The use of customer surveys for market definition and the competitive assessment of horizontal mergers

Journal of Competition Law and Economics

Volumn 4, Issue 2, 2008, Pages 411-431

  Reynolds, Graeme ^a   Walters, Chris ^a  

^a United Kingdom Competition and Markets Authority   (United Kingdom)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 50349090662     PISSN: 17446414     EISSN: 17446422     Source Type: Journal    
DOI: 10.1093/joclec/nhm036     Document Type: Article

References (45)

1. As well as merger inquiries, the CC also conducts other inquiries (for example, market investigations, regulatory appeals). The CC almost always uses customer surveys (generally both qualitative and quantitative) for market investigations, for which it also faces the issues that we discuss in this paper. Recent examples include Home Credit (2006), Store Cards (2006), Liquefied Petroleum Gas (2006), and Classified Directory Advertising Services (2006).
2. Principally but not exclusively, as the Secretary of State for Trade and Industry may refer mergers raising public interest issues.
3. Notification of mergers is not mandatory in the U.K. The period that the OFT has to review a merger essentially depends on whether it has been formally notified or not.
4. The annualized referral rate of mergers was around 17 percent for 2005/ 2006. That is, roughly one in six mergers that the OFT examined were referred to the CC. The OFT examines around 200 mergers a year.
5. The CC's inquiry into Deutsche Börse AG, Euronext NV, and London Stock Exchange PLC was two cases in one. The CC extended this case by eight weeks. The average, excluding this case, was 22 weeks.
6. The process of agreeing and implementing remedies with the parties has a separate timetable, as the last box in Figure 1 shows, and took 19 weeks on average for the four SLC cases in 2005/2006. The remedies timetable in the CC's Somerfield/Morrisons inquiry was extended because Somerfield applied to the Competition Appeal Tribunal (CAT) for an order quashing those parts of the CC's report that dealt with divestment. In a judgment given on 13 February 2006, the CAT dismissed the application. The average, excluding this case, was 15 weeks.
7. The CC may undertake further analysis in response to comments on its PFs but the bulk of its analysis will have been done before PFs.
8. For example, when further evidence was presented, the CC reversed its adverse finding of an SLC between PFs and the final report in its 2005 British Salt/New Cheshire Salt and Somerfield/Morrisons merger inquiries (in the latter case, the CC found an SLC in 14 local markets in its PFs but in only 12 local markets in its final report).
9. The CC sends the merging parties a "market questionnaire" for this purpose. The parties also sometimes submit the results of their own customer surveys, either existing market research work or (less often) those specifically commissioned for the CC's merger investigation.
10. For example, in its 2006 Heinz/HP and Cott/Macaw merger inquiries, the CC requested sales data from supermarkets that would not otherwise have been available from the merging parties.
11. As a public body, the CC submits invitations to tender to a predetermined list of market research organizations, selected every four years by advertising in the Official Journal of the European Union On occasion, the CC can and does commission market research from suppliers not on its predetermined list, should the need arise (for example, if they have specialist knowledge of surveying customers relevant to the CC's merger inquiry).
12. The CC uses such large-scale quantitative surveys more frequently than small-scale qualitative surveys, which tend to be focus groups or in-depth interviews. When qualitative surveys are used, they tend to precede quantitative surveys and allow the CC to understand customers and phrase and target its quantitative surveys more appropriately.
13. The response rate is the proportion of customers surveyed who answer the survey questions.
14. The CC does not disclose its pilot results but does give the merging parties the opportunity to comment on any changes that the CC makes to its survey in response to them. The CC pilots its surveys by pre-testing both the questions and the mode of data collection.
15. These stages are similar to those that Herbert Simon identified, who won the 1978 Nobel Prize in Economic Sciences for his pioneering research into the decision-making process within economic organizations. Simon identified five stages, with an implementation stage between the "matters of choice" and "matters of attitude" stages, for example, see Herbert Simon, The New Science of Management Decisions (New Jersey: Prentice Hall, 1977), but this stage does not usually concern the CC, as it is generally self-evident how customers made their purchases.
16. That is, telling the interviewer what the respondent believes to be the "right" answer, instead of what the respondent really believes.
17. Where it uses the HHI, the CC has regard to the threshold levels that the OFTused but only as one factor in its wider assessment of competition. The OFT regards an HHI of 1000 as high and an HHI of 1800 as very high. The OFT regards a merger as indicating possible cause for concern when there is an increment of 100 to a high HHI and of 50 to a very high HHI.
18. Broadly speaking, remedial action can restore the status quo ante market structure (for example, divestment), can increase the competition faced by the merged firm (for example, access to an essential input), or can prevent the abuse of market power (for example, a price cap).
19. For a discussion of these issues in the context of some recent CC merger cases, see Matt Hughes and Noel Beale, "Customer Surveys in UK Merger Control - the Art and Science of Asking the Right People the Right Questions," European Competition Law Review, 5 (2005), 297-303.
20. The CC's Heinz/HP (2006) merger inquiry is the most recent exception. Here, however, the parties implemented the SSNIP test, not the CC.
21. For a discussion of issues in estimating demand systems, see Daniel Hosken, Daniel O'Brien, David Scheffman, and Michael Vita, "Demand System Estimation and its Application to Horizontal Merger Analysis", in Jonathan Harkrider (ed.) Use of Econometric Analysis in Antitrust Investigation before the DoJ and FTC (American Bar Association Section on Antitrust, 2005).
22. Where a local newspaper is free to readers, it is said to be distributed. Where it is paid for, it is said to be circulated. Most of INM's 27 titles were free. Even for those that were paid for, advertising revenue dwarfed cover-price revenue.
23. In the end, the picture was mixed, given that the CC concluded that the geographic market was supralocal but not as wide as Greater London.
24. The price elasticity of residual demand will in general be much more elastic than the price elasticity of market demand.
25. This is not to say that more open or closed SSNIP-type questions may be inappropriate in other contexts.
26. Not all 579 respondents gave useable answers to this question. In particular, if respondents answered "nothing" to the question, the CC could not be sure whether they meant that a 0 percent price increase would make them switch (that is, they were planning to switch in any event), or whether they meant that no price increase would make them switch. However, even under the assumption that advertisers who answered "nothing" would not have switched for any price increase, 15 percent of customers would still be lost to a 5 percent price increase. The CC also investigated whether a strategy of price-discriminating price rises, targeting increases to avoid the most price-sensitive advertisers, would alter the results. This was important because of the practice of negotiating rates, but it did not reverse the initial conclusions.
27. A very small number of respondents were individuals advertising in categories like child care, private tuition, and health and beauty. The vast majority of respondents were businesses advertising in the property, motor, recruitment, and retail categories.
28. The chairman of the U.K.'s financial services regulator the Financial Services Authority suggested in a 2005 speech that one in five adults in Britain does not understand percentages. Although he was talking about percentage interest- and growth-rates for financial products, the point seems more general.
29. That is, by asking how much respondents paid, increasing it by 5 percent, and asking what they would have done had the price been higher.
30. Essentially, the CC sees four advantages in defining the research population from which to sample only from the merging parties' customers: (i) respondents are likely to be better informed about the issues that are relevant to the CC; (ii) respondents are likely to be affected by the merger; (iii) the CC can accurately assess whether the sample is representative; and (iv) the CC can accurately stratify the sample if necessary (for example, because the CC can ask the merging parties for revenue data).
31. Formally, the diversion ratio from A to B is the ratio of the cross-price elasticity of demand of A to B divided by the own-price elasticity of demand for A. For a formal derivation, see Chapter 10 of Simon Bishop and Mike Walker, The Economics of EC Competition Law: Concepts, Application and Measurement, 2nd edn (London: Sweet & Maxwell, 2002).
32. See, for example, the CC's recent Firstgroup/Scotrail (rail transport), National Express/Greater Anglia (bus and rail transport), and National Express/Thameslink (rail transport) merger inquiries.
33. The CC previously had identified these 56 as potentially problematic essentially by measuring concentration in plausible candidate local markets for each.
34. There are ways to measure diversion ratios other than through a customer survey. See, for example, Simon Baker, Andrea Coscelli, and Theon van Dijk "Unilateral Effects in Retail Chain Mergers", European Competition Law Review, 23:4 (2002), 180-92.
35. Morrisons had previously acquired these stores from Safeway and most of them had not been rebranded to Morrisons before their subsequent acquisition by Somerfield and rebranding to the Somerfield fascia.
36. The CC used confidential revenue-weighted diversion ratios as part of its analysis and not the nonconfidential customer diversion ratios given in Figures 4 and 5. This sentence refers to the revenue-weighted diversion ratios.
37. This is particularly relevant for two of the 12 SLC stores, where the nearest acquiring Somerfield store had shut (leaving just the acquired Morrisons) by the time of the survey. The CC estimated what the diversion ratio may have been had the Somerfield remained open and used this in its deliberations instead (see the Annex to Appendix D to the report).
38. The CC also surveyed respondents as they left the former Morrisons supermarket and not when they were at home choosing where to shop. This could have led respondents to prefer nearby stores but would not have systematically affected the diversion ratios between the merging parties unless Somerfield was consistently nearer to the former Morrisons supermarkets than were the other fascias in the right panel of Figure 4. This was not the case.
39. Specifically, that CC thought that it had some relevance but only in assessing the threshold level at which diversion ratios were high enough to give concern.
40. The CC did not know the identity of every possible, relevant nearby Somerfield store because verification of the isochrone analysis of candidate local markets that the parties submitted had not been completed at that stage.
41. Somerfield recognized that the simplicity of the question carried many advantages but called for care in making inference from it.
42. In contrast, for example, to the price elasticity of demand: A marginal customer's price elasticity of demand will exceed (in absolute terms) the average by definition.
43. For example, if the true proportion of marginal customers at a surveyed store was 10 percent - and if the CC maintained that a sample of 100 marginal customers per store was necessary for its survey results to be robust - then the CC would have had to survey 10 times as many shoppers as the 5444 that it actually surveyed, that is, over 54,000. This would not have been practical.
44. The average increase in advertising rates reported by respondents was around 10 percent.
45. There are (at least) three reasons to be cautious about asking survey respondents about previous price increases. First, data on actual price increases and switching may be collected from the merging parties more effectively rather than through a survey. Secondly, what is being surveyed is respondents' price sensitivity and not their ability to recall historical price changes and their reactions to them. And thirdly, there is no reason necessarily to expect respondents' previous reactions to an increase in a lower price to be the same as their future reactions to an increase in the higher current price.