198 4. Market Definition To help inform the analysis of the large amount of survey evidence considered in that case, a summary of the evidence is presented in table 4.6. In particular, note that the results of four separate surveys are reported. The surveys are called respectively Swift 1, ORC, Swift 2, and GfK after the survey companies which undertook them. The first two surveys were addressed toward customers who had recently stopped playing with a provider (“lapsed” customers) while the second two surveys involved current customers. In terms of the latter pair of surveys, the Swift 2 survey asked consumers directly how they would respond to a 10% price increase 29 while the GfK survey used show- cards to allow consumers to compare hypothetical product offerings. The CC has found that directly asking consumers about what they would do if prices went up by 10% can sometimes lead to results that are difficult to interpret. Show cards can also sometimes produce surprising results. For example, one part of the GfK survey used show-cards and suggested increasing demand schedules! Surveys aimed at capturing diversion ratios aim to directly estimate the substitu- tion effect between two products. These methods have the merit that they address directly the issue of interest in market definition and make few theoretical assump- tions. But they are heavily reliant on good-quality data obtained through high-quality surveys. Survey design in this area remains under development. Wherever our information on substitution patterns comes from, surveys or demand estimation, we will of course still need to use that information to evaluate the impor- tance of rival products as constraints on price-setting behavior. In section 4.6, we discuss strategies that can be used both quantitatively, when we have good-quality data, but also sometimes qualitatively when we do not. Before we do so we first turn briefly to one additional technique sometimes useful for geographic market definition. 4.5 Using Shipment Data for Geographic Market Definition Elzinger and Hogarty (1973, 1978) 30 proposed a two-stage test for geographic mar- ket definition. The two stages are known respectively as “little out from inside” (LOFI) and “little in from outside” (LIFOUT). Given a candidate market area, the LIFOUT test considers whether nearly all purchases come from within the region itself or whether there are substantial “imports.” Analogously, given a candidate market area, the LOFI test considers whether nearly all shipments go to the region itself or whether there are substantial “exports” from the region. Intuitively, import and export activities suggest competitive interconnectivity. LOFI is also sometimes 29 The Swift 2 survey asked: “If your pools company increased the cost of playing by 10%, what would you do?” 30 A nice description of the U.S. judicial history in this (and other) areas is provided by Blumenthal et al. (1985). See also Werden (1981, pp. 82–85). 4.5. Using Shipment Data for Geographic Market Definition 199 described as the “supply” element of the test, since it relates particularly to the des- tination of production coming from a candidate area, while LIFOUT is sometimes considered as the “demand” element of the test since it relates to purchases made by consumers in the candidate market area. The overall idea of the combined test (LIFOUT + LOFI) is to expand the candidate market areas until both “supply” and “demand” sides of the test are satisfied in a market area. To operationalize this test, we must first define what we mean by “little.” Elzinga and Hogarty suggested using benchmarks so that if only 25% (or they later suggested 10%) of production in an area is “exports” or “imports,” we would consider there to be respectively LOFI or LIFOUT. To apply the LOFI test, the authors suggest beginning with the largest firm or plant and finding the area where (say) 25% of that plant’s shipments goes to. The LOFI test then asks whether LOFI D 1  Shipments from plants in area to inside Production in candidate area D Exports Production in candidate area 6 0:25: If so, then the LOFI test is met, since “nearly all” of the sales from plants occur within the area. If the test fails, then area must be expanded to find an appropriate area where the test is indeed satisfied. One option is to find the minimum area needed to account for 75% of output from all plants within the previous candidate area. If the expansion of the area does not involve incorporating any new plants, then such a procedure clearly generates an area that will meet the LOFI test. On the other hand, expanding to capture more sales of the set of plants under consideration may sometimes also place additional plants within the candidate market area and we shall return to this observation in a moment. The LIFOUT test examines the purchase behavior of consumers within acandidate region, asking whether LOFOUT D 1  Purchases by consumers in area Production in candidate area 6 0:25: In some contexts, particularly commodity markets, the Elzinga–Hogarty test has been generally well received by government agencies, the courts, and the compe- tition policy academic community over the last thirty years. However, in the late 1990s the test came under renewed scrutiny after the U.S. agencies and state author- ities objected to seven out of a total of 900 hospital mergers between 1994 and 2000 and lost all seven of the cases! A number of these cases were lost because the courts accepted the merging parties’ application of the Elzinga–Hogarty test using patient flow data. A period of reflection and retrenchment followed with the Federal Trade Com- mission (FTC) and Department of Justice (DOJ) undertaking a major exercise of 200 4. Market Definition hearings and consultation, summarized in FTC and DOJ (2004). 31 DOJ and FTC concluded that “the Agencies’ experience and research indicate that the Elzinga– Hogarty test is not valid or reliable in defining geographic markets in hospital merger cases” (chapter 4, p. 5). Proponents of the test would no doubt argue that this is in fact a fairly limited conclusion, in particular perhaps noting that DOJ and FTC do not say that Elzinga– Hogarty is not valid and reliable, only that it is not valid and reliable in hospital mergers. However, at least these comments make the hospital context particularly interesting and so we focus on it. In addition, it is difficult to escape the observation that the primary critiques leveled at Elzinga–Hogarty in that context do appear to apply far more widely. To see how Elzinga–Hogarty was applied in hospital mergers, note that a patient who lives in a candidate market area but who goes to a hospital outside it for treatment is considered to be “importing” hospital services into the candidate area, and is measured as LIFOUT since she is inside the area and purchasing hospital services outside it. On the other hand, a patient who lives outside the candidate area and who comes into the area to the hospital is considered an “export” of services and so is measured as LOFI. The first critique of the Elzinga and Hogarty test is that existing “flows” of supply or demand need not be informative about market power. In particular, the fact that some consumers currently use hospitals outside the area does not imply that the level of “imports” would increase dramatically if hospitals within the market area increased prices by a small amount. The FTC and DOJ go on to note that patients travel for a number of reasons, including “perceived and actual variations in quality, insurance coverage, out-of-pocket cost, sophistication of services, and family con- siderations” (chapter 4, p. 8). If so, then the fact that some consumers travel does not immediately imply that those who are currently not traveling are price-sensitive. Capps et al. (2001) call this logical leap the “fallacy of the silent majority.” The second critique noted that if LIFOUT or LOFI fail with a given candidate region, the algorithm involves expanding the region and considering the wider can- didate market. However, doing so changes both the set of customers and the set of production facilities (patients and hospitals), so that the LIFOUT and FIFO tests may fail again in the wider region. In some examples, the resulting geographic market can expand without limit. The bottom line, as with many techniques we examine in this chapter, is that Elzinga and Hogarty’s test can provide a useful piece of evidence when coming to a view on the appropriate market definition. However, as the U.S. hospital experi- ence suggests, it may seriously mislead those who apply the test formulaically and we must be clear that we are finding evidence of interconnectivity which may, in particular, be substantively distinct from a lack of market power. 31 See, in particular, chapter 4 of FTC and DOJ (2004). 4.6. Measuring Pricing Constraints 201 4.6 Measuring Pricing Constraints One way to think about pricing constraints that restrict a firm’s ability to increase prices is that they arise directly from competitors who compete in the same market. Firms without competitors do not face pricing constraints, except to the extent that consumers decide not to purchase at all, and therefore will often have a unilateral incentive to increase prices. Turning these observations around suggests that one way to think about market definition is as a set of products which, if a firm were a monopolist, the constraints arising from weaker substitutes outside the market would be insufficient to restrict the monopolist’s incentive to increase prices. An antitrust market is then conceived as a collection of products “worth monopolizing.” This is the idea encapsulated in the hypothetical monopolist test (HMT). The focus of such tests is typically prices, but in principle they may equally be applied to relevant nonprice terms. That said, price is often the central dimension of short- run competition and so we will often consider whether a hypothetical monopolist has an incentive to implement a small, nontransitory but significant increase in price (SSNIP). In practice, the HMT is often applied quite informally when data or reliable estimates of relevant elasticities are not available. Informally, the HMT plays an important role in providing a helpful (though certainly imperfect) framework for structuring decision making in market definition. Next we provide a more formal description of the HMT test. 4.6.1 The Hypothetical Monopolist Test The price-based implementation of the HMT, the SSNIP test, is based on the idea that products within a market as a group do not face significant pricing constraints from products outside of the market. 32 Assume a market that includes all brands of still bottled water. The price of batteries is unlikely to exert a price constraint on the price of bottled water and can therefore be rapidly removed from consideration as a candidate for being in the relevant competition policy market. But what about the price of sparkling water? The SSNIP test calculates whether a monopolist of still bottled water could increase prices without losing profits to sparkling water produc- ers. If so, we would conclude that sparkling water is not in the same competition policy market as still water. If not, we would conclude that sparkling water must also be included in the market definition. A profitable monopolist would have to own both still and sparkling water production plants to be able to exercise market power. 32 We shall inevitably fall into the traditional activity of equating the HMT and SSNIP tests. However, the SSNIP is actually best considered as one particular implementation of an HMT test—one focused on the profitability of price increase. In some industries, advertising or quality competition may be the dominant form of strategic interaction and if so a narrowly focused SSNIP analysis may entirely miss other opportunities for a hypothetical monopolist to “make a market worth monopolizing.” 202 4. Market Definition The logic of a market as a collection of products that is “worth monopolizing” sug- gests that one approach to defining a market in antitrust investigations is to explicitly abstract from pricing constraints arising from competition within a proposed mar- ket, i.e., proposing a hypothetical monopoly over a set of products. A market can then be defined as the smallest set of products such that a hypothetical monopolist would have an incentive to increase prices. If we propose a candidate market which is too small, we will have a monopolist who faces a strong substitute outside the proposed market and so who will have no incentive to raise prices. Thus the hypothetical monopolist test tries to measure whether there is a sig- nificant price constraint on a given set of products that comes not from the intra- candidate market competition but from the availability of other products—outside the proposed market definition—that offer viable alternatives to consumers. 33 To do this, the HMT assumes that all products within the proposed market defini- tion are owned by one single producer which sets each of their prices in an attempt to maximize the total profits derived from them. If the hypothetical monopolist finds it profitable to increase prices, we will have found that constraints from goods outside the proposed market definition are not a sufficient constraint on producers within the market to render a price rise unprofitable. In other words, prices were kept down by the competition within the market. In practice, to operationalize this idea we must, among other things, be a little more precise about exactly what we mean by a “price rise.” To that end most jurisdictions apply the “SSNIP” test, which looks at whether a “small but significant nontransitory increase in prices” would be profitable for the hypothetical monopolist. Usually, “small but significant nontransitory” is assumed to mean 5–10% for a year. 34 4.6.1.1 Decision Making under the HMT Decision making when using the HMT can be represented by the algorithm represented in figure 4.12. We start with the narrowest product or geographic market definition which is usually called the “focal product” and actually usually also the focal product of the investigation. We then need to evaluate whether a monopolist of this product could profitably raise prices by 5–10% for a year. If so, that single product will then 33 A nice treatment ofthe SSNIP test is provided in the paper by the previous chairman of the U.K. Com- petition Commission, Professor Paul Geroski, and his coauthor, Professor Rachel Griffith (see Geroski and Griffith 2003). 34 This “tradition” in the competition policy world is potentially a dangerous one in the sense that in some markets a 5% price rise would correspond to an absolutely enormous increase in profitability. For example, in markets where volumes are high and margins are thin (e.g., 1%), a 5% increase in prices may correspond to a 500% increase in profitability. Relatedly, the consumer welfare losses associated with a 5% increase in prices may in some circumstances (particularly in very large markets) be huge. In such cases, it may be appropriate to worry about monopolization of markets even where monopolization only leads to an ability to increase prices by say 1% or 2%. As always, the key is for the analyst to think seriously about whether there are sufficient grounds for moving away from the normal practice of using 5–10% price increases for this exercise. 4.6. Measuring Pricing Constraints 203 Start with the narrowest product or geographic market definition. Is it profitable for a monopoly producer of that product to increase prices in a small but significant and nontransitory way (SSNIP)? There must be at least one good substitute excluded from the current market definition. Expand the market to include it. Now have a multiproduct monopolist. Could he/she profitably raise prices? Stop. Market definition is wide enough. No Yes No Yes Figure 4.12. The HMT decision tree. constitute our antitrust market. If not, we must include the “closest” substitute, that product which provides the best alternative to consumers facing the price increase. We then assume again a hypothetical monopolist, this time of each of the products in our newly expanded set of products in our candidate market and we repeat our question, will a 5–10% price increase for a year be profitable? This process continues as long as the answer to the question is “no.” A “no” indicates that we are missing at least one good substitute from our current candidate market definition and the omitted product is constraining the profitability of raising prices for our monopolist. We stop the process of adding products when we have a set of products that does indeed allow the hypothetical monopolist to profitably raise prices without losing customers to outside products. We define our antitrust market as the final set of products, the set of products which it is “worth monopolizing.” To illustrate further, suppose we face a situation in which three firms produce three products called, somewhat uninspiringly, products 1, 2, and 3. Each of these products is in fact a very good substitute; for the sake of argument, suppose they are perfect substitutes. Suppose also that there are two other products, products 4 and 5, which are rather poorer substitutes. Product 1 is the focal product. Table 4.7 demonstrates the step-by-step application of the HMT to this case. 204 4. Market Definition Table 4.7. Steps in a hypothetical monopolist test. PMD is proposed market definition. Step 1 Step 2 Step 3 PMD f1gf1; 2gf1; 2; 3g Q Does monopolization of product 1 give pricing power? Does a (hypothetical) monopolist of products 1 and 2 have pricing power? Does a (hypothetical) monopolist of products 1, 2, and 3 have pricing power? A No, because there are two perfect substitutes omitted from the proposed market. No ability to raise price of good 1. No, because there is still a perfect substitute omitted from the proposed market (product 3) that constrains the ability of our hypothetical monopolist of goods 1 and 2 to raise their prices. Yes, if products 4 and 5 are not good enough substitutes. If so, then the market definition of f1; 2; 3g is accepted. No, if either product 4 or 5 is a good enough substitute to constrain profitability of price increase. In that case, continue the test. Suppose we did not use the HMT at step 3 but just looked at the pricing power of three independent firms. Those firms would have no pricing power because of constraints that come from within the proposed market definition. For example, the firm producing 3 will have no market power because of the presence of producers of goods 1 and 2. Thus the HMT works by explicitly putting the focus on the constraints on pricing power that come from outside the proposed market definition. 4.6.1.2 Implementation of the SSNIP Test The SSNIP test consists of evaluating whether a 5–10% price increase for all the products in the candidate market will produce a profit. Consider the single-product candidate market. Recall that the firm’s profits are the total revenues minus the total variable and fixed costs: ˘.p t / D .p t  c/D.p t /  F; where, for simplicity, we have assumed a constant marginal cost. The change in profits due to an increase in prices from p 0 to p 1 can then be expressed as ˘.p 1 /  ˘.p 0 / D .p 1  p 0 /D.p 1 /  .p 0  c/.D.p 0 /  D.p 1 //; where the first term of the equality is the gain in revenues from the increase in prices on the sales at p 1 and the second term is the loss of margins due to the decrease in sales after the price hike. The core question is whether the drop in volume of sales at the new price, and consequent loss in variable profit, is big enough to outweigh the increased revenues obtained on goods still sold. This trade-off is shown graphically in figure 4.13. 4.6. Measuring Pricing Constraints 205 P 1 P 0 D c Q 1 Q 0 Gained revenue from higher price on goods still sold Lost margins on goods no longer sold + − Q P Figure 4.13. The trade-off when evaluating the profitability of a price increase. Evidently, the crucial assumption of the SSNIP test is that the fall in demand will be large when there are good substitutes available. In fact, we can show that it will be profitable for the monopolist to raise its prices as long as its margin is lower than the inverse of its own-price elasticity of demand. In our benchmark model, a hypothetical monopolist of a single product in a potentially differentiated product market will solve the profit-maximization problem: max p 1 ˘.p 1 Ip 2 ;:::;p J / D max p 1 .p 1  c/D.p 1 ;p 2 ;:::;p J /: A monopolist of product 1 will increase price as long as it raises their profits, i.e., as long as @˘.p 1 ;p 2 ;:::;p J / @p 1 D .p 1  c/ @D.p 1 ;p 2 ;:::;p J / @p 1 C D.p 1 ;p 2 ;:::;p J / > 0: We can rearrange the expression to obtain p 1  c p 1 6  D.p 1 ;p 2 ;:::;p J / p 1  @D.p 1 ;p 2 ;:::;p J / @p 1 à 1 D 1 Á 11 .p 1 ;p 2 ;:::;p J / : We will want to evaluate whether this inequality holds for all prices between p Comp 1 and p 5% 1 D 1:05p Comp 1 or p 10% 1 D 1:10p Comp 1 respectively depending on whether we use a 5% or 10% price increase. In this model, the data we need to perform the single- product variant of the SSNIP test are therefore (i) the firms’ margin information under competitive conditions and (ii) the product’s (candidate market’s) own-price 206 4. Market Definition elasticity of demand (again in the range Œp Comp 1 ;p 5% 1  or Œp Comp 1 ;p 10% 1 ). 35 For imple- mentation, the important aspect of this single-product variant of the test is that we do not need a full set of cross-price elasticities of demand. The pricing theory analysis of substitutability (usually associated with measuring cross-elasticities) turns into a problem which only involves an evaluation of the own-price elasticity of demand and a comparison of it with variable profit margins. (We will say more shortly.) A common shortcut for the SSNIP test in geographical market definition is to consider the cost of transporting goods from outside areas intothe candidate markets. This relies on the assumption that goods are homogeneous and buyers are indifferent as to the origin of the good. If transport costs are low enough that a price increase by up to 10% by the hypothetical monopolist is likely to be met by an inflow of cheaper product from elsewhere, the candidate market needs to be widened to include the area where the shipped goods are coming from. Evidence on existing shipping activity and transportation costs are therefore often used in practice to determine geographic market definitions. The purpose of the SSNIP test is to check whether the hypothetical monopolist would find it profitable to increase prices from the competitive level by a material amount (perhaps 5–10%) for a material amount of time (perhaps one year). Note that the reference price for this evaluation is usually described as the “competitive price.” This benchmark element of the test is crucial and sometimes it proves problematic as we will illustrate in the next section. In a formal application of SSNIP we may have an estimate of the marginal cost and also an estimate of a demand curve. This in turn gives us a description of the determinants of profitability so that we can directly evaluate whether ˘.1:05p Comp 1 Ip 2 ;:::;p J /  ˘.p Comp 1 Ip 2 ;:::;p J / D .1:05p Comp 1  c/D.1:05p Comp 1 ;p 2 ;:::;p J /  .p Comp 1  c/D.p Comp 1 ;p 2 ;:::;p J / > 0: At this point, we present a brief aside, aiming to note that there is a theoretical underpinning to the observation that the own-price elasticity of demand is informa- tive about substitution opportunities. In fact, we only need for income effects to be small enough to interpret own-price elasticities as the substitution effect. In most fast-moving consumer goods, the income effect will be relatively small, so when we look at the own-price elasticity of demand we are mostly talking about the sum of all the cross-price effects. Looking at own-price elasticity is appropriate when trying to assess the constraint of substitutes as long as we can be confident, as is generally the case, that the income effect is not playing a major role in the decision making. 35 It will often be very difficult to tell whether the own-price elasticity varies materially in the range, and it is usual to only report a single number estimated using the predicted change in quantity following a 5 or 10% change in prices. Such an elasticity estimated between two given points is also known as an arc-elasticity. 4.6. Measuring Pricing Constraints 207 When a function is homogeneous of degree zero, as is the case for an individual’s demand function for a product j , we can apply Euler’s theorem 36 J X kD1 p k @q j .p; y/ @p k C y @q j .p; y/ @y D 0: We then obtain 1 q j .p; y/ Ä @q j .p; y/ @ ln p j C X k¤j @q j .p; y/ @ ln p k C @q j .p; y/ @ ln y  D 0; which in turn can be written as Á jj C X k¤j Á jk C Á jy D 0 or  Á jj D X k¤j Á jk C Á jy : This relationship suggests that the own-price elasticity of demand will be large when either substitution effects are large or the income effect is large. The latter is caused by the fact that the increase in price reduces the customer’s real income and their income elasticity is high. Finally, note that the homogeneity property relies on us doubling the prices of all possible goods in the economy as well as income. In practice, we may treat one good as a composite good consisting of “everything outside the set of goods explicitly considered as potentially within the market,” or more simply the “outside good.” There will often not be any price data for the outside good, although we could use general price indices as an approximation. Substitution effects can occur to the outside good, so that if we doubled all inside good prices and income we will see that demand for the set of inside goods will fall. If so, then the own-price effects will be larger in magnitude than the sum of the substitution effects (to inside good products) plus the income effect. More generally, of course, we will want to evaluate whether a price increase for a collection of products is profitable.We discuss this case further in section 4.6.3. First, we consider a particular type of difficulty that often arises—even in a single-product context—when we apply the SSNIP test in practice. 4.6.1.3 The Cellophane and Reverse-Cellophane Fallacies and Other Difficulties The Cellophane Fallacy. In the U.S. v. DuPont case in 1956 37 it was crucial to determine whether cellophane (“plastic wrap”) represented a market. At that time 36 Assume a function homogeneous of degree r. By definition we have q j .p 1 ;p 2 ;:::;p J ;y/D  r q j .p 1 ;p 2 ;:::;p J ;y/: We obtain Euler’s results by differentiating both sides with respect to . 37 United States v. E. I. DuPont de Nemours & Co., 351 US 377 (1956). [...]... It also uses information about demand and in particular the own-price elastic40 This section draws on Harris and Simons (1989) and also the working papers by O’Brien and Wickelgren (2003) and by Katz and Shapiro (2003) 4.6 Measuring Pricing Constraints 211 ity of demand to make inferences about the price constraint exerted by substitute products The question asked in critical loss analysis is the... prices and the diversion ratios Following the logic of the single-product case, Baker and Bresnahan (1985) suggest solving the 2.J 2/ demand and pricing equations (for products j D 3; : : : ; J ) to provide a description of the equilibrium prices that would result for those products for any given level of prices for goods 1 and 2, the goods in the candidate market That is, suppose that we solve for the... is that proposed by Scheffman and Spiller (1987) for homogeneous product markets and Baker and Bresnahan (1985, 1988) for differentiated product markets The approach is known as the residual demand function approach and can be useful for evaluating the extent of market power or market definition in some particular circumstances However, these models are explicitly not implementing a standard SSNIP test... Ivaldi and Lorincz (2009) the market power of firms, the residual demand approach can be used for market definition in a fashion not unrelated to the FERM test To see why, we first recall the notion of a residual demand curve First, following Landes and Posner (1981) and Scheffman and Spiller (1987) consider the dominant-firm model In that model, the dominant firm faced a market demand D Market p/ and also... potentially solve for the 2J unknowns—equilibrium prices and quantities for all of the J goods in the market We discuss how to solve the full set of 2J equations explicitly for an arbitrary ownership structure and general demand systems in our discussion of merger simulation in chapter 8 The idea of the residual demand function approach is to solve the demand and supply equations for all of the goods... servers priced between €0 and €4,000 In addition the results also suggest there is a mid-range market for computers between €4,000 and €10,000 servers and a high-end market for computers above €10,000 Table 4.11 reports the analogous results applying the FERM test for market definition In doing so, Ivaldi and Lorincz obtain the same results for the competition policy market definition for low-end computer... estimate demand substitution effects by analyzing purchase patterns and conducting surveys Own- and crossprice elasticities can also be estimated econometrically, although doing so sufficiently robustly to withstand judicial scrutiny is by no means an easy task In order to formally evaluate a SSNIP test, it is not sufficient to estimate the own- and cross-price elasticities of demand Rather we need a standard... of firm 1 has sufficient market power to raise prices by 5% The residual demand function approach suggests that we could solve the 2.J 1/ demand and pricing equations for products j D 2; : : : ; J That is, we can solve for ln pj D Ej p1 I xŒ2WJ  ; wŒ2WJ  ; Œ2WJ  ; Á; ˇ; / for j D 2; : : : ; J; where we denote xŒ2WJ  D x2 ; : : : ; xJ / and wŒ2WJ  and Œ2WJ  are defined analogously These equations... “residual” demand elasticity The insight of the residual demand function approach is that the residual demand function captures all of the relevant information about the constraint implied by other firms and expresses it in terms of the residual demand elasticity Specifically, in considering the profitability of price rises for any firm (which for this example and without loss of generality we shall call firm 1)... that we solve for the equilibrium prices of goods outside the candidate market for a given set of prices of goods inside the candidate market: ln pj D Ej p1 ; p2 I xŒ3WJ  ; wŒ3WJ  ; Œ3WJ  ; Á; ˇ; / for j D 3; : : : ; J: Substituting these equations into the demand curves for products 1 and 2 gives us the “partial residual demand curves” for the two products: ln q1 D Á10 C Á11 ln p1 C Á12 ln p2 C J . draws on Harris and Simons (1989) and also the working papers by O’Brien and Wickelgren (2003) and by Katz and Shapiro (2003). 4 .6. Measuring Pricing Constraints 211 ity of demand to make inferences. to perform the single- product variant of the SSNIP test are therefore (i) the firms’ margin information under competitive conditions and (ii) the product’s (candidate market’s) own-price 2 06 4 the “demand” element of the test since it relates to purchases made by consumers in the candidate market area. The overall idea of the combined test (LIFOUT + LOFI) is to expand the candidate