The use of surveys for market definition: the CC experience Brasilia, may 2010
Introduction There does not appear to be much discussion in the literature of using customer surveys to assess the economic effects of mergers and cartels on competition. Yet, many competition authorities as the European Commission, the U.K. Competition Commission and the F.T.C. all make an extensive use of customer surveys (a statistic that I can tell you is that 14 out of 31 merger inquiries by the CC between 2004-2006 made use of a survey) 2
Available information Time and resource constraints limit the ability to collect large quantities of information that are specific to the case. There are four broad sources of such information: • official government statistics • industry sources • the merging parties and other firms in the relevant market, and customers and suppliers. • original data collected as a survey. 3
Available information (2) Usually, most official government statistics are collected at a level that is too aggregated for antitrust control purposes. This may also be the case for industry sources â€” e.g. commercial organizations that collect data on specific sectors to sell to firms wanting to benchmark â€” although the sectors for which such data are collected may be suggestive of how participants perceive the market. The case parties and their customers and competitors also routinely provide data. 4
Additional info from Surveys The regulator often may commission a customer survey, especially where there are a large number of customers in the market, the views of whom may be otherwise hard to recover. Typically, customer surveys are postal or selfcompletion questionnaires, telephone interviews, or face-to-face interviews. Each mode of data collection has its pros and cons, in terms of costs, time, coverage. (e.g. face-toface surveys tend to have higher response rates than telephone and self-completion surveys, but tend to take longer and cost more.)
Surveys To help ensure that surveys elicit meaningful answers to less-than-perfectly-tuned questions, they need to be structured in order to have respondents relive their purchasing decisions. This is done by asking questions in stages: • matters of fact • matters of behavior • matters of choice • matters of attitude. 6
Matters of Fact Survey questions first address simple factual points and the context in which the decision was made (for example, purchasing motives). For example, in its 2004 Archant/INM local newspaper merger inquiry, the U.K. Competition Commission (CC) asked advertisers â€œscreeningâ€? questions about whether they advertised in various print and non-print media. Similarly, in its 2005 Somerfield/Morrisons case, the CC asked shoppers what kind of shopping trip they were on (e.g. one-stop shop, top-up shop), and whether they normally shopped on this day and at this store. 7
Matters of Behaviour Survey questions next ask which alternatives were considered and which were thought to be most effective. For example, in the mentioned â€˜Archant/INMâ€™ case, the CC asked advertisers how important local newspaper advertising was to them and what attracted them to local newspapers, relative to the alternatives. Likewise, in Somerfield/Morrisons, the CC asked shoppers what attracted them to the store, relative to the alternatives. 8
Matters of Choice Survey questions ask what factors (e.g. price) led to one choice among the alternatives identified as matters of behavior. In Archant/INM, the CC asked advertisers which other newspapers, print, and non-print media they could have used instead. In Somerfield/Morrisons, the CC asked shoppers how much they had spent, how far they had travelled, and whether they always used the store for their shopping. 9
Matters of Attitude At last, survey questions ask what the respondent would have done under different circumstances, for example, what would the respondent do if the price rose by 5 percent? For example, in Archant/INM, the CC asked advertisers whether prices had increased (or nonprice factors worsened) and what advertisers would do if they did. In Somerfield/Morrisons, the CC asked shoppers what they would have done had the store been closed and how much worse this choice would have been.
Uses for customer surveys The most common uses for customer surveys are to assist with market definition and competitive assessment because customersâ€™ attitudes and behavior are the most relevant to them. This presentation will be especially dedicated to the use of surveys for market definition.
Market Definition a case study: Archant/INM (2004) In 2004 the U.K. Competition Commission scrutinized Archantâ€™s acquisition of Independent News and Mediaâ€™s (INM) 27 weekly local newspapers in the Greater London area. In this case surveys were used to establish the geographic and product markets The surveys provide also the occasion to consider some caveats and potential problems in their use.
Geographic market • INM’s 27 newspapers were circulated/distributed widely in Kent (to the southeast of London), east London and Essex (to the East of London), and also were circulated/distributed in north and northwest London • Archant’s (the acquirer) newspapers also were circulated/distributed widely in east London and Essex, but were circulated/distributed in north and northwest London only to a very limited extent, and were not circulated/distributed in Kent at all.
Geographic market – one scenario • Were the relevant market for advertisers in Archant/INM just weekly local newspapers in each local area (e.g. a postcode area), then there would have been many significant overlaps between the titles of Archant and INM in East London and Essex, at the very least. • Given that there were at most three publishers of weekly local newspapers in most of Greater London, these overlaps would have generated large increments in already concentrated markets.
Geographic market – alternative scenario • On the other hand, were the market for weekly local newspapers Greater London-wide, then the overlaps would not have been large enough to cause the CC concern. • This would have been the case even more so had the product market been wider than only weekly local newspapers (for example, if it included advertisingonly publications, London-wide daily newspapers, or other non-print media). 15
Geographic market (cont’d) In previous CC local newspaper merger inquiries, the narrowest candidate geographic markets considered have been the local contiguous postcode areas, usually known as circulation ‘footprints’. The CC begins the analysis of this candidate geographic market by making use of responses to a quantitative survey.
The survey The survey asked to Archant customers: • which Archant or former INM title(s) they had advertised in during the last 12 months and which was the most important. • which newspaper they would have turned to for their advertising needs had their main (or only) Archant/INM title not been available in 2003.
The survey (2) â€˘â€ˆ
Respondents were asked about four alternatives : (a) whether they would have advertised in another title in the same area; (b) whether they would have advertised in another title in a different area; (c) whether they would have used a different advertising medium (ie not local newspapers, for example radio, television, leaflets, outdoor etc); and (d) whether they would not have advertised at all. 18
The survey (3) The following tables reports the results of the survey.
The results There were 68 customers who most frequently advertised in east London, comprising 54 whose most important newspaper was an Archant title and 14 an ex-INM title. If the 54 Archant users could not use their most important title, • 12 would switch to another Archant title •
8 would switch to an INM title in the same area 20
The results (2) • 12 would switch to another publisher’s title in the same area • 9 said that they would switch within the area but did not know or could not recall the name of the title they would use. Thus, 41 of the 54 would switch to another local news-paper within the area. The 13 who would not, consisted of 5 who would use a title in a different area and 8 who would divert their budget into a medium other than local newspapers (6) or not advertise at all (2). 21
The results (3) The patterns of switching behaviour shown in the Table appear to show that the majority of Archant/ former-INM customers would use another local newspaper title in the same area of London were they unable to advertise in their current main/only Archant/ former-INM local newspaper. The inference that we draw from this is that the relevant geographic market could be local. Clearly Archant reacted saying there was ample evidence that the relevant geographic market was wider than the circulation footprint but said that the evidence did not indicate how wide it was. 22
The product market In this case there are two aspects to product market definition, which one can also use when considering competition within the market: • whether there are separate, relevant product markets for different categories of advertising; and • whether other printed media (for example, advertising-only publications, national and regional newspapers) and non-printed media (radio, Internet, etc) belong in the same relevant product market as local, free and paid-for, weekly newspapers. 23
The survey for the product market The quantitative survey asked respondents what level of increase (in %) in the cost of advertising (to them) in local newspapers, if any, would stop them advertising in local newspapers. Of 579 respondents, 355 answered the question: responses ranged from 1% (29 responses, i.e. 8.1%) to 100% (3 responses, i.e. 0.9%). The average (mean) reported critical price increase was 17% but the range of responses was skewed towards smaller critical price increases (the median critical price increase was 10%). 24
Some complications The following four sections discuss some complications in doing the assessment of the answers to the survey: • Distinguishing between residual and market elasticities of demand • Measuring total rather than partial substitution • Ensuring that respondents understand a SSNIP • The proportions of marginal and infra-marginal customers in the sample 25
Residual vs market elasticities of demand The CC was careful to distinguish between the elasticity of residual demand facing Archant or INM and the elasticity of market demand for all local weekly newspapers. That is, the CC asked about critical price increases across all local newspapers, not just for the respondentâ€™s relevant Archant or INM title(s). It is the elasticity of market demand that is relevant for market definition. 26
Total vs Partial substitution The CC asked whether the respondent would stop advertising after a critical price increase, not whether the respondent would reduce advertising, that is, the CC asked about total substitution, not partial substitution.
Total vs Partial substitution (2) Asking about total substitution has two effects on the SSNIP test: 1.â€ˆ customers who stop advertising may not be representative in expenditure terms, so losing 23 percent of customers may not imply that 23 percent of advertising revenues would be lost in response to a SSNIP â€” in Archant/ INM, the CC concluded that there was no reason to suppose that this was an issue because responses did not differ according to how much respondents spent on advertising; 28
Total vs Partial substitution (3) 1.â€ˆ customers who do not stop advertising in response to a SSNIP are assumed not to reduce their expenditure at all â€” the CC noted that the effect of this is to underestimate the revenue that would be lost to a SSNIP.
Total vs Partial substitution (4) The cumulative distribution of respondentsâ€™ critical price increases shows that 23 percent of customers said that they would stop advertising in local newspapers for a 5 percent price increase (a SSNIP). Using estimates of the costs saved from such a loss of advertising custom, the CC concluded that a SSNIP would not be profitable and that the relevant market was wider than only local newspapers in a given local area. 30
Ensuring that respondents understand a SSNIP Respondents were mostly business customers, so there were no problems in implementing the SSNIP test associated with the received market-research wisdom that â€œconsumers do not understand percentages.â€? In other contexts, the CC might avoid this problem by asking about actual price increases. However, with business customers, a different problem may arise in that the respondent of the survey may not be the decision maker, budget holder, and user of the product, and so answers may differ depending on who is the respondent within a business. 31
The proportions of marginal vs infra-marginal customers It is the behaviour of customers who are marginal to all firms in the market that is relevant for market definition. However, the CC and the other authorities frequently constructs the samples for its surveys from the customer lists of the merging parties. The proportion of marginal and infra-marginal customers that can be sampled from these customer lists may not be the same as the proportions in the population of customers. 32
The proportions of marginal vs infra-marginal customers (2) In Archant/INM, the CC supplemented 420 respondents who were customers of Archant or INM with 159 who were not, partly for this reason. In general, however, the CC does not sample socalled non-customers because this can introduce noise into a survey and cause problems with the integrity of the sample. Where the number of competitors in a market is large, there may also be problems in identifying noncustomers 33
www.planejamento.gov.br/gestao/dialogos firstname.lastname@example.org Departamento de Cooperação Internacional Secretaria de Gestão – SEGES
Ministério do Planejamento, Orçamento e Gestão Esplanada, Bloco K, 4° andar (61) 2020- 4906