Issue 11(1) January 2011 pp. 1-19 ISSN: 1567-7141 www.ejtir.tbm.tudelft.nl
Why Do Open Rail Freight Markets Fail to Attract Competition? Analysis on Finnish Transport Policy Miika MĂ¤kitalo1 Finnish Transport Agency, Helsinki, Finland
Since the beginning of 2007, EU legislation has encouraged member states to deregulate their domestic rail freight markets. While many countries have benefited from the resultant liberalisation, in others, markets are changing fairly slowly. The incumbent railway undertaking has a total market share e.g. in Lithuania, Ireland, and Finland. This paper analyses Finnish transport policy as well as the question why deregulation has not brought competition to the rail freight transport market. The research material was collected by using the Delphi technique. This paper illustrates that competition is expected in Finland, and there is a need for structural changes. Market inequality is manifested in access to incumbentâ€™s services, governmental inactivity on market entry as well as in competition in general, traffic control organisation, and personnel training. This paper suggests that governments and governmental authorities should have an active transport policy in order to create a level playing field which could be a basis for equal and stimulating competition.
Keywords: competition, entry barriers, railway, transport policy, Delphi technique, Finland
1. Introduction The transport policy of the European Union (EU) aims at developing an economically, socially, and environmentally sustainable transport system. The goal is to cut the growth of road and aviation transport and shift the balance towards environmentally friendly transport systems. According to the EU, railways should play an important role in the transport policy change, but the European Commission has not been pleased with either the performance of the railways or the quality of the rail transport services. Surely, there are plenty of reasons for the weak status, for example, the lack of interoperability and the poor condition of the infrastructure (European Commission, 2001; European Commission, 2008). However, the European Commission (2001) has stated that opening rail freight transport markets to competition is an obligatory precondition for revitalising rail transport and creating more competitive rail freight services, and that is why rail transport policy and rail legislation have progressed towards free markets. (European Commission, 2001; MĂ¤kitalo, 2007; Hilmola, 2007; Ludvigsen, 2009) Transport policies and policy definitions are very important, as those speak out the aims and necessary actions. That is also why policy analyses are important. (Tuominen and Himanen, 2007; Tuominen et al., 2008) 1
Kotinummentie 4 A 7, FI-00700 Helsinki, Finland, T: +358452053253, E: firstname.lastname@example.org
EJTIR 11(1), January 2011, pp. 1-19 Mäkitalo Why Do Open Rail Freight Markets Fail to Attract Competition? Analysis on Finnish Transport Policy
Even though domestic rail freight transport has been open to competition in most EU Member States since the beginning of 2007, transport markets have attracted few newcomers (Ludvigsen and Osland, 2009; Mäkitalo, 2007). However, some Member States have witnessed competition although their governments have supported incumbents and thus interfered the opening of the markets (Ludvigsen, 2009). New railway undertakings rarely reach a joint market share of more than 20 per cent which is why incumbent railway undertakings have endured and are expected to remain strong, and why markets are changing slowly. (European Commission, 2009; Laisi, 2009b; Slack and Vogt, 2007; cf. Ludvigsen and Osland, 2009.) Figure 1 illustrates the market shares of incumbent railway undertakings in the EU Member States at the end of 2008. The incumbent has a 100 per cent market share in Lithuania, Greece, Ireland, Slovenia, and Finland.
Figure 1. Incumbents’ market shares. (Adapted from source: European Commission, 2009) The United Kingdom and Sweden are often mentioned as pioneers in railway deregulation. Sweden reorganised the whole sector in the late 1980s and initiated action towards competition (see e.g. Nilsson, 2002; Mäkitalo, 2007), and, in the UK, the Conservative government decided to start privatisation of British Rail in 1992 (see, e.g., Nash, 2008; Woodburn, 2007). Despite the critics and process setbacks, competition has, in general, produced positive results (see e.g. Jensen and Stelling, 2007; Hilmola, 2007; Simola and Szekely, 2009). Ludvigsen and Osland (2009) argue that there is a great difference in legislative fulfilment grades, and their claim is based on their analysis of national implementation of the European Community (EC) legislation. The European Commission has not been satisfied with the inadequate or slow implementation, and it initiated official infringement procedures against nearly all Member States. However, the lack of EC legislation implementation does not mean that there would not be competition on railways (Ludvigsen and Osland, 2009; Ludvigsen, 2009.) Ludvigsen (2009) notes that deregulation was not conducted satisfactorily in the newly liberated countries, as governments safeguarded incumbent railway undertakings, incumbents’ employees, and service continuity. Despite partial
EJTIR 11(1), January 2011, pp. 1-19 Mäkitalo Why Do Open Rail Freight Markets Fail to Attract Competition? Analysis on Finnish Transport Policy
transport policy, new companies managed to enter the markets and make profits in, for instance, Poland, the Czech Republic, and Romania. (Ludvigsen, 2009.) National implementation of the EU regulation, white papers, and policies also form a broader question of the roles and powers of the Member States, the European Commission, the European Parliament, and the European Council of Ministers. The question is how to handle co-ordination, compatibility, and incompatibility at the EU level as well as national interests and preferences. (See e.g. Apostolopoulou and Pantis, 2009; Michelsen, 2009.) These kinds of conflicts have been seen also in transport policy and regulation implementation deficiencies (Mäkitalo, 2007; Ludvigsen and Osland, 2009; Ludvigsen, 2009). In Finland, the domestic rail freight market has been opened to competition according to EU legislation, and its implementation has been completely conducted (Ludvigsen and Osland, 2009; Mäkitalo, 2007). However, despite comprehensive legislative actions, there are no new railway undertakings in the market, and the incumbent railway undertaking remains the only company in the market. A couple of firms have commenced the market entry process by applying for a safety certificate, but, later, the process has come to a standstill. (Mäkitalo, 2009; Ludvigsen, 2009.) As already said, there is no competition in some EU Member States, but it has to be remembered that incumbents have a strong role in every Member State rail freight market. That is why rail freight market openness and transport policy are important issues in the whole of Europe. In spite of the prevailing situation in Finland, competition has been expected: According to the earlier studies (Iikkanen, 2007; Mäkitalo, 2009), competition could start first in wood transport, and, subsequently, in goods transport. It is anticipated that metal and chemical industries are also active in offering competitive tenders for their transports. Nevertheless, it would be important to understand why there is no competition in Finland. Deregulation has spread competition in many countries, but why not in Finland? The aim of this study is to assess reasons for the unattractiveness of the Finnish rail freight market, and to identify the status of the rail freight transport market equality and transparency as well as to find out what kind of changes are needed in transport policy. As already stated, Finland has implemented the European Community rail transport deregulation legislation in full, and rail freight competition would be welcomed by the Finnish industry. Therefore, the hypotheses of the study could be formed as follows: As the EU legislative adherence and desire for competition do not generate competition, national orientation and politics affect market rivalry. Because of this, national transport policy may create market entry opportunities or hinder competition. As the situation in Finland in not unique, this study can also contribute to the debate on the European level. In addition, this paper discusses also how EU level policies and policy definitions are seen from a different perspective in certain EU Member States. This research aims at answering the following questions:
Is there a level playing field in the Finnish rail freight transport market?
What kinds of actions are expected from transport policy?
The paper is structured as follows: Chapter 2 describes competition and welfare economics’ socioeconomic efficiency and discusses reasons for government interventions in market forces. Market entry barriers to rail transport are also presented in chapter 2. Short introduction of the Finnish transport sector and the rail freight market are provided in chapter 3. Chapter 4 describes research methodology of this study in detail, and empirical data analysis and the results of the study are presented in chapter 5. The last chapter, in turn, reviews and discusses the results and suggests a couple of ideas for further research.
EJTIR 11(1), January 2011, pp. 1-19 Mäkitalo Why Do Open Rail Freight Markets Fail to Attract Competition? Analysis on Finnish Transport Policy
2. Competition, Socioeconomic Efficiency, and Entry Barriers Deregulation is conducted in order to boost industry’s performance, as competition forces companies to take better care of their customers (see e.g. Mankiw, 2004; Porter, 1998). In this article, the focus is on the question of how to promote competition by government actions and transport policy. Encouraging competition has not materialised in all EU Member States, but, instead, some governments have hampered it (Ludvigsen, 2009). This paper discusses what the Government of Finland could do in order to create transparency and an equal playing field. In this section, transport policy will be put into a theoretical context in order to illuminate the role of government and transport policy. This is carried out by introducing economics framework and the structure–conduct–performance (SCP) model, which elucidates the role of the government policy in the industry’s performance. According to the neoclassical microeconomics theory, the allocation of recourses is most efficient in perfect competition. Competing markets work better and are socio-economically more efficient than a monopoly. (Mankiw, 2004.) This is a logical reason why vertical separation and competition have been promoted in the EU rail transport policy (see e.g. Mäkitalo, 2007; cf. Pittman, 2007b). Competitive markets entail better customer-focus, innovations, and lower prices, also in rail transport (Ludvigsen and Osland, 2009; Mäkitalo, 2009; Laisi, 2009a). Simola and Szekely (2009) state that rail freight transport competition has created good service solutions and has lowered the price level by 20–30 per cent in Germany (see also Slack and Vogt, 2007). Rail freight transport customers have received lower prices as well in Sweden, and new business concepts have been developed. This has shifted freight transports from road to rail. (Alexandersson and Hultén, 2008.) Porter (1998) argues that a company, its strategy and competitive situation ought to be viewed and analysed on the basis of the firm’s environment. The competitive situation and environment can be seen in five forces (dimensions or threats), which, in the railway context, are 1) rail transport competition; 2) new railway undertakings; 3) substitute services from other transport modes; 4) the power of service buyers; and 5) the power of suppliers (Porter, 1998; Mäkitalo, 2010a). Porter’s first two forces refer to dividing rail transport market amongst railway undertakings, and the third can enlarge or shrink market size. Should Porter’s five forces model be scrutinised alongside government’s transport policy, it could be concluded that the industry competitive situation may be affected in three model dimensions: Contributing to the first power by setting the competitive environment, and, to the second, by lowering entry barriers. Government policy can also influence the transport mode balance, for example, by taxation policy and infrastructure investment choices. Welfare economics state that imperfect competition and imperfect market forces function as grounds for the government to intervene in competition and the market functions in order to increase socioeconomic efficiency. Central elements for causing imperfection are external effects, monopolistic competition, and public commodities. (See e.g. Corchón, 2008; Tang and Wälde, 2001; Boadway and Wildasin, 1984.) In cases of a monopoly or monopolistic competition, the market incumbent may set the price above its marginal costs, which decreases socioeconomic efficiency as well as social welfare. The government has a couple of alternatives to reduce negative effects: regulating monopolistic markets operations, cutting the price gap between the price and marginal costs by taxes or charges, and preventing monopolisation of the market. (Corchón, 2008; Boadway and Wildasin, 1984; Kerosuo, 1987.) Governments may increase socioeconomic efficiency and social welfare also by intervening in the imperfect competition of rail transport. The means of intervention are explained in transport policy. External effects may be fixed or corrected within an industry, but it is possible to correct the external effects of the whole industry. For example, railways are used to correct socioeconomic efficiency distortions, which originate from external effects of road transport.
Therefore, it is rational to finance the potential shortfall of the whole railway sector – intending especially rail network management – in order to enhance the socioeconomic efficiency and public welfare. (Kerosuo, 1987; Corchón, 2008; Suvanto, 2004; Sonstegaard, 1992; Forkenbrock, 2001; see also Conrad, 2000.) Government’s role in the increase of socioeconomic efficiency and competition is illustrated in the structure–conduct–performance (SCP) model, which has been used in neoclassical studies on the industrial organisation and on market competition. The SCP model is used in assessing how market’s basic conditions, structure, and company’s competitiveness affect the performance and the efficient resource allocation of the industry. In the model, basic conditions equal industry’s specific characteristics. The SCP model’s market structure denotes buyers and sellers on the market. The market entry and market entry barriers are also linked to the market structure. The competition activity, in turn, signifies rivalry and competition actions as well as the behaviour of the companies. In the model, the next step is industry’s performance, meaning output, and efficiency. (Scherer and Ross, 1990 see also Slade, 2004.)
Figure 2. The SCP model (Adapted from source: Scherer and Ross, 1990) In the structure–conduct–performance model, the government may affect the market structure and competition activity, and the government may, if necessary, intervene in reducing the influence of imperfect competition. Market entry barriers, especially, have a great effect on competition, as they signify that competition is imperfect and non-competitive actions may be seen. (Scherer and Ross, 1990; Audretsch et al., 2001; see also Slade, 2004; Mäkitalo, 2007; Ng and Gujar, 2009; Porter, 1998.) Scherer and Ross (1990) have stated that competition policy and market regulation actions are the most forceful tools for a government to affect market structures and the competition intensity of industries. In this paper, the focus is on government competition policy. Hilmola and Szekely (2006) argue that government policy and the actions of governmental authorities have a massive effect on the performance of the rail transport industry – just as the SCP method illustrates. That is why it is important for governments to define the ends and means for railway sector in the society. Also, necessary actions are important in order to enable the sector to meet the objectives and to increase the performance of the industry. (Mäkitalo, 2007, 2009; Ludvigsen and Osland, 2009; Ludvigsen, 2009.)
Rail transport deregulation is not an old field of research interest, as the topic has come up in the late 80’s and early 90’s when the pioneers, the UK and Sweden, started restructuring the sector and encouraged competition. Consequently, rail transport competition studies have focused on these countries, especially on the UK. (See e.g. Brewer, 1996; Cowie, 2009; Nash, 2008; Alexandersson and Hultén, 2008; Woodburn, 2007; Nilsson, 2002.) In the European Union as well as in Finland, international rail freight transport was opened to competition in March 2003, in accordance with the first railway package. As far as intensive rivalry is concerned, opening domestic rail freight transport to competition from the beginning of 2007, as the second railway package intended, was more significant. Open competition has connoted that a company meeting the regulations may enter the market. (Directive 2001/14/EC; Mäkitalo, 2009.) After deregulation in the EU, rail transport competition research has also boomed, and recently, there have been studies concentrating on the countries, which have been slow to deregulate rail transport (see e.g. Hilmola & Szekely, 2007; Ludvigsen & Osland, 2009; Ludvigsen, 2009; Szekely, 2009, Mäkitalo, 2009, 2010; Laisi, 2009b). Previous studies conducted in the Finnish rail transport market argue that despite deregulation, only small changes in the rail transport sector structure and organisation have been carried out, and governmental authorities are claimed to have been passive (Mäkitalo, 2008, 2009, 2010a; Iikkanen, 2007; Laisi, 2009b). However, at the same time, the public declaration is nearly reversed as the official definitions of transport policies have been supporting competition (e.g. Prime Minister’s Office, 2007; Ministry of Transports and Communications, 2005, 2010). The Finnish Ministry of Transports and Communications has stated (2005, 2010) that several functions need to be changed in the railway sector, including the organisation of traffic control, education, and access to services. However, these actions have not materialised (Mäkitalo, 2010a, Ministry of Transports and Communications, 2010). Furthermore, there are several studies of the market entry barriers in rail freight transport which have identified the following as barriers: railway stock acquisition, access to the railway services owned by others, actions of the incumbent railway undertaking, and recruiting skilled staff. (Nash and Preston, 1992; Bergdahl, 2005; Alexandersson and Hultén, 2005; Nordenlöw and Alexandersson, 1999; Slack and Vogt, 2007.) The issues of the market entry and entry deterrence are important in economics theories and also in competition legislation and certainly in railways as entry barriers encourage imperfect competition in which case non-competitive actions may take place (Stehmann and Zellhofer, 2004; Mankiw, 2004; Mäkitalo, 2010a). In the European Union, market entry in rail transport consists of four steps, at least from the administrative perspective. Market entry begins with applying for a safety certificate from the national safety authority. The entrant needs also an operating licence, which is valid in the whole of the European Economic Area. After the acquirement of the safety certificate and the operating licence, the entrant may send a rail capacity application for the infrastructure manager. The last phase is an access contract with the infrastructure manager concerning rail network use and certain services. (Directive 2001/14/EC; Finnish Rail Administration, 2009b.) In his recent study, Mäkitalo (2007) suggests that market entry barriers may be divided into three dimensions: financial, technical, and administrative. Previous studies (Laisi, 2009b, 2010; Szekely, 2009; Simola and Szekely, 2009) have also underlined sector-related bureaucracy as an entry barrier. It is important for competition stimulation that government is active in lowering market entry barriers (Mäkitalo, 2007, 2010a; Ludvigsen, 2009). Anyhow, there are a lot of country-specific elements (see e.g. Laisi, 2009b; Simola and Szekely, 2009). Market entry phases and market entry barriers are presented in Figure 3. In the figure, the box size portrays the difficulty of the certain phase and/or the barrier. Even though market entry is said to be difficult in Europe, market entry barriers are greater in other parts of the world, for example, in the United States and in China where rail transport markets are vertically integrated. (See Hilmola et al., 2007; Pittman, 2004.) Russia is also taking
steps towards rail transport competition, as Russian railways RZD has launched a major reform programme. The aim is to increase competition, which is partly conducted by privatising some parts of the state-owned railway company RZD. In Russia, rail transport competition has a unique form, as only RZD operates rail transport, i.e., it is the only company with locomotives that run trains. RZD offers transport services to companies, which, in turn, sell rail freight transport services, capacity, and wagons to their own customers. (Pittman, 2007a; Laisi, 2010; Mäkitalo, 2010b.)
Figure 3. Market entry phases and entry barriers in rail freight transport (Source: Mäkitalo, 2010a.)
3. The Finnish Rail Transport Market The Finnish rail transport organisations are 1) the administrative sector of the Ministry of Transport and Communications, which grants operating licences for Finland-based railway undertakings; 2) the rail network infrastructure manager, i.e., the Finnish Transport Agency, which is in charge of accessing and the usage of the network and capacity allocation, and 3) Finland’s national safety authority – the one issuing safety certificates – called the Finnish Transport Safety Agency. There are also other offices in the sector: the Finnish Competition Authority, the Accident Investigation Board, and the Prime Minister’s Office, which is in charge of the corporate governance of the state-owned incumbent railway undertaking, VR ltd. Sector players are presented in Figure 4.
Figure 4. The railway sector structure of Finland According to an argument based on transport mode characteristics, rail freight service is best suited for heavy and long transports. In Finland, typical rail freight transports are for the metal, forest, and chemical industries. The majority of transport volumes hail from raw material transport to factories and production units as well as from product transports between factories and production units and harbours (Iikkanen, 2007; Mäkitalo, 2007, 2009). In 2008, total transport volume was 41.9 tonnes, and 10.8 billion tonne-kilometres (Finnish Rail Administration, 2009a). Transport volumes by product groups are presented in Figure 5. The Finnish rail transport market share of the whole transport market measured by tonne-kilometres is 25 per cent, as the EU-27 modal split for rail transport is around 10 per cent (Lahelma, 2010; European Commission, 2009).
Figure 5. Finnish rail transport volume development by product groups 1993–2009 (Source: Lahelma, 2010)
Rail transport volumes are affected largely by the financial position and the changes in basic manufacturing industry (Finnish Forest Industries, 2010; Mäkitalo, 2007). Economic crisis hit rail transport volumes in the end of 2008 – transport volumes dropped by almost a third, which can be seen also in Figure 5. The situation accelerates structural change, as huge forest industry factories are closing down and the production is moved to locations which are closer to the markets and where production costs are lower. This has, of course, a great impact on the Finnish rail transport market. International transport to and from Russia is playing a big role in Finland’s rail transport volumes, as it covers around forty per cent of total volume. A majority (approximately ¾) of this consists of Finnish industry’s raw material transport. (Iikkanen, 2007, Iikkanen et al., 2009; Mäkitalo, 2007.) Russia has announced that they will raise raw wood export tolls. The rise to such a level will shrink rail transport flows from Russia which has an effect on Finland’s domestic traffic flows. (Iikkanen et al., 2009.) Russian-bound transport has not been deregulated, even though the industry in Finland has expected it (Iikkanen, 2007.)
4. Research Methods Deregulation of rail transport and structural changes in order to increase competition have been sensitive issues in Europe as well as in Finland (Mäkitalo, 2007; Ludvigsen, 2009). Railway sector organisations, industrial companies, and interest groups may have strong publicly expressed positions, which may complicate the collection of research data, if it is done by interviews as interviewees can stick to their official positions. In Finland, there is only the incumbent railway undertaking on the market, and any other railway undertaking’s first-hand experiences of the level playing field could not have been collected. For these reasons, the Delphi technique fitted well to the research questions and was chosen to the data collecting methods. The Delphi technique is a research material collection method used in futures studies (see e.g. Popper, 2008; Vinnari and Tapio, 2009). The technique is used for studying indefinable or multidimensional questions and for the assessment of development and futures possibilities. The method was traditionally used as a tool for forming consensus, but lately the focus has been more on arguments and dissenting views. Research data is collected from an expert panel participating in consecutive questionnaire rounds. Latter rounds’ questionnaires may reveal arguments from earlier rounds. The names of the participants may be revealed amongst the panellists, but it is characteristic and essential in the method that experts remain anonymous. (Turoff, 1975, Linstone, 1978, Aligica and Herritt 2009, Kuusi 1999.) In this research, the Delphi expert panel comprised of 52 people from all rail transport interest groups: manufacturing industry, the Finnish Ministry of Transport and Communications, the Finnish Rail Administration (the infrastructure manager) which became part of the Finnish Transport Agency after 1.1.2010, safety authority’s personnel, which became a part of the Finnish Rail Agency – after 1.1.2010 part of the Finnish Transport Safety Agency, VR Group (the incumbent railway undertaking in Finland, former Finnish State Railways), the Finnish Competition Authority, labour organisations, and other logistics specialists. The aim of the study was to find out different arguments on rail transport competition, and that goal was considered when experts were chosen. The goal of the Delhi technique data collection was to acquire material and arguments on the rail freight transport market, assess the extent of the level playing field and the need for change. In the research, two consecutive Delphi rounds were conducted in 2005. As there are still no new railway undertakings in the market in Finland, Delphi material remains valid because of the market situation. Questionnaires’ questions and statements were based on theories and the goal of the research. The questionnaire consisted mostly of open-ended questions, but quantitative
material was also collected by seven-step Likert scale. The questionnaire response rate was 77 per cent on the first round. On the second round, it was 45 per cent. The drop-off in the response rate in the second round is considerable, and the reasons can only be speculated: it may be a consequence of boredom in answering the same kind of questionnaire again, somebody might have been provoked by the questions or first-round answers, the study subject was considered dangerous, or maybe the reason is the lack of time, since answering required about an hour on the first round. However, the drop-off did not have an effect on the material, as a gamut of views was just as well balanced as in the first round. That is important in a qualitative study. In the first questionnaire round, several questions regarding a level playing field were asked. Questions and statements were formulated in a way that they would collect arguments, counterarguments and reasoning for the expressed opinion, and thus, they were leading and provocative by nature. This also made it possible to check for the response consistency of the panellists. For example, a following statement on rail traffic control was presented to the panel: The existing organisation of traffic control under VR Limited does not constitute a problem pertaining to impartial competition. After the statement, the expert evaluated it on a seven-step scale with the following: fully disagree – disagree – disagree to some extent – neither agree nor disagree – agree to some extent – agree – fully agree. After the statement evaluation, the expert was asked to argue for his/her view. Also actions of the incumbent railway undertaking were asked: According to theories of market entry barriers, the market monopolist (the existing operator in the railway markets) can in many ways promote or prevent competition. Several statements with seven-step scale and supporting arguments were given for pondering: 1) The railway market monopolist will allow access for new operators to its existing services, e.g. depots and service points, at a reasonable price, 2) The opening of competition will not influence the monopolist’s pricing, 3) The monopolist will actively try to influence the opinions and decisions of the Finnish Rail Administration and the Ministry of Transport and Communications. Besides these, experts were also asked what other means the monopolist would exploit in promoting or preventing competition. As it is typical in the Delphi technique, in the second round, the questionnaire was formulated on the basis of answers in the first round and on the round’s themes with wide divergence and important for the research goal. Also some answers and arguments in the first round were presented to the panel. Experts could also assess their opinions again.
5. Empirical Data Analysis and Results The focus of the Delphi questionnaires’ data analysis and results, as well as the research questions, lies in transport policy, and the referential literature was described in chapter 2. The most important findings are presented under sections named as follows: passiveness of the governmental offices, traffic control organising, access to services and actions of the incumbent railway undertaking. According to this research, there is no consensus on market impartiality or on necessary government actions. On the contrary, there are a lot of conflicting views on virtually anything, as expected. Different organisations and individuals have different stances on rail transport deregulation and market change. This is why the Delphi technique is a good method for collecting this kind of data. In general, expert panellists may be divided into four different groups based on their expressed opinions (Figure 6). Panellists’ qualitative and quantitative answers were coded on a scale, which illustrates the need for governmental actions on creating a level playing field. When the scale is split into four, experts fall to these groups as Figure 6 shows.
Figure 6. Delphi panellists distribution based on their view on reform for equal preconditions. Figure 6 illustrates that instead of the two opposite views – promoting competition versus safeguarding the incumbent – there is a full scale of views on rail transport market openness and equality as well as on the need for active transport policy and changes. In fact, people see the expedient need for change differently. Therefore, transport policy faces a question when and how to act on market transparency issues. Even the moderates say that changes might be made, but only when there is rivalry and a real need for resolving an issue. Experts supporting progressive views argue that all changes should be made at once, no matter whether there are newcomers on the market or not. These views and the whole gamut are also seen in the following sections. 5.1 Passiveness of the Governmental Offices As described in chapter 2, the aim of deregulation is to promote competition, and naturally government actions and transport policy have an important role to play in this. The EU rail transport policy aims are clear, but the question is how this is manifested on the national level. Were all entities coherent, governments would act in accordance with mutual decisions and EU regulations. In an ideal situation, governmental offices would also stimulate competition in compliance with the EU level objectives. However, as described earlier, some EU Member States have hindered deregulation (Ludvigsen, 2009). As the SCP model illustrates, government policies – and governmental offices’ actions on operational level – have an effect on market structure and competition activity, and, thus, on industry’s performance. The Delphi material analysis reveals that Finnish governmental offices are partly accused of partiality and passiveness on competition related tasks. This is alarming, since it suggests that offices are reluctant to recognise their role in deregulation and in increasing the performance of the railway industry, as the SCP model specifies. This study indicates that the rail transport sector has not been organised in a way that it would be described as a level playing field. This has been seen as a signal on transport policy. It is expected that government offices would be active,
customer friendly, and swift. This kind of result underlines that transport policy actions speak louder than words (see also Ludvigsen, 2009; Slack and Vogt, 2007). However, despite general doubts, the Finnish Transport Agency is regarded as a neutral infrastructure manager in capacity allocation. In this case, the Finnish Transport Agency as a governmental office is expected to act impartially which would be a natural aim of deregulation and of the SCP model. Even though, according to this study, it seems impending that when the first entrant applies for capacity, it is the infrastructure manager who has to co-ordinate all capacity requests after which one – if not all – parties shall file a claim for rectification allocated by the Regulatory Body, which is the Finnish Transport Safety Agency. However, the results imply that it is likely that the process continues in the Administrative Court and in the Supreme Administrative Court. 5.2 Traffic Control Organising Traffic control is a function, which sets train paths on the operational level and ensures that trains are running safely and on time. Traffic control is part of infrastructure management, and therefore it is a public duty. In Finland, traffic control has been organised differently compared to most European Union Member States. The Finnish Transport Agency is in charge of traffic control, but it is outsourced for the incumbent railway undertaking. View and arguments on both ends were presented. Those who feel that traffic control is just an operational performer, view that traffic control only executes what the governmental infrastructure manager has decided earlier in the capacity allocation. One possibility would be to define rules for traffic control operations, and the Finnish Transport Agency could monitor traffic control actions. However, many experts view that the existing organising of traffic control is or may be in favour of the incumbent, and therefore it is problematic for transparency and impartiality. Traffic control is considered an important governmental task, and the organisation in question and corporate governance should be neutral. If traffic control were organised as it is typically done in EU Member States – as part of the infrastructure manager (the Finnish Transport Agency in Finland) – doubts about the traffic disruption solution bias would drop dramatically. In any case, traffic control is regarded as a governmental task, and therefore it would be presumable that the organisational dependence reflected that. Furthermore, the results of this study suggest that outsourcing traffic control to a railway undertaking is not considered a transparent solution as it is regarded as a task for a governmental office. As the current situation is seen problematic as far as competition neutrality is concerned, and traffic control organising is in the jurisdiction of the government policy, it would be advisable to re-organise traffic control in order to enhance transparency and equality, and to give the right kind of signal to the rail transport market. 5.3 Access to Services At the moment, the incumbent railway undertaking, VR ltd, owns most of the services related to railway transport production. Services include, for example, depots, maintenance facilities, and terminals. The Finnish Transport Agency owns railway yards, but the incumbent owns cranes and cargo handling equipment. The Finnish railway act says that the incumbent must allow access to its services if it has service capacity to spare. There were differing opinions about a reasonable price for an access to incumbent’s services. It was expected that new railway undertakings’ access is conceivable, as it is based on legal provisions and it would be assured with monitoring. It was expected that VR ltd
would open up access especially to companies, which are working with or for the incumbent. On the other hand, there were also great doubts of access pricing impartiality, as it is expected that companies in rivalry will defend their market share. Consequently, it was regarded that access to services is difficult or overpriced without extensive monitoring or strict regulation. However, blocking access may be interpreted as a misuse of dominant market position, and it may have negative consequences based on competition legislation (see e.g. Stehmann and Zellhofer, 2004). Earlier studies (e.g. Nash and Preston, 1992; Slack and Vogt, 2007; Mäkitalo, 2010a) have regarded inaccessibility to services as a major market entry barrier. In the SCP model, market entry barriers are part of market structure, which may be effected by government policy. This is also an important topic for the government policy to address. It is important for market transparency and neutrality that there is access to services. The lack of access has a great impact on industry’s performance, as the SCP model illustrates. For transport operations, it is vital to have basic services, whereas for small railway undertakings, it is impossible to invest in nationwide service network. From the point of view of socioeconomic efficiency and the SCP model as well as the results of this study, it would be advisable for the Government of Finland to solve the market inequality problems found in this study: Access to services should be made possible by enforcing access with some of the following methods, listed in accordance with the level of governmental intervening: 1) an access guarantee with a price ceiling, 2) giving an access-monitoring task for a governmental authority, for example the Finnish Transport Agency or by 3) changing ownership of central rail transport facilities to a governmental office. This is just a question of transport policy, as the Ministry of Transport and Communications and the Prime Minister’s Office are responsible for transport and ownership policies of the state-owned incumbent railway undertaking. 5.4 Actions of the Incumbent Railway Undertaking As well as in other countries (see e.g. Slack and Vogt, 2007; Ludvigsen and Osland, 2009; Simola and Szekely, 2009), it is foreseen also in Finland that the incumbent railway undertaking has several opportunities to hamper competition. This way the incumbent can have an effect on two of the Porter’s five forces: internal rail transport competition as well as new railway undertakings – if the incumbent can give a warning signal to the entrants. Study results imply that opening of rail freight transport to competition affects transport prices. It is expected that even the threat of competition will decrease price level, as entrants have to attract customers with better price– quality level and the incumbent may lower its prices in order to keep a high market share. This is in accordance with previous studies (see e.g. Laisi, 2009b; Simola and Szekely, 2009; Iikkanen, 2007). It can be concluded that in Finland the incumbent is seen as powerful, which may hinder and complicate the realisation of competition, as it has been in Germany (Slack and Vogt, 2007). Lobbying is another anticipated way of interfering competition: It is inevitable for the incumbent to communicate with the Finnish Ministry of Transport and Communications and governmental offices. Of course, it is normal that companies have active communication with its interest group. However, it was also suspected that the incumbent would try to ensure its market position by influencing market structure, decisions, and legislation preparation, in part directly and also on the political level. National incumbents are also contributing to lobbying at the EU level via their joint communities, for example, through the community of European railway and infrastructure companies CER. The study result implies that acquisition of staff may not be easy, as the incumbent employs skilled staff, since there are no other railway undertakings. Recruiting from the incumbent is of
course a possibility, but terms have to be good. However, earlier results have indicated that acquisition of staff has actually been an easy task (see e.g. Laisi, 2009b; Nordenlöw and Alexandersson, 1999). Training staff is also a choice, but the only training centre for train drivers, for example, is governed by the incumbent. The incumbent’s Training Centre’s services are argued to be equal, but there are also a lot of doubts and speculations according to which the Training Centre services may not be neutral. Railway undertakings need personnel, and there has to be available training. The training centre in the incumbent’s organisation may not need to be regulated, if there are several training schools competing for customers. It is typical in the EU Member States that there are plenty of training schools and that it is also possible for the railway undertakings to organise training by themselves. All together, based on this, it could be said that acquisition of staff may be more difficult in Finland than in other EU Member States. Access to train driver education is important for market neutrality, and governmental offices should assure that education is available for all railway undertakings.
6. Discussion and Conclusions The transport policy of the European Union is coherent: competitiveness of rail transport needs to be improved and, in this objective, competition is a focal means (European Commission, 2001, 2008). As the SCP model suggests, government policy has a big role in boosting industry’s performance. It can be concluded that the Finnish rail freight transport market has been opened to competition, but the market has not attracted new players in three years. In this respect, it could be said that Finland has competition only on paper. As a scientific contribution to the literature, this study result demonstrates that deregulation or the fully implemented European Community legislation do not automatically bring competition to the market. This is an important implication also at the European level: the Finnish example suggests that the European Community legislation does not extend to the required actions but allows too much space for national manoeuvres. Market entry barriers are high in Europe (Laisi, 2009b; Szekely, 2009; Simola and Szekely, 2009), and maybe even higher in Finland (Mäkitalo, 2007, 2010a). Besides, this study suggests that there is an unlevel playing field in Finland. This research suggests that, as far as government policy influence is concerned, inequality and market entry problems are considered to consist of governmental passivity on market entry, access to incumbent’s services, traffic control organisation, and training. It could be estimated that, due to these issues, the Finnish rail freight transport market is partial, and market structure distortion favours the incumbent railway undertaking. These kinds of results have been found also in earlier studies (see Ludvigsen, 2009; Laisi, 2009b; Iikkanen, 2007; Mäkitalo, 2010a; Simola and Szekely, 2009). These issues are again questions of transport policy and have an effect on market openness and competition activity. Due to the suggested partiality, this study suggests that the Finnish transport policy and governmental actions are not defined on the basis of the EU transport policy aims, and reform is moderate and slightly resembles competition interference and slow policy changes found in a couple of EU Member States (Ludvigsen, 2009). Welfare economics suggest that government may improve socioeconomic efficiency by reducing negative effects of imperfect competition. The SCP model also illustrates that government policy affects industry’s performance. To achieve fair competition prerequisite necessitates that governments and governmental authorities have active transport policy in order to create an equal competition environment and a level playing field. Also from this perspective, the Government of Finland has been dilatory in re-organising railway sector and creating an impartial market, even though that has not been the stated objective (Prime Minister’s Office, 2007; Ministry of Transports and Communications, 2005, 2010). Hereby
it seems that the Finnish rail transport deregulation policy resembles Sweden’s poorly focused policy in the early years. The Swedish policy was criticised by Nordenlöw and Alexandersson (1999) (see also Alexandersson and Hultén, 2008). Socioeconomic efficiency pursuit, together with the results of this study concerning competition equality problems, emphasise the importance of an active transport policy. As an important practical implication, it should be obvious that governmental authorities – the Finnish Ministry of Transport and Communications, the Finnish Transport Agency and the Finnish Transport Safety Agency – should be active, customer friendly, and swift in order to give a positive signal on their activities on deregulation and changed situation. Competition policy and market regulatory actions are the most powerful governmental means to influence market structures and competition (Scherer and Ross, 1990). This is of, course, the case in a railway context as well; Hilmola and Szekely (2006) argue that government policy and the actions of governmental authorities have a massive effect on the performance of the industry – just as the SCP method illustrates. Welfare economics state that government may improve the socioeconomic efficiency of an industry by intervening in incomplete competition forces, which would be a basis for equal competition. Therefore, it would be expected that after conducting tasks raised in this paper, it would be more likely to have real competition on the Finnish rail freight transport market. This is also why the European Commission has been active in finding means to lower market entry barriers (European Commission, 2008; Ludvigsen, 2009; see also Laisi, 2009a, Hilmola et al., 2007). The objective of governmental policy is to improve the performance of industry, which will create welfare for the whole of society. Governmental offices should be active in carrying out actions derived from the EU and national transport policy goals. Even though government policy has a huge effect on markets as the SCP model suggests, the realisation of rail freight transport competition is not in the hands of government or governmental offices. The industry has the key to market change, as it is the customer (Iikkanen, 2007; Mäkitalo, 2007; Laisi, 2009b). It must be remembered that despite some EU Member State governments’ obstructed deregulation and competition actions (Ludvigsen, 2009b), these countries received newcomers to the rail transport markets. This research has highlighted that governmental transport policy actions would make the market more transparent and equal, and would therefore stimulate competition. As the opposite – safeguarding incumbent railway undertaking and hampering competition – would decrease market attractiveness. In this sense, it could be concluded that transport policy actions have an effect on competition and economic development. It would be also assessed in general that government policies have an effect on economy and business, and on socioeconomic efficiency and social welfare. Thus, the research supports implications of welfare economics and the SCP method. The last scientific contribution of this research is the applicability of the research methods in solving the research problem. The Delphi technique worked very well in this study, as it displayed different views, raised and highlighted many issues on market inequality. It may be possible that these kinds of results would not have been achieved by interviews, since it may not be as suitable or easy to criticise and assess market functions and governmental authorities. Also, it can be argued that the Delphi technique is workable in environments that are defined by business operations and governmental or public administration. This study has also raised a few interesting topics for further research. It would be valuable to scrutinise what other reasons there are for the slow market change. This would be important in order to understand market dynamics. Correlation between market openness and the intensity of the market competition would also be a good subject.
References Alexandersson, G. and Hultén, S. (2005). Swedish Railways: from Deregulation to Privatisation and Internationalisation in a European Context. Third Conference on Railroad Industry Structure, Competition and Investment, Stockholm, 20–22 October 2005. Alexandersson, G. and Hultén, S. (2008). The Swedish Railway Deregulation Path. Review of Network Economics, Vol. 7, no. 1, pp.18-36. Aligica, P.D., and Herritt, R. (2009). Epistemology, social technology, and expert judgement: Olaf Helmer's contribution to futures research. Futures, Vol. 41, no. 5, pp. 253-259. Apostolopoulou, E. and Pantis J.D. (2009). Conceptual gaps in the national strategy for the implementation of the European Natura 2000 conservation policy in Greece. Biological Conservation, Vol. 142, no. 1, pp. 221-237. Audretsch, D.B., Baumol, W.J. and Burke, A.E. (2001). Competition policy in dynamic markets. International Journal of Industrial Organization, Vol. 19, no. 5, pp. 613-634. Bergdahl, P. (2005). Six deregulations. Liberalisation of the markets for electricity, postal services, telecommunications, domestic air traffic, rail and taxi services in Sweden. publications of Swedish Agency for Public Management 8/2005. Boadway, R.W. and Wildasin, D.E. (1984). Public Sector Economics, Boston, MA: Little, Brown & Company. Brewer, P.R. (1996). Contestability in UK Rail Freight Markets. Transport Policy, Vol. 3, No. 3, pp. 91-98. Conrad, K. (2000). Competition in Transport Models and the Provision of Infrastructure Services. Journal of Transport Economics and Policy, Vol. 34, no. 3, pp. 333-358. Corchón, L.C. (2008). Welfare losses under Cournot competition’, International Journal of Industrial Organization. Vol. 26, no. 5, pp. 1120-1131. Cowie, J. (2009). The British Passenger Rail Privatisation. Conclusions on Subsidy and Efficiency from the First Round of Franchises. Journal of Transport Economics and Policy, Vol. 43, No. 1, pp. 85–104. Directive 2001/14/EC of the European Parliament and of the Council of 26 February 2001 on the allocation of railway infrastructure capacity and the levying of charges for the use of railway infrastructure and safety certification. European Commission (2001). White Paper: European transport policy for 2010: time to decide, COM(2001) 370 final. European Commission (2008). Communication from the Commission to the Council and the European Parliament: The quality of rail freight services, COM(2008) 536 final. European Commission (2009). Second Report on Monitoring Development of the Rail Market from the Commission to the Council and the European Parliament, COM(2009) 676 final. Finnish Forest Industries (2010). Structural change affects Finland as well. Available at URL: http://www.forestindustries.fi/Infokortit/structuralchange/Pages/default.aspx, the Finnish Forest Industries Federation, Retrieved February, 2010. Finnish Rail Administration (2009a). Finnish Railway Statistics 2009, publications of the Finnish Rail Administration. Finnish Rail Administration (2009b). Finnish Network Statement 2011, publications of the Finnish Rail Administration F 6/2009.
Forkenbrock, D. J. (2001). Comparison of external costs of rail and truck freight transportation. Transportation Research Part A: Policy and Practice, Vol. 35, no. 4, pp. 321-337. Hilmola, O-P. (2007). European railway freight transportation and adaptation to demand decline – Efficiency and partial productivity analysis from period of 1980–2003. International Journal of Productivity and Performance Management, Vol. 56, no. 3, pp. 205-225. Hilmola, O-P. and Szekely, B. (2006). Deregulation of Railroads and Future Development Scenarios in Europe – Literature Analysis of Privatization Process Taken Place in US, UK and Sweden, Lappeenranta University of Technology, Department of Industrial Engineering and Management, Research Report 169, 2006. Hilmola, O-P., Ujvari, S. and Szekely, B. (2007). Deregulation of railroads and future development scenarios in Europe: analysis of the privatisation process taken place in the USA, the UK and Sweden. World Review of Intermodal Transportation Research, Vol. 1, no. 2, pp. 146-169. Iikkanen, P. (2007). Rautateiden tavaraliikenteen kilpailun kohdistuminen ja vaikutusten arviointi, Strategies and studies by the Finnish Rail Administration 1/2007. Helsinki: Finnish Rail Administration. [The Realisation of Competition in Railway Goods Traffic and an Evaluation of Its Effects, in Finnish] Iikkanen, P., Mukula, M., Kosonen, T. and Kiuru, T. (2009). Raakapuun terminaali- ja kuormauspaikkaverkon kehittäminen, publications of the Finnish Rail Administration A 4/2009. [Developing of raw wood’s terminal and loading area network, in Finnish] Jensen, A. and Stelling, P. (2007). Economic impacts of Swedish railway deregulation: A longitudinal study. Transportation Research Part E, Vol. 43, no. 5, pp. 516-534. Kerosuo, M. (1987). Valtion liikelaitoksen tulostavoitteen asettaminen ja yhteiskuntataloudellinen tehokkuus, Licentiate dissertation, University of Helsinki. [Setting a Profit Target to Public Enterprises, in Finnish] Kuusi, O. (1999). Expertise in the Future Use of Generic Technologies. Epistemic and Methodological Considerations Concerning Delphi Studies. Doctoral dissertation, Helsinki School of Economics. A159. Lahelma, H. (2010). Finnish Rail Freight Statistics. Presentation material by Senior Officer Harri Lahelma from the Finnish Transport Agency. Laisi, M. (2009a). Swedish and Polish railway freight operators’ market entry’, in Hilmola, O-P. and Korovyakovsky. E. (Eds.): North-European Logistics in the Era of Global Economic Turmoil, Fifth International Railway Logistics Seminar. Kouvola, Finland 2 June 2009, Lappeenranta University of Technology, Department of Industrial Management, Research Report 213, pp. 119-131. Laisi, M. (2009b). Market Entry Strategies and Confronted Barriers on Liberalized Railway Freight Markets in Sweden and Poland. publications of the Finnish Rail Administration A 11/2009. Laisi, M. (2010). Boosting Business Opportunities by Understanding the Russian Railway Freight Market’s Peculiarities. 16th International Working Seminar on Production Economics, 1-5 March 2010 Innsbruck, Austria. Linstone, H.A. (1978). The Delphi Technique’, in: Fowles, J. (ed.) Handbook of Futures Research, (pp.273–300). Greenwood Press, Westport, CT. Ludvigsen, J. (2009). Liberalisation of Rail Freight Markets in Central and South-Eastern Europe: What the European Commission Can Do to Facilitate Rail Market Opening. European Journal of Transport and Infrastructure Research, Vol. 9, no. 1, pp. 46-62.
Ludvigsen, J. and Osland, O. (2009). Liberalization of Rail Freight Markets in the Old and New EU-Member States. European Journal of Transport and Infrastructure Research, Vol. 9, no. 1, pp. 3145. Mankiw, G.N. (2004). Principles of Economics, Thomson South-Western, Mason, OH. Michelsen, J. (2009). The Europeanization of organic agriculture and conflicts over agricultural policy. Food Policy, Vol. 34, no. 3, pp. 252-257. Ministry of Transport and Communications (2005). Rautatiepolitiikan linjaukset, 5th December 2005, the Finnish Ministry of Transport and Communications. Available at http://www.mintc.fi/fileserver/upl646-Rautatiepolitiikan%20linjauksia.pdf Retrieved June 2010. [The Railway Policy Definitions, in Finnish] Ministry of Transport and Communications. (2010). Rautateiden henkilöliikenteen avaaminen kilpailulle: edellytykset ja etenemispolku, publications of the Finnish Ministry of Transport and Communications 17/2010. [Opening passenger rail transport to competition – requirements and further steps, in Finnish] Mäkitalo, M. (2007). Market Entry and the Change in Rail Transport Market when Domestic Freight Transport Opens to Competition in Finland. Doctoral dissertation, Tampere University of Technology, publication 702, http://URN.fi/URN:NBN:fi:tty-200810021074. Mäkitalo, M. (2009). The rise of competition in Finland’s rail transport’, in Hilmola, O-P. and Korovyakovsky. E. (eds.) North-European Logistics in the Era of Global Economic Turmoil, Fifth International Railway Logistics Seminar. Kouvola, Finland 2 June 2009, Lappeenranta University of Technology, Department of Industrial Management, Research Report 213, pp. 101-109. Mäkitalo, M. (2010a). Market entry barriers in Finland’s rail freight transport. World Review of Intermodal Transportation Research, Vol. 3, Nos. 1/2, pp.181-195. Mäkitalo, M. (2010b). Finland–Russia rail transport deregulation: future scenarios of market development. International Journal of Shipping and Transport Logistics, accepted, forthcoming. Nash, C. (2008). Passenger Railway Reform in the last 20 Years – European Experience Reconsidered. Research in Transportation Economics, Vol. 22, no.1, pp. 61-70. Nash, C. and Preston, J.M. (1992). Barriers to Entry in the Railway Industry, Working paper 354, Liverpool University. Nilsson, J-E. (2002). Restructuring Sweden’s railways: the unintentional deregulation. Swedish Economic Policy Review, Vol. 9, no. 2, pp. 229-254. Ng, K.Y.A. and Gujar, G.C. (2009). Government policies, efficiency and competitiveness: The case of dry ports in India, Transport Policy, Vol. 16 no. 5, pp.232-239. Nordenlöw, L. and Alexandersson, G. (1999). Standing in the Shadow of the Giants. Conditions for Entry and Survival of Small Businesses on the Deregulated Bus and Railway Markets in Sweden. Sixth International Conference on Competition and Ownership in Land Passenger Transport, Cape Town, South Africa, 19–23 September 1999. Pittman, R. (2004). Chinese Railway Reform and Competition. Lessons from the Experience in Other Countries. Journal of Transport Economics and Policy, Vol. 38, no. 2, pp. 309-332. Pittman, R. (2007a). Make or buy on the Russian railway? Coase, Williamson, and Tsar Nicholas II. Economic Change and Restructuring, Vol. 40, no. 3, pp. 207-221. Pittman, R. (2007b). Options for Restructuring the State-Owned Monopoly Railway. Research in Transportation Economics, Vol. 20, pp. 179-198. Popper, R. (2008). How are foresight methods selected. Foresight, Vol. 10, no. 6, pp. 62-89.
Porter, M.E. (1998). Competitive Strategy. Techniques for Analyzing Industries and Competitors, The Free Press, New York, NY. Prime Minister’s Office (2007). Government Programme of Prime Minister Matti Vanhanen’s second Cabinet, publications of the Finnish Prime Minister’s Office. Scherer, F.M. and Ross, D. (1990). Industrial Market Structure and Economic Performance, Boston, MA: Houghton Mifflin Company. Simola, M. and Szekely, B. (2009). The liberalization process in Europe. Market entry barriers versus competition stimulation – cases of Germany and Hungary, publications of the Finnish Rail Administration A 20/2009. Slack, B. and Vogt, A. (2007). Challenges confronting new traction providers of rail freight in Germany. Transport Policy, Vol. 14, no. 5, pp. 399-409. Slade, M.E. (2004). ‘Competing models of firm profitability’, International Journal of Industrial Organization, Vol. 22, no. 3, pp. 289-308. Sonstegaard, M.H. (1992). Balance and coordination for road and rail. Transportation Research Part A: Policy and Practice, Vol. 26, no. 5, pp. 419-432. Stehmann, O. and Zellhofer, G. (2004). Dominant rail undertakings under European competition policy. European Law Journal, Vol. 10, no 3, pp.327-353. Suvanto, T. (2004). Rautatieliikenteen haasteita, presentation at Ministry of Transport and Communications. [Challenges in Railway Policy, in Finnish] Szekely, B. (2009). The process of liberalizing the rail freight transport markets in Hungary: market entry conditions and competition dynamics. in Hilmola, O-P. and Korovyakovsky, E. (eds.): North-European Logistics in the Era of Global Economic Turmoil, Fifth International Railway Logistics Seminar, Kouvola, Finland 2.6.2009, Lappeenranta University of Technology, Department of Industrial Management, Research Report 213, pp. 59-71. Tang, P.J.G. and Wälde, K. (2001). International competition, growth and welfare. European Economic Review, Vol. 45, no. 8, pp. 1439-1459. Tuominen, A., Himanen, V. (2007). Assessing the interaction between transport policy targets and policy implementation – A Finnish case study. Transport Policy, Vol. 14, no. 5, pp. 388-398. Tuominen, A., Leonardi, J. and Rizet, C. (2008). Assessing the Fitness-For-Purpose of strategic transport research in support of European transport policy. European Journal of Transport and Infrastructure Research, Vol. 8, no. 3, pp. 183-200. Turoff, M. (1975). The Policy Delphi. in: Linstone, H.A., Turoff, M. (eds.) The Delphi Method, Techniques and Applications, Addison-Wesley, Reading, MA, pp. 84-101. Vinnari, M. and Tapio, P. (2009). Future images of meat consumption in 2030. Futures, Vol. 41, no. 5, pp. 269-278. Woodburn, A. (2007). Appropriate indicators of rail freight activity and market share: A review of UK practice and recommendations for change. Transport Policy, Vol. 14, no. 1, pp. 59-69.
Issue 11(1) January 2011 pp. 42-60 ISSN: 1567-7141 www.ejtir.tbm.tudelft.nl
Measuring Spatial Separation Processes through the Minimum Commute: the Case of Flanders Kobe Boussauw1, Ben Derudder2 and Frank Witlox3 Geography Department, Ghent University
average distance covered by individual commuting trips increases year after year, regardless of the travel mode. The causes of this phenomenon are diverse. Although increasing prosperity is often invoked as the main reason, the discipline of spatial planning also points to the relevance of land-use policies that enable processes of suburbanization and sprawl. By calculating time series of spatially disaggregated theoretical minimum commuting distances, this paper offers a method to identify and quantify the process of spatial separation between the housing market and the job market. We identify the detected spatial separation as one of the possible indicators for the contribution of spatial processes to the growth of traffic. In the case study area of Flanders and Brussels (Belgium), it is found that over time the minimum commuting distance increased in many municipalities, especially where population is growing faster than job supply, or where traditionally high concentrations of employment still increase. Decreases are noticed in suburban areas that are getting a more urban character by acquiring a considerable functional mix. For the study area in its entirety, we do indeed register an increasing spatial separation between home and work locations. However, this separation evolves less rapidly than the increase in commuting distances itself. Regarding the methodology, we find that the use of municipalities as a spatial entity is suitable for grasping regional transformations of the economy and intermunicipal forms of suburbanization and peri-urbanization. However, a similar methodology, applied at a more detailed geographical scale, could be used to detect processes of sprawl in the morphological sense. Keywords: excess commuting, Flanders, sustainable spatial development, urban sprawl
1. Introduction Over the last century, there has been a mutually reinforcing relation between rising prosperity and a general increase in individual mobility: the growth of traffic volumes are both an effect and a cause of mounting affluence. For instance, faster transport modes (especially but not exclusively, the car) have consistently gained ground because the individual budget spent on transport has in absolute terms continuously been growing. The net result is that the average distance covered per person has been systematically increasing (Bleijenberg, 2003). Krijgslaan 281/S8, 9000 Gent, Belgium, T: +3292644555, F: +3292644985, E: email@example.com Krijgslaan 281/S8, 9000 Gent, Belgium, T: +3292644556, F: +3292644985, E: firstname.lastname@example.org 3 Krijgslaan 281/S8, 9000 Gent, Belgium, T: +3292644553, F: +3292644985, E: email@example.com 1 2
EJTIR 11(1), January 2011, pp. 42-60 Boussauw, Derudder and Witlox Measuring Spatial Separation Processes through the Minimum Commute: the Case of Flanders
With regards to the commute, this means that, in general, workers have been looking for daily occupations increasingly further away from their home, or â€“ conversely â€“ that they have been moving house further away from their jobs. The logical result is a stronger separation between house and job location: if travel costs decrease, then travel consumption increases (Rietveld and Vickerman, 2003). This is especially true when the spent amount of time can be kept constant by increasing average travel speed, which seems generally to be the case in the Western world (Schafer, 2000). Interestingly, this evolution apparently has an important spatial component. That is, it looks like changing travel behaviour is partly materialized in suburban and peri-urban developments, implying that possible origins and destinations lost relative proximity to each other. In many regions in the developed world, this materialization may be responsible for a certain degree of irreversibility of the expansion of travel patterns, as this limits the possibility of travel distances being reduced again in the future, even if environmental or congestion policies would aim for this. Since the adverse external effects of traffic are linked to the distance travelled (especially when travel is by car), this evolution makes it harder to address traffic problems at the source. In addition, spatial separation of functions is probably a long term economic disadvantage, since it makes the economic system vulnerable to possible future circumstances where transport costs would increase, for example by rising oil prices (peak oil) (Boussauw and Witlox, 2009) or growing structural congestion. The perspective of this paper is the study of the mobility component of spatial separation of the job and housing market. Our research question is: how can spatial separation between residential locations and job locations be measured by means of a spatial proximity indicator, and what is the connection with the growth of observed commuting distances? We hereby hypothesize that the well-established line of inquiry on excess commuting (Ma and Banister, 2006) can provide a methodology to quantify evolutions of spatial separation in a functional sense, and to provide better insight in the contribution of spatial structure to the growth of traffic volumes over the last decades. In practice, this will be done through calculating spatially disaggregated evolutions in minimum commuting distances, defined as the theoretical minimum distance that each worker would have to cover in order to find a job as close to home as possible under the assumption that actual residential locations and job locations are maintained and the total distance travelled (by all workers together) is minimized. After applying the method to a case study area in northern Belgium, we evaluate the observed evolution based on both occurring sprawl and regionaleconomic shifts in the labour market, aiming to explain regional variations. Although assessing time series of the minimum commuting distance in order to describe spatial separation between the housing market and the labour market has been done before (Frost et al., 1998; Horner, 2007; Yang, 2008), the approach adopted in his paper is novel for two reasons. First, we adopt a study area which could be qualified as a suburbanized historically polycentric spatial structure - as there are many in Europe - and shows therefore significant spatial variations (i.e. it includes many historical urban areas, suburban and peri-urban developments, and major industrial zones), although it can be demarcated in a fairly consistent way based on economic and political criteria. Second, we refine the method in order to detect variations in the minimum commuting distance in a spatially disaggregated way, which certainly contributes to a critical understanding of the role of the present variations of spatial structure in comparison with Frost et al. (1998), Horner (2007) or Yang (2008) who define only one figure for the entire city or region. The remainder of this paper is structured as follows. First, we clarify the possible reasons behind evolutions in spatial proximity by putting this in the perspective of the existing literature. Second, we present an overview of the case study area (Flanders and Brussels), including the main socio-economic and spatial processes and changes that may play a role in our analysis.
EJTIR 11(1), January 2011, pp. 42-60 Boussauw, Derudder and Witlox Measuring Spatial Separation Processes through the Minimum Commute: the Case of Flanders
Then we develop our methodology, including calculation and evaluation methods, and discuss the available data. After applying the method, we present the results, discuss possible biases and draw some final conclusions.
2. Defining spatial separation processes One can think of several possible reasons leading to changes in spatial proximity between homes and jobs. For planning policies, the most relevant possible causes are suburbanization and sprawl. Ewing (1994) assumes a strong link between sprawl and increased traffic. But also thorough zoning policies may lead to increasing mutual distances. These phenomena are local in nature and manifest themselves at a small geographical scale, meaning that in quantitative analysis, these will particularly manifest when data are available at a detailed spatial division (e.g. traffic analysis zones). Newman and Kenworthy (1989) and Gordon and Richardson (1989) associate sprawl with a specific morphology, particularly consisting of monotonous suburban districts with a strict separation of functions, characterized by store strips, commercial architecture and large internal distances. However, the extent to which spatial separation leads to an effective increase of distances that need to be covered is less clear, since this cannot be derived from local morphological characteristics. For instance, a monotonous residential lot embedded in a major employment centre could possibly lead to a more sustainable travel pattern than a compact town that is immersed in a rural area. Therefore it is important to view spatial separation not only in a morphological way but also in a functional sense. Particularly in a context where average trip lengths, and in particular average commuting distances, have become very large in practice, it is hard to tell which kind of spatial developments are problematic in relation to mobility, and which are rather beneficial. Banister (1999) therefore observes the issue at hand in a non-morphological way, and suggests that it is of the utmost importance that new developments are sufficiently large and are located in or immediately subsequent to existing urban areas. As a consequence, local morphology, density and spatial diversity come only in second place. Another phenomenon that could cause this kind of functional separation processes, although particularly at a larger geographical scale level and consequently only partly related to suburbanization processes, consists of regional economic shifts within the labour market. An economic transformation, e.g. towards a more service-based industry, could lead to a different spatial distribution of jobs, e.g. through centralization. But also an absolute increase in the number of jobs in one zone can lead to a spatial distribution where work locations are on average closer to residences of potential employees. Since these kinds of regional transformations take place at a higher geographical scale than suburbanization processes, these will rather become clear when relatively large zones (e.g. districts or municipalities) are studied. If we restrict ourselves to commuter traffic, we see that trip lengths have increased systematically over the past decades. This trend was observed in Belgium (see below), the U.S. and the U.K., and we may assume that this evolution is manifest throughout the Western world (Aguilera, 2005). The basic mechanism that underlies this growth is an increase in travel speed. Basically, people do not spend more time commuting than they did before, nor do they spend a greater proportion of their income on transport (Schafer, 2000). Rather, it is the general prosperity growth that has led to an absolute increase in resources devoted to transport, resulting in more car ownership, more car use (at the expense of slower transport modes), an extension of the motorway network and a pushed up speed of public transport. In the surroundings of large agglomerations, congestion has slightly slowed down the overall increase of travel speed, although it did certainly not stop it (Van Wee et al., 2002).
EJTIR 11(1), January 2011, pp. 42-60 Boussauw, Derudder and Witlox Measuring Spatial Separation Processes through the Minimum Commute: the Case of Flanders
Even though spatial separation processes and increasing trip lengths appear to be associated, this does not necessarily imply a one-way causality. The fact that employees live on average further from their work location than before, a phenomenon which occurs partly in the form of sprawl, is caused by increasing individual mobility, i.e. a wealth-related phenomenon that is the basis of an autonomous growth of traffic. So, we might explain these spatial separation processes as a materialization of this increased mobility, which has itself a mutual reinforcing effect on the growth of traffic.
3. Measuring spatial separation by excess commuting characteristics The perceived increase in traffic volume due to spatial separation processes, whether or not perceived as sprawl, remains a main concern in most spatial policy plans (Sultana and Weber, 2007). Given the importance attached to mobility, it may be surprising that quantification of this phenomenon is usually confined to measuring morphological characteristics. However, in this paper we want to measure the process of spatial dispersion that results in further separating origins and destinations over the years, or, as might be the case in some areas, decreasing separation between origins and destinations. By measuring this development over a certain time interval, we have a good idea what proportion of the traffic growth in that period is due to spatial expansion. An important methodological issue concerns the definition of these origins and destinations. This obstacle disappears when restricting ourselves to the commute, where origins (residential locations of workers) and destinations (job locations) can be clearly defined. Cross-sectional analyses that compare travel behaviour of commuting residents or workers in suburban areas with commuting behaviour in areas with urban characteristics are common in the literature. Sultana and Weber (2007), for instance, recorded large differences in actual travel times and trip lengths, depending on the direction of the commuter flow. Flows within urban areas appear to be the shortest, both in distance and in time. Commuting trips within the suburban areas, however, appear shorter than commuting trips from the suburbs towards the city. In order to map the spatial separation between the housing market and the labour market, regardless of actual commuting patterns, other methods are needed. An elementary indicator of the spatial distribution of housing and jobs with an alleged impact on mobility is the jobshousing balance (the ratio between the number of jobs in an area and the number of workers living in the same zone). In its simplest form, this indicator is measured by zone, making it sensitive to variations in zoning and insensitive to effects of embeddedness in the surrounding region or the presence or absence of transport infrastructure. Peng (1997) tried to overcome these problems partly by calculating the jobs-housing balance based on the surrounding area of which every considered zone is the centre. A more advanced line of inquiry is the field of â€˜excess commutingâ€™, which originally focuses on the study of spatial structure in relation to commuting behaviour by comparing observed commuting distances with theoretical calculated minimum commuting distances. An overview of existing research on this topic is given by Ma and Banister (2006). One of the spatial characteristics, provided by the excess commuting literature, is the minimum commuting distance, which incorporates both the spatial distribution of jobs and housing, and the infrastructure network. It is related to the job-housing balance, but is a more sophisticated indicator as it also takes into account the embeddedness of the study area in the wider region, as well as the transport network (Horner and Murray, 2003). We consider the minimum commuting distance as a measure of the proximity of the housing market to the labour market (Horner, 2004), or as a variable that indicates to what extent the system could absorb shrinkage of commuting distances. The latter is relevant when we want to understand the potential impact of the
commute getting more expensive, e.g. under the influence of rising fuel prices (Boussauw and Witlox, 2009). The calculation of the minimum commuting distance implies that the existing residential and job locations are retained, but that workers are assigned to jobs in a way that leads to a minimization of the total distance travelled. More formally: minimize H
d t i 1 j 1
tij D j
t j 1
H = total distance travelled within the system by matching workers and jobs
n = number of zones Oi = number of workers living in zone i Dj = number of jobs in zone j dij = network distance between centroids of zone i and zone j tij = number of trips between zone i and zone j Ma and Banister (2007) published an in-depth study on the relationship between excess commuting and urban form. Horner (2007) and Yang (2008) used also the maximum commuting distance as a measure of spatial dispersion, while Charron (2007) and Yang (2008) introduced the terms "proportionally matched commute" and "random average commute", based on a more extensive theoretical exploration on possible commuting ranges, and proposed as more realistic alternatives to the maximum commuting distance. The maximum commuting distance should be seen as the conceptual inverse of the minimum commuting distance: it is the value obtained when all employees within the study area simultaneously would exchange jobs or houses aiming to live as far as possible from their jobs. In an urban system consisting of a single city that is isolated from neighbouring cities, this measure is a good complement to the minimum commuting distance. In the context of our assessment, however, the notion of maximum commuting distance is rather abstract. Not only will employees never be inclined to spontaneously maximize home to work trip length, also outcomes seem to be largely dependent on the considered region frontier and are highly correlated with the size of the study area (Charron, 2007). Moreover, the amount of traffic that is associated with a maximized commuting distance is not compatible with the existing transport infrastructure or any form of pursuing economic efficiency. In this paper, we will calculate the global value of the maximum commuting distance as an additional support of the observed trends in minimum commuting distance, but we will not elaborate on spatially disaggregated values. The contribution of a time series approach of the proportionally matched commute is mainly in quantifying the extent to which a regional system evolves from mono-centric to dispersed (Yang, 2008). Since the possible interpretation of the meaning of spatially disaggregated values of the proportionally matched commute in a polycentric region is subject to further research that is beyond the scope of this paper, we will no further elaborate on this. In the literature, the average minimum commuting distance is usually calculated as a single measure for an entire city or region (Ma and Banister, 2006). However, it is also possible to calculate this in a spatially disaggregated way in order to reveal regional variations. In the latter case, for each zone (traffic analysis zone, municipality, district ...) the distance outcome is calculated twice: once for the outgoing commute, and once for the incoming commute.
Niedzielski (2006) elaborated on a spatially disaggregated analysis of Warsaw on the basis of minimum commuting distances. When taken the locations of residence as a viewpoint, the largest minimum commuting distances were recorded in areas with suburban characteristics. Viewed from the perspective of job locations, the highest values are recorded downtown (CBD). Yang and Ferreira (2009) included spatially disaggregated values in their information system for planners in the Boston area. A similar analysis, albeit on a regional scale, was made for Flanders and Brussels (Belgium) by Boussauw et al. (2011). A spatially disaggregated calculation of the minimum commuting distance on cross-sectional data allows obtaining these values per zone for a given point in time. In a next step, a similar analysis based on time series data would then reveal trends, indicating for a certain zone if the housing market moved on average further away from the job market over the chosen time span, or if, maybe, the opposite has happened. Horner and Murray (2002) devoted considerable attention to the impact of the underlying zoning system on the calculated minimum commuting distance, both with regard to size and to location. This issue is known as the modifiable areal unit problem (MAUP). The use of larger zones means that more trips become intrazonal, leading to an overestimation of the minimum commuting distance (Horner and Murray, 2002). Both interzonal and intrazonal trips originating in large zones can only be simulated with limited precision, leading to significant inaccuracies. However, when excess commuting is studied, this problem can partly be accommodated by recalculating observed commuting distances using the same origin-destination matrix that is used to calculate the minimum commuting distance. For our research, however, the main problem associated with the selection of zone sizes is the geographical scale at which the studied developments occur. This means that the spatial separation between jobs and housing can both be caused by sprawl (at a small geographical scale) and by regional economic shifts within the labour market (at a large geographical scale). The first phenomenon will particularly manifest when detailed zoning (e.g. traffic analysis zones) is used, while the second phenomenon will become clear when relatively large zones (e.g. districts) are studied. While a considerable volume of research on excess commuting has been published yet, the literature is still fast evolving. Recent inquiries include attainability of trip length reductions (Oâ€™Kelly and Niedzielski, 2008), differentiation in excess rate depending on mode choice (Murphy, 2009), the relationship with the jobs-housing balance as a suitable indicator for measuring spatial expansion (Layman and Horner, 2010), the influence of uncertainties in travel time measurements when minimizing time distance (Horner, 2010) and optimization of spatially disaggregated calculation methods (Boussauw et al., 2011). Based on the properties of the excess commuting framework, we find a time series approach of the minimum commute useful to study processes of suburbanization. Existing research in this field is rare, however. Frost et al. (1998) described the evolution of the minimum commuting distance for a set of demarcated British cities. They compared the 1981 situation with 1991. Yang (2008) reported on similar research in the metropolitan areas of Boston and Atlanta, based on data from 1980, 1990 and 2000. Both studies find an increase of both minimum commuting distances and actual commuting distances, in all involved cities. Horner (2007) compared commuting data from Tallahassee, Florida, from 1990 and 2000. The average reported commuting distance and maximum commuting distance increased during this period, while the minimum commuting distance showed only a statistically non-significant increase. However, neither Frost et al. (1998) nor Horner (2007) or Yang (2008) applied a spatially disaggregated method, so it is not clear which parts of the urban areas are the most affected by spatial separation. Furthermore, no typical peri-urban areas, which are relatively far from the CBD, were incorporated in the analyses. As stated in the introduction, our approach consists of an extension to the empirical diversity of the three aforementioned authors by including suburban and rural areas and an extension of their methodology by applying a spatially disaggregated method.
4. Spatial development and commuting in Flanders and Brussels The case study area consists of the administrative regions of Flanders and Brussels, which together form the northern part of Belgium (Map 1). Historically, the area is built around three major cities (Brussels, Antwerp and Ghent), a dozen regional cities and several dozen smaller towns and central municipalities. The economic core is borne by the wide surroundings of the axis Brussels-Antwerp, a region known as the Flemish Diamond. The rapid development of an extensive railway network in the nineteenth century and the introduction of cheap season tickets formed the backbone of a policy aimed at an industrialization of the country based on limited urbanization (Verhetsel et al., 2010). The new working class was able to continue living outside major cities, in a house with a garden and under the watchful eye of the village priest, while commuting every day to the factory or the office. This form of institutionalized commute laid the foundation for the spatial separation between housing and work locations that has only developed more distinctly since the advent of the motorway and general car ownership. In the period before World War II, this phenomenon has materialized in the form of spatial developments that were strongly clustered around the railway stations. In the post-war period, however, this structure fanned out into a network of suburbanized and peri-urbanized areas around the ancient settlements. In this context, suburbanization refers to sprawl that is occurring in or adjacent to urban areas, while peri-urbanization consists of developments in the countryside, free-standing or in line with existing villages. These postwar developments gradually led to the emergence of a variety of forms of urbanization, ranging from historical densely populated cities and sparsely populated suburban residential belts to commuter areas and pure countryside (Vanneste et al., 2008). Depending on the definition, 57% (Vanneste et al., 2008) to 97% (Antrop, 2004) of the Belgian population lives in urbanized areas.
Map 1. Situation of infrastructure and urban areas The problem of spatial separation, supposedly mainly in the form of sprawl, is not a new issue in Flanders. In the period between 1980 and 2000 planning became a major political topic, as open space became scarce, and urban flight appeared to feed a wave of suburban and peri-urban
development. Newly emerging issues in the field of landscape ecology, water pollution, increasing distribution costs, road safety, congestion and social segregation were partly attributed to sprawl. The political response, in the form of the Spatial Structure Plan for Flanders (RSV, 1997/2004) had to wait until 1997, but did offer an answer in the form of the demarcation of urban areas and the focus on encouraging additional building in the cities and existing settlements. The policy measures that are imposed by this plan are based on a rather intuitive analysis of the problem, with often no thorough scientific rationale behind it. Since then, several quantitative spatial analyses have been carried out. The main research report on sprawl, which was issued by the Flemish Environmental Agency (Gulinck et al., 2007), approaches the phenomenon mainly from a landscape-ecological perspective. It focuses on morphological changes that are related to spatial fragmentation. In the process, sprawl detection is based on maps and satellite images from different periods. The main indicator is the overall total built-up area, which for Flanders increased over the period 1990-2000 from 13.2% to 14.6%. Other indicators were the length of the road network (increasing by 6.4% between 1991 and 2001), morphological grain size and global proximity to buildings and infrastructure. A time series analysis of these indicators points to a systematic increase in sprawl. The successive censuses show that commuting distances in Belgium systematically increase. Based on an assessment of the respondents, the perceived average distance between home and workplace evolved from 11.9 kilometres in 1970 (MĂŠrenne-Schoumaker et al., 1999) to 19.0 kilometres in 2001 (Verhetsel et al., 2007), an increase of no less than 60%. Although noncommuting trips and freight transport grew even faster over this period, it is clear that the increase in trip length is largely responsible for the overall growth in traffic.
5. Data The decennial census, organized by the Belgian federal government, assesses the municipality of residence and the work municipality of all working citizens. Furthermore, the (perceived) distance travelled every day to work is also registered. Commuting data for 1981, 1991 and 2001 is available in the form of origin-destination matrices, with the municipality as a spatial unit. The small municipalities of the Brussels Capital Region were combined into one zone, in order to avoid biases in relation to the two other metropolitan municipalities of Antwerp and Ghent. As a result, the study area is divided into 309 zones with a mean area of 44 kmÂ˛ (standard deviation: 29 kmÂ˛). For the particular purposes of our research, the matrices were cleaned by removing all trips that have their origin or destination outside the study area (Wallonia and the neighbouring countries). Apart from these inter-regional commuters, also home workers (including teleworkers) were removed from the matrices, as well as respondents with an itinerant occupation. The exclusion of these records is justified because we only want to measure the extent to which residential and labour structure separate within Flanders-Brussels. Although the method of data collection in the three survey moments was performed in a similar manner, there are rather large differences between the three data sets. Important variations in the number of commuting trips are found, as well as in the number of unknown or incomplete registrations. Moreover, the number of home workers declined dramatically over the three consecutive survey moments, and the structure of the inter-regional commute (between FlandersBrussels and the neighbouring region and countries) changed. Possibly more important is the influence of the regional economic transformations that occurred. In the period 1981-2001 the economy shifted towards a more service-based system, and lost a number of important industrial employment centres such as the Kempen coal mines in the east of Flanders and the Hainaut steel
industry south of the study area. Also, the port of Antwerp and the airport of Brussels went through an era of major growth, and a high standard service industry (especially in information technology and international public institutions) developed in the urban areas, in particular in the Brussels agglomeration. This restructuring of the labour market cannot be separated from the suburbanization of the residential structure: both transformations add to the spatial separation that we want to assess. Regarding the development of sprawl in the studied period, also the growth of the infrastructure network is undoubtedly important. Although most motorways that exist today where constructed in the period prior to 1981 (roughly between 1961 and 1981, see Map 1), it can be expected that many developments along the motorways have been built in the period 1981-2001.
6. Method To answer the research question, we investigate the evolution of the spatial distribution of residential locations of workers and job locations in Flanders and Brussels (Belgium), mainly relying on the minimum commuting distance concept, and comparing the trend found with the evolution of observed commuting distances. To this end we first take the study area as a whole, then we reiterate the calculation in a spatially disaggregated way (for each municipality separately). Regarding indicators for the study area as a whole, we will also provide the global maximum commuting distance. As we consider the minimum commuting distance as a measure of spatial proximity between job market and housing market, the possible mismatch between qualifications and job preferences of workers and requirements of employers is not taken into account, which is an important deviation from reality. Although in the literature attempts were made to disaggregate excess commuting characteristics based on a classification of workers and jobs (O'Kelly and Lee, 2005), this approach falls outside the scope of our paper. Nevertheless it is important to keep this constraint in mind when interpreting the results. O'Kelly and Lee (2005) found e.g. that service workers are subject to shorter minimum commuting distances than industrial workers, but that actual commuting behaviour of the former group gives rise to larger excess rates than the latter. However, it is unclear whether these findings also apply outside the surveyed American cities, e.g. in a Belgian context. Also, the unemployed population and home workers do not affect the results since they are excluded from the dataset, which is important to keep in mind when interpreting the results. An increase of the minimum commuting distance towards or from a certain area is usually indicative of the spatial expansion of the residential market or a trend of concentration in the labour market, while a decline usually will point to a greater functional mix. The underlying causes may be both of spatial development nature (sprawl), and of regional-economic nature (e.g. the economic shift towards service industries, increased tertiarisation). According to our hypothesis, the rate at which the minimum commuting distance increases is an indicator for the impact of spatial transformations on the actual average commuting distance. The difference between the growth rate of the minimum commuting distance and the growth rate of the actual commuting distance is a measure of the non-spatial component in the overall growth of the commute. Consequently, an increase in commuting distances that would develop faster than the increase of the minimum commuting distances points to a mobility growth that is relatively independent of changes in spatial structure. In the opposite case, we would be dealing with a spatial evolution that is in and by itself responsible for the entire growth in commuter traffic volume.
To calculate the minimum commuting distance, most authors make use of one of the minimization algorithms for the so-called "transportation problem", as available in various software packages. However, we applied an alternative method that has been used before for a cross-sectional analysis of the same study area, and is explained in Boussauw et al. (2011). This method is conceived as an iterative process that simulates the behaviour of commuters who simultaneously look for a job closer to home. The disadvantage of this method is that the achieved “optimum” remains slightly higher than the mathematical minimum. The advantage of the method is that the produced origin-destination matrix holds a much greater realism, since it is not influenced by algebraic tricks that are applied to find the mathematical minimum but do not make sense in a real world approach. An example is the allocation of all resident workers from one zone to jobs in only one corresponding zone. Moreover, the structure of origindestination matrices that are produced by the iterative process on the basis of various data sets (representing different points in time) is similar, meaning that comparing these matrices using statistical methods (such as Student's t-test) makes sense. This is not necessarily true when applying a standard “transportation problem” algorithm, where the structure of the resulting origin-destination matrix depends on the used software package. Confronting the minimum commuting distance for an area with the observed (calculated) commuting distance is known as the study of excess commuting. In practice we implement this confrontation by calculating the excess rate, defined as the quotient of the mean observed commuting distance and the mean minimized commuting distance. The mean observed distance used in the analysis is actually a calculated (thus not perceived by the respondent) distance. For each municipality, the centre of gravity (centroid) is determined. Then a shortest-path matrix is calculated by means of a skeleton file of the Belgian road network, making the shortest network distance between each possible pair of municipalities available. Intra-municipal distances are simulated by taking for every municipality half of the network distance to the centroid of the nearest municipality. Given that the road network in 1981 was already very dense, we use the same skeleton file for the three time periods.
7. Results 7.1 Evolution of the mean observed commuting distance We calculate for each municipality the mean trip length for both outgoing (outbound mean calculated distance: omcd) and incoming (inbound mean calculated distance: imcd) commuting trips. A weighted average can be found in Table 1, along with the commuting distance as perceived by the respondents (mean perceived distance: mpd). Table 1. Observed commuting trips and commuting distances number of commuting trips (matrix) mcd (matrix) mpd (Belgium) a
source: Mérenne-Schoumaker et al. (1999), p. 80
source: Verhetsel et al. (2007), p. 60
1981 1,950,477 12.2 km 14.6 kma
1991 2,331,090 12.6 km 17.2 kmb
2001 1,907,197 14.7 km 19.0 kmb
The important differences between the calculated (mcd) and the observed mean commuting distance (mpd) can be explained as follows:
mcd is calculated based on the shortest network distance while the shortest route is usually not the fastest, and thus not the path that is chosen by the commuter;
mcd is calculated from centroids of municipalities, which is an important simplification, particularly with regard to intra-municipal trips;
mcd ignores interregional, usually relatively long, commuting trips;
mpd includes commuting trips in the south of Belgium, which are longer on average.
Other possible biases in perceived trip length are discussed by Witlox (2007).
Map 2. Evolution of the commuting distance (1981-2001) based on municipality of residence
Map 3. Evolution of the commuting distance (1981-2001) based on municipality of workplace
Taken together, the mcd variable should therefore be regarded as a theoretical measure, which only makes sense when compared to similar calculations, such as the minimum commuting distance. For comparison with survey data from other countries, mpd will naturally be better suited. More important than the absolute figure is the trend that both quantities exhibit. A closer look reveals that over the period 1981-2001 mcd increases in 280 of the 309 municipalities when considering the outbound commute (Map 2), and increases in 291 of the 309 municipalities when looking at the incoming commute (Map 3). Although municipalities with a relative growth show a certain spatial clustering, the virtually general upward trend suggests that at least some of the causes of the growth in trip length are of non-spatial nature. 7.2 Evolution of the mean minimum commuting distance Given the significant variation in the number of recorded trips between the three snapshots (1981, 1991 and 2001), we first examine whether this sample variation affects the calculated mean minimum commuting distance (mmid). We did this by weighing commuting trips for the year 1981 and 1991 so that the three tables on which the minimization procedure was then performed all contain 1,907,197 trips (equalling the number of trips in the origin-destination matrix for 2001, see Table 1). The obtained results are thereupon compared with the results that are based on the unweighted tables. No significant differences were found, and Pearson's correlation coefficient between the two sets of spatially disaggregated results is 1.00. Next, the global mmid for the entire study area is calculated for the three different points in time. The obtained figures are presented in Table 2, along with the evolution of mcd, and the maximum commuting distance (mmad). Table 2. Evolution of the global minimum commuting distance and excess rate Mmid evolution mmid (base 1981) Mcd evolution mcd (base 1981) excess rate (mmid/mcd) evolution excess rate (base 1981) Mmad evolution mmad (base 1981)
1981 9.0 km 12.2 km 1.36 19.5 km
1991 8.9 km -1.4% 12.6 km +3.4% 1.43 +4.8% 18,6 km -4.5%
2001 9.4 km +4.3% 14.7 km +20.1% 1.57 +15.1% 21.6 km +10.8%
A salient element in Table 2 is the negative evolution of mmid over the period 1981-1991 (-1.4%), which is associated with an increase of mcd (+3.4%). So, the studied residential locations and job locations would have come 1.4% closer together over this period, while commuting distances continuously increased. Nevertheless, over the entire period under investigation (1981-2001) we find, as expected, an increase in mmid. The second striking result is the growth rate of mmid: this is much lower (+4.3%) than the growth rate of mcd (+20.1%). Logically, this trend is in line with an increase of the excess rate. This result suggests that over the studied period the average worker became less inclined to seek a job close to home, or to look for a home close to work. This also means that only a small part of the increase in commuter traffic can be attributed to the expansion of the spatial structure. The evolution of mmad confirms the observed trend of mmid, although the differences are more pronounced. For a theoretical approach to the possible interpretation of the differences between mmid and mmad, which is beyond the scope of this paper, we refer to Charron (2007) and Yang (2008).
Note that in our literature overview only three papers (i.e. Frost et al., 1998, Horner, 2007 and Yang, 2008) studied time series of minimum commuting distance. The results of Horner (2007) and Yang (2008) are analogous with our findings: in Tallahassee, in Boston and in Atlanta both minimum commuting distances and observed commuting distances grew, while the former increased more slowly than the latter. Frost et al. (1998), however, found for all analysed British cities that the growth of the minimum commuting distance was faster than the growth of the observed commuting distance, suggesting that changes in the spatial structure could be held fully responsible for the increase in commuter traffic volume, along with an improved efficiency in the commute itself (given the reduction of the excess rate). In contrast to our research, however, Frost et al. (1998) limited their study area to demarcated cities, to which only incoming commuting trips were added. So, the spatial structure of the commuter area around the considered cities was not fully grasped. Within these demarcated cities urban sprawl hardly occurs, and also the growth in traffic itself is far more constrained by congestion than is the case in the suburban areas. While this alternative approach may explain the difference in results, apparently great caution in interpreting the results is needed. 7.3 Evolution of the spatially disaggregated minimum commuting distance A second step in the analysis is the calculation of spatially disaggregated values of mmid. For each municipality and for each of the three points in time mmid was calculated twice: once for the outgoing commute (ommd) (these are the outgoing and internal trips together) and once for the ingoing commute (immd) (these are the incoming and the internal trips together). So, for each municipality we obtain two time series. Then, for each time series the existence of a clear trend for the period 1981-2001 was examined. First, a Student's t-test was applied to compare columns and rows of the produced origindestination matrices. Non-significant differences (significance level: p < 0.05) are considered as a status quo. The decision rules that were used to determine the existence of a trend are presented in Table 3. As a general principle it is assumed that a trend is only acknowledged if the evolution over the period 1981-1991 is not contradictory to the evolution over the period 1981-2001. Moreover, the differences have to be statistically significant in at least one of the two periods. Table 3. Decision table trend evolution 1981-2001 + + + not sig.
evolution 1981-1991 + not sig. + not sig. + / - / not sig.
evolution 1991-2001 + / - / not sig. + + + / - / not sig. + / - / not sig.
conclusion + 0 + 0 0
For the municipalities where a significant trend was found, the absolute differences in mmid over the period 1981-2001 were mapped (Map 4 and Map 5). The following applies: and
ommd81-01 = ommd01 - ommd81
immd81-01 = immd01 - immd81
7.4 Outbound minimum commuting distance by municipality Map 4 shows the evolution of the minimum commuting distance over the time frame 1981-2001 for those municipalities that contain more working residents than jobs. Municipalities with a job surplus are omitted since the applied method implies that the minimum commuting distance for workers living in such a municipality remains constant as long as the job surplus exists. Following issues on the map stand out:
Most municipalities in the economic core around the triangle Antwerp-Brussels-Leuven (A) show a decrease of ommd. Exceptions are some municipalities with a more rural character that are located on the edge of the conurbation and have received a large share of housing suburbanization (Kapellen (B), Nijlen and Berlaar (C), Zemst (D), Oud-Heverlee (E), Beersel and Sint-Pieters-Leeuw (F)).
The former mining region in the Kempen (G) shows a sharp increase of ommd, perhaps due to economic transformation processes that have led to a decline of job supply in this area.
In some areas apparently the construction of the motorway has indeed led to a suburban housing development that entailed only few jobs. This is clearly the case along the E40 between Ostend and Ghent (H), and in some municipalities in the Voorkempen (Brecht and Zoersel (I)) and Limburg (such as Maasmechelen (J), Riemst (K), Heers (L)).
In other municipalities also a lot of employment developed in the proximity of the motorway. This is the case in Leie valley south of Ghent (M), or in Lummen (N).
Map 4. Evolution of the minimum commuting distance (1981-2001) based on municipality of residence 7.5 Incoming minimum commuting distance by municipality Map 5 shows the evolution of the minimum commuting distance over the period 1981-2001 for those municipalities that contain more jobs than working residents. In this case, municipalities with a surplus of workers are omitted since the applied method implies that the minimum
commuting distance for workers employed in such a municipality remains constant as long as the worker surplus exists. Following issues on the map stand out: ď‚ˇ
The value of immd increases in almost all non-peripheral cities, and the Brussels region has the strongest growth. This means that these cities need to cover an expanding recruitment area to have their available jobs occupied.
A number of cities in the more remote areas, however, show a reduction of the minimum commuting distance. Thus, in these places the concentration of employment is proportionally shrinking. This is the case in the cities Kortrijk (O), Ostend (P) and Bruges (Q) in the west, and in the cities of Hasselt and Genk (R), Sint-Truiden (S) and Tongeren (T) in the east.
Map 5. Evolution of the minimum commuting distance (1981-2001) based on municipality of workplace 7.6 General interpretation and possible biases The results of our analysis seem to indicate that at the macro level the distance between job locations and residential locations of the working population did not dramatically increase, and that commuter traffic shows an autonomous growth that can only partially be explained by changes in the spatial structure. There are two, possibly complementary, explanations for the relatively limited detected increase of the average distance between home and work locations. First, jobs have partly followed the suburbanization process of housing. Sprawl does not simply consist of new residential allotments, but equally of new business parks. Moreover, many new jobs engrafted onto the additional residential areas, e.g. new employment that developed in schools, childcare, local health care, supermarkets and public services in growing municipalities. At the macro level, suburban areas are relatively multifunctional: sprawl does not necessarily imply the absence of a functional mix, even if it is at a lower density in comparison with the city. Nevertheless, the regional job market remains more spatially concentrated than the housing market. This is the case in many industrial activities, but in particular the growth of specialized services (such as financial services, technology and consultancy) led to an increase in the number of jobs in the Brussels
region. This growth explains a significant portion of the found increase in the general minimum commuting distance. A second, perhaps equally important reason is the large-mesh nature of the used dataset. Our analysis cannot possibly grasp the sprawl that occurs within a municipal boundary. Although, given the rather small area of an average municipality, suburbanization certainly plays a role too at an intermunicipal scale level, it appears that the intramunicipal share of these transformation processes cannot be detected from a low resolution data set. To identify sprawl on a lower scale level time series data at a much denser spatial aggregation level is needed. Thus, the discrepancy between our analysis and the morphological analyses mentioned above is mainly due to the difference in scale. It is likely that the increase in the minimum commuting distance, and therefore the expansion of the functional space, would be better reflected in the figures when examining this micro level. This is due to the modifiable areal unit problem (MAUP) of which the consequences for the minimum commuting distance were discussed above. What is certain is that the use of a fine-mesh data set leads to lower absolute figures for the minimum commuting distance. On the basis of cross-sectional approximate data for 2007 available for the same study area (Flanders and Brussels) at the level of (fine-meshed) traffic analysis zones, and using the same algorithm as we did, Boussauw et al. (2011) found a mean minimum commuting distance of 6.9 kilometres (to be compared with our calculated 9.4 kilometres (2001)). Unfortunately, time series data are not available at this microscopic scale level, so it is not possible to verify our assumption that spatial separation may be mainly taking place at this smaller geographical scale level. The inability to outline developments based on fine-mesh data might lead to an overrepresentation of the impact of regional economic transformations in the results, pushing the effect of the spatial shift in the housing stock to the background. The decline of some industrial activities, the development of logistics in the port areas and the general shift towards service industries play perhaps a more important role at the studied scale level with respect to spatial proximity between home and work locations than the suburbanization of the housing stock does. When we consider the spatially disaggregated evolution of the minimum commuting distance, we notice that the major conurbations in the economic core of Flanders and Brussels gain importance in terms of employment, while a reduction of jobs in more peripheral industrial sectors has led to a local increase of the minimum commuting distance. The phenomenon of residential suburbanization in the municipalities along the motorways has led only here and there to spatial separation, particularly where jobs did not follow the spatial shift in housing.
8. Conclusions The applied method detects spatially disaggregated evolutions in minimum commuting distance, identifying local increases or decreases in spatial separation between home and work locations at the level of the municipality. The results show that there is indeed a general loss in spatial proximity between housing and jobs for the study area in its entirety, although the pace of this separation process is on average much lower than the growth in observed commuting distances. It is found that the minimum commuting distance increased in many municipalities, especially where population is growing faster than job supply, or where traditionally high concentrations of employment still increase. Furthermore, there are also municipalities where a decrease is noticed, especially in suburban areas that are getting a more urban character by acquiring a considerable functional mix. The development of the motorway network has undoubtedly contributed significantly to the growth of actual commuting distances. In those municipalities where mono-functional planning practice has facilitated residential development with access to the motorway, the minimum
commuting distance increased, with negative consequences for the spatial proximity between residential and work locations. In municipalities where a better balance between different functions was achieved, the development of multifunctional sprawl has not necessarily led to an increase of the minimum commuting distance at the supra-municipal scale level. Nevertheless, the influence of sprawl on commuting behaviour seems to be only secondary to the effects of regional-economic transformations, which for example led to the loss of employment in the Kempen region and an increase in employment concentration in Brussels and (to a lesser extent) in Antwerp. Still, in spatial and economic planning it is important to ensure the local balance between jobs and inhabitants as good as possible. Horner and Murray (2003) argue that the most effective way to do this is raising residential density in areas with a decent job supply: through the deliberate reallocation of workersâ€™ residences a significant decrease of the minimum commuting distance can be attained. However, Yang (2008) shows that job decentralization may also be responsible for the growth of the excess rate itself, and thus for a weakening land usetransport connection. From this finding Yang (2008) argues that a policy of ensuring a good jobhousing balance is insufficient: concentrations of employment both in cities and in suburbs should take the form of compact, yet relatively large, centres and subcentres. In order to draw valid conclusions, the degree of detail and the consistency of the used data is crucial. We find that the use of municipalities as a spatial entity is suitable to grasp regional transformations of the economy, but is far from perfect to detect sprawl in the morphological sense. If the same analysis could be repeated at a lower scale level, probably more concrete guidelines for spatial planning could be given. This is particularly true when mode choice would be taken into account too. In that case, concentration of activities around the stops of public transport would remain equally important. For cyclists, at the other hand, proximity and the availability of infrastructure are also of interest at the micro level.
Acknowledgements This research has been made possible within the Flemish Policy Research Centre for Housing and Space, funded by the Ministry of the Flemish Community. Also, we would like to thank the anonymous referees and the editor for their useful comments which led to significant improvements of the paper. All remaining errors are ours.
References Aguilera, A. (2005). Growth in commuting distances in French polycentric metropolitan areas: Paris, Lyon and Marseille. Urban Studies, Vol. 42, no. 9, pp. 1537-1547 Antrop, M. (2004). Landscape change and the urbanization process in Europe. Landscape and Urban Planning, Vol. 67, no. 1-4, pp. 9-26 Banister, D. (1999). Planning more to travel less. Town Planning Review, Vol. 70, no. 3, pp. 313-338 Bleijenberg, A. (2003). The driving forces behind transport growth and their implications for policy. in OECD (ed) Managing the Fundamental Drivers of Transport Demand: Proceedings of the International Seminar, OECD Publishing, Paris. Boussauw, K., Neutens, T. and Witlox, F. (2011). Minimum commuting distance as a spatial characteristic in a non-monocentric urban system: the case of Flanders. Papers in Regional Science, Vol. 90, no. 1, doi:10.1111/j.1435-5957.2010.00295.x
Boussauw, K. and Witlox, F. (2009). Introducing a commute-energy performance index for Flanders. Transportation Research Part A, Vol. 43, no. 5, pp. 580-591 Charron, M. (2007). From excess commuting to commuting possibilities: more extension to the concept of excess commuting. Environment and Planning A, Vol. 39, no. 5 , pp. 1238-1254 Ewing, R. H. (1994). Characteristics, causes, and effects of sprawl: a literature review. Environmental and Urban Studies, Vol. 21, no. 2, pp. 1-15 Frost, M., Linneker, B. and Spence, N. (1998). Excess or wasteful commuting in a selection of British cities. Transportation Research Part A, Vol. 32, no. 7, pp. 529-538 Gordon, P. and Richardson, H. W. (1989). Gasoline consumption and cities: a reply. Journal of the American Planning Association, Vol. 55, no. 3, pp. 342-346 Gulinck, H., Meeus, S., Peymen, J., Stalpaert, L., Bomans, K., Dewaelheyns, V. and Heremans, S. (2007) Milieurapport Vlaanderen, Achtergronddocument Versnippering (Environmental Report for Flanders, Background Document on Fragmentation), Flemish Environment Agency, Aalst. Horner, M. W. (2004). Spatial dimensions of urban commuting: A review of major issues and their implications for future geographic research. The Professional Geographer, Vol. 56, no. 2, pp. 160-173 Horner, M. W. (2007). A multi-scale analysis of urban form and commuting change in a small metropolitan area (1990-2000). The Annals of Regional Science, Vol. 41, no. 2, pp. 315-332 Horner, M. W. (2010). Exploring the sensitivity of jobs - housing statistics to imperfect travel time information. Environment and Planning B, Vol. 37, no. 2, pp. 367-375 Horner, M. W. and Murray, A. T. (2002). Excess commuting and the modifiable areal unit problem. Urban Studies, Vol. 39, no. 1, pp. 131-139 Horner, M. W. and Murray, A. T. (2003). A multi-objective approach to improving regional jobshousing balance. Regional Studies, Vol. 37, no. 2, pp. 135-146 Layman, C. and Horner, M. W. (2010). Comparing methods for measuring excess commuting and jobs-housing balance: an empirical analysis of land use changes. Transportation Research Board Annual Meeting 2010, Washington DC. Ma, K. R. and Banister, D. (2006). Excess commuting: A critical review. Transport Reviews, Vol. 26, no. 6, pp. 749-767 Ma, K. R. and Banister D. (2007). Urban spatial change and excess commuting. Environment and Planning A, Vol. 39, no. 3, pp. 630-646 MĂŠrenne-Schoumaker, B., Van der Haegen, H. and Van Hecke, E. (1999.) Algemene Volks- en Woningtelling op 1 maart 1991: Werk- en Schoolpendel (General Census, 1st March 1991: Work and School Commuting), Statistics Belgium, Brussels. Murphy, E. (2009). Excess commuting and modal choice. Transportation Research Part A, Vol. 43, no. 8, pp. 735-743 N.N. (1997/2004). Ruimtelijk Structuurplan Vlaanderen - gecoĂśrdineerde versie (Spatial Structure Plan for Flanders - consolidated version). Ministry of the Flemish Community, Brussels. Newman, P. and Kenworthy, J. (1989). Cities and Automobile Dependence. A Sourcebook, Gower, Aldershot. Niedzielski, M. A. (2006). A spatially disaggregated approach to commuting efficiency. Urban Studies, Vol. 43, no. 13, pp. 2485-2502
O'Kelly, M. E. and Lee, W. (2005). Disaggregate journey-to-work data: implications for excess commuting and jobs-housing balance. Environment and Planning A, Vol. 37, no. 12, pp. 2233-2252 O’Kelly, M. E. and Niedzielski, M.A. (2008). Efficient spatial interaction: attainable reductions in metropolitan average trip length. Journal of Transport Geography, Vol. 16, no. 5 , pp. 313-323 Peng, Z. R. (1997). The jobs-housing balance and urban commuting. Urban Studies, Vol. 34, no. 8, pp. 1215-1235 Rietveld, P. and Vickerman, R. (2003). Transport in regional science: The “death of distance” is premature. Regional Science, Vol. 83, no. 1, pp. 229-248 Schafer, A. (2000). Regularities in travel demand: An international perspective. Journal of Transportation and Statistics, Vol. 3, no. 3, pp. 1-31 Sultana, S. and Weber, J. (2007). Journey-to-work patterns in the age of sprawl: evidence from two midsize southern metropolitan areas. The Professional Geographer, Vol. 59, no. 2, pp. 193-2 Van Wee, B., Rietveld, P. and Meurs, H. (2002). A Constant Travel Time Budget? In Search for Explanations for an Increase in Average Travel. Vrije Universiteit Amsterdam, Amsterdam. Vanneste, D., Thomas, I. and Vanderstraeten, L. (2008). The spatial structure(s) of the Belgian housing stock. Journal of Housing and the Built Environment, Vol. 23, no. 3, pp. 173-198 Verhetsel, A., Thomas, I., Van Hecke, E. and Beelen, M., (2007). Pendel in België. Deel I: de woonwerkverplaatsingen (Commuting in Belgium: Part I: Home-to-Work Travel). Statistics Belgium, Brussels, 166 pp. Verhetsel, A., Thomas, I. and Beelen, M. (2010). Commuting in Belgian metropolitan areas: The power of the Alonso-Muth model. Journal of Transport and Land Use, Vol. 2, no. 3, pp. 109-131 Witlox, F. (2007). Evaluating the reliability of reported distance data in urban travel behaviour analysis. Journal of Transport Geography, Vol. 15, no. 3, pp. 172-183 Yang, J. (2008). Policy implications of excess commuting: Examining the impacts of changes in US metropolitan spatial structure. Urban Studies, Vol. 45, no. 2, pp. 391-405. Yang, J. and Ferreira, J. (2009). Informing the public of transportation-land use connections. Transportation Research Part C, Vol. 17, no. 1, pp. 27-37.
Issue 11(1) January 2011 pp. 61-79 ISSN: 1567-7141 www.ejtir.tbm.tudelft.nl
Conflict between Independent Scrutinisers of Transport Megaprojects: Evidence from Australia Stuart Kells12 Centre for Regulatory Studies, Monash University, Melbourne, Australia
If, in the context of an ‘audit explosion’, public sector projects are subject to multiple and uncoordinated forms of independent scrutiny, then the different scrutinisers could reach contradictory conclusions, adopt incompatible methods, and ultimately provide less effective oversight. In the Australian state of Victoria, three independent scrutiny mechanisms operate concurrently on transport megaprojects. The three mechanisms – performance auditing, probity auditing and gateway reviews – are not coordinated and yet have overlapping goals relating to integrity and value for money. This paper describes the three scrutiny mechanisms, before presenting evidence that the mechanisms can generate contradictory conclusions that remain unreconciled, and that the mechanisms conflict in ways that affect the viability of at least one of the mechanisms. The paper concludes with a discussion of implications for Europe and directions for future research. Keywords: performance auditing, gateway reviews, probity auditing, transport megaprojects
1. Introduction The role played by independent auditors, evaluators, inspectors and reviewers in public administration has a long history and, since ancient times, has been the subject of much discussion and analysis. In recent decades, major contributions to the academic literature on independent scrutiny institutions include Head et al., 2008; Hood et al., 1999; OECD, 1996; and Pollitt et al., 1999. The main themes in this literature concern the goals of oversight institutions (including efficiency, process integrity and accountability); the methods they use; their institutional form, and what degree of independence they enjoy; the status of their findings and recommendations; and their impact on public accountability and agency performance. The overwhelming majority of studies of independent scrutiny mechanisms assume, most often implicitly, a ‘vertical’ or bilateral relationship between a single scrutiniser and one organisation, program or project (or a set of organisations, programs or projects) that is scrutinised.3 The term that is most often used to describe public auditors – ‘Supreme Audit Institutions’ (SAIs) – reflects this singularity. Drawing on a standard German public finance textbook (Blankart, 2006), Blume The author is grateful to three anonymous reviewers for helpful comments on an earlier version of this paper. Centre for Regulatory Studies, Faculty of Law, Monash University, Victoria 3800, Australia, T: +61399054135, E: firstname.lastname@example.org 3 For discussions of standard conceptual models of auditing, see Leeuw 1996 and Lonsdale 1999. 1 2
EJTIR 11(1), January 2011, pp. 61-79 Kells Conflict between Independent Scrutinisers of Transport Megaprojects: Evidence from Australia
and Voight (2007) claimed that ‘SAIs are monopolists...since competition between SAIs seems to be a rare exception’ (page 7). It may not be so exceptional, however, for public auditors (and public sector reviewers and inspectors more generally) to face competition. In this era of an ‘audit explosion’ and an ‘audit society’ (Power, 1999; Power, 2000; Bowerman et al., 2000) some advanced countries have experienced an increase in the number of bodies with independent audit and review functions. In several countries, multiple audit and review bodies now have overlapping mandates. For example, Williams et al. (2010) studied a British government accommodation project that was subject to scrutiny by independent ‘gateway’ reviews, the Home Office Audit and Assurance Unit, the National Audit Office (NAO) and the Parliamentary Accounts Committee. Kells and Hodge (2010) analysed jurisdictional overlaps between four public institutions with performance audit functions in a single Australian jurisdiction. As a consequence of the growing institutional richness, new research questions have emerged regarding the conduct and impact of public sector oversight bodies. Multiplication of such bodies within a jurisdiction means that there are important institutional relationships beyond the vertical or bilateral ones. When different bodies inhabit the same institutional space, there is an additional horizontal dimension to their conduct. For example, scrutiny bodies that cohabit the same space can interact with each other (Henkel, 1991; Kells and Hodge, 2010). Possible modes of interaction include cooperation, competition, ‘convergence’ (where the bodies mimic each other) and ‘divergence’ (whether the bodies seek to differentiate themselves) (Henkel, 1991; Kells and Hodge, 2010). In Britain, Bowerman et al. (2003) documented ongoing ‘turf battles’ between overlapping public audit agencies such as the NAO and the Audit Commission. In principle, interaction between scrutiny bodies might generate benefits for society. If, for example, multiple bodies are authorised to review the same projects and agencies, there are ‘more hands to the pump’ in the scrutiny task, and therefore scrutiny may be more thorough. Also, multiple bodies might engage in benign competition, to become more efficient, to exert more effort, or to innovate by adopting new techniques. Another set of potential consequences, however, is less sanguine. With multiple scrutiny bodies, coordination problems might arise. There may be wasteful duplication of effort, or the reviewers could reach contradictory conclusions (a possibility given the importance of judgement in auditing and evaluation; see Barzelay, 1996; Keen, 1999; and Lindeberg, 2007), and therefore pull reviewees in opposing directions. The results might be higher participation costs for reviewees, and less effective oversight for the community. These dangers have led to calls for greater coordination of public sector audit and review bodies. In its 2010 report on ‘Assurance for high risk projects’, the NAO called for greater integration across the various mechanisms, including gateway, that provided assurance to government and the community about the delivery of major public-sector projects (NAO, 2010). In Australia, SSA (2010) recommended the establishment of an ‘integrity coordination board’ to better integrate the activities of public sector ‘watchdog’ bodies such as the Victorian Auditor-General and the Victorian Ombudsman. The potential pitfalls from the lack of coordination identified by the SSA included making poor use of the available information, duplicated effort, contradictory findings and public confusion (SSA, 2010). Less obvious is the risk that scrutiny bodies will adopt incompatible methods, or undermine each other’s effectiveness. For example, some scrutinisers adopt more discreet and ‘constructive’ methods – Williams et al. (2010) called these the ‘friendly’ scrutinisers – while others do not. Suppose there are two review bodies, A and B, who have overlapping jurisdiction over an agency, ‘TransCo’, and who have the power to disclose each other’s findings. Body A adopts ‘friendly’ methods. This means body A works constructively and discreetly with TransCo, and delivers its findings, including any that are adverse, directly to TransCo, without publicly
EJTIR 11(1), January 2011, pp. 61-79 Kells Conflict between Independent Scrutinisers of Transport Megaprojects: Evidence from Australia
disclosing them. Accordingly, there is a high level of trust between TransCo and body A. TransCo’s staff are candid and unguarded in their disclosures to body A, which is therefore uniquely able to identify correctly TransCo’s shortcomings. Body B, however, adopts a less ‘friendly’ approach. This body is dogged in the pursuit of TransCo’s failings, and their publication. TransCo’s staff are therefore very reserved when dealing with body B, and reluctant to offer up any more information than is strictly required. If body B has the power to discover and then publish body A’s findings, however, it may do so, and thereby undermine the ‘friendly’ approach of body A. This would affect the relationship between TransCo and body A, and could have a transformative effect on body A’s methods, or even its viability. These are just a few of the potential complications that could arise from the multiplication of independent overseers of public sector projects. In light of the potential complications, this paper examines three independent scrutiny mechanisms which, in the Australian state of Victoria, operate concurrently on ‘transport megaprojects’ (Altshuler and Luberoff, 2003; Kay, 2009; Priemus et al., 2008) that are procured by public-sector agencies via competitive tendering. There is no single definition of ‘megaproject’, but the term generally refers to discrete projects that are multi-faceted or complex, and that involve large expenditures ranging from hundreds of millions of dollars into billions of dollars. The concept encompasses large-scale engineering and capital works projects, information-technology projects, and major services contracts. For the purposes of the present paper, a ‘megaproject’ is valued in excess of A$100 million. In the transport sphere, relevant megaprojects include new railway infrastructure, freeways, major port infrastructure, public transport ticketing systems and long-term public transport system operating contracts. Relevant modes of procurement include ‘public-private partnerships’, alliancing, design and construct contracts, and franchise contracts. The three scrutiny mechanisms are performance auditing, probity auditing and gateway reviews. The mechanisms have overlapping goals relating to probity and value for money. The paper describes the three mechanisms, before considering whether there is evidence of the kinds of pitfalls identified above. Transport megaprojects are an appropriate site to study interactions between the three mechanisms because such projects constitute a discrete sample of projects that have all been subject to the three scrutiny mechanisms extensively. The projects face similar risks, and are of high public interest around the world (Cour des Comptes, 2005; Cox, 2003; Flyvberg et al., 2003; Kay, 2009; Ware et al., 2007). Also, in Victoria, transport megaprojects have been an area of procurement innovation: Victoria’s largest public-private partnership was a freeway project (Eastlink), and transport was the first Victorian government portfolio to adopt alliancing on a large scale as a method of major project procurement. The principal contribution of the paper is in analysing empirically how three overlapping audit and review processes operate together. The paper seeks to deepen the literature on interactions between audit and review mechanisms. There are few analyses of this kind,4 and none from Australia that relate specifically to transport megaprojects. The paper is also relevant to the literature on megaproject governance. For recent analyses of governance regimes for megaprojects, see Flyvberg et al. (2003), Klakegg et al. (2008, 2009), Miller and Hobbs (2005), Miller and Lessard (2001), Priemus et al. (2008) and Williams et al. (2010); for analyses specifically related to transport megaproject governance, see Priemus (2010) and other papers in the March 2010 special issue of the European Journal of Transport and Infrastructure Research, on ‘Large Transport Infrastructure Projects: Improving Institutions and Decision Making’. Williams et al. (2010) examined several modes of project scrutiny but did not explore how they worked (or did not work) in concert. Klakegg et al. (2008, 2009) studied similar instances of project assessment, but also did not focus on coordination of or interaction between the various scrutiny mechanisms to which the projects were subject.
EJTIR 11(1), January 2011, pp. 61-79 Kells Conflict between Independent Scrutinisers of Transport Megaprojects: Evidence from Australia
The paper is structured as follows. Section 2 describes the scope and method of the empirical study, and the context of transport megaproject procurement in Victoria. Section 3 describes three mechanisms of scrutiny of transport megaprojects that are used there. In Section 4, evidence of contradictory findings and conflict between the three mechanisms is presented. Section 5 concludes the paper with a discussion of implications for Europe and directions for future research.
2. Scope and method of the study Three independent review mechanisms in the Australian state of Victoria were analysed. Data were collected from the Parliament of Victoria (from legislation, parliamentary reports and transcripts), the Victorian Auditor-General’s Office (VAGO), Victorian Government departments and agencies, and professional standards bodies. Only publicly available sources were used. Performance audits, probity audits and gateway reviews are the principal broad-scope review mechanisms that operate regularly on procurement of transport megaprojects in Victoria. Transport projects are also subject to investigations by the Victorian Ombudsman, but those investigations occur irregularly and display considerable variability in their subject matter and scope. The State Services Authority (SSA) and the Victorian Competition and Efficiency Commission have also conducted a small number of reviews relevant to transport, but these have not generally been procurement-focused or project-specific. Some issue-specific and narrowscope review bodies (in areas such as sustainability and equity) have also examined aspects of transport projects, but the projects are not routinely subject to these reviews, and the reviews have not focused on value for money or probity in procurement. Eleven transport megaprojects were examined: the Eastlink freeway project; the ‘regional fast rail’ project; the Tullamarine and Calder Interchange (TCI) project (freeway interchange); the M1 Upgrade (freeway upgrade); the Port of Melbourne Channel Deepening project; the New Ticketing System (NTS) (for public transport); the Southern Cross Station redevelopment; the metropolitan train franchise; the metropolitan tram franchise; the ‘Improved local and cross-town bus services’ project; and the South Morang rail extension project. The total value of the projects was approximately A$16 billion in 2010 dollars. All the projects were above the A$100 million size threshold, and the average size was A$1.45 billion. The rationale for focusing on transport megaprojects was outlined in the previous section. The period of study was 2005–06 to 2009–10. During that time, all 11 projects were subject to VAGO performance audits. The results of the audits were published in the form of printed reports. The study involved an examination in detail of each of the reports to identify aspects of the performance audits’ scope and method. Also during the study period, probity auditing and gateway reviews became widespread in Victoria. Those types of scrutiny also resulted in reports being produced, but those reports were not published. Nor was the public routinely informed about whether specific projects were subject to gateway reviews or probity audits. Accordingly, indirect means were used to determine if projects had been subject to these scrutiny mechanisms. VAGO’s performance audit reports frequently refer to probity auditing and gateway reviews, and the detailed review of those reports was therefore also a source of information about the other scrutiny mechanisms. Using the performance audit reports as well as other public information, the author determined that no fewer than ten of the 11 projects in the study were subject to either probity auditing or gateway reviews (as well as performance auditing), and no fewer than five of the projects were subject to all three mechanisms. Precisely how many of the 11 projects were subject to all three mechanisms is likely to be significantly higher than five, given the extent of the obligation to undertake gateway reviews and probity auditing in Victoria, which is detailed in the next section.
There are several reasons why Victoria is a suitable location in which to study scrutiny of megaprojects. A state within a federation, Victoria is a first-world jurisdiction with a mixed economy. It has many institutional and cultural similarities (and some geospatial ones) with other advanced jurisdictions in Europe and North America. Victoria has a bi-cameral parliament and Westminster-style government. The state of Victoria is highly urbanised. The majority of the population (of five million people) live in Melbourne the capital, and most of the remainder dwell in regional municipalities such as Geelong, Ballarat and Bendigo, with radial transport links extending from Melbourne to these centres.5 Victoria’s public-sector scrutiny institutions and methods are similar to Britain’s and, to a lesser degree, those of other European countries. For example, performance auditing (also known as ‘value for money auditing’) is an important part of public administration in Victoria as well as in Sweden, Finland, the Netherlands and France. Another reason to study Victoria is the current level of transport infrastructure investment there. Across Victoria, a range of major road, rail and other transport projects are underway, with the goals of easing congestion and accommodating future growth. In 2008, the Victorian government released a ‘Transport Plan’ which featured projects to the value of A$38 billion. Before outlining the framework of megaproject scrutiny in Victoria, it is useful to introduce some other features of public administration there. The state’s principal minister (the premier) chairs a committee of senior ministers (the cabinet) which is the principal decision-making body in the executive branch of the government. With an annual operating budget of approximately A$50 billion, the Victorian government is responsible for delivering healthcare, education, police and transport services.6 The executive branch of the Victorian government is organised into large, multi-portfolio departments. One of those departments, the Department of Treasury and Finance (DTF), promulgates whole-of-government procurement policies and requirements, and provides guidance and ‘best practice advice’ to agencies undertaking major tenders. Another department, the Department of Transport, has primary responsibility for transport policy and for the oversight of major transport projects. Other transport agencies in the state include VicRoads (Victoria’s principal roads management agency), the Port of Melbourne Corporation, VicTrack (which owns much of the state’s rail infrastructure) and the Transport Ticketing Authority.
3. Description of the three scrutiny mechanisms 3.1 Performance auditing The Victorian Auditor-General is an independent officer of the state parliament. The AuditorGeneral is appointed for seven years (eligible for re-appointment) and may employ staff and delegate statutory powers and functions to those staff. Under Victoria’s constitution, ‘the Auditor-General has complete discretion in the performance or exercise of his or her functions or powers and, in particular, is not subject to direction from anyone in relation to whether or not a particular audit is to be conducted; the way in which a particular audit is to be conducted; [and] the priority to be given to any particular matter’. Since the 1980s, the Victorian Auditor-General’s Office (VAGO) has conducted performance audits of public-sector agencies, projects and programs. The adoption of a performance audit function reflected an international trend toward a more holistic approach to public sector auditing (Kells and Hodge, 2009). VAGO’s current performance audit mandate is provided by Melbourne has the world’s largest tram network (245 km), and extensive road and rail infrastructure. Melbourne’s metropolitan train and bus networks feature almost 400 km of track and more than 4000 km of road routes respectively. 6 Defence, immigration and border protection are national responsibilities. 5
section 15(1) of the Audit Act 1994, which states that ‘The Auditor-General may conduct any audit he or she considers necessary to determine – (a) whether an authority is achieving its objectives effectively and doing so economically and efficiently and in compliance with all relevant Acts; or (b) whether the operations or activities of the whole or any part of the Victorian public sector…are being performed effectively, economically and efficiently in compliance with all relevant Acts.’ The performance audit mandate covers all departments and agencies in the Victorian public sector. The Auditor-General has considerable discretion regarding the use of this mandate. For example, he has discretion regarding how many performance audits he undertakes, what he audits, the scope of his audits, and what matters are disclosed in his audit reports. The selection of performance audit topics is based on criteria such as materiality, risk and ‘concerns raised by Members of Parliament, government agencies or the community’ (VAGO, 2009). Transport megaprojects typically fit one or more of these criteria, and have therefore been an area of focus for VAGO in recent years. Approximately 30 VAGO performance audits are tabled in the state parliament each year. VAGO’s performance audits vary in scope and focus from one audit to the next, but the range of matters considered is broad. It includes project planning and funding, the procurement strategy, the conduct of the procurement (including the integrity of tender processes), negotiation with selected proponents, and all aspects of project delivery including value for money. Like all audits, VAGO’s performance audits are conducted ‘after the fact’, either subsequent to the completion of a project, or subsequent to the completion of a major component or phase of a project. Performance audits take a considerable length of time to complete (typically at least six to twelve months, and sometimes considerably longer). The principal direct cost of performance audits is the cost of salaries of VAGO staff, and the cost of consultants who are used from time to time to provide specialist input. The Audit Act requires VAGO to comply with Australian auditing standards in undertaking performance audits. The current standards of most relevance are: Standard on Assurance Engagements: ASAE 3500 Performance Engagements; and Standard on Assurance Engagements: ASAE 3000 Assurance Engagements Other than Audits or Reviews of Historical Financial Information (both issued by the Australian Auditing and Assurance Standards Board). The standards explicitly recognise that judgement plays an important role in the auditor’s decision making. Each performance audit results in a report that is tabled in state parliament, and is simultaneously published in printed form and on VAGO’s website. The reports feature a standardised format that includes findings and recommendations, which are not binding on the government or its agencies. 3.2 Probity audits ‘Probity’ is a flexible term that has connotations of integrity, honesty, uprightness, adherence to proper process, and being open to scrutiny. In Australia, the terms ‘probity auditor’ and ‘probity advisor’ have become part of the standard vocabulary for procuring public infrastructure (Shead, 2001). A number of different definitions of ‘probity auditor’ and ‘probity advisor’ exist. Under one interpretation, the role of a probity advisor in a major tender is to establish administrative arrangements for managing conflicts of interest, communication with tenderers and the evaluation of bids, while the role of the probity auditor is to independently review these arrangements and how they are applied, and to furnish an opinion to the procuring agency about the overall probity of the tender process (VAGO, 2007). Under DTF’s procurement policies, ‘A probity auditor may be used if the procurement is $10 million or more; sensitive; or complex’ (DTF, 2009c). As transport megaproject tenders meet
all three criteria (size, sensitivity and complexity), probity auditors are normally engaged to provide independent assurance about the tenders. According to DTF, ‘The main task of the probity auditor is to provide to the Secretary, or his or her nominee, an independent and appropriate signoff on probity requirements, in hindsight, at designated milestones in the process’ (DTF, 2009b). Probity auditors must be independent of the project team. In most cases, and for all the largest projects, the role of probity auditor is performed by a private sector firm. DTF maintains a panel of suitable firms that departments and agencies can engage to provide probity practitioner services. These firms include lawyers, accountants, auditors and specialist probity firms. In Victoria, government staff and agencies are subject to a number of probity obligations which are articulated in several instruments including: DTF ‘Good Practice Guidelines’ on the conduct of commercial engagements; other probity and procurement guidance issued by DTF and the Victorian Government Purchasing Board; the Public Administration Act 2004; the Code of Conduct for Victorian Public Sector Employees; the ‘Conflict of interest policy framework’ established by the Public Sector Standards Commissioner; and guidance issued by the Victorian Ombudsman on ‘Conflicts of interest in the public sector’. In these instruments, the concept of probity is defined broadly and includes steps taken to achieve value for money. Probity auditors are required to apply the Victorian government’s probity principles and to monitor adherence to other relevant probity requirements. Probity auditors are given full access to project documentation and project team personnel. The probity auditors document their findings in probity audit reports, which typically are not published, but are provided to the head of the project or the head of the department or agency procuring the project. 3.3 Gateway reviews Gateway reviews are a form of independent peer review that is increasingly being used around the world to improve the delivery of major public-sector projects. The concept was first developed by the British government, and a modified version of Britain’s gateway program was adopted by the Victorian government in 2003. Victoria’s gateway program initially applied to all ‘major capital investments’ in the state, but has since been expanded to encompass ‘all high-risk programs and projects’ including policy, infrastructure, organisational change and acquisition programs and projects, and ICT-enabled business changes (DTF, 2009a). The gateway program applies to all departments and agencies in Victoria. Gateway reviews are undertaken by specially convened review teams at six key project stages or ‘gates’ (see Figure 1). At each stage, the review team asks a standard set of questions to determine if the project team has taken all the necessary steps, and has achieved good practice. The benchmarks are sourced from good practice standards and documents such as the international ‘Project Management Body of Knowledge’ (PMBOK) and guidance developed by DTF. The scope of gateway encompasses: pre-project planning; business case development; project approval processes; market analysis and market soundings; procurement planning; management of the tender process; selection of one or more preferred proponents; negotiation with the preferred proponents; contract award; ongoing contract management; and evaluation of the project’s benefits (value for money). The purpose of gateway reviews is to ensure that ‘good practice’ is achieved at each project stage. Gateway review teams usually consist of three or four people (DTF, 2010b) and comprise a mixture of government and private sector staff. All members of each team are independent of the project being reviewed. The gateway unit in DTF appoints review teams as follows:
EJTIR 11((1), January 20011, pp. 61-79 Kells between Indep pendent Scrutiinisers of Tran nsport Megaprojects: Conflict b Evidence from Australia
for ‘‘high-risk’ projects, p the gateway u nit appointss an indepe endent revieew team and d review team m leader from m outside th he procuring g departmen nt or agency; and
for ‘‘medium rissk’ projects,, the gatew ay unit app points an in ndependent review team m leader from m outside th he department or agen ncy to lead a review te eam consistting of inde ependent depaartmental/aagency staff who work o outside the project p team m (DTF, 20100b).
Accordiing to DTF, ‘the ‘ indepen ndence of th he review tea am from the e project, an nd in the case of high risk reviiews, the independence e of the revi ew team fro om the depa artment, is th the key to delivering objectivee, high quallity reviews and reportss’ (DTF, 2010 0b).
Figure 1.. Project stagges and the coorresponding gateway ‘gattes’ Source: A Adapted from DTF website, accessed 7 Sep ptember 2009.
The gateeway review w team documents its ffindings, an nd uses thesse to develoop a report which is provided to the prroject directtor or the ‘‘senior resp ponsible own ner’ of the project. Th he report includess recommen nded actionss, along with h green, amb ber or red ‘ttraffic light’ ratings (DT TF, 2005): Green – ‘The pro oject is on target to succeed bu ut may ben nefit from the uptake e of the recomm mendations’; Amber – ‘T The project sh hould go forward with actions on rrecommend dations to be carrieed out befo ore the next gate’; and R Red – ‘To achieve a succcess the proj oject should take the recomm mended actio on immedia ately’. Red light rating gs are partiicularly sen nsitive becau use they imply siignificant criticism of th he reviewed project. The resu ults of gatew way reviewss are not pub blished. DTF F argues tha at the confid dentiality of gateway reports, and recomm mendations,, is a key feaature of thee program, as a it encouraages reviewees to be forthcom ming about issues i and progress, p and d it encoura ages reviewe ers to be can ndid in their findings (DTF, 20005; DTF, 20009a). As at Jan nuary 2010, 326 gatewa ay reviews h had been con nducted in Victoria. V Moore than 165 5 projects were reeviewed. Th he total vallue of the projects wa as more than A$30 biillion (DTF, 2010a). Currenttly, around 50 5 gateway reviews r are conducted each e year in Victoria.
EJTIR 11((1), January 20011, pp. 61-79 Kells between Indep pendent Scrutiinisers of Tran nsport Megaprojects: Conflict b Evidence from Australia
3.4 Discu ussion of the three scrutin ny mechanism ms Figure 2 shows in n a simplifie ed way how w the threee scrutiny mechanisms m s relate to the t state parliameent and to the t various components c s of the execcutive brancch of govern nment (the executive e branch iis the shadeed area in th he figure). A All three mecchanisms ha ave the same me target in common, c but diffeer in the naature of theiir relationsh hip with thee public sector. Other ccharacteristics of the three m mechanisms are summa arised in Taable 1. The table shows that the m mechanismss have a number of other sim milarities and differencees. The threee mechanissms differ with w regard tto their timing and periodicity. Gatteway review ws occur immediaately beforee key projecct decisions,, whereas performance p e audits and d probity au udits are retrospeective. Performance aud dit reports in n particular are publishe ed a consideerable length h of time after thee events they y analyse.
Figure 2.. Principal in nstitutional reelationships Source: O Original figuree.
Perhapss the most important difference is that gatteway and probity aud dit reports are not publisheed, whereass performan nce audit rreports are.. Publication has a nu umber of effects: e it contribu utes to the au uditeesâ€™ acco ountability tto parliamen nt and the wider w commu munity, subje ecting the perform mance audit findings to open scrutiiny, includin ng scrutiny by the med dia; it may influence i what the auditors are a told and how they aare told it; and a the prosspect of open en scrutiny may m lead auditeess to make im mprovementts they otherrwise would d not make.
In part because of the publication of performance audit results, and in part because, of all three types of scrutiniser, performance auditors are institutionally the most distant from the auditees (Figure 2), the relationship between performance auditor and auditee is more adversarial than that between gateway reviewer and reviewee, and that between probity auditor and auditee. From auditees, VAGO seeks formal responses to its draft performance audit reports, and publishes those responses in its final reports. The reports and responses indicate that the auditees are sometimes uncooperative with the audit process, and are frequently sensitive about the auditor’s findings and conclusions. Compared with the other two types of review, performance auditing leans toward the ‘blame culture’ in the sense of Klakegg at al. (2008). A pertinent question to ask is whether the three mechanisms are effective. This is a difficult question to answer definitively because data on the impact and effectiveness of the three mechanisms are not published. Even if such data were published, a problem of attribution would arise: because the three mechanisms have similar goals and features, it would be difficult to disentangle their separate influences on the success or failure of transport megaprojects and the integrity of tender processes. However, some tentative comments and inferences about the impact of the mechanisms can be made. Victoria rates highly on indices of government transparency and integrity (SSA 2010). In transport, some probity issues have arisen, but there have been no major scandals in recent years, and no significant instances of corruption. This points to the review mechanisms being successful with regard to their probity and integrity goals. However, it is difficult to draw causal links between the three mechanisms and the state’s integrity outcomes. Victoria has many other integrity-enhancing features, such as strong legal institutions; cultural values that underpin a respect for the rule of law; a public sector code of conduct supported by sanctions; an independent judiciary; free media; and other checks and balances including parliament and the police. It is not clear whether the three mechanisms have a separate beneficial impact over and above the other integrity-enhancing factors at work in Victoria. The review mechanisms have goals that extend into matters of value for money. While some projects (such as the TCI project) were delivered on time and under budget, delays and costoverruns have occurred on other projects, as have other problems relating to value for money (Infrastructure Australia, 2008; VAGO, 2010; VAGO, 2006). It is therefore possible to conclude tentatively that the three mechanisms have not been entirely successful in overcoming these problems. Several causes of this are possible. Addressing probity and integrity may be easier than achieving value for money, perhaps because the former goal may be amenable to more generic or ‘boilerplate’ solutions, while the latter goal may require more complex and idiosyncratic actions for each project. Arguably, the incentives of tenderers and other private sector participants are better aligned with the government’s integrity goals than they are with the goal of maximising value for the community, sometimes at tenderers’ expense. These differing alignments may encourage firms to cooperate with probity standards, while engaging in strategic behaviour to maximise private profits.
Table 1. Summary of the three independent scrutiny mechanisms Mechanism
Formal legislative mandate
All aspects of projects including probity and value for money
After completion of a project or a major stage of a project
Ongoing public sector staff (audit office staff). Supported by external consultants and specialists as needed
Must adopt performance audit standards. Scope for discretion and judgement
Standardised report style, with some flexibility in reporting approach. Report is tabled in parliament and published
Initiative of executive branch
Probity (including value for money) in the procurement process
Ongoing oversight through the procurement process. Periodic review reports at discretion of auditor
Usually private sector probity specialists (consultants)
Applies probity standards and state government probity principles. Considerable discretion about methods used
Probity report including a conclusion (‘sign off’) about the probity of the process. Report is typically not published
Initiative of executive branch
Six key ‘gates’ in the procurement process, including ‘benefits realisation’
Before each gate or key decision point in the procurement process
Specially convened mixed teams of public sector staff and private sector experts and consultants (for high risk projects)
Rigid review framework built around the six gates
Gateway review report including ‘traffic light’ rating. Provided to ‘senior responsible owner’. Report is not published
Source: Original table.
4. Evidence of conflict between the three scrutiny mechanisms This section explores interactions between the three mechanisms, and presents evidence, firstly, of contradictory findings, and, secondly, of the mechanisms adopting conflicting methods. 4.1 Contradictory findings In the period of study, the three mechanisms had overlapping scopes: all three explicitly looked at ‘value for money’, as well as compliance with the rules covering tendering in Victoria. The mechanisms had similar goals and made similar claims about what they did and why. There were many examples where VAGO’s performance audit reports covered the same ground as gateway reviews and probity audits. This included compliance with procurement rules and guidelines, integrity of tender processes, and the achievement of value for money. For example, VAGO’s Eastlink performance audit covered ‘whether the tender process was well managed’, whilst in relation to Train and Tram Franchising, VAGO considered ‘whether probity requirements were observed’. These are core matters of interest for gateway reviews and probity audits. The scope of VAGO’s Fast Rail interim performance audit included ‘the tender evaluation and selection process for the project’ (ie. elements of gates 3 and 4). The final performance audit report for that project, ‘Delivering regional fast rail services’, encompassed project planning, procurement and delivery (ie. elements of all six gates).7 As VAGO falls outside the executive branch of government, the Victorian parliament is the only institution that has authority over all three scrutiny mechanisms. With regard to private-sector probity auditors, this authority is especially limited and indirect. While parliament is the source of relevant legislation and the executive branch’s authority to operate, parliament does not perform a coordinating role or a directing role in relation to the three mechanisms. Rather, the mechanisms evolve and operate at different speeds and in a decentralised way. For example, the evolution of performance auditing reflects trends in public sector auditing, and the AuditorGeneral’s use of his discretion with regard to how to perform his statutory roles, while probity audit methods and conventions are partly driven by commercial and professional trends outside government. Because the three mechanisms covered the same matters in similar ways, and were not coordinated, there was the potential for the mechanisms to reach contradictory conclusions on those matters. One way to avoid contradictory findings would be for the scrutiny mechanisms to ‘selforganise’, such as by treating each other’s findings as substitutable for, or as acquitting the need to provide, their own findings. In other words, they could cede turf to each other in a limited way. In the period of study, however, the mechanisms did not do so, and nor did the formal arrangements for the mechanisms envisage such ‘self organisation’. In the gateway program, gateway review teams were not afforded latitude to use probity audit reports to acquit part of the scope of the gateway reviews. Likewise, probity audit arrangements did not envisage using gateway reports to address probity audit criteria. Moreover, because probity audits and gateway reviews occur before performance audits, it is not feasible for them to cede ground to VAGO. Conversely, it could in principle be practicable for VAGO, which has been undertaking performance audits since the 1980s, to cede ground to probity auditing and to gateway. ‘The audit examined the adequacy of: the feasibility studies to properly inform the government’s decision to proceed; the development phase to put in place contracts which provided a sound basis for the delivery of the agreed infrastructure upgrade [and] the delivery phase to manage all the fast rail components to deliver the planned improvements within the planned time lines and budgets’ (VAGO, ‘Results of special audits and other investigations’, August 2006).
Nevertheless, this has not occurred. The emergence of probity auditing, and the implementation of the gateway program, were not associated with any reduction in the Auditor-General’s role. VAGO’s legislation was not revisited, for example, to make space for the new mechanisms. Nor did the Auditor-General voluntarily retreat from the subject matter covered by probity auditors and gateway reviews, or use probity auditing and gateway results to replace or reduce any of its own work. (VAGO did, however, use them to develop performance audit criteria and lines of inquiry.) Accordingly, the potential for contradictory findings was not addressed either by coordination ‘from above’, such as by parliament, or coordination in the form of self-organisation among the scrutiny mechanisms. In this context, the study found evidence of conflict between performance auditing and probity auditing. During the period of study, VAGO was frequently critical of the conduct and findings of probity auditors. For example, with respect to the TCI project, VAGO criticised the probity audit report on the grounds that it did not include ‘an audit opinion, nor [was] there any reference to the fact that probity audit services were provided.’ In VAGO’s performance audit of Melbourne’s New Bus Contracts, it criticised the brevity of the interim and final probity audit reports. In the Train and Tram Franchising projects, VAGO criticised the probity auditor’s role in managing conflicts of interest. The most intense area of criticism related to the boundaries and definition of probity auditor roles. Throughout the study period, there was pointed disagreement between VAGO and probity practitioners regarding the boundaries of probity audit and probity advice services. The heart of the disagreement was whether it was proper for a single probity practitioner to establish and advise on the probity framework for a megaproject, and then audit adherence to that framework. VAGO criticised the combining of probity auditor and probity advisor roles (or the lack of a clear distinction between these roles) in the following six transport megaprojects: train franchising; tram franchising; NTS; TCI project; M1 Upgrade; and Melbourne’s new bus contracts. This conflict was not a minor matter. A significant part of VAGO’s NTS report, for example, was devoted to this issue, as was a significant part of the parliamentary hearings that were subsequently held regarding that report. VAGO used the failure to separate probity auditor and probity adviser roles as the basis of a series of negative conclusions about the probity arrangements for transport megaprojects in Victoria. In this disagreement, parliament’s Public Accounts and Estimates Committee (PAEC 2008) sided with VAGO, but the Government and DTF refused to endorse VAGO’s position, instead issuing contrary guidance for agencies engaging probity practitioners. In the NTS performance audit report, VAGO recommended that DTF amend the ‘Policy for the Conduct of Commercial Engagements’ to provide that, for major and complex tenders, probity auditor and probity advisor functions should be provided by different parties. DTF rejected the recommendation, arguing that a probity auditor could prepare the probity plan and subsequently report on whether there had been compliance with the plan, and that it was a matter for departmental heads to determine, depending on the particular transaction, whether the separation of probity auditor from probity advisor was warranted, and whether these practitioners should be engaged externally or internally to Government. This disagreement was not conclusively resolved. In December 2008, the Minister for Finance stated that DTF noted VAGO’s recommendation that probity auditor and advisor roles be kept separate, but went on to state that ‘the current policy titled Government Purchasing Board Policy
for the Conduct of Commercial Engagements [did] not require amendment.’8 At the time of writing, the Auditor-General and DTF continue to hold opposing positions regarding probity auditing. Probity auditors who continue to provide probity advice are therefore exposed to criticism from the Auditor-General, while agencies that appoint probity auditors in accordance with DTF guidance are similarly exposed to adverse findings from the parliamentary auditor.
4.2 Conflicting methods In addition to tension between performance auditing and probity auditing, the study also found evidence of tension between performance auditing and gateway reviews. Specifically, there was tension between, on the one hand, the gateway goals of encouraging candidness in reviewees and frank advice from gateway reviewers, and on the other hand the Victorian AuditorGeneral’s wide powers to examine gateway review content and to publish some or all of that content. The goals of gateway are supported by the confidentiality of gateway review reports, but the Auditor-General is in no way bound to preserve that confidentiality, and has increasingly chosen not to do so. In principle, there are several ways in which VAGO can disclose gateway content beyond merely mentioning that a gateway review has been conducted. For instance, VAGO can disclose: the number of findings or recommendations in a gateway report; the general subject matter of those findings and recommendations; specific details of the findings and recommendations; and the gateway ‘traffic light’ ratings. A variety of these modes of disclosure can be seen in VAGO’s performance audit reports. The NTS project was subject to two gateway reviews, and VAGO’s NTS performance audit report (October 2007) included detailed information about the matters covered by the reviews. In the report, however, VAGO did not disclose the gateway findings or recommendations. Indeed, before December 2007, VAGO did not disclose any findings and recommendations from gateway reviews of transport megaprojects, despite the fact that gateway had been in operation since 2003. During this period, the Auditor-General effectively tied his own hands and adhered to a voluntary convention of not disclosing such gateway review content. A change occurred in December 2007, when VAGO reported on its performance audits of the TCI project and the M1 Upgrade. The TCI project was subject to a gateway review, and VAGO noted in its performance audit report that the gateway review had identified ‘exemplar’ practices within the alliance structure for the project, and that all recommendations arising from the gateway review were agreed by management. Furthermore, VAGO recommended that the project ‘should undergo a “Gate 6 – Benefits Realisation” review’.9 VAGO’s reporting for the M1 Upgrade went further by disclosing a specific gateway finding as well as two gateway recommendations. In the performance audit report, VAGO stated that ‘a Gateway Review in June 2006 recommended that the management of the project to strengthen the West Gate Bridge should be brought into the overall governance and management framework of the M1 project’. VAGO also noted that ‘A Gateway Review undertaken in June 2006 identified VicRoads’ limited experience with alliance contracting, and the resulting need to put in place a program of activities for key project team members to ensure that they have the required capabilities for their roles’. ‘Response by the Minister for Finance to the Auditor-General’s Reports issued during 2007–08’, Government of Victoria, December 2008. 9 In the TCI performance audit report, VAGO also published an agency response which referred to gateway findings in support of the decision to adopt an alliance approach: ‘The risk analysis supported some form of relationship contract rather than the traditional “hard dollar” contract. This decision was supported by the DTF Gateway Review undertaken at the time.’ 8
In May 2009, VAGO’s performance audit of the Port of Melbourne Channel Deepening project disclosed five gateway review findings, whilst VAGO’s June 2010 performance audit report on ‘Management of Major Rail Projects’ detailed gateway findings and recommendations relating to the South Morang Rail Extension project. Importantly, that performance audit report indicated for the first time the status assigned by the gateway review team to some of the gateway recommendations. The South Morang Project was subject to gateway reviews in February and August 2009. VAGO reported that the February 2009 review found that the absence of a consolidated risk management plan and detailed risk register had: contributed to a lack of structure for managing project risks; resulted in overlaps between risk allowances and contingencies for some cost components; and made it difficult to accurately estimate the impact of risks on costs and timelines. VAGO further reported that the August 2009 gateway review found the project was in a ‘critical situation’ because the department had not made sufficient progress given its scale, significance and planned timelines. VAGO reported that the gateway reviewers found that the department ‘needed to act immediately on several issues’. In other words, VAGO reported that the gateway review had given the project ‘red lights’. These findings indicate that the Victorian Auditor-General has shown an increasing willingness to disclose gateway report content in his transport megaproject performance audits. VAGO’s relationship with gateway marks out a progression along the spectrum of increasing disclosure of gateway findings, that began with mentioning that a gateway review was conducted, then progressed to mentioning the matters covered in the review, then disclosing findings, then findings and recommendations and, finally and most sensitively, disclosing the criticality of the gateway recommendations. This progression has several potential implications for the gateway review program. The confidentiality of gateway review reports is a mainstay of the program, and one that distinguishes it from performance auditing. Paradoxically, as the Auditor-General increasingly uses his powers to disclose gateway findings, gateway may become more like performance auditing: if an expectation arises that gateway results will ultimately be published, then reviewees are likely to be less candid, and the review relationship is likely to become more adversarial and less ‘friendly’ in the sense of Williams et al. (2010). The methods used by VAGO and gateway in relation to gateway review content are in conflict. Could the conflict between VAGO and gateway be removed or prevented? Yes, but only with changes that would alter fundamentally the character of performance auditing or gateway reviews as they are practised in Victoria. Routinely publishing gateway review reports, for example, would transform gateway, while removing the Auditor-General’s discretion to publish information from gateway reports would be a critical diminution of his powers, as would any attempt to reduce the Auditor-General’s jurisdiction in matters of megaproject tendering.
5. Conclusion and implications for Europe Transport megaproject procurement is one of the most reviewed aspects of public administration in Victoria. The multiplication of scrutinisers of transport megaprojects can be seen as further evidence of an ‘audit explosion’. The multiplication of scrutiny mechanisms has a number of implications for public administration and for society as a whole. This paper explored an instance where multiple independent scrutinisers examining the same projects have reached contradictory conclusions and adopted conflicting methods. While the scrutinisers were independent of the projects they examined, they were also in some respects interdependent.
There are a number of ways to interpret the paper’s findings. One is to think of them as evidence of an SAI asserting its supremacy over newcomers, or over ‘small fry’ reviewers operating in its patch. Another is in the context of the evident trend toward greater public sector transparency. It is a commonplace to claim that appetites for government transparency and accountability are increasingly ravenous. Those appetites are being fed by more extensive whistleblower laws and freedom on information laws, as well as unauthorised WikiLeaks-style disclosures. In the context of that trend, ‘friendly’ mechanisms that rely on secrecy may become less palatable and, ultimately, less viable. Yet another interpretation of the results is that they are one more skirmish in the long war between an executive branch and its external auditor. As was noted in Section 2, a number of European countries have adopted similar scrutiny mechanisms to those studied here; it is notable for example that ‘friendly’ review approaches have been used in defence procurement in Norway (Klakegg et al. 2008; Williams et al., 2009). Indeed, some countries in Europe have two or more mechanisms operating in concert that are analogous to the mechanisms studied here: in Britain, for instance, the National Audit Office conducts value-for-money audits of megaproject tenders10, the government commissions gateway reviews, and international accounting and audit firms offer probity audit services. The paper’s findings are therefore relevant to the design of megaproject scrutiny in Europe, and to European public sector scrutiny regimes more generally. One clear lesson for Europe is that there is a place for coordination of multiple scrutiny mechanisms, or at least for establishing a clear and workable hierarchy between them. SSA (2010) recommended greater coordination of oversight bodies authorised by the legislature; the evidence presented here suggests that the need for coordination also extends to scrutiny mechanisms that originate from beyond the legislature. A further lesson is that, in designing oversight mechanisms, attention should be paid not just to vertical or bilateral relationships between reviewers and reviewees, but also to horizontal relationships between multiple reviewers. The study’s results have a number of implications with regard to megaproject governance. First of all, insofar as independent review is an aspect of the megaproject governance, the study is further evidence of the dynamic nature of that governance (Miller and Hobbs 2005, Williams et al. 2010). Secondly, to the extent that external review mechanisms provide useful input for megaproject decision-makers, contradictory recommendations would make the mechanisms less useful for avoiding project pitfalls. Thirdly, if there is reason to believe that gateway-style reviews are more useful in megaproject governance (because they provide more candid and timely input) and yet this type of review is most likely to lose out in conflicts with SAIs like VAGO, then megaproject governance may be adversely affected. Underpinning the paper’s findings are several cultural factors. For example, the idea of making candid disclosures to ‘friendly’ reviewers (Williams et al., 2010) such as gateway review teams is imbued with assumptions and relationships that may be culturally specific. Similarly, the willingness of the Auditor-General to disclose otherwise secret gateway results may depend on a particular cultural context; elsewhere, SAIs may not be inclined to disclose these results, or they may move much more quickly along the disclosure progression than did Victoria’s AuditorGeneral. A question for future research is how these cultural elements and dynamics would play out differently outside the countries from the Anglo-American cultural sphere. The paper provides a number of other possible lines of inquiry for future research. Such research would desirably characterise more fully the types of interaction that occur between multiple scrutiny mechanisms, including between gateway reviews and probity audits (possibly using
10 The scope and method of the UK NAO’s value for money audits are substantively equivalent to VAGO’s performance audits.
survey evidence or interviews with practitioners); examine the extent to which gateway reviews and probity auditing can be seen as strategies of the executive branch to claim the megaproject scrutiny ‘turf’, or to otherwise influence independent performance auditing; examine how the scrutiny mechanisms have been applied with regard to a wider range of projects in the transport arena and beyond; and seek to better understand the costs and benefits of the three mechanisms individually and collectively.
References Altshuler, A. and Luberoff, D. (2003). Mega-projects. The changing Politics of Urban Public Investment, Washington D. C., The Brookings Institution. Barzelay, M. (1996). Performance Auditing and the New Public Management: Changing Roles and Strategies of Central Audit Institutions. in: OECD (1996), Performance Auditing and the Modernisation of Government, Paris, OECD. Bowerman, M., C. Humphrey and Owen, D. (2003). Struggling for supremacy: the case of UK public audit institutions. Critical Perspectives on Accounting, Vol. 14, pp. 1–22. Cour des Comptes (2005). Les transports publics urbains. Paris, Cour des Comptes. Cox, W. (2003). Competitive Participation in U.S. Public Transport: Special Interests Versus the Public Interest. Paper Presented to the 8th International Conference on Competition and Ownership in Land Passenger Transport, Rio de Janeiro, September 2003 (revised). DTF [Department of Treasury and Finance] (2005). The Victorian Gateway Experience. Presentation by Wayne Sharpe, 2005. DTF (2009a). Overview: Gateway Review Process. Melbourne, Government of Victoria. DTF (2009b). Good Practice Guidelines: Conduct of Commercial Engagements. Strategy and Policy, Government Services Group, DTF, Melbourne, Government of Victoria. DTF (2009c). Template: Procurement Process Report – Tenders. Government Services Group, DTF, Melbourne, Government of Victoria. DTF (2010a). Gateway Newsletter. January 2010, Melbourne, Government of Victoria. DTF (2010b). Gateway website, www.gatewayreview.dtf.vic.gov.au, accessed 14 June 2010. Flyvbjerg, B., N. Bruzilius and Rothengatter, W. (2003). Megaprojects and Risk: An Anatomy of Ambition. Cambridge: Cambridge University Press. Head, B., A. J. Brown and Connors, C. (eds.) (2008). Promoting Integrity: Evaluating and Improving Public Institutions. Farnham and Burlington, Ashgate Publishing. Henkel, M. (1991). The New ‘Evaluative State’. Public Administration, Vol. 69, No. 1, pp. 121–36. Hood, C., C. Scott, O. James, G. Jones and Travers, T. (1999). Regulation inside government: Waste watchers, quality police, and sleaze-busters. Oxford, Oxford University Press. Infrastructure Australia (2008). Review of Major Infrastructure Delivery. Sydney, Infrastructure Australia, December 2008. Kay, M. A. (2009). Transportation megaproject procurement: benefits and challenges for PPPs and alternative delivery strategies, and the resulting implications for Crossrail. Thesis (Master in City Planning and Master of Science in Transportation), Massachusetts Institute of Technology, Department of Civil and Environmental Engineering.
Keen, J. (1999). On the Nature of Audit Judgements: The Case of Value for Money Studies. Public Administration, Vol. 77, No. 3, pp. 509–25. Kells, S. and Hodge, G. (2009). Public sector performance auditing: Reconceptualising the task. Journal of Contemporary Issues in Business and Government, Vol. 15, No. 2, pp. 33–60. Kells, S. and Hodge, G. (2010). Redefining the performance auditing space. Asia Pacific Journal of Public Administration, Vol. 32, No. 1, pp. 1–26. Klakegg, O. J., T. Williams, O. M. Magnussen and Glasspool, H. (2008). Governance frameworks for public project development and estimation. Project Management Journal, Vol 39, No. S1, pp. S27–S42. Klakegg, O. J., T. Williams and Magnussen, O. M. (2009). Governance frameworks for public development and estimation. Newtown Square, Project Management Institute. Klemperer, P. (2002). What Really Matters in Auction Design. Journal of Economic Perspectives, Vol. 16, No. 1, pp. 169–89. Leeuw, F. L. (1996). Auditing and Evaluation: Bridging a Gap, Worlds to Meet? New Directions for Evaluation, Vol. 71, Fall, pp. 51–60. Lindeberg, T. (2007). The Ambiguous Identity of Auditing. Financial Accountability and Management, Vol. 23, No. 3, pp. 337–50. Lonsdale, J. (1999). Impacts. in: Performance or Compliance? Performance Audit and Public Management in Five Countries, Oxford, Oxford University Press. Miller, R. and Hobbs, B. (2005). Governance regimes for large complex projects. Project Management Journal, Vol. 36, No. 3, pp. 42–50. Miller, R. and Lessard, D. (2001). The Strategic Management of Large Engineering Projects: Shaping Risks, Institutions and Governance, Cambridge, MIT Press. NAO [National Audit Office] (2010). Assurance for high risk projects. London, NAO. OECD (1996). Performance Auditing and the Modernisation of Government. Paris, OECD. PAEC [Public Accounts and Estimates Committee] (2008). Review of the findings and recommendations of the Auditor-General’s reports tabled July 2006 – February 2007. Melbourne: Parliament of Victoria. Pollitt, C., X. Girre, J. Lonsdale, R. Mul, H. Summa and Waerness, M. (eds) (1999). Performance or Compliance? Performance Audit and Public Management in Five Countries, Oxford, Oxford University Press. Power, M. (1999). The Audit Society: Rituals of verification, Oxford, Oxford University Press. Power, M. (2000). The Audit Society – Second Thoughts. International Journal of Auditing, Vol. 4, No. 1, pp. 111–19. Priemus, H. (2010). Decision-making on Mega-projects: Drifting on Political Discontinuity and Market Dynamics. European Journal of Transport and Infrastructure Research, Vol. 10, No. 1, pp. 19– 29. Priemus, H., B. Flyvbjerg and van Wee, B. (2008). Decision-Making on Mega-Projects. Cost-Benefit Analysis, Planning and Innovation. Cheltenham, Edward Elgar. Shead, B. (2001). Probity Auditing: Keeping the Bureaucrats Honest? Australian Journal of Public Administration, Vol. 60, No. 2, pp. 66–70.
Sloan, N. (1996). The Objectives and Performance Measurement of Performance Audit, in: OECD (1996), Performance Auditing and the Modernisation of Government, Paris, OECD. SSA [State Services Authority] (2010). Review of Victoria’s integrity and anti-corruption bodies. Melbourne, SSA. VAGO [Victorian Auditor-General’s Office] (2006). Results of special audits and other investigations, Melbourne, VAGO. VAGO (2007). The New Ticketing System tender. Performance audit report. Melbourne, VAGO. VAGO (2009). Annual Plan 2009–10. Melbourne, VAGO. VAGO (2010). Management of major rail projects. Performance audit report. Melbourne, VAGO. Ware, G. T., S. Moss, J. E. Campos and Noone, G. P. (2007). Corruption in Public Procurement: A Perennial Challenge. in: J. Edgardo Campos and Sanjay Pradhan (eds.) The Many Faces of Corruption: Tracking Vulnerabilities at the Sector Level, Washington D.C., World Bank. Williams, T., O. J. Klakegg, O. M. Magnussen and Glasspool, H. (2010). An investigation of governance frameworks for public projects in Norway and the UK. International Journal of Project Management, Vol. 28, pp. 40–50.
Issue 11(1) January 2011 pp. 80-95 ISSN: 1567-7141 www.ejtir.tbm.tudelft.nl
Do Planners Get it Right? The Accuracy of Travel Demand Forecasting in Norway Morten Welde1 Norwegian Public Roads Administration, Oslo, Norway; Norwegian University of Science and Technology, Trondheim, Norway
James Odeck2 Norwegian University of Science and Technology, Trondheim, Norway
This paper deals with the accuracy of travel demand forecasts among Norwegian road projects. We use data collected from tolled roads and toll free roads. The results reveal that while traffic forecasts of tolled schemes are fairly accurate, traffic forecasts among toll free roads have a higher degree of inaccuracy and are generally underestimated. An explanation for the observed discrepancy between estimated and actual traffic among toll free roads is that road planners may have ignored the existence of induced traffic and that the standard national traffic growth rates used in the transport models has been too low. For tolled roads, an explanation for the higher degree of forecast accuracy is that planners over the years have been scrutinized to provide careful estimates. Our recommendation is that traffic forecasts provided by planners should constantly be subjected to scrutiny by independent consultants before being presented to the decision makers. Aspects that need to be specifically examined include: (1) the extent to which a road project may lead to induced traffic, (2) the extent to which transport models accommodate appropriate factors and, (3) the extent to which forecasts made address uncertainties by providing confidence intervals of estimates. Keywords: Traffic-forecast accuracy, toll roads, toll free roads
1. Introduction Risk and uncertainty are issues of increasing concern in transport planning, and it is generally acknowledged that inaccurate travel-demand forecasts represent a major source of risk in the planning of infrastructure projects. International experience suggests that bias, or deliberately skewed forecasts, may play a role in the planning of road-infrastructure projects and that risks are often downplayed.
P.O. box 8142 Dep, NO 0033, Oslo, Norway, T: +4797730031, E: email@example.com NO 7491, Trondheim, Norway, T: +4748025940, E: firstname.lastname@example.org
EJTIR 11(1), January 2011, pp. 80-95 Welde and Odeck Do Planners Get it Right? The Accuracy of Travel Demand Forecasting in Norway
The over- or underestimation of traffic levels can have severe implications. Traffic forecasts are used to determine the capacity of transport infrastructure, and inaccurate traffic forecasts can therefore result in inefficient and inaccurate sizing of the road. Accurate forecasts are also important from a socioeconomic point of view. All road projects in Norway and most other countries of Western Europe are subjected to traditional cost-benefit analyses, which rely heavily on the accuracy of the forecasts being used. If traffic levels turn out to be significantly lower than estimated, this can affect total benefits derived from time savings, reduced accidents or lower vehicle-operating costs. In the case of traffic underestimation, the capacity relief on the congested links could turn out to be lower than planned. This may distort the social viability of such projects and result in non viable projects being implemented. The end result may be inefficient resource allocation. For toll projects, the implications of inaccurate forecasts are even more serious. Whether a road project can (completely or partly) be financed using tolls or not depends largely on the traffic level (i.e., the number of paying vehicles). Thus, in addition to the consequences for toll free roads, toll roads on which traffic levels fail to meet expectations also risk financial default. Furthermore, toll roads are often financed through non-recourse loans that are secured against future toll revenue only and with no other collateral. Bondholders and lenders should therefore require proposed toll roads to be subjected to a thorough risk assessment before investing in projects where the repayment of loans relies on precise traffic estimates. Over the years, several toll projects have experienced financial difficulties due to traffic shortfalls, cost overruns and/or increased interest rates. The Ă…lesund Tunnels project in Norway experienced payment difficulties soon after opening in 1987, and the main creditor, SunnmĂ¸rsbanken, eventually collapsed. Despite a restructuring of the loans, the project was, in effect, bankrupt. The debt continued to increase, and when the project finally was terminated in October 2009, the remaining debt was still some â‚Ź 165 million, which had to be covered by the government. To date, however, it is the only Norwegian toll project that has gone into default. With over 100 projects financed by tolls, the success rate of Norwegian tolling must hence be considered high. Internationally, the Hungarian M1/M15 represents a well-known example of overestimation. The project opened on time and within budget, but the traffic soon turned out to be only about half of what was projected. As the concessionaire relied solely on the traffic revenue, guarantees from both the shareholders and the state had to be drawn, and eventually, the concession was nationalised and toll rates halved. The shareholders suffered substantial losses and received no compensation (Joosten, 1999). More recently, the M6 toll road outside Birmingham, UK, is now being used by less than half the number of vehicles for which it was intended, and haulers have called for the road to be subsidised to ease congestion on the main M6, which has no tolls (BBC, 2008). The aim of this study is to provide new evidence on the magnitudes of traffic forecast inaccuracies using Norway as a case study. We provide explanations for the accuracies and inaccuracies and, based on these explanations, give recommendations for improving road-forecast practices. The differences in the forecast accuracies between toll and toll free road projects are specifically examined. The paper is organised into the following sections. Section 2 discusses the forecast uncertainties on toll roads versus toll free roads. Section 3 presents the data and methodology used in the analysis. In Section 4, the results are presented, and in Section 5, some conclusions are drawn.
EJTIR 11(1), January 2011, pp. 80-95 Welde and Odeck Do Planners Get it Right? The Accuracy of Travel Demand Forecasting in Norway
2. Forecast inaccuracies for toll roads vs. toll free roads The practice of financing new infrastructure through user fees is increasing worldwide. For roads, cost recovery through tolls is becoming ever more common as total tax revenues are often insufficient to cover the requisite infrastructure investments. Traffic forecasting is a complex issue, and adding tolls to the calculation normally increases the uncertainty of the forecasts. Road users respond to tolls in various ways, not all of which are rational, and the models used to forecast traffic are not necessarily designed to incorporate these reactions. The financial viability of a toll project relies heavily on the number of paying vehicles passing through the toll stations, and overestimation of traffic can potentially have severe financial implications. Thus, it is expected that planners treat toll projects with a higher degree of caution and, even if uncertainties in the estimates cannot be eliminated, use conservative traffic estimates to avoid overly optimistic forecasts. Forecast inaccuracy is not necessarily a problem. If the errors for various projects are randomly distributed around the true mean, there is a possibility that they would cancel each other out for a given project portfolio. When the forecasts are systematically biased, however, with averages significantly different from zero, perhaps due to over optimism or downright dishonesty on the part of the planners, the problem should be taken more seriously. Whereas transport models can be improved through increased computing power, improved data quality and other factors, deliberate human error is much harder to completely avoid. The concept of optimism bias or risk denial has been the focus of several studies by Flyvbjerg (2005) and Flyvbjerg et al. (2005, 2006). Based on the data from transport projects around the world, the authors concluded that planners in the transport industry do a poor job of estimating demand. For roads, the actual traffic was found to be, on average, 9.5% higher than forecasted. The actual and forecasted traffic differed by more than Âą 20% in over half of the road projects in the sample. Based on these rather disappointing results, Flyvbjerg suggested that planners and decision makers should take traffic forecasts, especially rail forecasts, which do not properly deal with uncertainty with â€œa pinch of saltâ€? (Forster, 2006, p. 9). Furthermore, Flyvbjerg used these results to make allegations regarding the professional honesty (or dishonesty) of the planners and argued that the end result was often that the most misrepresented projects were built rather than the best ones (Flyvbjerg, 2007). This was opposed by Osland and Strand (2010), who found no general support for the theory of strategic misrepresentation and argued that there are other mechanisms at work that could better help to explain the variations in the forecast accuracies that were often observed. It is often assumed that planners have become better at predicting traffic levels due to improvements in transport models and computing power. Flyvbjerg et al. (2006) did not support this proposition. In fact, the opposite seems to be the case for Danish road projects, as forecasts there seem to have become more inaccurate over time. Odeck et al. (2009), however, reached different conclusions. By investigating the accuracies of the national and regional traffic forecasts, they found that the forecasts have become more accurate since 2001, when the regional and national transport models were improved. Although their findings relate to forecasts at the macro- and regional levels rather than project-specific forecasts, it is still of interest to compare their results with ours. Traditionally, the studies of forecast accuracy have been based on toll free roads. With the use of toll financing increasing, however, toll projects have come under increasing scrutiny, especially from credit-rating agencies that routinely gauge the financial viability of such projects on behalf of potential investors. Perhaps the first comprehensive study of toll road traffic-estimation performance
EJTIR 11(1), January 2011, pp. 80-95 Welde and Odeck Do Planners Get it Right? The Accuracy of Travel Demand Forecasting in Norway
was conducted by the investment bank JP Morgan (1997), revealing that 13 of the 14 newly implemented US toll roads displayed traffic levels below forecasts. In four of the projects, the opening-year traffic was 30% below what was expected. The bank concluded that traffic-forecasting inaccuracy represents one of the major sources of risk in toll road projects. The credit-rating agency Standard & Poor’s (S&P) have performed risk studies of the traffic forecasts in toll projects since 2002 (Bain and Wilkins 2002; Bain and Plantagie 2003; Bain and Plantagie 2004; Bain and Polakovic 2005) and reported consistent findings. Their conclusions were that toll projects throughout the world suffer from extensive optimism bias and error. The performance of the projects studied ranged from actual traffic being only 15% of the forecasts to actual traffic exceeding forecasts by more than 50%. From the perspective of potential investors, these results are alarming. Even worse, it is likely that Standard and Poor’s sample, like the other samples of misleading forecasts, was biased because the toll facilities with a higher credit quality were over-represented. The worst cases of traffic underestimation were probably not included in the sample. As stated by Bain and Polakovic (2005, p. 68): ‘(…) very poorly performing assets will remain under-represented in the sample and the results derived from our case studies are likely to be flattered in comparison with average, global toll road forecasting performance’. Forecasts for complex road schemes with intricate traffic patterns are hence likely to be vague or non existent, making follow-up studies more difficult. The concept of demand ramp-up is often considered to be an argument against using the first whole year of operation as the basis for measuring the inaccuracy in forecasts because the demand for travel often depends on variables that might take years to spread through the system. It may thus take a few years before a new road reaches its full traffic potential. In the S&P 2005 study, however, Bain and Polakovic (2005) investigated the concept of demand ramp-up and found no such effect, as there was no systematic improvement in the traffic forecasting accuracy after Year 1. The underestimation of the traffic in Year 1 was likely to persist during Years 2-5, meaning that the forecasts did not become more accurate over time. Similar conclusions were reached by Fitch Ratings (George et al., 2003), who found the actual performance in US toll projects to be heavily skewed downward. However, unlike the other studies mentioned above, George et al. found clear evidence of ramp-up and that traffic tended to gravitate back towards and even exceed the original forecasts over time. Given these rather disappointing results, one might ask why the toll-financing share of total road financing annually increases if traffic revenues regularly fail to meet expectations in the first critical years of operation. A probable reason is that a high proportion of user-financed projects actually do meet expectations. Mauchan and Bates (2007), of the transport-planning consultants Steer Davies Gleave (SDG), studied 15 privately funded toll projects and found that the forecasts showed a distribution around the expected value, with no evidence of optimism bias. In fact, for the majority of the projects, the traffic was within 5% of forecasts, which in many ways is extraordinarily accurate. Their sample was small, however, and even included seven shadow-toll projects, making them, in effect, toll free projects, so the transferability of the results may be limited. Users do not pay at the point of use in shadow-toll projects, and including such projects in a toll road sample could be considered dubious. However, Bain (2009b) argued that shadow-toll projects share the same error characteristics as traditional toll projects because of the private financing mechanism. The SDG study showed, however, that no general conclusions can be drawn regarding the accuracy of traffic forecasts for toll roads and that research in the industry would benefit from a more case-specific approach, focusing on one country or region at a time. However, the general impression of tollproject forecasting accuracy is of overestimation. Further examples from the US, Spain and Australia (TRB, 2006; Vasallo, 2007; Li and Hensher, 2009 – cited in Bain, 2009b) have all suggested consistent over optimism and/or optimism bias of toll road traffic forecasts.
Studies of forecast accuracies in toll roads versus toll free roads have been rare to date. However, Bain (2009a) provided a comparison of toll and toll free roads based on the data from the S&P studies referred to above and the sample presented in Flyvbjerg et al. (2005). The comparison showed that the toll roads and toll free roads suffered from the same uncertainties. The forecast distributions for the two categories of roads were similar (the same error) but centred around different means; this is a sign of potential bias. The traffic on toll roads was found to be generally lower than the forecast, whereas the traffic on toll free roads was found to be higher than the forecast. The consequence of a similar distribution is that the observed bias can be corrected for, and the potential for error and economic losses can be reduced. Thus, there is no evidence to support the theory that forecast error is reduced when drivers are not required to pay tolls. NĂŚss et al. (2006) reached the same conclusions with similar forecasting accuracy in terms of the absolute error between the two classes of roads. The studies cited above show that while traffic on ordinary road projects often turns out to be higher than the forecast, toll road traffic is generally overestimated. What these studies have in common though, is that the data often have been collected from secondary sources and from different countries on different continents. Some observations even date back decades. Given that different countries inevitably have different planning traditions and tools and place different emphasis on forecasting accuracy, we argue that the conclusions reached should be interpreted with care. A data set from one country and one data source only would, in our opinion, yield much more reliable results due to the greater opportunity for quality control of the data. Accordingly, the focus of this study is the accuracy of travel-demand forecasts for Norwegian road projects.
3. Data and methodology 3.1 Data The data for this study consisted of observations from 25 toll projects and 25 toll free road projects in Norway. The data from toll projects are often more available and generally of better quality than for other road projects because all toll projects require a specific approval from the Norwegian parliament. The parliamentary bill in which the project is presented includes all financial assumptions, including the forecasts for the average annual daily traffic through the toll stations. The critical test for traffic forecast accuracy is thus how the actual traffic relates to what was presented to decision makers at the time of the decision to build. The Norwegian Public Roads Administration (NPRA) collects data annually on the traffic levels, costs and revenues in all toll projects throughout the country. The data set includes 12 fixed-link crossings (bridges and tunnels), 11 ordinary highway projects and 2 toll cordons. The tolling in the projects started in the years 1990 to 2007. In projects where the traffic patterns were difficult to forecast, the data are often unavailable. This is consistent with the sampling bias that was observed in the studies mentioned above. Although our sample consisted of relatively few observations, we still consider it to be representative of the population. During the years 1990 to 2007, 33 toll projects were implemented. Thus, our sample comprised 76% of the total projects in the analysed period. The criteria on which the sampling was based were data availability and quality. We acknowledge that using the projects where data was not available or of a sufficiently high quality for inclusion in the data set would increase the precision in the various property estimates of the population. However, due to the high sample/population ratio and the fact that the quality of the observations was considered to be very high, we still expected to be able to draw some valid conclusions regarding the accuracy of the forecasts in the Norwegian toll road industry.
An important distinction between toll projects and toll free projects is that paying traffic will always differ from (and be lower than) ordinary traffic because of various discounts and exceptions. However, given that the information on the fare and discount system was known before the start of the project, we still expected planners to be able to estimate their effects with a reasonable degree of accuracy. For the toll free projects, the data situation was somewhat more complicated. Although these projects are approved by parliament in the same way as the toll projects, less data are presented to the decision makers, and the quality of post opening data are generally less reliable. However, the parliamentary bill includes the net present value (NPV) estimated in the cost-benefit analysis (CBA) that relies heavily on the forecasted traffic levels. To find the original estimates, we thus had to consult the original CBAs in which the NPV estimates presented to the decision makers were found. The previous CBAs were not stored in a single database, and even when the estimates were present, they were often on an overall level, so access to the original detailed calculations was necessary. In the Norwegian case, impact assessment, including CBA, is carried out by use of the EFFEKT software. This was a rather demanding process and required collecting data from several sources. The next step was determining the actual traffic. The NPRA collects traffic data from 9,000 sites on all roads based on permanent and temporary monitoring. Among these, 600 sites are so-called Level 1 sites where the traffic is counted continuously. Unfortunately, no system is in place that requires traffic data to be automatically collected on new roads. This means that traffic data were not available for several new roads and could not be included in our data set. We were nevertheless able to find 25 toll free roads where both reliable estimates and actual traffic levels were available. The data were from the years 2001 to 2007 and consisted primarily of projects outside the major urban areas. We often found that less emphasis was placed on traffic forecasts for small road projects such as the straightening of curves. The sample thus consisted mainly of larger projects. 3.2 Methodology To estimate the accuracy of the traffic forecasts, we compared the actual traffic with forecasted values:
U (( X a X f ) 100) / X f where U is percent inaccuracy, X a is the actual traffic and X f is the forecasted traffic. With this estimation, perfect accuracy is indicated by zero, and for example, -20% would imply that the actual traffic was 20% lower than expected. For forecast values, we used the estimated traffic in the first calendar year of operation. This is normally presented in the parliamentary bill where the decision to approve the project is made. This means that if a project opens for traffic in August, the basis for comparison would be January to December the next year. In addition, we examined Years 3 and 5 to test whether any improvement in the forecast accuracy occurred over time. One might argue that focusing merely on the first year of operation does not allow for the long-run nature of many forecasting models. However, the principles of discounting suggest that the first years of operation are crucial for both financial and social viability. If a toll project with a pay-off period of 15 years fails to meet revenue expectations in the first five years, the risks of default increase considerably, even if the forecasts become more accurate in, for example, 8–10 years. It is probably unreasonable to expect planners to be able to predict values with perfect accuracy, especially for projects with complex traffic patterns. However, no acceptable level of forecasting accuracy is defined, and it must thus be regarded as an empirical matter. For construction costs, the
Ministry of Transport and Communications requires cost estimates to be in the range of Âą10%; with no specific requirement for demand-forecast accuracy, we used this as a benchmark and regarded the demand estimates that were within Âą10% of the actual traffic to be within an acceptable range.
4. Results The purpose of this study was to assess divergences between the forecasted and actual traffic for toll and toll free projects and to investigate whether there were differences in the forecast accuracies for the two types of projects. In this section, we present the results of our findings. 4.1 Forecast accuracy: Toll roads As with the international studies referred to above, we found the forecasted traffic on Norwegian toll projects to be higher than the actual traffic. However, with the actual traffic being 2.5% less than forecasted on average, the scale of overestimation was much less than that revealed in the studies in other parts of the world. Summary statistics for the forecast inaccuracies with the Norwegian toll projects are presented in Table 1. Table 1. Summary statistics for forecast inaccuracies for toll roads. Statistic Number of cases
Std. error of mean
These results appear to be encouraging. A mean of -2.5% was well within what we defined as an acceptable range. However, a closer look at the data revealed that a majority of the projects experience traffic overestimation, as in the international studies reported above. Additionally, 24% of the projects had over 20% less traffic than expected. Clearly, a traffic overestimation of up to 35% in the first whole year of operation can potentially have severe financial implications for the viability of a project. There is a significant risk that projects with traffic shortfalls of this magnitude could experience financial difficulties that necessitate loan refinancing, a prolonged payment period, increased tolls or a combination of alternatives. Luckily, the Norwegian economy has been blessed with the rare combination of high economic growth and low interest rates for some time. If this were to turn into a recession with increasing interest rates and demand shortfalls, as seen in the late 1980s and early 1990s, the risk of default would increase considerably. The standard deviation was 22%, indicating a rather large variation between the projects. Table 2 provides the distribution of projects by percentage inaccuracy.
Table 2. Distribution of projects by percentage inaccuracy. Number of projects
Projects with overestimation larger than -20%
Projects with overestimation 0 to -20%
Projects with underestimation 0 to +20%
Projects with underestimation larger than +20%
A histogram showing the distribution of these observations is provided in Figure 1. 6
5 4 3 2 1 0
20.0 0.0 Forecast accuracy
Figure 1. Inaccuracies of the toll road traffic forecasts. Figure 1 reveals a curve that is close to a normal distribution. A Kolmogorov-Smirnov test of normality (D(25) = 0.159, p > 0.15) confirmed that the forecasts for the traffic levels in the Norwegian toll projects were normally distributed around the mean or that the assumption of a normal distribution was not rejected. A t-test for deviation from zero revealed a test statistic of -0.57 and a significance value of 0.58, which meant that the mean forecast inaccuracy was not significantly different from zero and that we could not conclude that the underestimation was more common than overestimation for the Norwegian toll roads. From a credit perspective, it is worrying that the majority of the forecast errors for Norwegian toll projects were overestimations. However, a toll project can sometimes struggle to reach its full traffic potential in the first whole year after opening. This could potentially mean that the overestimation is more severe in Year 1 than in subsequent years and that the traffic better fits the forecasts as time
progresses. The international evidence of ramp-up has been inconclusive, as different studies have shown different results. However, as shown in Table 3, even though the number of observations decreased over time (N = 22 in Year 3 and N = 19 in Year 5), there were signs of ramp-up in the Norwegian toll projects. Table 3. Demand ramp-up. Year since opening
Although the traffic in Year 1 was overestimated, it increased over time, and after five years, the average traffic exceeded the original forecasts. Although the financial implications of forecast error in Year 1 might be severe, there is less need to worry if the traffic soon increases to or even exceeds the necessary levels. This is contrary to Bain’s (2009b, p. 37) claim that “…projects that under perform in their early years may never catch up with their original forecasts in later years”. In our sample, among the 13 projects with an overestimation greater than the sample mean, four exceeded their original forecasts, five exhibited traffic growth that may well soon put them in the above-forecast figures and four continued to under perform at Year 5. From a financial perspective, the failure to meet revenue predictions in the first five years of operation is, of course, potentially alarming, but our results nevertheless provide a more nuanced picture than that painted by Bain. 4.2 Forecast accuracy: Toll free roads For the toll free roads, for which we had 25 reliable observations from the last nine years, we noted that the traffic was, on average, higher than forecasted. The mean underestimation was 19.0%, but the range was large, from -14.6% to +76.1%. This was consistent with the pattern observed by Flyvbjerg et al. (2005) and Næss et al. (2006). Only six projects had traffic levels below the forecasts, and 13 projects exhibited traffic overestimation above the sample mean. In seven projects, the actual traffic was over 30% higher than predicted. This is clearly unacceptable. We would expect the forecast accuracy to be higher for toll free roads, but this was not the case for the Norwegian roads. The picture that emerged was that the traffic forecasts for the toll free Norwegian roads were skewed to the right. The summary statistics for the toll free roads are presented in Table 4. Table 4. Summary statistics for the forecast inaccuracies for toll free roads. Statistic Number of cases
Std. error of mean
Despite a high standard deviation, we found that the mean was significantly different from zero at the 99% level (t(24) = 4.64, p < 0.05). Thus, we concluded with a high level of certainty that the traffic on toll free Norwegian roads has been underestimated. The spread in the distribution was alarmingly high, which indicated a high level of general error. The shape of the distribution, as illustrated in Figure 2, indicated that the observations were normally distributed around the mean but with a slight positive skew (Kolmogorov-Smirnov: D(25) = 0.098, p > 0.20). As illustrated by the standard deviations for both classes of roads, the internal variations with both the toll and the toll free projects were huge. However, the difference in the means between the two categories of roads was -21.5, and a t-test of the difference revealed that the difference in the mean forecast accuracy between the two categories was highly significant and not a result of coincidence (t(48) = -3.58, p <0.01).
20.0 Forecast accuracy
Figure 2. Forecast inaccuracies for the toll free roads. 4.3 Do planners get it right? Our results suggest that the Norwegian transport planners should not be satisfied with the accuracies of their forecasts. On average, the planners do not get it right even if the toll road forecasts were, on average, within an acceptable range. Here, the results presented by Flyvbjerg (2005) and Bain (2009a) were confirmed because the toll road forecasts were more accurate than the forecasts for the toll free roads and because the countries with more toll road experience produced more accurate forecasts. However, the ranges that both the toll roads and toll free roads forecasts fell within were alarmingly high and should be a cause for concern.
The traffic on toll free roads was significantly higher than forecasted. This is worrying and would be doubly disturbing in a situation where the road capacity was limited and where the motivation for new road construction was to relieve congestion. But this is usually not the case in countries where the congestion is still limited to the peak-period traffic in the large cities. The higher traffic levels will often, as argued by Kjerkreit and Odeck (2009), lead to a higher NPV than originally estimated. However, this is obviously not a satisfactory situation in the long term. If the traffic is generally underestimated, then this could lead to inefficient resource allocation or the implementation of the wrong projects and a shorter relief period from congestion for the roads in urban and congested areas. Transport models for the Norwegian roads sector have improved since 2001, when regional transport models replaced a wide range of locally developed models, and Odeck et al. (2009) claimed that the forecasts have since improved. Today, the traffic on ordinary toll free roads is estimated through the use of national transport models for trips longer than 100 km and regional transport models for trips shorter than 100 km. Both models are traditional four-step models based on the fixed-trip matrix approach. Induced traffic is thus not taken into account. The traffic on toll roads, however, is estimated using elasticity models where the effects of tolls are calculated specifically. Hence, there are two different models used, one for toll roads and one for toll free roads. The use of the regional transport models has made it possible to identify the factors that lead to inaccurate forecasts. If the distributions of the forecasts between regions are similar, it is easier to isolate the cause of the error if the forecasts have been based on the same models than if different models have been used. Thus, it was interesting to test whether the assumption of increased accuracy after the introduction of regional transport models holds true using the data to which we had access. The toll road sample included 13 projects implemented in the years 1990 up to and including 2000 and 12 projects implemented in the years 2001–2007. For the toll free roads, we had 13 projects from 2001 to 2004 and 12 projects that opened for traffic in the years 2005–2007. Table 5 shows the differences in the mean accuracy for the two time periods for the two road categories. The number of observations is so low that caution should be taken when interpreting the results, but there appears to have been little or no improvement in the forecast accuracy over time either for the toll roads or for the toll free roads. However, the weaknesses in the Norwegian transport-demand models were first identified in the work leading up to the National Transport Plan for the years from 2002 to 2011, and because the planning process for roads often takes years, it has not been until very recently that we can expect to see real improvements caused by the improvements in the transport models. Table 5. Traffic-forecasting inaccuracies over time. Road category/opened for traffic
Toll roads 1990–2000
Toll roads 2001–2007
Toll free roads 2001–2004
Toll free roads 2005–2007
The concept of induced traffic is often used to explain traffic levels in excess of what was originally predicted. A study by Goodwin (1996) found the traffic in 151 UK highway schemes to be 10% higher on average than the forecasts in the short term and 20% higher in the long term. Thus, the forecasts for these projects were not able to fully include the extra traffic created by the network improvements leading to retiming, redistribution, mode shifting or change of frequency. Goodwin
also suggested that the addition of the capacity itself, regardless of the changes in the travel time, could help explain the increases in traffic flow. However, the changes in traffic brought by the improvements in the pleasantness of travel, such as a smoother ride from better surfaces, remains an under-researched area (Goodwin and Noland, 2003). Traditionally, transport planning has been based on the traffic levels and independent of the supply conditions and the quality of the road network. The growth demand has been largely attributed to economic factors such as income, population growth, the prices of petrol and other input factors. This is normally referred to as the fixed-trip matrix approach and is still in use in Norwegian road planning (except for the straight-crossing projects). Because the traffic on toll free roads is generally higher than forecasted, there are indications that this approach should be abandoned and that the induced traffic should be dealt with explicitly. Another potential explanation for higher traffic levels than estimated is the long period of economic growth that Norway has experienced over the last decade. Because transport is a derived demand, forecasting traffic relies on the forecasts of a range of other parameters (Boyce and Bright, 2003). Thus, if the income estimates in the transport models are underestimated, traffic may also be underestimated. The same pattern was observed during the 1990s, when the average national traffic growth over the years 1992â€“2002 was higher than all the forecasts that had been produced (Larsen et al., 2004). The recession early in the decade was followed by an economic boom that was accompanied by strong traffic growth. De Jong et al. (2003) distinguished between input uncertainty, or difficulties in producing good forecasts for transport model input variables, and model uncertainty. Because Norway has experienced unprecedented, strong economic growth over the last decade, there are clear indications that while more emphasis has been put on improving the transport models, the main causes of the observed error are input error rather than model error. This fits well with the observations of Larsen et al., who found the standard national traffic growth rate, which has been a mandatory input in the transport models, to be too low. Strategic behaviour and bias are often cited when no other explanations for forecast inaccuracy can be found. Wachs (1987, 1989) argued that because planners are concerned with having their projects financed and built, they deliberately produce overly optimistic forecasts for both capital costs and traffic. Because governments operate under budget constraints, the projects compete with each other for funding. Planners could thus be tempted to underestimate costs and overestimate benefits to meet a specific benefit-cost ratio (BCR) cut off. Although intuitively appealing, we do not necessarily agree that this serious allegation can be used as a general explanation for traffic-forecast inaccuracy. First, our results showed no evidence of such behaviour, which in itself was an indication that these forces are not occurring in Norway. Second, as the traffic on ordinary roads was underestimated, the benefits were also underestimated and not overestimated, all other things being equal (Kjerkreit and Odeck, 2009). In uncongested conditions, traffic in excess of what was forecasted will increase the overall benefits, and planners will thus have little to gain from underestimating the traffic, as this would mean presenting projects with a lower NPV than what they later turn out to produce. In fact, the funding for Norwegian road projects does not always rely on a positive BCR at all. Odeck (1996, 2010) studied whether Norwegian decision makersâ€™ ranking of road projects was explained and/or positively influenced by a positive BCR. Contrary to expectations, he found that the BCR was not a significant explanatory variable for the selection of projects and that more emphasis was placed on non monetised impacts. The projects with a positive BCR were sometimes not put on the priority list at all, whereas the projects with a negative BCR were sometimes given a very high ranking. Nilsson (1991) found similar results in Sweden. Although clearly unsatisfactory from a socioeconomic point
of view, placing less emphasis on socioeconomic profitability and monetised impacts could reduce the risk of optimism bias in producing the traffic forecasts. We do acknowledge, however, that the bidding process by which the toll road contracts are awarded may play a role in explaining the optimism bias. As argued by Flyvbjerg (2005), Vasallo (2007) and Bain (2009b), awarding toll road contracts based on a bidding process where the bidder with the highest revenue projections or the lowest capital cost projections wins could reward over optimism rather than accuracy. Here, the Norwegian framework for toll financing provides an alternative framework that might reduce the risk of overoptimistic forecasts. Norwegian toll projects are initiated locally, usually because much-needed road investments cannot be realised in the near future within the government budget. The proposal is then evaluated by the NPRA, who closely scrutinises all major assumptions before it is forwarded to the Ministry of Transport, which prepares a bill to be tabled in Parliament. The project might have a positive or a negative NPV, but all toll projects are stress-tested for financial robustness to ensure that the risk of financial default is low, even in worstcase scenarios. Once passed by Parliament, the operation of the toll road is managed by a non profit toll company operating as a financial vehicle on behalf of the NPRA, which remains the ultimately responsible party for the project (for a detailed presentation of the organisational framework of Norwegian tolling, see Welde and Odeck, 2009). Although the system is not without its flaws, there is less incentive for appraisal optimism than in alternative frameworks, and the system of quality control and the emphasis on conservative estimates has so far prevented any major financial scandals in the Norwegian toll road industry. The absence of any major scandals due to inaccurate traffic forecasts should not, however, lead us to conclude that this is not an area that warrants continuous attention. The huge variation in forecasting accuracies continues to be a major source of risk in the planning of Norwegian road projects. However, merely pointing out the problem will not make it disappear. The increasing range of international studies focusing on this issue has apparently not contributed to any major improvements in terms of forecasting accuracy. However, the knowledge generated from studies such as this will hopefully facilitate learning and lead to improvements in the forecasting methodologies. Furthermore, we strongly suggest that that the fixed-trip matrix approach (i.e., assuming a zero elasticity of demand) be abandoned for all road projects, as this is very likely a cause of the poor estimation of traffic levels and, ultimately, total economic benefits. In addition, because the process of project bidding or requiring projects to pass a certain BCR threshold to receive financing clearly increases the risk of deliberate over optimism, a system where NPV/BCR is only one input variable in the decision-making process should be considered. It would be interesting to see if countries that apply alternative appraisal frameworks, such as Multi-Criteria Analysis suffer from the same inaccuracies as countries where the decision makersâ€™ preferences are more determined by the outcome of the traditional cost-benefit analysis. Finally, the signs of good practices should act as encouragement and as an incentive for further research into why some projects are more successful than others.
5. Conclusions In this study, we examined the accuracy of traffic forecasts made in the Norwegian road sector. Two types of road projects were studied and compared: (1) toll roads and (2) toll free roads. This distinction was made because the consequences of inaccuracy with toll projects are considered to be more serious because they may lead to financial difficulties and the bankruptcy of toll companies.
We found that the traffic forecasts in the Norwegian toll projects have been fairly accurate. A likely explanation is that the planners over the years have been scrutinised and pressed to provide careful estimates for these projects. The results prove that the planners have been careful, but the number of projects where the actual traffic was significantly below what was forecasted suggest that the tollproject traffic forecasts should continue to be closely scrutinised. However, the variation in the forecast accuracies was high, and we suggest that the results from this study be used to identify the causes of why some projects have performed better than others. For the toll free roads, the results showed a clear underestimation of the traffic. Some of the projects studied had underestimations so great that there is reason to suspect that the projects may have experienced induced traffic for which the planners failed to account. This underestimation may have lead to inefficient resource allocation. Thus, the decision makers may have been misled into foregoing projects that were beneficial in favour of less beneficial ones. The results from the Norwegian road sector are slightly better than some of those presented in studies from other countries, especially for the toll roads, where the actual traffic was, on average, very close to the forecasted traffic. Even though the number of projects with traffic levels significantly different from the forecasts was high, the mean forecast accuracy and the relatively high share of projects with traffic levels close to perfect accuracy is a source of some encouragement. Our findings should be of interest to planners and policy makers in Norway and elsewhere. First, the planners need to reconsider their traffic-forecasting models, at least for the toll free projects, to ensure that all relevant factors are captured and forecast inaccuracy thus reduced. Second, the issue of induced traffic in particular must be considered. The Norwegian models for traffic forecasts do not consider induced traffic explicitly, and this may well be a reason why underestimation is prevalent. Third, with the high uncertainty revealed in such a crucial variable as the traffic level, presenting decision makers with single-point estimates for the NPV might potentially be misleading. This suggests that the presentation of social surplus through the NPV should be done through a confidence interval illustrating the inherent uncertainty in a project evaluation. Finally, the care taken when estimating traffic in toll projects demonstrates good practice but, even here, there is a potential for improvement.
Acknowledgements The authors wish to thank our research assistant Sigrid Johanne BĂ¸ckmann, a student at the Norwegian University of Science and Technology, for her role in collecting the data needed to complete this study. We also thank the three anonymous referees for providing valuable comments.
References Bain, R. (2009a). Error and optimism bias in toll road traffic forecasts. Transportation, 36 (5), pp. 469482. Bain, R. (2009b). Toll Road Traffic & Revenue Forecasts. An Interpreters Guide. Robert Bain: ISBN 978-09561527-1-8. Bain, R. and Wilkins, M. (2002). Traffic Risk in Start-Up Toll Facilities. Standard & Poorâ€™s, McGraw-Hill International (UK) Ltd.
Bain, R. and Plantagie, J.W. (2003). Traffic Forecasting Risk: Study Update 2003. Standard & Poor’s, McGraw-Hill International (UK) Ltd. Bain, R. and Plantagie, J.W. (2004). Traffic Forecasting Risk: Study Update 2004. Standard & Poor’s, McGraw-Hill International (UK) Ltd. Bain, R. and Polakovic, L. (2005). Traffic Forecasting Risk Study Update 2005: Through Ramp-Up And Beyond. Standard & Poor’s, McGraw-Hill International (UK) Ltd. BBC (2008). M6 Toll road marks fifth birthday [online]. Available at: http://news.bbc.co.uk/ [accessed on January 15, 2010]. Boyce, A.M. and Bright, M.J. (2003).Uncertainty in traffic forecasts: Literature review and new results for the Netherlands. Proceedings from the European Transport Conference. Strasbourg, France, 8-10 October, 2003. de Jong, G., Daly, A., Pieters, M., Miller, S., Plasmeijer, R. and Hofman, F. (2003). Reducing or managing the forecasting risk in privately funded projects. Proceedings from the European Transport Conference. Strasbourg, France, 8-10 October, 2003. Flyvbjerg, B. (2005). Measuring inaccuracy in travel demand forecasting: methodological considerations regarding ramp up and sampling. Transportation Research Part A, 39 (6), pp. 522-530. Flyvbjerg, B. (2007). Policy and planning for large-infrastructure projects: problems, causes, cures. Environment and Planning B: Planning and Design, 34 (4), pp. 578-597. Flyvbjerg, B., Skamris Holm, M., Buhl, S.L. (2005). How (In)accurate Are Demand Forecasts in Public Works Projects: The Case of Transportation. Journal of the American Planning Association, 71 (2), pp. 131-146. Flyvbjerg, B., Skamris Holm, M., Buhl, S.L. (2006). Inaccuracy in Traffic Forecasts. Transport Reviews, 26 (1), pp. 1-24. Forster, A. (2006). Does transport planning have a ‘dark side’? Local Transport Today, 438, p. 9. George, C., Streeter, W. and Trommer, S. (2003). Bliss, Heartburn and Toll Road Forecasts. Fitch Ratings Project Finance Special Report. Goodwin, P. (1996). Empirical evidence on induced traffic. Transportation, 23 (1), pp. 35-54. Goodwin, P. and Noland, R.B. (2003). Building new roads really create extra traffic: a response to Prakash et al. Applied Economics, 35 (13), pp. 1451-1457. Joosten, R (1999). M1/M15 – How a successful project ended unnecessarily in tears. Private Finance International, 179, pp. 52-55. Kjerkreit, A. and Odeck, J. (2009). The accuracy of ex-ante benefit cost analysis – a post opening evaluation in the case of Norwegian road projects. International Transport Economics Conference (ITrEC).University of Minnesota, Minneapolis Minnesota, June 15-16, 2009. Larsen, O.I., Rekdal, J., Johansen, K., Berge, G., Kleven, O.A., Alfheim, L. and Odeck, J. (2004). Prognoser i Statens vegvesen [In English: Forecasts in the Norwegian Public Roads Administration]. Norwegian Public Roads Administration, Report no. 5, 2004. Mauchan, A. and Bates, P. (2007). How accurate are toll road forecasts? Steer Davies Gleave’s news & reviews 32, September 2007, p. 18. Morgan, J.P. (1997). Examining toll road feasibility studies. Municipal Finance Journal, 18 (1), pp. 1-12.
Nilsson, J.-E. (1991). Investment Decisions in a Public Bureaucracy: A Case Study of Swedish Road Planning Practices. Journal of Transport Economics and Policy, 25 (2), pp. 163–175. Næss, P., Flyvbjerg, B. and Buhl. S. (2006). Do Road Planners Produce More ‘Honest Numbers’ than Rail Planners? An Analysis of Accuracy in Road-traffic Forecasts in Cities versus Peripheral Regions. Transport Reviews, 26 (5), pp. 537-555. Odeck, J. (1996). Ranking of regional road investments in Norway. Does socio economic analysis matter? Transportation, 23 (2), pp. 123-140. Odeck, J. (2004). Cost overrun in road construction – what are their sizes and determinants? Transport Policy, 11 (1), pp. 43-53. Odeck, J. (2010). What Determines Decision-Makers' Preferences for Road Investments? Evidence from the Norwegian Road Sector. Transport Reviews, 30 (4), pp. 473-494. Odeck, J., Kleven, O.A. and Johansen, K. (2009). Prognoser treffer bedre nå enn før. [In English: Forecasts more accurate than before]. Samferdsel, 9, pp. 10-11. Osland, O. and Strand, A. (2010). The Politics and Institutions of Project Approval – a CriticalConstructive Comment on the Theory of Strategic Misrepresentation. European Journal of Transport Infrastructure Research, 10 (1), pp. 77-88. Vasallo, J.M. (2007). Why traffic forecasts in PPP contracts are often overestimated? EIB University Research Sponsorship Programme. Luxemburg: European Investment Bank. Wachs, M. (1987). Forecasts in urban transportation planning: Uses, methods, and dilemmas. Climatic Change, 11 (1-2), pp. 61-80. Wachs, M. (1989). When planners lie with numbers. Journal of the American Planning Association, 55 (4), pp. 476-479. Welde, M. and Odeck, J. (2009). The efficiency of Norwegian road toll companies. 8th Conference on Applied Infrastructure Research. Berlin Germany, October 9-10, 2009.
Issue 11(1) January 2011 pp. 96-97 ISSN: 1567-7141 www.ejtir.tbm.tudelft.nl
Book review: Anders af Wåhlberg: Driver Behaviour and Accident Research Methodology; Unresolved Problems Karel Brookhuis1 Department of Clinical and Neuropsychology, Faculty of Behavioural and Social Sciences, University of Groningen; Section of Transport and Logistics, Faculty of Technology, Policy and Management, Delft University of Technology
Anders af Wåhlberg has apparently written this book starting from a negative attitude. He has his own ideas about what is acceptable in traffic psychology research and what is not. Throughout the book he sets his face against the established order such as the rather standard way of using accident statistics, against self-reports, well-known and widely accepted relationships such as exposure and accidents, against traffic conflict studies, and even experimental studies in simulators and instrumented vehicles. Accident statistics nowadays, based on extensive (mostly police) accident reports, have evolved from short texted information in pretty uninformative headings on difficult forms, to balanced digital information sources that turn out to be of unmatched value (e.g. Evans, 2003). afWåhlberg, however, grossly discards that value and seeks to set out for culpability issues as the major source of information concerning the basics of accidents. In line with this thinking is his defence of accidentproneness, revived from the past. To my idea it would be more fruitful to integrate personal characteristics and inclinations such as sensation seeking within the field of personality psychology, in stead of bringing up high correlations in high-number accidents statistics which always produce significant effects. His arguments for accident proneness, with the aid of linear regression techniques, are not very convincing, and rather odd-looking. Even more puzzling is his attack on the so-called low-mileage bias. It has been reported a number of times that accident rate per distance covered is related to driving experience. In other words, drivers that drive a lot tend to have a lower accident rate than drivers that do not drive much. afWåhlberg argues pretty cynically against this proposition with the argument that “… the usually over-learned and automatic skill of driving must deteriorate rather fast if the amount of driving is reduced …”. However, while this argument would apply to the very basic skills, which indeed do not deteriorate rather fast since they are over-learned by nature such as handling the wheel and pedals, it does not apply to other skills. The less frequent events, situations and matching actions are not so quickly internalized and automated because it needs repetitive occurrence within a reasonable time frame before controlled processing shifts to automatic processing (Shiffrin and Schneider, 1977), or, before rule-based actions turn into skill-based actions (Rasmussen, 1985). The less frequent the event or situation the more likely a high-mileage driver would profit in that sense. 1
Grote Kruisstraat 2/1, 9712 TS Groningen, The Netherlands, T: +31503636772, E: email@example.com
EJTIR 11(1), January 2011, pp. 96-97 Brookhuis Book review: Anders af W책hlberg: Driver Behaviour and Accident Research Methodology; Unresolved Problems
I found the book rather difficult to read, or perhaps I should say rather difficult to finish. His lengthy defense of using the outcome of accidents in terms of who-dun-it, supported by rather complicated (meta)analyses, is not very exciting unless you like an in-depth treatise about methodology. The sometimes utterly negative treatment of what other (well respected) traffic psychologists consider useful techniques and measures is partly responsible for that. He is rather bitter about the establishment that has not accepted some of his submitted manuscripts that offend rather than reason closely. I had the honour of reviewing one of his, admittedly thoroughly written, manuscripts. The rather hostile lecture against the DBQ (Driver Behaviour Questionnaire) overshadowed his arguments. Naturally we should be receptive to arguments against weak methodologies, some of us are perhaps a bit too much inclined to be sloppily applying methods and analyses that are not warranted given the questions raised or objectives pursued. The most disappointing aspect of the book is that it justly discusses the validity of a wide variety of premises and assumptions that are usually adopted in driving behaviour studies but offers no decent alternative to study safety of driver behaviour. It can not be that all the attacked research efforts (in the western countries) are totally unrelated to the grossly improved traffic safety in the past 40 years. Since research efforts in traffic and transport were substantially increased starting in the seventies, among others, traffic safety enhanced tremendously. For example, in the Netherlands the number people killed in traffic dropped from more than 3000 to less than 1000 in 40 years time. Perhaps a more distant view would help afW책hlberg to climb out of the details and numbers that he seems to be entangled in.
Reference Evans, L. (2003). A New Traffic Safety Vision for the United States. American Journal of Public Health, 93, 1384 -1386. Rasmussen, J. (1985). The role of hierarchical knowledge representation in decisionmaking and system management. IEEE Transactions on Systems, Man and Cybernetics, SMC-15, 234-243. Shiffrin, R.M. and Schneider, W. (1977). Controlled and automatic information processing: II. Perceptual learning, automatic attending and a general theory. Psychological review, 84, 127-190.
Editorial Board EJTIR Bert van Wee (Editor in Chief) Delft University of Technology Caspar Chorus (Co-Editor in Chief) Delft University of Technology Vincent Marchau (Secretary) Delft University of Technology Serge Hoogendoorn Delft University of Technology Dominic Stead Delft University of Technology L贸ri Tavasszy Delft University of Technology Rob van der Heijden Radboud University Nijmegen Eddy van de Voorde University of Antwerp Frank Witlox Ghent University
Advisory Board EJTIR Peter Nijkamp (Chairman) The Netherlands Stef Weijers The Netherlands Erik Verhoef The Netherlands Jos van Ommeren The Netherlands Daniel Schefer Israel Frank Sanders The Netherlands Piet Rietveld The Netherlands
Aura Regiani Italy Robert Paaswell United States of America Ricco Maggi Switzerland Martin Lanzendorf The Netherlands Veli Himanen Finland Ingo Hansen The Netherlands Marina van Geenhuizen The Netherlands Piet Bovy The Netherlands Michel Beuthe Belgium David Banister United Kingdom
EJTIR Volume 11, Issue 1 Table of Contents Editorial board Why Do Open Rail Freight Markets Fail to Attract Competition? Analysis on Finnish Transport Policy Pages 1-19 Miika M채kitalo Barriers to Entry in Rail Passenger Services: Empirical Evidence for Tendering Procedures in Germany Pages 20-41 Arne Beck Measuring Spatial Separation Processes through the Minimum Commute: the Case of Flanders Pages 42-60 Kobe Boussauw, Ben Derudder and Frank Witlox Conflict between Independent Scrutinisers of Transport Megaprojects: Evidence from Australia Pages 61-79 Stuart Kells Do Planners Get it Right? The Accuracy of Travel Demand Forecasting in Norway Pages 80-95 Morten Welde and James Odeck Book review: Anders af W책hlberg: Driver Behaviour and Accident Research Methodology; Unresolved Problems Pages 96-97 Karel Brookhuis