Defining Markets for Ex Ante Regulation using the Hypothetical Monopoly Test

Abstract

The EU is gradually systematizing its approach to ex ante regulation (notably with the publication of the recent regulatory framework for Electronic communications). Market definition is viewed as an essential first step in this, with the ‘Hypothetical Monopoly Test’ a useful conceptual tool for use in the assessment. This paper examines the use of this test for market definition when there are several differentiated goods or services under consideration. It sets out an analytic framework, discusses the pros and cons of using the test and illustrates its use in a Telecom context.

Key Words:

• Market Definition
• Hypothetical Monopoly Test
• SSNIP Test
• JEL Classifications: L41, L42, L5

Notes

1. For example, Ofcom (2002, 2003b,c). See also Monti (2001) and the reports published by the Office of Fair Trading; OFT (1992, 1998, 1999, 2001). Note that throughout this paper, the UK telecoms regulator is referred to as ‘Ofcom,’ the Office of Communications regulation. In practice, earlier ‘Ofcom’ reports were produced by an previous incarnation of that NRA, namely Oftel, the Office of Telecommunications Regulation. See the Ofcom website at http://www.ofcom.org.uk/ for details of the reorganization and its rationale.

2. See for example CAT (2004).

3. It is also possible to include supply side substitution in the consideration of profitability – if other potential producers can quickly enter the market consequent on a price increase, this will reduce or eliminate the profitability of it. In the short run however, the main focus tends to be on demand side substitution.

4. See for example Hausman et al. (1994), Werden and Froeb (1994), Nevo (2001), Slade (2006).

5. This assumption rules out the possibility that grouping products might facilitate alternative modes of production; clearly economies of scale or scope would tend to reduce costs. However, for many regulatory market boundary assessments, the time horizon for the test is as short as 1–2 years, and on this horizon, the assumption probably holds fairly well.

6. There is some logic to this as it attempts to define market boundaries independently of the actual nature of competition in the market. The idea being, presumably, that the latter is assessed in the subsequent SMP assessment. Earlier versions of the merger guidelines appeared to allow the possibility of induced price changes outside the set of prices under the control of the HM. See Elroy (1995) for further discussion of this issue.

7. Of course, this holds for Hicksian demands at the level of the individual consumer. However, here, the interest is in Marshallian demands, and in addition, the demand system under consideration comprises only a (probably fairly small) sub‐set of ‘all products.’ This is so because any regulatory or anti‐trust investigation will be concerned with some ‘industry grouping’ (such as ‘telecommunications’). It follows that properties which arise at the individual level imply no useful restrictions on the structure of the above functions. Notwithstanding this, diagonal dominance can still be regarded as a reasonable empirical assumption – that is, one which is valid in most empirical applications.

8. If ≤ 0, then no price increase is profitable and α_max =0; no price increase is profitable.

9. This is discussed in detail in the next section. See also footnote 4 above.

10. There are a range of other issues not addressed here which have been discussed in the literature, notably the issue of competitive price reactions, the geographic dimension and also the issue of price discrimination (market segmentation through tariff design – see Hausman et al. (1996). Also, space limitations preclude consideration of alternative qualitative or quantitative assessment approaches to market definition; these are usefully reviewed in OFT (1992, 1998, 1999, 2001), ICP‐ANACOM (2003). See also Slade (1986), Massey (2000).

11. See e.g., Morris and Mosteller (1991), Massey (2000), Evans and Schmalensee (2001).

12. See Dobbs and Richards (2004).

13. In merger analysis, it may be possible to avoid the need for market boundary analysis through a direct approach to the assessment of market power – by analysing the extent to which the merger would lead to price increases etc. post‐merger – see e.g., Nevo (2001), Slade (2004).

14. Clearly if an α%SSNIP test is passed, then the HM can profitably raise each individual price by at least α%. If a proportionate price increase is profitable, this is sufficient to establish that a more general price adjustment could generate even greater profit. It is possible to revise the form of the HMT to allow for non‐proportionate price adjustments, although this is beyond the scope of the present work.

15. Using actual price when this reflects monopoly power is often referred to as the ‘cellophane fallacy,’ following a US supreme court case concerning the market for cellophane wrapping: US v. E.I.duPont de Nemours&Co., 353 US 586 (1957).

16. Section 6 provides a good illustration of the problems raised here.

17. There is also an argument that the relevant costs should be short‐run avoidable costs since the price increase induces a volume reduction.

18. Particularly for new services, where the norm is to ‘build ahead of demand’ in anticipation of a growing market, in which case short run marginal costs (associated with short run excess capacity) can be quite low (and persistently low).

19. Undertaking sensitivity analysis is computationally feasible using hand calculations, but for more than three products, this is singularly tedious and a computer program or spreadsheet is really needed to undertake the exploration for larger product groupings (the program used for this example is available at http://www.staff.ncl.ac.uk/i.m.dobbs/).

20. Only fairly simple forms of sensitivity analysis are discussed here. A possible extension would be to embed the market definition algorithm in a simulation model. Thus, suppose the interest is in a particular subset L. Given the covariance matrix associated with the elasticity estimates, and perhaps with distributional assumptions made for marginal costs, one could envisage taking drawings from these distributions, computing – and then repeating this process a number of times. The proportion of the runs for which >0 could thus be used to back up statements such as ‘L is a relevant market at the 95% level of significance.’

21. No attempt has been made to gather econometric evidence – in part no doubt because of the emergent nature of the market in this case – however there is some scope to move in this direction; see Nankervis (2004) for example.

22. This section is based on Dobbs (2004), comments submitted to Ofcom as part of the consultation process during the wholesale Broadband market review. This material is available from Ofcom at web address http://www.ofcom.org.uk/consultations/past/wbamp/response/unt.pdf. The original Ofcom surveys are available at web address http://www.ofcom.org.uk/static/archive/oftel/publications/research/2003/.

23. Thus for example, Table 2 panel 1 columns 2,3,4,5,0 figures are divided by ⟨1−(col6/100)⟩. For example, the Table 3, line 1, col. 2 figure of 12.66% is given by Table 2 line 1 col. 22 figure of 10% divided by (1−0.21).

24. For example, Table 3, panel 1, line 1, column 1 figure of −29.11% equals the sum of 12.66+10.13++0.00+0.00+6.33% and so on.

25. For example, in Table 3, panel 2, line 3, the 10.42% has been allocated to columns 2,3,0 in line 4 in the same proportions as were found in the earlier survey (at line 2). This process was repeated at panel 4, line 4.

26. For example, Table 4, panel 1, line 1, column 1 figure of −10.42% corresponds to Table 3, panel 2, line 4, column 1 (and so on).

27. See Collins (2004) and Millward Brown (2004) for assessments of error margins in the Ofcom survey work.

28. There are several other issues; for a more detailed commentary, see Dobbs (2004).

29. This is of course fairly contentious. No significant analysis has been undertaken by any of the parties involved; the above assessments are based on ‘first principles’ reasoning..

30. For other concerns regarding the surveys, see Collins (2004).

31. See website http://146.101.202.226/legacy_regulators/oftel/oftel_internet_broadband_brief/?a=87101#9.

32. Another issue, discussed in section 4, but not addressed here, concerns whether these constitute competitive price levels.

33. Note that only relative shares are required in order to compute the measure , as only the sign matters, not the absolute figure.

34. The above calculation illustrates the fact that it is erroneous to simultaneously subscribe to the HMT methodology and to focus solely on cross price effects (ignoring own price effects) when deciding on whether services are in the same market or not (and Ofcom, whilst subscribing to the HMT methodology, continues to emphasize cross price effects rather than own price effects; see Ofcom (2004, e.g., para. 2.70).