«Abstract The appropriate deﬁnition of the relevant market is the main task in competition cases. But this deﬁnition, and its application, has ...»
Price Tests for Market Deﬁnition with an Application
to the Beer Market in Uruguay
June 3, 2009
The appropriate deﬁnition of the relevant market is the main task in competition cases.
But this deﬁnition, and its application, has proved difﬁcult in abuse of dominance cases,
mainly because of the cellophane fallacy. Also, in countries with less developed competition advocacy and less transparent private information, structural econometric analysis is unfeasible. Some authors have proposed price tests to overcome some of these difﬁculties, although they have rarely been systematically used. I apply price test to deﬁne the beer market in Uruguay and analyze its key results. In doing this, some new theoretical interpretations are offered for further development.
Keywords: antitrust, relevant market deﬁnition, monopolization, price tests, beer market.
JEL Numbers: C22, C32, L40, L66.
∗ AssistantProfessor Universidad de Montevideo and General Directorate of Commerce, Ministry of Economy and Finance. Email: email@example.com. The opinion of the author do not compromise those of the Ministry of Economy and Finance. I thank Mario Bergara, Fernando Borraz, Walter Cont, Helena Croce, Serafín Frache, Fernando Lorenzo and Jorge Ponce for comments on earlier versions of this paper. All remaining errors and omissions are my own.
1 Introduction Competition policy sets a framework to avoid actions from ﬁrms with market power which may reduce social welfare. As market power is difﬁcult to measure directly, competition agencies rely on indirect evidence to evaluate the competition effect of mergers or other business practices by ﬁrms. In brief this procedure starts by deﬁning the relevant market where ﬁrms compete, taking into account barriers to entry, and ﬁnally measuring ﬁrms market shares. The result of this various measures (narrow markets, high barriers to entry, high market shares of the ﬁrms involved, high switching costs for consumers, atomistic competitors, and the like) should presume the existence of market power, or dismiss it.
The “Horizontal Merger Guidelines” issued by the U.S. Department of Justice (DOJ) and the Federal Trade Commission (FTC) in 1982, and its revisions in 1984, 1992 and 1997, has set the standard to evaluate relevant antitrust markets in merger analysis. Although this framework is highly
to be applied literally, -in particular, it requires a very sophisticated econometric analysis of markets- it also establishes a very detailed framework to delimit markets in antitrust analysis.
Nevertheless,in abuse of dominance or monopolization cases there is not a general rule for relevant market delineation. Abuse of dominance cases are different in nature from merger cases. In the latter the focus is made on the enhancement or creation of market power through mergers between ﬁrms. In abuse of dominance cases, the focus is on actions by ﬁrms designed to sustain existing market power, or to create it if ﬁrms do not already have it. In the last setting, applying the “Merger” provisions is difﬁcult because of the well known “cellophane fallacy”.
Instead, some authors have proposed price tests to deﬁne antitrust markets. These tests are associated with the law of one price (LOP), and study the behavior of different product prices in order to establish integration between markets. With a few exceptions these analysis have hardly been used in antitrust cases for either relevant product or geographical market deﬁnition, although it requires less information than structural analysis.
This paper brieﬂy reviews different approaches to antitrust market deﬁnition, establishing their advantages and limitations. Then it reviews price tests, and applies them to the Uruguayan beer sector. The focus is on their systematic application, looking for uniform interpretations. As different tests convey different information about markets, a careful interpretation of the results is needed. Also, some theoretical interpretations of empirical tests are advanced.
The paper is organized as follows. In the next section, traditional deﬁnitions of relevant markets are reviewed. Next, price tests are presented, showing their limitations and introducing some theoretical interpretations. In section 4, price tests are applied to the Uruguayan beer sector, and results cast evidence that beer is an antitrust market for Uruguay. Section 5 shows the main conclusions.
2 Traditional Deﬁnitions of Relevant Market and their Lim
The methodology set out in the “Horizontal Merger Guidelines” is now standard for relevant market delineation in antitrust cases (see, among many others, Church and Ware (2000) chapter 19, and Motta (2004) chapter 3). The Guidelines established a framework for market delineation in antitrust analysis, called antitrust markets, which differs from traditional market deﬁnition in economics (Church and Ware (2000), Slade (1986), among others). In the ﬁrst one, emphasis is on establishing whether ﬁrms have -or have not- the ability to raise prices, whether in traditional economic analysis markets are deﬁned so as to study how prices are being set.
The Guidelines deﬁne a market as “a product or group of products and a geographic area in which it is produced or sold such that a hypothetical proﬁt-maximizing ﬁrm, not subject to price regulation, that was the only present and future producer or seller of those products in that area likely would impose at last a ’small but signiﬁcant and non transitory’ increase in price (SSNIP test), assuming the terms of trade of all other products are held constant” (U.S.
Department of Justice and Federal Trade Commission (1997) page 4). This idea, and further elaborations included in the Guidelines are the main reference for relevant market deﬁnition for various antitrust agencies.1 1 For the UE see the “Commission Notice on the deﬁnition of the relevant market for the purposes of the This analysis, however, could be misguided if used on monopolization or abuse of dominance cases, as shown in the 1956 Du Pont case,2 which give rise to the label “cellophane fallacy”. The case was dismissed because the Supreme Court agreed with the ﬁrm that the market was ﬂexible packaging materials, instead of cellophane. The ﬁrm argued that there was high substitution between cellophane and other ﬂexible packaging materials. But high substitution could be the result of ﬁrm already having market power: as its pricing policy approach that of monopoly, substitution tend to become higher and substitute goods tend to show up as prices increases.
The cellophane fallacy gives a word of caution when the SSNIP test is applied in monopolization cases, as it should not be applied to the prevailing market price, as in merger cases, but to more competitive ones. This solution advances further problems such as how to establish which should be the “competitive” price, as it is not observed in the market.3 As a result, the SSNIP test is not suitable for monopolization cases. This was established either by defenders or detractors of the Guidelines (see White (2005), Werden (2000), Forni (2002), and Geroski and Grifﬁth (2003), among many others).4 Also in merger cases, the strict version of the SSNIP test in either its qualitative or quantitative version are not widely used for relevant market deﬁnition. Copenhagen Economics (2003) reports that only 4% of merger cases in the UE between 1990 and 2001 used this method as a framework for geographic market deﬁnition, and 11% for product market deﬁnition. In USA, Coate and Fischer (2007) study 116 cases of relevant market deﬁnition by the FTC, and report that in 55% of those cases the relevant market analysis was relatively easy and did not require any sophistication to be determined. Only in one case the SSNIP test was used as established in the Guidelines.
The detailed information needed to perform the SSNIP test (prices and quantities of involved Community competition law” available at http://ec.europa.eu/comm/competition/antitrust/relevma_en.html (accessed the 6 of September, 2008). For the UK, see the Ofﬁce of Fair Trading ”Market Deﬁnition”, available at http://www.oft.gov.uk/shared_oft/business_leaﬂets/ca98_guidelines/oft403.pdf (accessed the 22 of June 2008).
2 U.S. v. E.I. du Pont de Nemours & Co., 351 U.S. 377 (1956). See Church and Ware (2000) pages 599 - 600.
3 Motta (2004) page 105.
4 Justice been said that the Guidelines were though to be applied in merger cases.
products, and cost and demand shifters for a long enough period of time),5 and the statistical sophistication implied, made this analysis the exception rather than the rule in antitrust analysis.
The consensus is that the Guidelines set a framework to study market deﬁnition issues, rather than set the SSNIP test as a rule for quantitative analysis.
Taking into account these limitations, some authors’ advice is to use price tests to establish if products or geographical areas belong to the same relevant market. Stigler and Sherwin (1985) deﬁne a market as the area in which the price is determined, and two products should be in the same market if their prices tend to co-move. This methodology has its roots in the LOP which establish that if two commodities belong to the same market arbitrage should equate their prices in the long run or. If we allow for transportation costs, the difference should be stable.
Nonetheless, Geroski and Grifﬁth (2003) point that there is no obvious relationship between market limits obtained applying this deﬁnition and the one set in the Guidelines.6 Caution should be the main advice in using price tests, as the results are complementary to qualitative evidence at hand. As the next section shows, price tests also have their problems and as such its conclusions should be carefully evaluated.
3 Price Tests
The literature points to four price tests, which can be divided into descriptive tests and analytical ones. Descriptive tests are correlation analysis, and unit root tests. Analytical test are those associated with vector error correction: cointegration, weak exogeneity and Granger causality.
A detailed exposition of each of these tests can be found in Haldrup (2003).
Starting with Stigler and Sherwin (1985) price tests have been advocated either as an alternative test for the one proposed in the Guidelines, or as a means to implement the SSNIP methodology. Haldrup (2003) notes that the Guidelines establish an attractive framework for thinking about markets in antitrust, but this methodology is not operative. Although price tests 5 See the analysis of Hausman, Leonard, and Zona (1994) for the USA beer market, Nevo (2000) and Nevo (2001) for the ready-to-eat breakfast cereal, and Scheffman and Spiller (1996) for margarine and butter.
6 See also Froeb and Werden (1993) and the reply of Sherwin (1993) and the discussion in Haldrup (2003) pages 4 - 7.
have received criticism, he points out that they are convenient tools for market delimitation.
At the onset analysis is carried out the integration order of the price series should be tested.
Haldrup (2003) established that a necessary condition, although not sufﬁcient, for goods to belong to the same relevant market is that price series have the same integration order. The idea is very simple; if one series is a white noise and the other is a random walk, then there is not arbitrage between them that bind the path of these goods, and they could not belong to the same market.
3.1 Descriptive tests
The ﬁrst test was proposed by Stigler and Sherwin (1985) who study the correlation between the logarithm of the price of goods candidates to be in the same geographic market and their ﬁrst differences. If correlation is not high enough, one could presume that goods should not belong to the same product market.7 But in order to have coherent results when using a correlation test series need to be stationary.8 As the distribution of the t-statistic is derived from the normal distribution -which is by deﬁnition stationary-, if series are non stationary they need to be differentiated enough times so that they become stationary. Then, it could be checked if correlation exists or not between them. But even if they are stationary, correlation between two products could be high because a third one, like a common input, moves the prices of the other two. In this case, the correlation is spurious in the sense that is not driven by arbitrage between substitutes, but by movements in the common input. This problem could appear either if series are stationary or not, although when series are non stationary its effects are exacerbated because of the time trend in the series.
A related problem is that if the adjustment between product prices occur with delay, the correlation coefﬁcient might not capture this inﬂuence especially if data is of high frequency; e.g.
Although the literature points towards the use of this test in order to exclude goods from the 7 See also the discussion in Motta (2004) pages 107 - 109.
8 Fernando Lorenzo pointed me this issue.
market when series are not stationary (see Motta (2004) pages 108-9), the only conclusion that can be draw from this test is if there is or not a lineal relationship between stationary series.
Besides the statistical threshold for rejection of the zero value of the correlation, a high or low value is just a matter of interpretation.
A second test was proposed by Forni (2002) who analyzes the stationarity of the log of the price ratio of products candidates to belong to the same relevant market. If two goods belong to the same market we would expect that their price ratio should be stationary. If not, prices are driven away from each other and there is no arbitrage mechanism that binds them. This test is similar to the cointegration test of the log of nominal prices with an [1, −1] cointegration vector, but if deﬂated series are used instead of nominal ones, then it is useful to check the consistency of results from both tests.
Hosken and Taylor (2004) make the point that the lack of relevant information in the unit root test as proposed by Forni (2002), could result in misguided results. The authors advocate for collecting adequate institutional information, and use this test as a complement of qualitative information.