Ratio (p-value) 0.3967 0.698 LLPtot -0.2101 LLPtot Order 1 0.8350 0.8553 0.84758 -1.12920.0929 -1.1181 0.0920 -4.5866 -4.5986 (p-value) 0.8336 0.3869 0.5080 0.4684 0.4816 0.6894 -1.1303 0.6757 -1.0975 0.6801-3.6585 0.3387 ( 0.6397 )-3.6532 as well. Given0.3395 the similar result obtaine Order 1

0.8553

0.84758

0.386

(3)0.6988 (4)0.6395 (1)0.0000 (2)0.0000Ratio (3)0.2583 (4)0.2724 (5)0.0003 (6)0.0003 ( (p-value) ) as(1)0.3923 well. (2) Given similar result(5)obtained for the Coverage of Bad Loans, it seems 0.3967 the 0.2588 (6)0.2635 Order 2 -0.208 -0.2101 0.661 Note:1 Arellano zero autocorrelation first differenced : no autocorrelation Order 0.8553 - Bond 0.84758test for 0.3869 0.4684 -1.1292 on -1.1181 -4.5866 errors. -4.5986 H0-1.1303 -1.0975 -3.6585 -3.6532 quite banks 0.8336 which 0.508 hav (p-value) 0.8350 quite clear that0.3967 cooperative banks which have a portfolio of loans 0.4123 with0.2724 a0.4681 higher level clear of that cooperative (p-value) 0.6988 0.6395 0.2635 0.0000 0.0003 Order 2 0.3923 -0.208 -0.2101 0.6619 0.7037 0.2588 -1.6805on -1.6848 -0.39960.0000 -0.4183 0.2583 -0.95680.0003 -.95513 Note: Arellano Bond test for zero autocorrelation first differenced errors. H : no autocorrelation collateralization tend to maintain a lower level Note: Arellano Bond test for zero au collateralization a lower0.0929 level of0.0920 Coverage probably to the0.3387 fact that (p-value) 0.8350 tend 0.8336to maintain 0.5080 0.4816 0.6894Ratios, 0.6757due00.6801 0.6397 0.3395 Order 2 Arellano -0.208 - Bond -0.2101 0.6619 0.7037 -1.6805 -1.6848 -0.3996 -0.4183 0.4123 0.4681 -0.9568 -.95513 credit default risk decreases in presence of loans Note: test for zero autocorrelation on first differenced errors. H : no autocorrelation 0 credit default decreases presence0.0929 of loans that are totally guaranteed. This (partially) (p-value) 0.8350 risk 0.8336 0.5080 in 0.4816 0.0920 0.6894 0.6757 0.6801 0.6397could 0.3387 0.3395 The negative relations

The negative relationship with is confirmed for the Impaired Loans Coverage Ratio explain lower average Coverage Ratios exper (explain )theas-lower well. Given the similar result obtained the Ratio of Badthe Loans, it seems Note: Arellano Bond test for zero autocorrelation on first differenced errors. no autocorrelation average Coverage Ratios experienced byHCCBs withCoverage respect to the Italian banking 0: for negative relationshipwith with is forfor thethe Impaired LoansLoans Coverage Ratio Ratio ( The The negative relationship is confirmed confirmed Impaired Coverage system. system. quite clear that cooperative banks which have a portfolio of loans with a higher level of) as well. Given the quite clear that cooperativ relationship withobtained isfor confirmed forCoverage the Impaired Coverage ( The ) negative as well. Given theresult similar result obtained forCoverage the of Bad Loans, itRatio seemsquite results Estimation also suggest that ascollateralization well. Given the similar the RatioRatio of Loans Bad it to seems clear that Estimation results also suggest that seems toCoverage be pro-cyclical withLoans, respect the to the fact tend to maintain awhich lower level of Ratios, due probably that collateralization tend to ma quite) asclear that cooperative banks have atheportfolio ofRatio loansofwith aLoans, higher level before of earnings taxes and loan loss provisions (ov (earnings well. Given the similar result obtained for Coverage Bad it seems before taxes and loan loss provisions (over Total Assets). cooperative banks which ain portfolio of loans withthat a higher level ofprobably collateralization tend maintain credit default risk decreases of of loans are totally guaranteed. This couldto(partially) credit adefault risk decreases collateralization tend to have maintain apresence lower have level Coverage Ratios, due to the factof that   quite clear that cooperative banks which a portfolio of loans with a higher level   C explain the lower average Coverage Ratios byguaranteed. CCBs with respect to that the Italian credit riskAUTOCORRELATION decreases in due ofDIFFERENCE loans thatfact totally This (partially) ABLE 11 TEST FORbanking AUTOCORRELATION OFaverage FIRST DIFF lower level Coverage Ratios, probably toexperienced the that credit default risk decreases inB.presence ofexplain loans the lower collateralization tend to maintain apresence lower level of Coverage Ratios, probably tocould theTfact TABLE 11Bdefault . Tof EST FOR OF FIRST – are CCB SAMPLE –due EQUATION 7-10 system. explain the risk lower average Coverage Ratios experienced by CCBs with respect to could the Italian banking system. credit default decreases in presence of loans that are totally guaranteed. This (partially) – Bond) that are totally guaranteed. This– Bond) could (partially) explain the lower average Coverage Ratios experienced (Arellano by (Arellano (Arellano – Bover) system. Estimation results al explain the lower average Coverage Ratios experienced by CCBs with to respect to the Italian banking Estimation results also suggest that seems be pro-cyclical with respect to the (7) (8) before taxes (9) CCBs respect to the(8)also Italian banking system.seems(7)to be pro-cyclical earnings and lo( system.with (7) (9) that (10) (8) (9) respect (10)to the Estimation results suggest with Order 1 -4.6168 -4.6186 -3.4581 -3. earnings before taxes and loan loss provisions (over Total Assets). Order 1 -4.6168 -4.6186 -3.4581 -3.4762 -4.9193 -4.9333 -5.7568 -5.7734   earnings beforeresults taxes and loss provisions (over Total seems to be with respect to the the earnings before Estimation also suggest that (p-value) 0.0000 0.0000 0.0005 0. Estimation alsoloan suggest that toAssets). be pro-cyclical pro-cyclical with respect to 0.0000results 0.0000 0.0005 0.0005seems 0.0000 0.0000 0.0000 0.0000  (p-value) T ABLE 11 B . T EST FOR AUTOCOR   earnings before taxesprovisions and loan loss provisions (over Total Assets). taxes and loan loss (over Total Assets). Order 27-10 0.92156 0.93488 -0.23131 -0. TABLE 1111BB. .TTEST FORAUTOCORRELATION AUTOCORRELATION OF FIRST DIFFERENCE – CCB– E SAMPLE EQUATION TABLE EST FOR OF FIRST DIFFERENCE – CCB SAMPLE QUATION–7-10  Order 2

0.92156

0.93488

-0.23131

-0.1720

0.8038

0.8198

0.1087

0.1429 (p-value)

0.3568

0.3499

0.8171

(A 0.

(p-value) 0.3568 0.3499 0.8171 0.8634 0.4215 0.4123 0.9134 0.8864 T ABLE 11B. TEST FOR AUTOCORRELATION FIRST DIFFERENCE – CCB SAMPLE – EQUATION (Arellano –OF Bond) (Arellano7-10 – Bover) Note: (Arellano – Bond) (Arellano – Arellano Bover) - Bond test for zero autocorrelation on first differe Note: Arellano - Bond test for zero autocorrelation on first differenced errors. H0: no autocorrelation (Arellano – Bond) (Arellano – Bover) (7) (8) (9) (10) (9) (10) (7) (8) (9) (10) (7) (7) (8) (8) (9) 5. Conclusions 5. Order Conclusions 1 -4.6168 -4.6186 -3.4581 -3.4762 -4.9193 -4.9333 -5.7568 -5.7734 (7) (8) -4.6186 (9) -3.4581(10) (7) (8) (9) (10)-5.7568 Order 1 -4.6168 -3.4762 -4.9193 -4.9333 (p-value) 0.0000 0.0000 0.0005 0.0005 122 0.0000 0.0000 0.0000 0.0000 125 Order 1 -4.6168 -3.4762 -4.9193 -4.9333 -5.7568 -5.7734 (p-value) 0.0000 -4.61860.0000 -3.4581 0.0005 0.0005 0.0000 0.0000 0.0000paper This JEOD Issue 1 (2014) (p-value) 0.0000 0.0000 0.0005 0.0005 0.0000 0.0000 0.0000 This examines Loan Loss Provisions and- Vol.3, Coverage Ratios determinants for the0.0000 Italian Order 2 paper 0.92156 0.93488 -0.23131 -0.1720 0.8038 0.8198 0.1087 0.1429

Order 1 (10) (p-value)

(7) -4.6168 0.0000

(8 -4.61 0.00

Order 2

0.92156

0.934

-5.7734

0.0000 Loan Loss Provisions examines (p-value) 0.3568 0.34 banking system over a 7-year (2006-201 Note: Arellanoperiod - Bond test for zero au banking system over a 7-year period (2006-2012), using financial statements and balance sheets (p-value) 0.3568 0.3499 0.8171 0.8634 0.4215 0.4123 0.9134 0.8864 Order2 2 0.92156 0.93488 0.93488-0.23131-0.23131 -0.1720 0.8038 0.8198 0.1087 0.1429 Order 0.92156 -0.1720 0.8038 0.8198 0.1087 0.1429 from the Association database. Note:the Arellano - Bond test for zero autocorrelation on0.8171 first differenced errors. H0: no autocorrelation from Italian Banking Association database. We also provide an0.4215 analysis for 0.4123 a sub sample ofItalian Banking (p-value) 0.3568 0.3499 0.8634 0.9134 0.8864


( ) as well. Given the similar result obtained for the Coverage Ratio of Bad Loans, it seems quite clear that cooperative banks which have a portfolio of loans with a higher level of collateralization tend to maintain a lower level of Coverage Ratios, due probably to the fact that credit default risk decreases in presence of loans that are totally guaranteed. This could (partially) Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence explain the lower average Coverage Ratios experienced by CCBs with respect to the Italian banking Alessi M.; Di Colli S.; Lopez J.S. system. Estimation results also suggest that seems to be pro-cyclical with respect to the earnings before taxes and loan loss provisions (over Total Assets).  

Table 11b. Test for autocorrelation of first difference – CCB sample – Equation 7-10 TABLE 11B. TEST FOR AUTOCORRELATION OF FIRST DIFFERENCE – CCB SAMPLE – EQUATION 7-10 (Arellano – Bond)

(8) -4.9333 0.0000

(9) -5.7568 0.0000

(10) -5.7734 0.0000

Order 2 0.92156 0.93488 -0.23131 -0.1720 0.8038 0.8198 (p-value) 0.3568 0.3499 0.8171 0.8634 0.4215 0.4123 Note: Arellano - Bond test for zero autocorrelation on first differenced errors. H0: no autocorrelation

0.1087 0.9134

0.1429 0.8864

Order 1 (p-value)

(7) -4.6168 0.0000

(8) -4.6186 0.0000

(Arellano – Bover)

(9) -3.4581 0.0005

(10) -3.4762 0.0005

(7) -4.9193 0.0000

Note: Arellano - Bond test for zero autocorrelation on first differenced errors. H0: no autocorrelation

5. Conclusions

This paper examines Loan Loss Provisions and Coverage Ratios determinants for the Italian banking system over a 7-year period (2006-2012), using financial statements and balance sheets 5.from Conclusions the Italian Banking Association database. We also provide an analysis for a sub sample of cooperative credit banks. We investigate not only the determinants of Total LLP, but we also try to model the main explanatory variables for Bad Loans and Impaired Loans dynamics. This and paperdetect examines Loan Loss Provisions and Coverage Ratios determinants for the Italian banking Along with the standard explanatory variable commonly used in empirical literature, we testItalian the system over a 7-year period (2006-2012), using financial statements and balance sheets from the impact of guaranteed loans, as an additional factor included in the non-discretionary component of Banking Association database. We also provide an analysis for a sub sample of cooperative credit banks. provisioning strategies. We investigate only the determinants of Total but we with also generalized try to modelmethod and detect the main For the not empirical analysis, we perform theLLP, estimation of moments explanatory variables Bad Loans and Impaired Loans Along with thedeviations standard explanatory (GMM) using first for differences (see Arellano and Bonddynamics. 1991) and orthogonal (Arellano and Bover 1995). used in empirical literature, we test the impact of guaranteed loans, as an additional variable commonly factor included in the non-discretionary component of provisioning strategies. For the empirical analysis, we perform the estimation with generalized method of moments (GMM) 14   orthogonal deviations (Arellano and Bover, 1995). using first differences (see Arellano and Bond, 1991) and   Empirical results suggest that the provisioning mechanism in Italian banks is mainly driven by non-discretionary behavior. Discretionary behavior of bank managers and the economic cycle do not appear to be relevant, as well as expectations about future potential losses and credit risk. A specific analysis conducted on the sub sample of cooperative credit banks pointed out that their Loan Loss provisioning is less pro-cyclical than that of the full sample of banks; moreover, a higher level of collateralized loans, which can reduce credit risk and future losses, has a negative and greater (if compared with other banks) influence on the amount of provisions.

References Altunbas, Y., Fazylov, O., Molyneux, P. (2002). Evidence on the bank lending channel in Europe. Journal of Banking And Finance 26. pp. 2093-2110. http://dx.doi.org/10.1016/S0378-4266(02)00201-7 Anandarajan, A., Hasan, I., Mccarthy, C. (2007). Use of loan loss provisions for capital, earnings management and signaling by Australian banks. Accounting and Finance 47(3), pp. 357-379. http://dx.doi. org/10.1111/j.1467-629X.2007.00220.x Arellano, M., Bond, S.R. (1991). Some tests of specification for panel data: Monte Carlo evidence on employment equations. Review Of Economic Studies 58, pp. 277-297. http://dx.doi. org/10.2307/2297968 Arellano, M., Bover, O. (1995). Another look at the instrumental variables estimation of error-components models. Journal Of Econometrics 68, pp. 29-51. http://dx.doi.org/10.1016/0304-4076(94)01642D 126 123 JEOD - Vol.3, Issue 1 (2014)


Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence Alessi M.; Di Colli S.; Lopez J.S.

Arpa, M., Giulini, A., Ittner, A., Pauer, F. (2001). The influence of macroeconomic developments on Austrian banks: Implications for banking supervision. Bis Paper n. 1, pp. 91-116. Asea, P.K., Blomberg, B. (1998). Lending cycles. Journal of Econometrics 83, pp. 89-128. http://dx.doi. org/10.1016/S0304-4076(97)00066-3 Balla, E., Rose, M.J., Romero, J. (2012). Loan loss reserves accounting and bank behaviour. Federal Reserve Bank of Richmond Economic Brief, EB12-03. Beaver, W.H., Engel, E. (1996). Discretionary behavior with respect to allowances for loan losses and the behavior of security prices. Journal of Accounting and Economics 22, pp. 177-206. http://dx.doi. org/10.1016/S0165-4101(96)00428-4 Berger, A.N., Udell, G.F. (2002). Small business credit availability and relationship lending: The importance of bank organisational structure. The Economic Journal 112, F32-F53. http://dx.doi. org/10.1111/1468-0297.00682 Bernanke, B.S., Lown, C.S. (1991). The credit crunch. Brooking Papers on Economic Activity 2, pp. 205248. http://dx.doi.org/10.2307/2534592 Bhat, V. (1996). Banks and income smoothing: An empirical analysis. Applied Financial Economics 6, pp. 505-510. http://dx.doi.org/10.1080/096031096333953 Bikker, J.A. (2004). Cyclical patterns in Pro.Ts, provisioning and lending of banks. In competition and efficiency in a unified European banking market. pp. 241-273, Edward Elgar. Bikker, J.A., Metzemakers, P.A. (2005). Bank provisioning behaviour and prociclicality. Research Series Supervision n. 50, pp. 1-21. Bouvatier, V., Lepetit, L. (2008). Banks procyclical behavior: Does provisioning matter?. Journal of International Financial Markets, Institutions and Money. 18, pp. 513-26. http://dx.doi.org/10.1016/j. intfin.2007.07.004 Brinkmann, E.J., Horvitz, P.M. (1995). Risk-based capital standards and the credit crunch. Journal of Money, Credit And Banking 27(3), pp. 848-863. http://dx.doi.org/10.2307/2077755 Caporale, G.M., Di Colli, S., Lopez, J.S. (2013). Bank lending procyclicality and credit quality during financial crises. DIW Discussion Papers 1309. Chami, R., Cosimano, T.F. (2001). Monetary policy with a touch of basel. IMF Working Paper 01/151. Clerc, L., Drumetz, F., Jaudoin, O. (2001). To what extent are prudential and accounting arrangements pro or countercyclical with respect to overall â&#x20AC;Ś financial conditions?. BIS Papers 1. Collins, J., Shackelford, D., Wahlen, J. (1995). Bank differences in the coordination of regulatory capital, earnings and taxes. Journal of Accounting Research 33, pp. 263-292. http://dx.doi. org/10.2307/2491488 Curcio, D., Hasan, I. (2013). Earnings and capital management and signaling: The use of loan-loss provisions by European banks. European Journal of Finance. http://dx.doi.org/10.1080/135184 7X.2012.762408 Dewenter, K.L., Hess, A.C. (2003). Are relationship and transactional banks different? Evidence from loan loss provisions and write-offs. EFMA 2004 Basel Meetings Paper. http://dx.doi.org/10.2139/ssrn.478101 Di Colli, S., Girardi, A. (2010). Restrizione creditizia durante la crisi del 2007-2009 e il ruolo anticiclico delle banche di credito cooperativo. Cooperazione di Credito n. 205-206. Di Giulio, D. (2009). Finanziamenti bancari al settore produttivo: Credit crunch o extra-credito. Temi di Economia e Finanza Abi Novembre, n. 1. Ehrmann, M., Gambacorta, L., Martinez-Pagès, J., Sevestre, P., Worms, A. (2003). Financial system and the role of banks in monetary policy transmission in the euro area. in Angeloni, I., Kashyap, A., Mojon, B. (eds.). Monetary policy transmission in the euro area. Cambridge University Press. 127 124 JEOD - Vol.3, Issue 1 (2014)


Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence Alessi M.; Di Colli S.; Lopez J.S.

Fernandez De Lis, S., Martinez-Pagès, J., Saurina, J. (2001). Credit growth, problem loans and credit risk provisioning in Spain. Bis Papers 1. Fonseca, A.R., González, F. (2008). Cross-country determinants of bank income smoothing by managing loan-loss provisions. Journal of Banking and Finance 32, pp. 217-228. http://dx.doi.org/10.1016/j. jbankfin.2007.02.012 Fudenberg, D., Tirole, J. (1995). A theory of income and dividend smoothing based on incumbency rents. Journal of Political Economy 103(1), pp. 75-93. http://dx.doi.org/10.1086/261976 Furfine, C. (2001). Bank portfolio allocation: The impact of capital requirements, regulatory monitoring and economic conditions. Journal of Financial Services Research 20, pp. 33-56. http://dx.doi. org/10.1023/A:1011147609099 Gambacorta, L. (2005). Inside the bank lending channel. European Economic Review 49, pp. 1737-1759. http://dx.doi.org/10.1016/j.euroecorev.2004.05.004 Gambacorta, L., Mistrulli, P.E. (2004). Does bank capital affect lending behavior?. Journal of Financial Intermediation 13, pp. 436-457. http://dx.doi.org/10.1016/j.jfi.2004.06.001 Greenwalt, M.J., Sinkey, Jr. (1988). Bank loan loss provisions and the income smoothing hypothesis: An empirical analysis. Journal of Financial Services Research 1, pp. 301-318. Guttentag, J.M., Herring, R.J. (1984). Credit rationing and financial disorder. Journal of Finance 39, pp. 1359-82. http://dx.doi.org/10.1111/j.1540-6261.1984.tb04912.x Guttentag, J.M., Herring, R.J. (1986). Disaster myopia in international banking. Essays in International Finance 164, International Finance Section, Princeton University. Hancock, D., Wilcox, J.A. (1998). The credit crunch and the availability of credit to small businesses. Journal of Banking and Finance 22, pp. 983-1014. http://dx.doi.org/10.1016/S0378-4266(98)00040-5 Jiménez, G., Saurina, J. (2006). Credit cycles, credit risk and financial regulation. International Journal of Central Banking 2, pp. 65-98. Jordan, J., Peek, J., Rosengren, E. (2002). Credit risk modeling and the cyclicality of capital. Federal Reserve Bank of Boston, Paper Prepared for the Conference “Changes in Risk Through Time: Measurement and Policy Options”, Bis, Basel. Kanagaretnam, K., Lobo, G.J., Yang, D.H. (2005). Determinants of signaling by banks through loan loss provisions, Journal of Business Research. 58, pp. 312-320. http://dx.doi.org/10.1016/j. jbusres.2003.06.002 Kashyap, A.K., Stein, J.C. (1995). The impact of monetary policy on bank balance sheets. Carnerige Rochester Conference Series on Public Policy 42, pp. 151-195. http://dx.doi.org/10.1016/01672231(95)00032-U Kashyap, A.K., Stein, J.C. (2000). What do a million observations on bank say about the transmission of monetary policy. The American Economic Review 90, pp. 407-428. http://dx.doi.org/10.1257/ aer.90.3.407 Keeton, W.R. (1999). Does faster loan growth lead to higher loan losses?. Federal Reserve Bank of Kansas City Economic Review 84, pp. 57-75. Kishan, R.P., Opiela, T.P. (2000). Bank size, bank capital, and the bank lending channel. Journal of Money, Credit and Banking 32(1), pp. 121-141. http://dx.doi.org/10.2307/2601095 Laeven, L., Majnoni, G. (2003). Loan loss provisioning and economic slowdowns: Too much, too late?. Journal of Financial Intermediation 12(2), pp. 178-197. http://dx.doi.org/10.1016/S10429573(03)00016-0 Lown, C., Morgan, D.P. (2006). The credit cycle and the business cycle: New findings using the loan officer opinion survey. Journal of Money, Credit, and Banking. http://dx.doi.org/10.1353/mcb.2006.0086 128 125 JEOD - Vol.3, Issue 1 (2014)


Loan Loss Provisioning and Relationship Banking in Italy: Practices and Empirical Evidence Alessi M.; Di Colli S.; Lopez J.S.

Ma, C.K. (1988). Loan loss reserve and income smoothing: The experience in the U.S. banking industry. Journal of Business Finance and Accounting 15(4), pp. 487-497. http://dx.doi. org/10.1111/j.1468-5957.1988.tb00150.x Packer, F., Zhu, H. (2012). Loan loss provisioning practices of Asian banks. Bis Working Papers n. 375. Pain, D. (2003). The provisioning experience of the major UK banks: A small panel investigation. Bank of England, Working Paper 177. Peek, J., Rosengren, E.S. (1995). Bank regulation and credit crunch. Journal of Banking and Finance 19, pp. 679-692. http://dx.doi.org/10.1016/0378-4266(94)00148-V Peek, J., Rosengren, E.S., Tootell, G.M. (2003). Identifying the macroeconomic effect of loan supply shocks. Journal of Money, Credit, and Banking 35(6), pp. 931-946. http://dx.doi.org/10.1353/ mcb.2003.0046 Rajan, R.G. (1994). Why bank credit policies fluctuate: A theory and some evidence. The Quarterly Journal of Economics 109, pp. 399-441. http://dx.doi.org/10.2307/2118468 Ronen J., Sadan, S. (1981), Smoothing income numbers: objectives, means and implications. Addison-Wesley Educational Publishers Inc. Shrieves, R.E., Dahl, D. (2002). Discretionary accounting and the behavior of Japanese banks under financial distress. Journal of Banking and Finance 27(7), pp. 1219-1243. http://dx.doi.org/10.1016/ S0378-4266(02)00252-2 Trueman, B., Titman, B. (1988). An explanation of accounting income smoothing. Journal of Accounting Research 26 (Supplement), pp. 127-139. http://dx.doi.org/10.2307/2491184 Wagster, J.D. (1999). The Basel Accord of 1988 and the International Credit Crunch of 1989-1992. Journal of Financial Services Research 15, pp. 123-143. http://dx.doi.org/10.1023/A:1008023803152 Zicchino, L. (2005). A model of bank capital, lending and the macroeconomy: Basel I versus Basel II. Bank of England Working Paper n. 270.

129 126 JEOD - Vol.3, Issue 1 (2014)


AT T R I B U T I O N 3 . 0

You are free to share and to remix, you must attribute the work

17 |June 2014 | Vol.3, Issue128-161 1 (2014) 131-163 Publication date: ??? Vol.3, Issue 1 (2014)

AUTHOR MITJA STEFANCIC Faculty of Economics, University of Ljubljana, Slovenia mitja.s@hotmail.it

Investigating Management Turnover in Italian Cooperative Banks ABSTRACT Management turnover is a common tool for disciplining top managers both in corporations and in financial intermediaries. This paper examines the turnover of top managers in Italian banks. By applying a survival analysis to a dataset covering all Italian banks, the study tests the hypothesis that management turnover differs according to different types of banks; and, specifically, it attempts to assess whether top managers in non-commercial banks are more likely to stay on in their managerial position longer than those in commercial banks. Results confirm that the juridical form of banks is significantly related to management turnover as top managers in cooperative banks show a higher survival probability. Similarly, banks’ history and institutional legacy have a significant influence on both management turnover and on the disciplinary mechanisms for top managers. Managers in cooperative banks tend to survive longer even when bank performance, measured as return on assets, is below average.

KEY-WORDS ITALIAN COOPERATIVE BANKS; JURIDICAL FORM OF BANKS; TURNOVER OF TOP MANAGERS; CORPORATE GOVERNANCE; INSTITUTIONAL SETTINGS

Acknowledgements An earlier version of this paper was presented at the third Euricse conference on “Cooperative Finance and Sustainable Innovation” in Trento (June 2012), and as a poster during the “Potential and Limits of Social and Solidarity Economy” conference, organised by the UNRISD in Geneva (May 2013). The author is thankful to Clara Graziano (University of Udine), Erich Battistin and Bruno M. Parigi (University of Padova) for valuable data on top managers in Italian banks. The author also thanks Matteo Dimai for research assistance, Aleksandra Gregorič (CBS Copenhagen), Yiorgos Alexopoulos (Agricultural University of Athens) and Silvio Goglio (University of Trento) for valuable comments. The usual disclaimer applies.

JEL Classification: D21; G21; G28; G34; J63 | DOI: http://dx.doi.org/10.5947/jeod.2014.007

131 128 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

1. Introduction Management turnover is a common tool for disciplining top managers both in corporations and in financial intermediaries. Top managers are subject to pressures to act in the interest of shareholders and stakeholders more broadly. Often, however, disciplinary forces do not appear to be entirely effective, and managers are frequently in a position to counter said disciplinary forces â&#x20AC;&#x201C; for instance by entrenching themselves and making themselves costly to replace. In order to shed light on these issues, it is crucial to investigate whether top management turnover as a disciplinary mechanism differs in different types of corporations. The present study investigates how management turnover differs in Italian banks by means of a survival analysis. Although there may be various factors and issues that affect management turnover, the focus here is on external pressure and, thus, on the disciplining mechanisms applied to top managers. Addressing such issues is important as the quality of management is essential for an effective governance of banks. By contrast, in a highly competitive market, poor management often results in poor bank performance and inadequate risk management. In order to prevent this, the owners (board of directors on behalf of the owners) have the right to replace the managers of the bank when performing poorly. The question is, however, whether the owners have the necessary information and incentives to act against existing management, and to what extent they are able to impose such actions in practice. In fact, it can be safely assumed that managers do not like outside interference on their management (such issues are investigated, for instance, in Fama and Jensen, 1983; and in Dewatripont and Tirole, 1992). A number of developments resulting from the process of organizational restructuring in Italian banking have in part blurred the differences among different types of banks (Zazzaro, 2004; Bonaccorsi di Patti et al., 2005; Ayadi et al., 2010; Gallo et al., 2011). Nonetheless, differences in the governance of banks persist. Differences in the ownership structure and types of owners are among the most important differences between commercial and cooperative banks. In commercial banks, the owners invest with an aim to obtain the required return on their investments, and have a number of mechanisms through which they can push the bank managers to achieve such a goal. For example, they can strengthen their control over the management by increasing their ownership share and, thus, gain control on the board. By contrast, the owners of cooperative banks are cooperative members and bank customers. They are less interested in the bottom line of such banks and exert a lower pressure over the bank managers. Formally, they cannot increase their power as each of the owners only holds one vote, regardless of his/her stake in the bank. The core hypothesis is therefore that management turnover differs according to different types of banks. Precisely, I assume that Italian non-commercial banks, for instance cooperative banks, are less exposed to external pressure by the cooperative members and stakeholders. A number of scholars have focused on management turnover both in banks operating in the Italian market and in those operating in other European markets and internationally (see for instance Crespi et al., 2004; Ä&#x152;ihĂĄk et al., 2009). The paper that initially most influenced the design of the present study is a working paper by Battistin et al. (2006)1. The authors show that local managerial and political connections have a significant effect on Italian non-commercial banks. They argue that although disciplinary mechanisms are in place in every bank, connected top managers and top executives have a significantly lower turnover in such clusters of banks. Results from the present paper may be viewed as supportive to their findings. The present paper also tends to confirm findings from scholarly research (Ferri, Masciandro and Messori, 2001; Bongini and Ferri, 2007) on the fact that managers and boards are generally more stable in Italian cooperative banks compared to commercial banks. In fact, holding a top management 1

The final version is published in the European Economic Review (see Battistin et al., 2012). 132 129 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

position in cooperative banks decreases the risk of a manager leaving that position. Furthermore, results in the present study clearly show that the juridical form of banks as well as the institutional legacy and history matter in terms of governance specifics in banks. The remaining part of the paper is structured as follows: section 2 provides a descriptive summary of the Italian banking system with a concise account of policies aimed at improving the competition of banks and consolidating the market. Section 3 provides an account of both the research methodology used in this paper and the dataset. Section 4 provides results on management turnover obtained with an exploratory analysis at first, and then substantiated with both parametric and semi-parametric regressions. Sections 5 and 6 further discuss disciplinary mechanisms as well as specifics in the corporate governance of banks by reference to bank performance and business cycles. Finally, section 6 sets the conclusions.

2. Italian banking: institutional setting Although contemporary banking is often considered to be an industry that operates globally, it is nonetheless important to focus on specific banking groups at a country level as, in fact, some types of banks operate only locally or, at best, nationally (and “not” internationally). The specifics of banks at a country-level continue to be very important and, thus, need to be properly acknowledged (Pines, 2003). A number of scholarly accounts show that during the 1990s Italian banking was subject to significant market improvements and regulatory changes. Several banks have been privatised, and Government ownership of banks in Italy has decreased sharply – that is, from 68 percent in 1992 to less than 10 percent in 20032. The goal driving such changes was to increase the competitiveness of Italian banks, both within the country and at the European level (Angelini and Cetorelli, 2003; Messori, 2004; Carletti et al., 2005; Chiaramonte, 2007; Bini Smaghi, 2007). Among the various changes, one may focus on the following. First, an anti-trust policy was developed in order to secure and improve market competition. In particular, anti-trust for banking and market supervision was assigned to the Bank of Italy in 1990. In the same year the Amato Law – named after former Prime Minister Giuliano Amato – was signed, aimed at securing diversity in banking and thereby increasing the competitiveness of Italian banks. Finally, the Italian banking regulation (the so-called Testo Unico in materia bancaria) was signed in 1993. The outcomes from these innovations in banking regulation were a process of banking consolidation and a rearrangement in the ownership of banks. As a result of regulatory changes, mergers among banks and some technological innovations, the number of banks decreased from 1,061 to 769, in more-or-less one decade. Also, such developments increased some differences and actually drew a sharp line between banks and non-financial enterprises (Messori, 2004; Panetta, 2004). In the early 1990s, the Italian banking landscape was characterized by a large number of small banks with strong local connections and by only a small number of large banks operating at a national level. Despite some market improvements (such as a general increase in bank’s profitability without significant impairments e.g. in the availability of banking services and loans), the overall changes have been quite limited: at the end of 2005, the Bank of Italy classified most of the banks as “small” (605 out of 784) or “very small” (124) whereas those classified as “medium-sized” were relatively few (33), and large banks

2

Estimates provided by the Bank of Italy (1998-2004). 133 130 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

(11) or banking conglomerates even fewer (11)3. Therefore, it can be argued that during the 1990s and 2000s differences among banks and different banking groups persisted. This is particularly relevant for the present paper as the aim is to provide the reader with an account of the main differences between banks, particularly those between commercial and non-commercial banks in Italy. Banks, like other financial intermediaries and financial markets in general, are fundamental for the existence of a market economy. Banks traditionally perform important activities such as the reduction of transaction costs from direct finance, the transformation of short-term liabilities into long-term loans essential for firms, and the provision of payment mechanisms (Gurley and Shaw, 1960; Diamond and Dybvig, 1983; Dewatripont and Tirole, 1992). However, banks differ in the goals they pursue and in terms of how they organise their main business. In drawing to the reader’s attention that the present paper focuses on the Italian banking market, it is particularly important to distinguish between commercial and non-commercial or cooperative banks. Cooperative banks in Italy as well as in other countries are characterized by a democratic governance model. Italian cooperative banks are characterized by conservative development policies and by the dominance of relationship banking (see, for instance, Berger and Mester, 1997; Ayadi et al., 2010; Giagnocavo et al., 2012; Manetti and Bagnoli, 2013). It is argued that the evaluation of such banks should focus not only on their profit-making and capacity to make a surplus. Instead, they should be praised for effectively channelling funds to local enterprises and for securing quality of life to local communities and societies4. Italian cooperative banks, particularly mutual cooperative banks, are characterized by the following features: - they are rooted in local economies (though some popular banks operate nationally) and their model is encouraged to flourish at the local levels; - they are based on the notion of cooperative membership in which members are the primary customers of the bank5; - the voting mechanism in such banks is based on the principle “one person, one vote”. In fact, their model of governance is based on democratic member control and a rather conservative profit allocation policy. Nonetheless, management and overall governance in such banks is not free of problems. For instance, it is often argued that in Italy they are not excluded from political influence (Stefancic, 2010). Similarly, drawing on a discussion dating back to Alfred Marshall (1920), one may suggest that, as in the past, cooperative banks currently face problems in selecting and retaining the best managers. It may be argued that banks with lower management turnover have adopted more conservative management practices, development policies and business models6.

3

See the document Relazione Annuale della Banca d’Italia del 2006 sul 2005; and Chiaramonte (2007, particularly pp. 97-101).

4

Refer for instance to the principles and values of mutual cooperative banks as listed in the Carta dei valori del credito cooperativo signed in 1999. For a discussion on the focus on local economies of such banks, see also Pagano and Panunzi (1997).

5

As observed by Alexopoulos et al. (2013, p. 392), the fact that members in mutual cooperative banks can be both depositors and borrowers implies that managers should represent the contrasting interest of both.

6

In some limited cases, certain banks might simply have no alternatives to that. 134 131 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

3. Methodology and data 3.1. Survival analysis In order to study the management turnover in Italian banks, a survival analysis has been performed. Survival analysis, which is popular in biology and medicine, is a branch of statistics applied to the study of death in biological organisms or failure in mechanical systems (see for example Lee and Wenyu Wang, 2003; Alisson, 2004; Jenkins, 2005; Matter, 2012). It includes a variety of statistical methods designed to “describe, explain or predict the occurrence of events” (Alisson, 2004, p. 369). As such, it can be applied to the study of banks and their governance systems, and is particularly suitable for the study of specific issues in economics and business such as, for instance, management turnover and the duration of managerial tenure. In the present study, a turnover is interpreted primarily as a failure to keep a top position, or the result of a top manager being fired from his/her position (by contrast, a turnover resulting from retirement or voluntary step down may be rare). Survival analysis is quite similar to “duration analysis” or duration modelling used in the social sciences. Even if it is not very popular in economics and business studies, it is nevertheless gaining popularity as it has been recently applied to a number of fields including innovation strategies and the survival of new firms (Audretsch, 1991; Audretsch and Talat, 1995; Helmers and Rogers, 2008), the survival or shut-down of manufacturing plants (Bernard and Jensen, 2007), the study of bank failure and financial distress (Lane et al., 1986; Gepp and Kumar, 2008), and the study of specific topics such as recidivism (Rossi et al., 1980) to list but a few. Such a method appears to be particularly useful for the goals and purpose of the present study. Indeed, even though it is impossible to carefully distinguish between a voluntary departure (or quit) from a bank and non-voluntary turnover, the study takes into account the amount of time in which a manager stays in his/her position in the same bank. The inability to clearly distinguish the cause for the turnover is due to the fact that collected data and published sources do not evidence whether it is forced resignation, death of the manager, or retirement. Previous research suggests that this is a common problem in such types of research7. A question naturally arises whether the problem could be tackled with other, more common approaches. Common time series models (i.e. ARIMA) must be ruled out because the phenomenon we are observing is whether a manager keeps his/her position, and using a time series could only be described with a binary variable (manager keeping the position/losing the position). There are other options, though. One could think of a time series model with a binary observable variable (manager in position/not in position), dependent on a latent unobservable variable (such as “trust” by the owners). This could be interpreted as a Hidden Markov Model (HMM) or another time series model with latent variables, according to the state transition rules. It should be noted, however, that such a model would require an estimation of numerous parameters (at least a state transition matrix dependent on the covariates) and it would require timedependent covariates relative to the bank performance that are not generally available for all years and all banks in the sample for the time-frame considered. Also, some variables (i.e. bank type), whose significance is exactly what I intend to measure in the present study, would be fixed and their interpretation would thus

7

For a discussion refer to – amongst others – Brunello et al. (2003). Anecdotal evidence confirms that voluntary resignation from top management positions in Italian banks are quite rare or exceptional. 135 132 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

be problematic8. One should instead proceed to simpler models until simplicity can be traded for greater explanatory power: survival analysis provides straightforward models that abide to the statistical principle of parsimony. The core assumption is that all managers enjoy the same level of confidence from the owners/board when they are nominated into a position, and each year they risk losing their jobs. In other words, each year they are subject to risk (we may define it as “termination risk”). The levels of risk may vary in time9 and may vary depending on bank- and manager-specific effects. The magnitude of these specific effects can then be estimated, which is the main aim of the study. It must be also taken into consideration that the data available is limited to 10 years. Therefore, a non-negligible subset of the data is censored, that is, managers were already in position (for an unknown amount of years) when the observation window began and remained in position after the end of the observation window (again, for an unknown number of years). Excluding censored data would lead to bias (as censored data is more likely to be related to managers holding their position for a large number of years), and ignoring the censoring would therefore lead to bias as well (as censored data would be underestimated). Although one could model the probability of a manager stepping down or being fired as a logit/probit model, including the length of his term as a covariate, neither a logit model nor a time series framework (like a HMM) would provide a clear and simple way to deal with censored data as survival analysis models do. The distribution of the variables studied in survival analysis are not usually defined in terms of the probability density function f(t) or of the cumulative distribution function F(t). Instead, the distributions used are usually characterized by their survival function S(t) and their hazard function h(t). Both can framework (like a HMM) wouldfrom provide clearF(t). and simple way to deal be derived f(t)a and The survival function equals the probability that a component, or device, vival analysis models do. until time are t. Itnot thus equals the reliability the variables studiedsurvives in survival analysis usually defined in terms function, namely the probability that the component is function f(t) or of the distribution F(t). Instead, the the probability that a manager keeps his position from stillcumulative working at time t. Infunction our setting, it describes ally characterized by their survival function S(t) and their hazard function time 0 to time t. The survival function S(t) is defined as follows: d from f(t) and F(t). The survival function equals the probability that a

urvives until time t. It thus equals the reliability function, namely the S(t) 1 – F(t) onent is still working at =time t. In our setting, it describes the probability position from time 0 to time t. The survival function S(t) is defined as

(1)

where F(t) is the lifetime distribution (or cumulative distribution function). ilana.bodini 26/2/14 13:55 Commenta [2]:distributions posizione quote from Gepp and Kumar, “lifetime with a nonnegative random S(t) = To 1 – F(t) (1) distributions are variable that represents the lifetimes of individuals (or businesses) in some population. Lifetime distributions etime distribution (orcan cumulative distributionby function). be characterized a number of descriptor functions, the most commonly being the survival or hazard p and Kumar, “lifetime distributions are distributions with a nonnegative The survival function S(t) represents resents the lifetimesfunction. of individuals (or businesses) in some population. the instantaneous rate of failure at a certain time t. The be characterized byinterpretations a number of descriptor functions, the most of these two functions iscommonly very different, but either one can be derived from the other” (Gepp rd function. The survival function S(t) represents the instantaneous rate of and Kumar, 2008, p. 3). The interpretations of these two functions is very different, but either one The hazard ther” (Gepp and Kumar 2008, p. 3). function, on the other hand, is a measure of the risk of failure/death, and is defined as , on the other hand, is a measure follows: of the risk of failure/death, and is defined ilana.bodini 26/2/14 13:56

h(t ) =

f (t ) S (t )

(2)

Commenta [3]: pos

(2)

as the instantaneous failure rate or h(t ) = lim

Δt →0

S (t ) − S (t + Δt ) . Δt × S (t )

Effects dependent on bank characteristics could be either modelled as trend effects or as multiplicative effects. Both would mimic a natural rate of trust decay depending on bank/manager fixed characteristics, but in doing so, a survival analysis reference frame implicitly assumed. wn from the database provided by theisItalian Banking Association (ABI) 9 Annuari ABI, containingConstant information oninbanks bank managers. The function, is a special case in survival analysis where survival time is risk, or other and words, a constant hazard an 11-year dataset, which includes data to onanvirtually all distribution. available top distributed according exponential Italian banks from 1993 to 2003. The dataset contains information both on managers in Italian banks, and on the banks where they were employed. 136 133 he ABI codes for each bank, their juridical form, and the province of their JEOD - Vol.3, Issue 1 (2014) ct to the juridical form of banks, Italian banks are classified in four large cial banks (PLCs), people’s banks (Banche popolari), cooperative mutual cooperativo), and special purpose banks10. Furthermore, these typologies 8


Lifetime distributions can be characterized by a number of descriptor functions, the most commonly being the survival or hazard function. The survival function S(t) represents the instantaneous rate of failure at a certain time t. The interpretations of these two functions is very different, but either one can be derived from the other” (Gepp and Kumar 2008, p. 3). The hazard function, onInvestigating the otherManagement hand, is a Turnover measure theCooperative risk of failure/death, and is defined in of Italian Banks Stefancic, M. as follows:

h(t ) =

f (t ) S (t )

It can be interpreted as instantaneous the instantaneousfailure failurerate rate or or h(t ) = lim It can be interpreted as the Δt →0

(2)

ilana.bodini 26/2/14 13:56 Commenta [3]: pos

S (t ) − S (t + Δt ) . Δt × S (t )

3.2. Dataset 3.2. Dataset Databeen havedrawn been drawn the database provided theItalian Italian Banking Banking Association Data have fromfrom the database provided bybythe Association(ABI) (ABI) and from and from the so-called Annuari ABI, containing information on banks and bank managers. The the so-called ABI, on containing information on banks anddata bank analysis analysisAnnuari is performed an 11-year dataset, which includes on managers. virtually allThe available topis performed positions in Italian banksdata fromon 1993 to 2003.allThe dataset contains information positions both on in Italian on an management 11-year dataset, which includes virtually available top management the positions of 2,725 top managers in Italian banks, and on the banks where they were employed. banks Data fromon1993 2003.ofThe dataset contains both on theand positions of 2,725 top managers bankstoconsist the ABI codes for eachinformation bank, their juridical form, the province of their headquarters. With to thewhere juridical form of employed. banks, ItalianData banks classified in four large in Italian banks, and onrespect the banks they were onare banks consist of the ABI codes for groups or types: commercial banks (PLCs), people’s banks (Banche popolari), cooperative mutual each bank, their juridical form, and the province of their headquarters. With respect to the juridical form banks (Banche di credito cooperativo), and special purpose banks10. Furthermore, these typologies of banks, Italian banks are classified in four large groups or types: commercial banks (PLCs), people’s banks of banks are then divided into two main groups: commercial banks (PLCs and those special purpose banks which are largely commercially-oriented); and cooperative banks (Banche popolari and (Banche popolari), cooperative mutual banks (Banche di credito cooperativo), and special purpose banks10. Furthermore, these typologies of banks are then divided into two main groups: commercial banks (PLCs 10 Thespecial classification of banks is aswhich follows.are Thelargely special purpose banks typology includes and commercial land banks, and those purpose banks commercially-oriented); cooperative banks (Banche leasing, finance, medium and long term credit banks; PLCs are banks with the “SPA” denomination in Italian law: these popolari and Banche di credito cooperativo). Cases of mergers and acquisitions (M&As) of banks are mostly commercial banks, saving and loans which are not classified as special purpose banks; Banche popolari are have been profit-oriented banks numberto of prevent governanceerrors specificities (e.g.,outputs one vote per capita of M&As the numbercan of be, in fact, carefully accounted forwith inaorder in the from theirrespective analysis. shares held di by the shareholder) and thus, by reference to Bongini and and Ferri (2007, p. 20), such banks can of be classified as Banche credito cooperativo). Cases of mergers acquisitions (M&As) banks have been “cooperatives with a limited mechanism propensity to mutuality”; finally, Banche diand credito cooperativo rural (see and Jensen and thought of as a accounted disciplining or as a way of the selecting retaining top(formerly managers carefully for in order to prevent errors in the outputs from the analysis. M&As can be, in artisan banks) are mutual credit banks which are aimed to serve local communities. Ruback, 1983). Therefore, I will later focus on the6 or process of M&As in theand period undertop observation fact, thought of as a disciplining mechanism as a way of selecting retaining managersand (see Jensen and Ruback 1983). for Therefore, I will later focus banks. on the process of M&As in the period its potential disciplining mechanism top managers in Italian

under observation and its potential disciplining mechanism for top managers in Italian banks.

Table 1. Bank sample, general information

TABLE 1. BANK SAMPLE, GENERAL INFORMATION

ilana.bodini 26/2/1

Bank sample and classification codes Total number of banks in the sample

770

Number of ABI codes in the sample in 1993

716

Number of ABI codes in the sample in 2003

572

Number of independent banks in 1993

649

Number of independent banks in 2003

402

on top managers, alongthe with the name, surname of birth, include: Data onData top managers, along with name, surname and and yearyear of birth, include:

-

-­‐

the position (level of responsibility) of each listed person, namely: (a) CEO, (b) managing

director, and of (c)responsibility) (honorary) president; the position (level of each listed person, namely: (a) CEO, (b) managing director, and -­‐ the level of education of the person; (c) (honorary) president; -­‐ the starting year in managerial position X as well as (possibly) the last year in managerial - the level of education of the person; position X are specified. - the starting managerial position X as well as (possibly) the lastdata yearforin 375 managerial For year someinvariables, some data were missing. For example, positionsposition were X with reference to the “age” variable while similar problems were encountered with respect aremissing specified. to “education”. On the other hand, a number of potential variables which have been easily Forextrapolated some variables, some data have were been missing. For simply example, for 375 positions were missing from the dataset dropped duedata to their statistical insignificance: suchwith reference to case the “age” similar problems were encountered respect to “education”. is the of thevariable variable while denoting the gender of top managers, as thewith number of women present in On 11 the dataset is too low . the other hand, a number of potential variables which have been easily extrapolated from the dataset have To further substantiate the analysis and account for the performance of banks, the dataset with been dropped simply due to their statistical insignificance: such is the case of the variable denoting the

10

information on top managers has been merged with financial data obtained from the Bankscope database, specifically with data on Return on Average Assets (ROAA). Since ROAA is a measure of bank performance, it made it possible to investigate and cast some light on the disciplinary mechanisms in different groups of Italian banks. The ROAA variable has been preferred to others as The it classification of banks is as bank follows.size Thewhereas special purpose banks typology includes commercial land banks, leasing, finance, is independent from for instance the size of deposits or assets is not. Finally, medium termGDP creditfor banks; areunder banks with the “SPA” denomination in Italian law: these are mostly commercial dataand on long Italian the PLCs period investigation, obtained from the Italian national statistical banks, saving and loans which are not classified as special purpose banks; Banche popolari are profit-oriented banks with a institute ISTAT,specificities has been taken intoperaccount to assessof the the number impactof of theheld business cycle on and number of governance (e.g., one vote capita irrespective shares by the shareholder) management turnover. thus, by reference to Bongini and Ferri (2007, p. 20), such banks can be classified as “cooperatives with a limited propensity to

mutuality”; finally, the Banche di credito cooperativo (formerly rural and artisan banks) are mutual credit banks which are aimed to serve local communities. 4. Empirical analysis

Research shows that a specific feature of the Italian Banche popolari is the longer tenure of 137 134 their board members. Some scholars tend to agree on the fact that, since the focus in such banks is JEOD - Vol.3, Issue 1 (2014) mainly on longer-term business horizons, they have more stable boards of directors (Ferri, Masciandro and Messori 2001; Bongini and Ferri 2007). By reference to these papers, the

Commenta [4]: pos


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

gender of top managers, as the number of women present in the dataset is too low11. To further substantiate the analysis and account for the performance of banks, the dataset with information on top managers has been merged with financial data obtained from the Bankscope database, specifically with data on Return on Average Assets (ROAA). Since ROAA is a measure of bank performance, it made it possible to investigate and cast some light on the disciplinary mechanisms in different groups of Italian banks. The ROAA variable has been preferred to others as it is independent from bank size whereas for instance the size of deposits or assets is not. Finally, data on Italian GDP for the period under investigation, obtained from the Italian national statistical institute ISTAT, has been taken into account to assess the impact of the business cycle on management turnover.

4. Empirical analysis Research shows that a specific feature of the Italian Banche popolari is the longer tenure of their board members. Some scholars tend to agree on the fact that, since the focus in such banks is mainly on longerterm business horizons, they have more stable boards of directors (Ferri, Masciandro and Messori, 2001; Bongini and Ferri, 2007). By reference to these papers, the hypothesis that there is, indeed, a difference between the group of Italian commercial and cooperative banks which originates from their institutional setting (among other factors), is here tested. If such were the case, the institutional and juridical dimension of banks would clearly matter in terms of how we conceive and understand governance in banks as I assume that the juridical form of banks also implies different mechanisms for disciplining managers on the one hand and, on the other, different ways of selecting (or keeping) managers for top positions. I investigate the main variables explaining for the turnover of top managers in Italian banks by focusing, for instance, on their level of education, age, as well as a number of specific bank features (such as the tradition of a bank and its institutional legacy as some banks have been transformed from savings banks to PLCs). To generate and analyse survival data means to observe a sample of subjects (in our case top managers in Italian banks) over a predefined period of time, and recording whether and when the individuals experience the event, which in the present study is a step-down from the position. The main variable is measured in years as the difference between the year the person has been nominated into the position and the year the person has left the position. A continuity correction of +0.5 years has been applied. Performed analyses include: - Kaplan-Meier estimates of the general survival function for the whole sample and depending on various dichotomical and polytomical variables. These non-parametric estimates of the survival function are used as an exploratory analysis to identify the most influential factors that will be tested in the regression models. - Kernel estimates of the hazard functions. Assessing whether the distribution of survival times conforms to a known distribution is necessary for subsequent parametric models. As described in the previous chapter, distributions are characterized by their hazard functions just as they are characterized by a probability density function. - Log-rank test on the main hypothesis, namely that there is a significant difference in the duration of terms for commercial and cooperative banks. The test is meant to assess whether the main hypothesis holds, in which case it is sensible to explore the data further with regression models. 11

This is not surprising: a recent paper by the Bank of Italy confirms that the low presence of women on Italian bank boards is a persistent problem. Refer to Del Prete and Stefani (2013). 138 135 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

-

Parametric regression (Accelerated Failure Time model) on the dataset to estimate influences of bankand manager-specific factors on mean survival time; model selection based on Bayesian Information Criterion (BIC); and, finally, Cox regression as a semi-parametric alternative to AFT models, relaxing the assumptions of AFT models.

4.1. Exploratory analysis: non-parametric estimates of survival functions To begin with, general trends in terms of employment of the top managers in the sample are investigated. Figure 1 shows that around half of the terms in the sample have lasted throughout the whole observation period 1993-2003. This suggests that in the period under observation, more or less half of the top managers remained in their position in the same bank, while half of them either shifted position or have changed their employment conditions. In either case, however, they “did” step down from the original position. Figure 1. Kaplan-Meier estimates of the survival function FIGURE 1. KAPLAN-MEIER ESTIMATES OF THE SURVIVAL FUNCTION

0.6 0.4 0.0

0.2

Survival probability

0.8

1.0

Survival function

0

2

4

6

8

10

Years

An insightful question posed at the beginning of the study is whether the macro-region in which a insightful question posed at the beginning of top the managers. study is whether macro-region in bank is An located influences the probability of survival for its Figure 2the shows that the survival which a bank is located influences the probability of survival for its top managers. Figure 2 shows probability decreases more rapidly if themore mainrapidly branchiforthe headquarters legale) of the(sede bank legale) is situated that the survival probability decreases main branch(sede or headquarters inofItaly’s southern regions to central andcompared northern to regions. good explanation relates to the the bank is situated in compared Italy’s southern regions centralAand northern regions. A good explanation relates to the disappearance an independent system Italy and, higher management disappearance of an independent banking of system in southernbanking Italy and, thus,inasouthern thus, a higher management in asouthern regions as a result structures of changesand in governance management turnover in southern Italian turnover regions as result ofItalian changes in governance structures and management teams and the subsequent novelties resulting from M&As. According to teams and the novelties resulting from M&As. According to scholars such as Zazzaro, between scholars suchsubsequent as Zazzaro, between 1990 and 2000, M&As have been particularly frequent among 1990 and 2000, M&As have been particularly frequent among commercial though it may be commercial banks, even though it may be argued that cooperative banksbanks, have even not been excluded from this (Zazzaro see also Giannola argued thatprocess cooperative banks2003 haveand not2004; been excluded from this2002). process (Zazzaro, 2003 and 2004; see also Giannola, 2002).

139 136 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

Figure 2. Survival function by area of bank’s main branch

FIGURE 2. SURVIVAL FUNCTION BY AREA OF BANK’S MAIN BRANCH

0.6 0.4 0.2

Survival probability

0.8

1.0

Survival function by area of the banks' main branch

0.0

North Center South 0

2

4

6

8

10

Years

In subsequent analysis, northern and central Italian regions have been merged together in order to In subsequent analysis, northern and central Italian regions have been merged together in focus order the analysis country This is not anyto shortcoming ininthe to focus at theaanalysis at alevel. country level. This due is nottodue any shortcoming the dataset dataset oror any uneven any uneven distribution of banks. Instead, the choice is supported by the fact that analysis at a level is most distribution of banks. Instead, the choice is supported by the fact that analysis at a national national level is most interesting to economists and most relevant12to policy-makers12. interesting to economists and most relevant to policy-makers . Figure 3. FJuridical of the bank IGURE 3. form JURIDICAL FORM OF THE BANK

12

0.6 0.4 0.2

Survival probability

0.8

1.0

Survival function by juridical form of the bank

0.0

Research on cooperative banks conducted at a country level is possibly the best option for promoting a discussion on such banks with a communitarian policy framework.PLC By addressing relevant issues on cooperative banks within a Special purpose banks national banking system framework, one is able to capture the specifics and peculiarities of such banks. Indeed, regional People's banks and macro-regional differences are often significant for the Italian Cooperative banksbanking system. In addition to that, it should be recalled that Italian cooperative banks do not engage in international operations (as some commercial banks do) since their business focuses largely 0on local and2 regional markets. 4 6 8 10

10 Years

The core hypothesis in the present study, namely that there are differences in survival times of top The core hypothesis in the present study, namely that there are differences in survival times of managers cooperative and commercial banks, is outlined 3. The TheKaplan-Meier Kaplan-Meier estimate of topin managers in cooperative and commercial banks, is outlinedininFigure Figure 3. 12

estimate of the survival function clearly confirms this hypothesis: Top managers in cooperative banks are those that have the highest survival probability; by contrast, top managers in commercial banks are those that show the lowest survival probability. Top managers in people’s banks and Research on cooperative banks conducted at a country level is possibly the best option for promoting a discussion on such special purpose banks are somewhat in between. In subsequent analysis, people’s banks will be banks with a communitarian policy framework. By addressing relevant issues on cooperative banks within a national banking merged into the cooperative banks group and special purpose into the commercial banks group system framework, one is able to capture the specifics and peculiarities of such banks. Indeed, regional and macro-regional together with PLCs in order to limit the analysis to the two major banking categories. differences are often significant for the Italian banking system. In addition to that, it should be recalled that Italian cooperative Before doing that, it is interesting to first observe that there are some differences even within banks do not engage in international operations (as some commercial banks do) since their business focuses largely on local and the vast group of PLCs themselves (see Figure 4). Precisely, managers in banks which were regional markets. formerly savings banks (Casse di risparmio) have a higher probability of survival. 140 137 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

the survival function clearly confirms this hypothesis: Top managers in cooperative banks are those that have the highest survival probability; by contrast, top managers in commercial banks are those that show the lowest survival probability. Top managers in peopleâ&#x20AC;&#x2122;s banks and special purpose banks are somewhat in between. In subsequent analysis, peopleâ&#x20AC;&#x2122;s banks will be merged into the cooperative banks group and special purpose into the commercial banks group together with PLCs in order to limit the analysis to the two major banking categories. Before doing that, it is interesting to first observe that there are some differences even within the vast group of PLCs themselves (see Figure 4). Precisely, managers in banks which were formerly savings banks (Casse di risparmio) have a higher probability of survival. IGURE 4.status PLCS BY Figure 4. PLCs byFformer as FORMER savingsSTATUS banksAS SAVINGS BANKS

It is therefore helpful to reclassify the juridical form variable in order to account for the differences It is therefore helpful to reclassify the juridical form variable in order to account for the between PLCs differences with and between withoutPLCs a past a savings is shown next figure (Figure 5). withasand without abank. past asThis a savings bank. in Thisthe is shown in the next figure (Figure 5).

Figure 5. Juridical form of the bank

FIGURE 5. JURIDICAL FORM OF THE BANK

0.6 0.4 0.2

Survival probability

0.8

1.0

Survival function by juridical form of the bank

0.0

PLC Special purpose banks People's banks Cooperative banks PLC with a past as savings bank 0

2

4

12

6

8

10

Years

While top managers in cooperative banks still show a higher probability of survival, the survival probability of managers in PLCs which were formerly savings banks are somewhat closer to those in peopleâ&#x20AC;&#x2122;s banks. By contrast, top managers 141 in PLCs without a past as a savings banks and 138 special purpose banks have the lowest survival probability. This result confirms the validity of JEOD - Vol.3, Issue 1 (2014) arguments provided by Ferri et al. (2000) on the fact that turnover for top managers is lower in commercial banks which were formerly savings banks, particularly when performance is either negative or below average. It also points to the fact that not only the juridical form and the


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

While top managers in cooperative banks still show a higher probability of survival, the survival probability of managers in PLCs which were formerly savings banks are somewhat closer to those in people’s banks. By contrast, top managers in PLCs without a past as a savings banks and special purpose banks have the lowest survival probability. This result confirms the validity of arguments provided by Ferri et al. (2000) on the fact that turnover for top managers is lower in commercial banks which were formerly savings banks, particularly when performance is either negative or below average. It also points to the fact that not only the juridical form and the institutional setting, but also banks’ history and tradition influence management turnover and, presumably, their corporate governance. Managers in cooperative and formerly cooperative banks tend to remain in their position longer than managers in commercial banks. Arguably, they are exposed to lower external pressure and are subject to looser disciplinary mechanisms – something that will be further investigated later on in this paper. FIGURE form 6. JURIDICAL FORM OF THE BANK Figure 6. Juridical of the bank

In Figure 6, cooperative banks include cooperative credit banks and people’s banks; commercial banks In Figure 6, cooperative banks include cooperative credit people’s include all PLCs and special purpose banks. Differences between thesebanks two and groups appearbanks; to be rather commercial banks include all PLCs and special purpose banks. Differences between these two significant. Indeed, the figure clearly shows that the survival probability for top managers in cooperative groups appear to be rather significant. Indeed, the figure clearly shows that the survival probability for top compared managers intocooperative is higher compared to that of top managers in commercial banks is higher that of topbanks managers in commercial banks. banks.

142 139 JEOD - Vol.3, Issue 1 (2014)

14


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

Figure 7. ByFIGURE age of 7. top BYmanagers AGE OF TOP MANAGERS

ilana.bodini 26/2/14 14:1

Commenta [5]: if it’s poss of the table below, it should be by age when nominated into p

0.6 0.4 0.2

Survival probability

0.8

1.0

Survival function by person age when nominated into position

0.0

18-49 years 50-64 years 65+ years 0

2

4

6

8

10

Years

As shown in Figure 7, age is “not” a particularly significant factor when considering the survival probability As shown in Figure 7, age is “not” a particularly significant factor when considering the of top managers. The group with the lowest survival probability, aged 50-64 years when they took the survival probability of top managers. The group with the lowest survival probability, aged 50-64 position, years is alsowhen the largest one. is also important to acknowledge theimportant fact thattoforacknowledge these variables they took theItposition, is also the largest one. It is also the there fact that for these variables there are 375 cases with missing data. From the above figure, one may are 375 cases with missing data. From the above figure, one may conclude that being aged between 50 and conclude that being aged between 50 and 64 years increases the managers’ probability of leaving 64 years increases managers’ is probably due toatthe fact his or her the position. This isprobability probably dueoftoleaving the facthis thatorinher Italyposition. this is theThis average retirement age, leastthis forisbanking managers. Managers that older to keep their position longer that as onare theolder one seem that in Italy the average retirement age, atare least forseem banking managers. Managers hand they are probably viewed as valuable human resources with a significant amount of experience to keep their position longer as on the one hand they are probably viewed as valuable human resources with and knowledge; and, on the other hand, as a result of the emotional attachment that managers may a significant of experience and knowledge; haveamount for the bank after many years of service. and, on the other hand, as a result of the emotional attachment that managers may have for the bank after many years of service. IGURE 8. BYprofile EDUCATIONAL PROFILE Figure 8. By Feducational

0.6 0.4

15 0.2

Survival probability

0.8

1.0

Survival function by person's educational profile

0.0

Tertiary education Secondary education Honorific title No data 0

2

4

6

8

10

Years

Next, in Figure 8 the focus is on the connection between the educational profile and honorific titles of the manager and the probability of remaining in the position throughout the examined 143 and honorific titles have been extrapolated from 140 period. Information about the managers’ education JEOD - Vol.3, 1 (2014) the dataset. Managers that have an honorific titleIssue – such as Knights of Labour, Barons, members of the Italian Parliament, have a higher probability of remaining in their position. According to Battistin et al. (2006, p. 11), such titles are “bestowed over people that have distinguished


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

Next, in Figure 8 the focus is on the connection between the educational profile and honorific titles of the manager and the probability of remaining in the position throughout the examined period. Information about the managers’ education and honorific titles have been extrapolated from the dataset. Managers that have an honorific title – such as Knights of Labour, Barons, members of the Italian Parliament, have a higher probability of remaining in their position. According to Battistin et al. (2006, p. 11), such titles are “bestowed over people that have distinguished themselves for their service to the country or to their business”13. On the other hand, managers with only secondary education have a lower survival probability. Differences between groups, however, do “not” appear as significant, perhaps due to a number of missing data for this variable (precisely, 703 out of 2,725). FIGURE 9. BY TOP MANAGEMENT Figure 9. By top management position POSITION

ilana.bodini 26/2/14 14:12

Survival function by position type

0.6 0.4 0.2

Survival probability

0.8

1.0

Commenta [6]: posizione

0.0

President CEO Managing director Honorary president 0

2

4

6

8

10

Years

In the above figure (Figure 9) the top management positions are outlined and divided as follows: (a) the above figure(CEO); (Figure (c) 9) Managing the top management outlined president. and dividedResults as President; (b) InChief executive director;positions and (d) are Honorary from follows: (a) President; (b) Chief executive (CEO); (c) Managing director; and (d) Honorary the analysis suggest that CEOs have the most power in their hands. Therefore, becoming a CEO warrants president. Results from the analysis suggest that CEOs have the most power in their hands. Therefore, a CEO warrants the highestpositions. survival By probability top managing the highest survivalbecoming probability among top managing contrast,among honorary presidents have a positions. By contrast, honorary presidents have a low probability of survival, possibly due to the low probability of survival, possibly due for to such the age at which getmeaningful nominatedto for such athis position. It age at which people get nominated a position. It ispeople therefore reclassify variable into a dichotomous variable CEO/not CEO (Figure 10): is therefore meaningful to reclassify this variable into a dichotomous variable CEO/not CEO (Figure 10):

13

The dataset used in the present analysis is comprised of titles denoting the educational level obtained in the Italian formal education, and honorary appointments. These include the following titles: dr., dr. ing., geom., rag., cav., etc. Such is the honorary title Cavaliere del lavoro which can be translated as Knight of Labour. This title is normally awarded for excellence in industry, commerce, agriculture and in related fields.

17 144 141 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

FIGURE 10. BY TOP MANAGEMENT POSITION

Figure 10. By top management position

FIGURE 10. BY TOP MANAGEMENT POSITION

Figure 10, the difference between the two groups is quitesignificant. significant. As shown As by shown Figureby10, the difference between the two groups is quite 4.2. Testing for statistical significance

4.2. Testing for statistical significance

To further substantiate the statistical validity of the analysis performed so far and formally assess whether a difference among the two major banking groups exists, the log-rank test is To further substantiate statistical validity of the so fargroups and formally assess performed. Thebytest isthe performed on the null hypothesis thatgroups the performed twois main under As shown Figure 10, the difference between theanalysis two quite banking significant. have thethe same test is atest hypothesis test to The test whether ainvestigation difference among twosurvival major function. banking Stated groupsotherwise, exists, thethislog-rank is performed. compare distributions of two samples â&#x20AC;&#x201C; that is, the cooperative banks on the one hand, 4.2. Testingthe forsurvival statistical significance is performed on the null hypothesis two mainthe banking groups under whether investigation have the same and commercial banks on the that other.the Specifically, test aims at verifying a difference exists the survival probabilities of managers in the of twothe main of banks under survival function. Stated otherwise, thisstatistical test is a hypothesis test togroups compare the survival distributions Toinfurther substantiate the validity analysis performed so farinvestigation. and formally of two H0: survival probabilities of commercial banks and cooperative banks are the same. assess whether a difference among the two major banking groups exists, the log-rank is samples â&#x20AC;&#x201C; thatH1: is, the cooperative banks on the onebanks hand,and and commercial banks on the other.test Specifically, of on commercial cooperative banks arebanking different.groups under performed. survival The testprobabilities is performed the null hypothesis that the two main the testinvestigation aims at verifying whether a difference existsStated in theotherwise, survival probabilities of managers have the same survival function. this test is a hypothesis testintothe two TABLE 2. LOG-RANK TEST, OUTPUTS main groups of the banks underdistributions investigation. compare survival of two samples â&#x20AC;&#x201C; that is, the cooperative banks on the one hand, Nr Observed Expected and commercial banks on the other. Specifically, the test aims(O-E)^2/E atbanks verifying whether H0: survival probabilities of commercial banks and cooperative are(O-E)^2/V the same.a difference Commercial banks 1379 396 of banks 37.1 exists in the survival probabilities of managers in the 517 two main groups under investigation. H1: survival probabilities of commercial and 338 cooperative banks areare different. Cooperative banks 1346 banks 459 32.0 H0: survival probabilities of commercial banks and cooperative banks the same. Chi-squared = 90.8 on 1 of degrees of freedom banks and cooperative banks are different. H1: survival probabilities commercial p-value = 0 Table 2. Log-rank test, outputs TABLE 2. LOG-RANK TEST, OUTPUTS

H0 is rejected. As can be observed in Table 2,Expected the test shows that there is,(O-E)^2/V in fact, a difference Nr Observed (O-E)^2/E in the survival probability banking groups. Commercial banks of top managers 1379 in the two517 396 37.1 Cooperative banks

1346

4.3. Parametric and semi-parametric regressions

338

459

32.0

Chi-squared = 90.8 on 1 degrees of freedom p-value = 0

18 H0 is rejected. As can be observed in Table 2, the test shows that there is, in fact, a difference H0inisthe rejected. As can be observed in Table 2, the test shows that there is, in fact, a difference in the survival probability of top managers in the two banking groups.

survival probability of top managers in the two banking groups. 4.3. Parametric and semi-parametric regressions

18

145 142 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

4.3. Parametric and semi-parametric regressions Considering that the main hypothesis (that survival probabilities in commercial and cooperative banks differ) has been confirmed, with the null hypothesis rejected with a p-value almost indistinguishable from zero, it makes sense to explore further relationships between other variables and survival probabilities. This can be done with the use of regression models. that the main hypothesis (that survival modelling probabilities and by Gepp In survivalConsidering analysis, several statistical methods for regression areinat commercial hand. As noted cooperative banks differ) has been confirmed, with the null hypothesis rejected with a p-value and Kumar (2008), the basic difference between various survival analysis models is in the assumptions almost indistinguishable from zero, it makes sense to explore further relationships between other the set of explanatory variables. about the relationship between the hazard survival) function variables and survival probabilities. This (or can be done with the use 14 of and regression models. In survival analysis, several statistical methods for regression modelling are at hand. As noted models, namely: a) Traditionally, survival analysis has been divided into two main types of regression by Gepp and Kumar (2008), the basic difference between various survival analysis models is in the 14 Coxâ&#x20AC;&#x2122;s PH model (Cox accelerated failure time (AFT) models, and b) proportional hazards (PH), of which assumptions about the relationship between the hazard (or survival) function and the set of variables.InTraditionally, survivalthe analysis has been divided into twoand main of selection 1972) isexplanatory the most famous. the present study, AFT model is estimated first, thetypes model regression models, namely: a) accelerated failure time (AFT) models, and b) proportional hazards process is(PH), based the Coxâ&#x20AC;&#x2122;s progressive inclusion of significant covariates, final selection between of on which PH model (Cox 1972) is the most famous. with In thea present study, the AFT models is estimated first,based and theonmodel selection process is based on the progressive inclusion with all model significant covariates the Bayesian Information Criterion (BIC). The BIC isofa standard a final models with all significant based on 15 . Subsequently, the covariates Cox regression is used as a criterionsignificant for modelcovariates, selectionwith among a selection finite setbetween of models the Bayesian Information Criterion (BIC). The BIC is a standard criterion for model selection 15 semi-parametric among a alternative. finite set of models . Subsequently, the Cox regression is used as a semi-parametric alternative.

4.3.1. Hazard function 4.3.1. Hazard function AFT model is a fully parametric model. It is therefore necessary to assess the distribution The AFTThe model is a fully parametric model. It is therefore necessary to assess the distribution of the of the dependent variable (survival time) before attempting regression. This could be done with dependent variable (survival time) before This could be done by with non-parametric non-parametric estimates of the hazardattempting function, asregression. each distribution is characterized a specific hazard function. estimates of the hazard function, as each distribution is characterized by a specific hazard function. FIGURE 11. KERNEL ESTIMATES OF THE HAZARD FUNCTION

Figure 11. Kernel estimates of the Hazard function

0.15 0.10 0.00

0.05

Hazard Rate

0.20

0.25

Hazard function

2

4

6

8

10

Follow-up Time

14

As noted in the previous chapters, the hazard function and the survival function characterize a distribution much like the probability density function and the cumulative distribution function. 15 The BIC was developed by Gideon E. Schwarz in the late 1970s. See Schwarz (1978, pp. 461-464).

14

15

19 the survival function characterize a distribution much like the As noted in the previous chapters, the hazard function and probability density function and the cumulative distribution function. The BIC was developed by Gideon E. Schwarz in the late 1970s. See Schwarz (1978, pp. 461-464). 146 143 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

The general hazard function is monotonically declining and is hinting at a Weibull distribution16 for the survival time with shape parameter k<117. This means that the accelerated failure time model is appropriate for the present analysis. Additionally, it means that the results will be interpretable both as accelerated failure times and as proportional hazards. The general hazard function is monotonically declining and is hinting at a Weibull

4.3.2. AFT models16with based on with the Bayesian Information Criterion distribution for selection the survival time shape parameter k<117 . This means that the accelerated

failure time model is appropriate for the present analysis. Additionally, it means that the results will be interpretable both as accelerated failure times and as proportional hazards.

In this section, the actual AFT models are presented and briefly discussed. To start with, the independent variables4.3.2. usedAFT in the models are listedbased in the below. Information All variables except for “age_at_position” are models with selection on table the Bayesian Criterion The general hazard function is monotonically declining and is hinting at a Weibull binary/categorical, and the lowest index is used as a baseline, so the coefficient shows the effect of the 16 17 distribution for section, the survival time with parameter k<1 . and Thisbriefly means discussed. that the accelerated In this the actual AFTshape models are presented To start with, the failure time model is appropriate for thevs. present analysis. Additionally, it means resultsThe will higher indexed (i.e. variables cooperative commercial banks for table juridic_form_new2). variables, independent usedbanks in the models are listed in the below.that Allthevariables except for listed be interpretable both as accelerated failure times and as proportional hazards. are binary/categorical, lowest their indexgeographic is used as location, a baseline, the in Table“age_at_position” 3, include information on the juridical and formthe of banks, andsoinformation coefficient shows the effect of the higher indexed (i.e. cooperative banks vs. commercial banks for on top managers – that is,selection what some refer toInformation as the “demographic 4.3.2. AFT models with basedscholars on the Bayesian Criterion variables” such as age and level juridic_form_new2). The variables, listed in Table 3, include information on the juridical form of of education (Huselid, 1995). banks, their geographic location, and information on top managers – that is, what some scholars In this section, the actual AFT models are presented and briefly discussed. To start with, the refer to asvariables the “demographic such as age andtable levelbelow. of education (Huselid 1995). independent used in thevariables” models are listed in the All variables except for TableThe 3. Variables used infunction the general hazard is monotonically declining and lowest is hinting at a isWeibull “age_at_position” are models binary/categorical, and the index used as a baseline, so the 16 17 TABLE 3. V ARIABLES USED MODELS k<1 . This means that the accelerated distribution for the survival time with IN shape coefficient shows the effect ofTHE theparameter higher indexed (i.e. cooperative banks vs. commercial banks for failure time model is appropriate for the present analysis. Additionally, it means that the results will VariableThe variables, listed in Table 3, include information Definition on the juridical form of juridic_form_new2). be interpretable both as accelerated failure times and as proportional hazards. banks, their geographic location, and information on top managers – that is, what some scholars 1=commercial banks juridic_form_new2 refermodels to as with the selection “demographic as ageCriterion and level of educationbanks (Huselid 1995). 4.3.2. AFT based onvariables” the Bayesiansuch Information 2=cooperative 1=commercial banks InTABLE this section, the actual AFTINmodels are presented and briefly discussed. To start with, the 3. VARIABLES USED THE MODELS juridic_form_new3 2=commercial banks formerly independent variables used in the models are listed in the table below. All variables except for a savings bank banks (including Variable “age_at_position” are binary/categorical, and the lowest index is3=cooperative used as Definition a baseline, so the people’s banks) coefficient shows the effect of the higher indexed (i.e. cooperative banks vs. commercial banks for 1=North+Centre 1=commercial banks juridic_form_new2). juridic_form_new2 The province2 variables, listed in Table 3, include information on the juridical form of 2=South 2=cooperative banks banks, their geographic location, and information on top managers – that is, what some scholars 0=the(Huselid main branch is not in a province’s capital 1=commercial banks refer to as the “demographic variables” such as age and level of education 1995). main_branch 1= thebanks main formerly branch isa in a province’s 2=commercial savings bank capital juridic_form_new3 3=cooperative banks (including people’s banks)

TABLE 3. VARIABLES USED IN THE MODELS

position2

Variable

Definition

province2

1=commercial banks Age 2=cooperative banks

juridic_form_new2 age_at_position

position2 province2

Commenta [7]: pos

ilana.bodini 26/2/14 14:16 Commenta [7]: pos

ilana.bodini 26/2/14 14:16 Commenta [7]: pos

when promoted to a top management position

0=the main branch is not in a province’s capital education 1=commercial 1=banks the main branch is1=tertiary in a province’s capital

main_branch juridic_form_new3 education3

1=not CEO

1=North+Centre 2=CEO 2=South

ilana.bodini 26/2/14 14:1

2=commercial banks formerly a savings bank 2=secondary education 3=cooperative banks (including people’s banks) 1=not 3=honorific CEO title

2=CEO

1=North+Centre 2=South

age_at_position

Age when promoted to a top management position

main branch is not in a province’s capitalsimple to more complex (namely, with a larger Next, the models are0=the introduced, starting from 1= the main branch is in a province’s capital 1=tertiary education

main_branch

Next, models introduced, from simple to more complex with ahas larger set the of parameters being included).starting Since the null2=secondary hypothesis that the juridical(namely, form of banks no set of education education3 are 1=not CEO position2 3=honorific title rejected by the log-rank test, it is the 2=CEO effect on the survival probabilities of top managers has been parameters being included). Since the null hypothesis that the juridical form of banks has no effect on age_at_position promoted to a top management position first term to be includedAgeinwhenthe models. As we fit the model using a Weibull distribution, the 1=tertiary education the survival probabilities of introduced, topis managers been rejected by in thethelog-rank it is the first term Weibull scale parameter estimated ashas well and is included tables. Ittest, should noted that in to be Next, the models are starting from simple to more complex (namely, with abelarger 2=secondary education education3 3=honorific title set all ofin parameters included). Since the null hypothesis that the juridical formthe of Weibull banks hasscale no parameter is cases, it is being closeAs to 1we and therefore ofusing little interest. included the models. fitisthe model a Weibull distribution, effect on the survival probabilities of top managers has been rejected by the log-rank test, it is the estimated as well and is included in the tables. Itfitshould be noted that in all cases, it is close to 1 and is Next, models introduced, starting simple to more complex with ABLE 4.toMare ODEL 1 JURIDIC (COMMERCIAL -COOPERATIVE BANKS )a larger firstTthe term be included in FORM thefrom models. As we the(namely, model using a Weibull distribution, the set of parameters being included). Since the null hypothesis that the juridical form of banks has no therefore of scale littleprobabilities interest. Weibull parameter as well and is by included in the tables. be noted that in effect on the survival of is topestimated managers has been rejected is theIt should Variable Value the log-rank Std.err. test, it z-score p-value all cases, it is close to models. 1 and isAstherefore little using interest. first term to be included in the we fit theofmodel a Weibull distribution, the Weibull scale parameter is estimated as well and is included in the tables. It should be noted that in all cases, it is close to 1 and therefore of little interest. -COOPERATIVE BANKS) TABLE 4. MODEL 1 JisURIDIC FORM (COMMERCIAL

λ )k p-valuea shape parameter and λ a scale The survival function of the Weibull distribution is Std.err. , with k being S t = e −( x z-score

16 1 JURIDIC FORMVariable TABLE 4. MODEL (COMMERCIAL-COOPERATIVE BANKS ) Value

Variable

Value

Std.err.

z-scorek −1

()

p-value

k ⎛ x ⎞ ⎜ ⎟ and it is monotonically declining with k between 0 and 1. λ ⎝ λ ⎠ − ( x λ )k 16 16 The survival function the Weibull with kkand being aashape andand λ a lscale (t ) = aeshape ) is The survival function of of the Weibull is Sfound ,, with being shapeparameter parameter a scale parameter. 16 Examples Weibull hazard be the Appendix. The survival function ofof thethe Weibull distribution is functions , with k being in parameter λ a scale (distribution Sdistribution t ) = e −( x λcan parameter. The hazard function is

h(t ) =

k

17

k −1 k −1 for the juridic The hazard functions estimated k ⎛ x ⎞ forms of banks are in the Appendix. Results confirm the viability and k ⎛ x ⎞ Theparameter. hazard function is is⎜monotonically monotonically with kdeclining 0 and 1. Examples parameter. The hazard function is and with k between 0 between and 1. (t ) =method The hazard function anddeclining it isdeclining monotonically with k between 0 andof1.the Weibull hazard ( hand t )for =itit isthe ⎜ is⎟ used appropriateness of hthe analysis. ⎟ λ ⎝ λ ⎠ λ ⎝ λ ⎠

20 Examples of the Weibull hazard functions can be found in the Appendix. Examples of estimated the Weibull functions can be found in the Appendix. 17 functions can be found inhazard the Appendix. The hazard functions for the juridic forms of banks are in the Appendix. Results confirm the viability and 17 The of hazard functions appropriateness the method used forestimated the analysis.for the juridic forms of banks are in the Appendix. Results

17

confirm the viability and

of the method used for Theappropriateness hazard functions estimated forthe theanalysis. juridic forms of banks are in the Appendix. Results confirm the viability and 20 appropriateness of the method used for the analysis. 20

147 144 JEOD - Vol.3, Issue 1 (2014)


1=tertiary education 2=secondary education 3=honorific title

education3

Next, the models are introduced, startingTurnover from simple more complex (namely, with a larger Investigating Management in Italian to Cooperative Banks M. set of parameters being included). Since the Stefancic, null hypothesis that the juridical form of banks has no effect on the survival probabilities of top managers has been rejected by the log-rank test, it is the first term to be included in the models. As we fit the model using a Weibull distribution, the Weibull scale parameter is estimated as well and is included in the tables. It should be noted that in all cases, it is close to 1 and is therefore of little interest. Table 4. Model 1 Juridic form (commercial-cooperative banks) TABLE 4. MODEL 1 JURIDIC FORM (COMMERCIAL-COOPERATIVE BANKS)

16

Variable

Value

Std.err.

z-score

p-value

Intercept

1.2572

0.0988

12.7

4.22e-37

juridic_form_new2

0.6963

0.0672

10.4

3.80e-25

Log(scale)

0.0461

0.0256 − ( x λ )k

1.8

7.16e-02

The survival function of the Weibull distribution is Scale= 1.05 Weibull distribution

h(t ) = Log-likelihood(model)

parameter. The hazard function is Intercept

S (t ) = e

, with k being a shape parameter and λ a scale

k −1

k ⎛ x ⎞ monotonically declining with k between 0 and 1. ⎜ ⎟ and it is0.0988 -3313.8 12.7 4.22e-37 λ ⎝ λ ⎠1.2572

Log-likelihood(intercept) -3369.6 juridic_form_new2 0.6963 0.0672 10.4 3.80e-25 Examples of the Weibull hazard functions can be found in the Appendix. 17 Chi-squared 111.67 on 1 degr. freedom The hazard functions estimated for the juridic forms of banks are in the Appendix. confirm the viability and Log(scale) 0.0461 0.0256 1.8 ofResults 7.16e-02 appropriateness of thep-value= method Scale=0used 1.05 for the analysis.

20 As shown in Table 4, all parameters in the first model are significant, and the model as a Log-likelihood(model) -3313.8 As shown Table 4, all parameters modelratio aretest significant, and the model as a whole is whole in is significant as well, as showninbythe the first likelihood between Log-likelihood(intercept) -3369.6 the null model (intercept only) and the model. significant as well, asfull shown by the likelihood ratio test between Chi-squared 111.67the on 1 null degr. ofmodel freedom (intercept only) and the The multiplicative effect of the covariates on the risk of losing the position is as follows (the p-value= 0 full model. intercept value should be ignored): The multiplicative effect of the4,covariates on the riskfirst of losing as follows (theasintercept As shown in Table all parameters in the model the are position significant,is and the model a Intercept 0.3010368 value should be ignored): whole is significant as well, asjuridic_form_new2 shown by the likelihood ratio test between the null model (intercept 0.5143221 ilana.bodini 27/2/14 11:33 only) and the full model. Commenta [8]: pos Intercept 0.3010368 multiplicative effect ofbank the covariates theof risk of losing the position position in is as follows (the A The manager in a cooperative has half theonrisk leaving his/her a given year juridic_form_new2 0.5143221 18 intercept value should be ignored): compared to the same manager in a commercial bank . Weibull distribution

Intercept TABLE 5. MODEL 2 JURIDIC FORM AND AREA OF THE MAIN BRANCH (0.3010368 NORTH+CENTER, SOUTH)

0.5143221 A manager in a cooperative bank hasjuridic_form_new2 half the risk of leaving his/her position in a given year comparedilana.bodini 27/2/14 11:3 Variable Std.err. z-score p-value Commenta [8]: pos 18 Value to the same manager in aIntercept commercial bank . has half the 0.1307 13.25 his/her 4.39e-40 A manager in a cooperative bank1.7325 risk of leaving position in a given year 18 0.7216 0.0673. 10.72 compared to thejuridic_form_new2 same manager in a commercial bank -0.3993 0.0701 south) -5.69 Table 5. Model 2 Juridic form province2 and area of the main branch (north+center,

8.28e-27 1.25e-08

TABLE 5. MODEL 2Log(scale) JURIDIC FORM AND AREA OF THE MAIN 0.0255 BRANCH (NORTH , SOUTH) 0.0433 1.69+CENTER 9.02e-02 Scale= 1.04 Variable Weibull distribution Intercept

Value

Std.err.

z-score

p-value

1.7325

0.1307

4.39e-40

Log-likelihood(model) juridic_form_new2 Log-likelihood(intercept) province2

0.7216

13.25 -3298.1 10.72

-0.3993

0.0673 8.28e-27 -3369.6 0.0701 -5.69 1.25e-08 143.01 on 2 1.69 degr. of freedom 0.0255 9.02e-02

Chi-squared Log(scale) p-value= Scale=01.04

0.0433

Weibull distribution

All parametersLog-likelihood(model) are significant (the non-significance of the scale parameter is meaningless). -3298.1 The multiplicative effect of the covariates is as follows: Log-likelihood(intercept) -3369.6 Chi-squared p-value= 0

143.01 on 2 degr. of freedom 0.1903095 0.5010334 1.4657601

Intercept juridic_form_new2 province2

ilana.bodini 27/2/14 11:33 Commenta [9]: pos

All parameters are significant (the non-significance of the scale parameter is is meaningless). meaningless). The All parameters are significant (the non-significance of the scale parameter shown by the multiplicative effectsis of the covariates, a top manager in a bank with the TheAs multiplicative effect of the covariates as follows: multiplicative effect in of one theofcovariates is as follows: main branch the southern regions has a 46.5% higher risk of leaving his/her position. Intercept

0.1903095

juridic_form_new2 0.5010334 Intercept 0.1903095 province2 1.4657601 juridic_form_new2 0.5010334 18 In layman’s terms, an acceleratedprovince2 failure time model should be interpreted in terms of time and a proportional hazards 1.4657601 shown inbyterms the of multiplicative of the covariates, a top manager in a bankthewith should beAs interpreted hazard. In this effects specific case, a manager in a cooperative bank experiences samethe branchhis/her in one of thewithin southern hashe/she a 46.5% riskinto of leaving his/her riskmain of leaving position two regions years since was higher nominated that position as a position. manager in a

ilana.bodini 27/2/14 11:3 Commenta [9]: pos

commercial bank within the first year (or, if you prefer, four versus two years). In the proportional hazards settings

As(which shown byspecial the multiplicative ofdistribution the covariates, a top manager in athebank in this case coincides sinceeffects a Weibull is being used), it can be argued that risk of with leavingthe main position within a given year for a manager in a commercial bank is twice that of a manager in a cooperative bank. branch intheone of the southern regions has a 46.5% higher risk of leaving his/her position. 18 18

21 be interpreted in terms of time and a proportional hazards In layman’s terms, an accelerated failure time model should should be interpreted in terms of hazard. In this specific case, a manager in a cooperative bank experiences the same risk of leaving his/her position within two years since he/she was nominated into that position as a manager in a commercial bank within the first year (or, if you prefer, four versus two years). In the proportional hazards settings In layman’s(which terms,in an accelerated failure time model should be interpreted in terms of time and a proportional hazards should this special case coincides since a Weibull distribution is being used), it can be argued that the risk of leaving be interpreted in terms of hazard. this case, inbank a cooperative bank same risk of leaving his/ the position within a givenIn year forspecific a manager in aa manager commercial is twice that of a experiences manager in athe cooperative bank.

her position within two years since he/she was nominated into that 21 position as a manager in a commercial bank within the first year (or, if you prefer, four versus two years). In the proportional hazards settings (which in this special case coincides since a Weibull distribution is being used), it can be argued that the risk of leaving the position within a given year for a manager in a commercial bank is twice that of a manager in a cooperative bank. 148 145 JEOD - Vol.3, Issue 1 (2014)


Investigating Management Turnover in Italian Cooperative Banks Stefancic, M.

Table 6. Model 3 Juridic form and level of education

TABLE 6. MODEL 3 JURIDIC FORM AND LEVEL OF EDUCATION Variable

Value

Std.err.

z-score

p-value

Intercept

1.4169

0.1303

10.872

1.57e-27

juridic_form_new2 Variable

0.6099 Value

0.0787 Std.err.

7.747 z-score

9.40e-15 p-value

TABLE 6. MODEL 3 JURIDIC EDUCATION0.0595 InterceptFORM AND LEVEL OF 1.4169 0.1303 education3 -0.0432

10.872 -0.726

1.57e-27 4.68e-01

7.747 0.688 z-score -0.726 10.872 0.688 7.747

9.40e-15 4.92e-01 p-value 4.68e-01 1.57e-27 4.92e-01 9.40e-15

TABLE 6. MODEL 3 JURIDIC FORM AND LEVEL OF EDUCATION

juridic_form_new2 Log(scale) Variable education3 Scale= 1.02 Intercept

Log(scale) Weibull distribution juridic_form_new2 Scale= 1.02 Log-likelihood(model) education3

Weibull distribution Log-likelihood(intercept) Log(scale) Log-likelihood(model) Chi-squared Scale= 1.02

0.6099 0.0199 Value -0.0432 1.4169 0.0199 0.6099

0.0787 0.0289 Std.err. 0.0595 0.1303 0.0289 0.0787

-0.0432

0.0595

-2518.2 -0.726 4.68e-01 -2550.4 0.0289 0.688 4.92e-01 -2518.2of freedom 64.49 on 2 degrees

0.0199

Log-likelihood(intercept) p-value= 9.9e-15 Weibull distribution Chi-squared Log-likelihood(model) p-value= isLog-likelihood(intercept) “not” 9.9e-15 significant in

-2550.4 64.49 on 2 degrees of freedom -2518.2

EducationEducation is “not” significant in explaining the the turnover top thecoefficient coefficient has a explaining turnoverofof-2550.4 top ma