Critical Loss Analyses in Korean Liquor Mergers

The SSNIP(Small but Significant Nontransitory Increase in Price) test is a well-known conceptual framework of market definition for competition policies in most countries. Critical loss analysis is a practical method that implements the principle of SSNIP test in a quantitative way to determine whether the relevant market for an antitrust case should be enlarged or not. The method and empirical results have been successfully adopted in defining markets relevant to Korean merger cases in soju and beer industries (Moohak-Daesun in 2002 and Hite-Jinro in 2005), providing useful references for the Korean Court and Fair Trade Commission. This paper introduces the actual applications of critical loss analyses in these cases and remarks on several issues brought in the course of applications.


Introduction
The SSNIP test is a well-known conceptual framework of market definition for the purpose of competition policies in most countries. Critical loss analysis is a practical method which implements the principle of SSNIP test in a quantitative way to determine whether the relevant market for an antitrust case should be enlarged or not. The method has been referred to in many antitrust court cases as well as US and UK competition authorities' guidelines of market definition. Critical loss analysis is now popular among experts on competition policies in Korea since it has been successfully adopted in economics analyses of recent two merger cases in soju and beer industries. 1 1 Soju is popular liquor in Korea which is a kind of spirit with alcoholic content of about 20-22%. There are two kinds of soju-distilled and diluted. Popular one in Korea is the diluted, which are made by diluting alcohol essence extracted from grains-ethanol made from rice, barley, corn, etc. The sales of diluted soju in 2004 were about 2.34 trillion won. Total liquor sales were 6.64 trillion won in 2004. Hence, diluted soju accounts for 35.2% of total liquor markets. On the other hand, beer whose sales were 3.45 trillion won in 2004 accounts for 52%.
The following section discusses the SSNIP test as a well-known principle of market definition for competition policies and explains critical loss analysis that implements the SSNIP test practically. In Sections 3 and 4, actual applications of critical loss analyses in the Moohak-Daesun and Hite-Jinro cases are introduced. The author was involved with the two cases as a principal economic expert and submitted the relevant economic reports to the court in the first case and KFTC in the second case. The court and KFTC agreed upon the author's arguments. The two sections in the paper are based on the author's analyses and the court and KFTC's judgments on the two cases. The last section concludes by remarking on several issues brought on in the course of the applications.

SSNIP test and critical loss analysis
Article 2.8 of the Korea Monopoly Regulation and Fair Trade Act (KMRFTA) defines the relevant market as "the range of transactions where there exist or may exist competitive relations in terms of objects, stages, or regions of trade," and KFTC merger guideline articulates it as "the set of products or the whole geographic area into which a representative consumer can switch his/her purchases in response to the small but significant and nontransitory increase in prices of the specific products or regions holding other prices constant." The spirit is the same with SSNIP test on which antitrust enforcement agencies in the USA, EU, and many other countries base their market definition. 5 According to the US horizontal merger guideline in [4], a market is defined as "a product or a 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 or future producer or seller of those products in that area likely would impose a small but significant and nontransitory increase in price, assuming the terms of sale of all other products are held constant." EU guideline in [5] specifies the question to be answered as "whether the parties' customers would switch to readily available substitutes or to suppliers located elsewhere in response to a hypothetical small (in the range 5-10%), permanent relative price increase in the products and areas being considered. If substitution would be enough to make the price increase unprofitable because of the resulting loss of sales, additional substitutes and areas are included in the relevant market. This would be done until the set of products and geographic areas is such that small, permanent increases in relative prices would be profitable." There exists a subtle difference between market definition in Korea and that in USA. In the former, a market is defined in a consumer's perspective as the largest range where he/she can switch his/her purchases in response to SSNIP ("a representative consumer version"). In the latter, it is defined in a producer's perspective as the smallest range where he/she can make profits by SSNIP ("a hypothetical monopolist version"). The representative consumer version may imply a smaller market than the hypothetical monopolist version. The reason is as follows. The representative consumer version considers only substitution effect and includes products or regions which are close substitutes in the relevant market. On the other hand, the hypothetical monopolist version considers price effect, i.e., both substitution and income effects. Even if all close substitutes are included, a hypothetical monopolist may not be able to increase price profitably because of the negative income effect in case of normal goods. Hence, the range of the relevant market should be enlarged further under the hypothetical monopolist version. 6 However, it seems that the hypothetical monopolist version is more appropriate from antimonopoly perspective, since we are concerned about the price increase regardless of its sources. Another advantage of the monopolist version is its practicality; the definition implies the critical level of sales loss that can be compared with actual sales loss after a price increase, which is the idea of critical loss analysis.
While SSNIP test is a conceptual framework of market definition for competition policies, critical loss analysis is a practical method of implementing it in real cases. The analytical method, since it was first introduced by [6], has been tried in many US antitrust cases. 7 Among enforcement agencies, UK OFT and US DOJ and FTC mention explicitly the method as a useful tool in their market definition guidelines. In Korea, Jeon introduced critical loss analysis first in an economic analysis of the geographic market definition relevant to Moohak-Daesun merger (see [1] The idea of critical loss analysis is simple. Profitability of a price increase depends on the amount of consequent sales loss. "Critical loss for X-% price increase" is defined as the maximum percentage loss of sales volume that would not result in profit loss for X-% price increase. To put it another way, critical loss is the minimum percentage loss of sales volume 6 I presume here normal goods with negative income effect. Discussions must be reversed in case of inferior goods. 7  ]. Especially in Swedish match, both enforcement agencies and defendants tried to define the relevant market based on their own critical loss analyses, and the court, reviewing them, suggested its own interpretations. See [7]. that would result in profits decrease. That is, if actual sales loss is larger (smaller) than critical loss, then the price increase will lead to profit decrease (increase). 8 Table 1 summarizes the method of market definition using critical loss analysis. 9 If actual sales loss after a SSNIP is less than critical loss corresponding to the SSNIP, then a hypothetical monopolist can make more profits by such a SSNIP, which implies that the relevant market should be confined there with no need of further expansion. It is because there are no closely substitutable products or regions to which consumers can switch their current purchases in response to a SSNIP. On the other hand, if actual sales loss after a SSNIP is more than the critical loss, then a hypothetical monopolist cannot make more profits by such a SSNIP. This implies that the relevant market should be expanded to include next available substitutes. Market definition according to critical loss analysis starts from considering a set of products and regions for which anticompetitive concerns are raised. Compare actual sales loss with critical loss repetitively until there is no need for further expansion where actual loss is less than critical loss. That is, the relevant market is the smallest market for which a hypothetical monopolist can make more profits by a SSNIP.
Denoting critical loss by CL, we have a very simple relationship of such dependence as follows: CL = (X/(X + M)). 10 Critical loss corresponding to X-% price increase depends on X. The larger a price increase is, the greater sales loss a hypothetical monopolist can endure without incurring profits loss. Another important determinant price-cost margin, M = ((P − C)/P) where P is unit price and C is marginal or incremental cost. For a high rate of margins, each sale lost entails a relatively large loss of profit. Hence, a high rate of margins implies a small level of critical loss. 11

Brief introduction of Moohak-Daesun case
In January 2003 the KFTC made an order that Moohak dispose all stocks of Daesun it purchased in 2002, 12 since the acquisition would restrain competition seriously in local soju markets of Busan and Kyungnam regions and infringe Article 7.1 of KMRFT Act. Daesun, the acquired one, commanded monopolistic position in Busan area with 84.4% market share in 2001, while Jinro, the largest soju producer in the country, had only 7.2% share in the region. On the other 8 Notice that I adopt a breakeven version of critical loss rather than a profit-maximization version. The former is more often used in practices because of its simplicity and independence of the shape of demand curve. 9 If we reinterpret sales loss due to price change as price elasticity of demand, critical loss analysis becomes critical elasticity analysis. 10 For the derivation, see [8]. 11 In interpreting the implication of a high rate of margins, we should be careful of the well-known cellophane fallacy. That is, if high current margins are due to monopoly power or collusion, a simple conversion of critical loss may mislead to an enlarged market definition. Then the starting price in calculating margins should be adjusted to a counterfactual competitive price. 12  hand, the acquirer, Moohak, was in a monopolistic position in Kyungnam area with 84.3% market share in 2001, while other producers' market shares were insignificant except Daesun's 12.9% share in the region.
The KFTC defined the geographic market relevant to Moohak-Daesun merger as diluted soju in Busan and Kyungnam. 13 Given this market definition, the combined company achieves 91.5% share in Busan and 97.2% share in Kyungnam, and the merger meets the conditions for presumption of competitive concerns in Article 7.4.1 of KMRFT. Moreover, according to KFTC, there exist de facto entry restrictions, although not de jure ones, in the soju industry; it takes long time and enormous costs in building up brand recognition in soju market. In addition, soju producers are not allowed to engage in wholesaling. Hence, new entrants have difficulties in establishing distribution channels, since incumbent distributors already maintain long-term relationships with Moohak and Daesun in the areas.
14 Moohak appealed against the KFTC decision to Seoul High Court. But the second court reaffirmed it. The case is now regarded as a landmark in antitrust enforcement in Korea in that two parties submitted economic analyses which supported their own views on the relevant geographic market, and the court resorted to economic analyses in reaching the decision (see [1]).
In order to address Moohak-Daesun case properly, we should understand the consumers' strong loyalty to their local products in Busan and Kyungnam. Daesun in Busan gained market share by more than 20% in 1997, while Jinro, which is a nationally dominant producer, lost market share by almost 20%. Daesun strengthened its market dominance afterward and maintained around 85% share in 2000s.   To understand the persistent dominance of local products and the huge shifts of market shares in these as shown in Tables 2 and 3, we should consider the history of regulation in local soju markets and the political power shift among regions in Korea. A regulation of mandatory purchase of local products more than 50% onto wholesalers had been introduced in order to protect local soju producers since mid-1970s. The regulation was abolished in 1992 and revived in June 1996 and finally declared illegal in December 1996 by the constitutional court. Interestingly, local characteristic has strengthened after the final abolishment of mandatory purchase of local products. First of all, local producers, confronted with more competitive pressures since mid-1990s, made greater survival efforts. Especially Daesun initiated such efforts by lowering prices and introducing new products with less alcoholic contents. 15 Moreover, the regionalism in Busan and Kyungnam became stronger after the shift of political power from their regions to another. 16 It seems that people in these regions became more cohesive in their regional compassion, and such political atmosphere strengthened the consumers' loyalty to local soju products. 17 Generally the small but significant price increase in SSNIP test is in the range of 5-10%. But when there are a large number of consumers who are loyal to a given product, we may have to consider higher price increases as well. To be more precise, consider the following numerical example. There are 100 consumers in a region who have unit demands for a product. The current price for the product is $100, and the cost is $70. All consumers are now using the product. There are two groups of consumers: 30 price-sensitive consumers who will switch to another substitutable product if the price increases by 5% and 70 loyal consumers who stick to the product unless the price increases by more than 20%. In this situation, a hypothetical monopolist of the product cannot make more profits by a price increase of 5% or 10%: Of course, the monopolist will increase the price by 20%. In that case, the market should be confined to the product and not be extended further. However, if we considered only 5-10% range of price increases in the SSNIP test, we might end up with extending the market by including next available substitute products or regions. Given the possibility that the hypothetical monopolist can exploit market power by a large price increase such as 20%, the danger of expanding the relevant market is serious.

Critical loss analysis
To determine whether Busan and Kyungnam are the geographic market of diluted soju relevant to Moohak-Daesun merger, we have to estimate and compare the critical and actual loss of regional sales corresponding to various levels of SSNIP.

Estimation of margins and critical loss
Proper margins, in an economic sense, are the difference between price and marginal cost. But average variable cost is used for marginal cost in practice because of measurement difficulties: The rate of operating profits, 27.1%, is considerably high in comparison with the food and drink industry average of 7.3% as well as the whole manufacturing industry average of 6.7%.
To convert operating profits into margins, we should deduct fixed parts from sales costs and marketing and administration costs in Table 4. The fixed components in sales cost are "rent" and "depreciation," and those in marketing and administration cost are "rent," "depreciation," "intangible assets deduction," "taxes and charges," "insurance," and "membership fees." Deducting these fixed costs from total costs, we can estimate the margins of Moohak and Daesun in 2002 (  Source: Companies' annual report. The other party on Moohak's side commented that the estimated ratio of margins, 31.9%, was misleadingly too low. They contended that fixed components should be defined as those which do not depend on operation level in one year and accordingly regarded "wages and salaries," expenses for "training," "advertising," and "maintenance" should be regarded as fixed costs as well as "rent," "depreciation," "intangible asset deduction," "taxes and charges," "insurance," and "membership fees." 19 But Seoul High Court decision in 2004 made it clear that those costs such as labor, advertising, and maintenance are not easily regarded as fixed during the significant period over which monopoly power can be exercised (the relevant time horizon should not be confined to the period of one year).
We will consider high levels of SSNIP such as 15% and 30% as well as conventional 5% and 10%. Such consideration is warranted by heterogeneous composition of consumers with loyal majority and price-sensitive minority in Busan and Kyungnam. As shown by the previous numerical example, dominant local companies may disregard price-sensitive consumers and employ a high price strategy for loyal consumers in the circumstance. The possibility of high price strategy has significant implications on geographic market definition discussed in the following.

Estimation of actual sales loss and geographic market definition
We used a survey data of consumers' choice of soju products in Busan and Kyungnam, estimated consumers' purchase substitution in response of price changes, and consequent actual sales loss. The sample size is 1042, and the sampling error is 3.03. 20 The number of consumers whose favorite soju was either Daesun or Moohak was 945. They were questioned whether they would switch consumption into Jinro if the price of "Dasesun's C1" or "Moohak's White" relative to the price of "Jinro's Chamiseul" increased in each of two places-"dining/ drinking houses" (e.g., restaurants or pubs) and "retailing shops" (e.g., convenience stores or supermarkets). The amounts of soju consumption in two places are said to be comparable in terms of quantities. But soju prices in the former are almost three times higher than those in the latter. Also it may be the case that consumption behavior might be different in two places since people usually drink together with others in the former while buying for in-house consumption in the latter. However, it turned out that the results based on the data of the former were very similar to those on the data of the latter. This analysis starts from an integrated region of Busan and Kyungnam, and consider whether the relevant geographic market should be enlarged further or not. If a hypothetical monopolist in Busan and Kyungnam could increase profits by an SSNIP, then the geographic expansion of the relevant market is not necessary. If that is the case, it is not necessary to consider whether the market should be separated into each of Busan and Kyungnam since anticompetitive concerns on Moohak-Daesun merger are serious enough as long as the relevant market is local, regardless of whether the relevant market is Busan and Kyungnam separate or combined. We obtain similar estimates of actual losses and the same results of critical loss analyses, with data of retailing shops.
The above results show that a hypothetical monopolist in Busan and Kyungnam region could not increase profits by a low SSNIP of 5%, but could do so by higher SSNIPs such as 10%, 15%, and even 30%. This is because there are two groups of consumers in the region-a large group of price-insensitive consumers who are loyal to local products and a small group of pricesensitive ones; a monopolist can opt for a high price strategy to exploit loyal consumers, taking the risk of losing price-sensitive customers. To conclude our critical loss analyses, our results indicate that the geographic market of diluted soju product relevant to Moohak-Daesun merger is confined to the local area within Busan and Kyungnam province and not extended to the country as a whole.
How sensitive are the above results to the breakdown of loyal and price-sensitive consumers? To address this question, it is helpful to interpret the critical analyses above in a reverse way. From Tables 7 and 8, we know that the critical percentage of "strongly" loyal consumers who would stick to local soju unless the price increase is higher than 30% is 51.5%. On the other hand, the actual percentage of such loyal consumers is 65.5% with the data of dining/drinking houses and 56.6% with the data of retailing shops. If we regard the consumers who would stick to local soju unless the price increase is higher than 10% as "broadly defined" loyal consumers, then the critical percentage of such loyal consumers is 76.1%, while the actual percentage of loyal consumers is 80.1% with the data of dining/drinking houses and 79.0% with the data of retailing shops. Hence, regardless of the criterion of loyalty, and the place of consumption, the hypothetical monopolist has the sufficient percentage of consumers to exploit their loyalty with price increases higher than 10%.
We can confirm the locality of the relevant geographic market with the complimentary analysis of LIFO-LOFI indexes. 21 LIFO defined by regional production's share in regional total consumption for a give product is 94.6%, while LOFI defined by regional consumption's share in regional production for a give product is 99.3%. A high ratio of LIFO means "Little In From the Outside," while a high ratio of LOFI means "Little Out From the Inside." It is a rule of thumb that LIFO and LOFI about as high as 75-90% are regarded as implying the establishment of regional market.

Implications for anticompetitive effects of Moohak-Daesun merger
The market definition in antitrust cases is a starting point of analyzing anticompetitive effects of mergers and consequent dominant position. The Korean competition law presumes "a combination of enterprises" as "practically suppressing competition in any particular business area" if all of the following conditions in Article 7.4 are met: a. The combined company is in a dominant position in the relevant market; i.e., its market share is 50% or more, or CR-3 is 75% or more except that its market share is less than 10%.
b. The combined market share is the largest in the relevant market.
c. The difference between the combined market share and the next largest market share is 25% or more.
Denoting the largest, the second largest, and the third market share by s 1 , s 2 , and s 3 , respectively, we can recapitulate the above presumptive conditions as follows: the combined market share should be s 1 and should satisfy either (i) or (ii):   If the relevant market is confined to Busan, Kyungnam, or the integrated region of BS + KN, it is obvious that the merger meets the conditions of presumption on suppression of competition according to KMRFTA. In fact, the market gets close to monopoly. On the other hand, if the market is national, then the merger does not belong to even the range of concerns about possible restraint on competition according to KFTC's guideline. Competitive concerns about Hite-Jinro merger and problems of the relevant market definition were twofold: one for a merger between Hite beer and Jinro soju and another for a merger between Hitejujo soju and Jinro soju. For the first one, the parties opposing the merger, especially OB beer which almost halves Korean beer market with Hite, alleged that beer and soju are substitutes so close that they constitute a single product market relevant to the merger, the so-called pub alcoholic drink. If that is the case, the merger would be very difficult to go through since it would be a horizontal merger with a significant increase in concentration ( *Hite beer + Hitejujo soju. **OB beer + Cass beer (both brands belong to the same company). ***All others are local soju producers such as Moohak, Daesun, and Keumbokju. Source in [11]. Sales are in terms of net sales amount before taxes. Table 10. "Pub alcoholic drink" market in 2004. 22 Hite and Jinro sell spring water, but the market is very competitive in that they are just two among many producers with market shares of 6.2% and 10.5%, respectively. No competitive concerns were raised in that regard. On the other hand, if beer and soju are regarded as separate products, the merger is basically conglomerate. In that case, there still may remain some anticompetitive concerns since both beer and soju companies share the same channels of wholesale distribution. But the anticompetitive allegations will be sagging.

Market definitions in
The second horizontal part of the merger was not a big issue since Hitejujo was a relatively small company. But an interesting issue here was whether the relevant geographic market was confined to the local province of Chonbuk or extended nationwide. Recall that the geographic soju market relevant to Moohak-Daesun merger was confined to the local areas of Busan and Kyungnam. If the geographic market was defined locally in this case too, e.g., as Chonbuk, then the horizontal merger would create a virtual monopoly in the region; the combination of Hitejujo and Jinro would have a market share of 92.8% in Chonbuk. On the other hand, if the relevant market is national, market concentration does not change consequentially.
In this case, Jeon et al. submitted to KFTC economic analyses on behalf of Hite (see [2]). They defined the relevant product market for the first conglomerate case as two separate markets of beer and soju and the relevant geographic market for the second horizontal one as the national market in soju excluding some southern regions. On the other hand, Ryu and Yi in [10] in behalf of OB Beer Company contended that the relevant product market was a single market of beer and soju, the so-called pub alcoholic drinks. Interestingly, both parties applied the same method of market definition-critical loss analysis. However, their estimates of actual and critical losses were different, which led to conflicting conclusions on the relevant product market definition.
Reviewing both parties' economic analyses, Korea Fair Trade Commission adopted the market definitions by Jeon et al. KFTC final decision in 2006 was to allow the merger with some behavioral remedies attached. The remedies included a price cap of RPI + 5% on Hite and Jinro's beer and soju, the division of marketing workforce and organization of Hite and Jinro for five years, and the provision of some arrangements by Hite itself that would ensure it not to commit exclusionary practices in the future.

Critical loss analysis for product market definition
In the following, I summarize the part of Jeon et al.'s analyses on the definition of product market relevant to Hite-Jinro merger.

Estimation of margins and critical loss
We can apply the same method of calculating margins of soju and beer industries as was introduced in the previous section. The rate of margins in soju industry was calculated with income and cost statements of Jinro in 2003 and 2004 ( Source: Jinro's annual report. Source: Companies' annual report. The estimated rate of margins for beer industry, 27.8%, is slightly lower than that for soju industry, 29.9%.
On the other hand, Ryu and Yi in [10] estimated the rates of margins for soju and beer industries as 52.6% and 53.2%, respectively. The difference comes from their classification of fixed costs and, more fundamentally, their perspective of "nontransitory" period in SSNIP. They contended that fixed components should be defined as those which are fixed regardless of operation level within the period of one year, and accordingly regarded "wages and salaries," expenses for "training," "advertising," and "maintenance" should be regarded as fixed costs as well as "rents," "depreciation," "intangible assets deduction," "taxes and charges," "insurance," and "membership fees." As noted before, Seoul High Court in regard to Moohak-Daesun case decided that those costs such as labor, advertising, and maintenance are not easily regarded as fixed during the significant period over which monopoly power can be exercised. Concurring to this decision, KFTC rebutted the high estimates of margins by Ryu and Yi; "they made an error of overestimating margins by regarding the time horizon dividing variable and fixed costs as one year rather than a significant period for exercising monopoly power, and consequently overestimating variable costs." Given rate of margins (M) and price increase (X), critical loss (CL) is derived from CL = (X/ (X + M)). For soju industry with M = 29.9%, critical losses corresponding to X = 5% and 10% are ( Table 13): Critical loss 14.3% 25.0% Table 13. Critical losses for soju industry.
For beer industry with M = 27.8%, they are ( Table 14): Critical loss 15.3% 26.5% Notice that we considered only 5% and 10% levels of SSNIP in this case. This is because we expect that consumer loyalty is usually associated to specific brands within a product, but not to a product as a whole. That is, consumers may be loyal to some specific brand in comparison with all other brands in a product, but not loyal to a specific product in comparison with all other products. Hence we do not expect that there are a small group of price-sensitive consumers and a large group of loyal consumers for soju product or beer product as a whole. Moreover, we have to consider high level of SSNIP only if a hypothetical monopolist cannot make profits with 5-10% price increases. But the hypothetical monopolist can make profits with the normal 5-10% SSNIP in the present context, and we do not have to consider a higher level of price increase.

Estimation of actual sales loss and geographic market definition
Jeon et al. used Gallup Korea's survey data in estimating actual sales losses that result from increases in prices of soju and beer. Ryu and Yi instead used weekly date of sales in discount stores with bar codes. As KFTC decision noted, the data have a serious sample selection bias in that discount stores account for only 5% of total sales of soju and beer, and consumers who purchase soju and beer in discount outlets do not represent the whole population. Especially, the characteristics of consumers using discount outlets such as Carrefour may be different from those of usual consumers who buy soju and beer in drinking/dining places and other retail shops. On the other hand, a survey data can avoid such selection bias by constructing a sample which reflects the population of soju and beer consumers in terms of sex, age, education, job, income, region, etc. Gallup Korea conducted the survey with a sample of 1603 soju consumers and 1547 beer consumers in such a way that avoided selection bias. These results show that product markets relevant to Hite-Jinro merger are two separate ones of soju and beer, not an integrated one of "pub alcoholic drink." A hypothetical monopolist of soju (beer) product can implement 5% and 10% price increases profitably. This means that soju and beer are not so close substitutes that both constitute a single product market in antitrust perspective.
Reinterpreting the actual losses in Tables 15 and 16 in terms of elasticities, the price elasticity of soju is about 1.1 and that of beer is about 2.2-2.6. Previous empirical studies using actual data of soju and beer consumption shows the robustness of the results based on survey data. 23 Chung in [12] found that the own price elasticity of soju is in the range of 0.58-1.18 and that of beer is in the range of 0.94-1.31. And another Chung in [13] estimated the price elasticities of soju and beer as about 0.75-0.85 and 1.3-1.6, respectively. Given these estimates, our survey results seem to overstate consumers' response to price increases a little bit. However, the differences may not be so huge as to discredit the survey results after all. Furthermore, the critical loss analyses with the empirically estimated elasticities would support the conclusion even more strongly that soju and beer markets should be defined separately.

Critical loss analysis for geographic market definition
Given that soju and beer are separate products from a perspective of competition policy, Hite-Jinro merger is basically conglomerate. However, the merger contains a horizontal part since Hite has a subsidiary soju company, Hitejujo, in Chonbuk. Competitive evaluation of the horizontal part hinges on definition of the relevant geographic market. In the following I summarize the part of Jeon et al. on the definition of geographic soju market relevant to the horizontal merger between Hitejujo and Jinro. The starting point of analysis is Chonbuk area for which competitive concerns about the merger might be raised, and then it will be checked whether the relevant market should be enlarged further.

Estimation of margins and critical loss
We estimated the margins in the same way as before, in this case with Hitejujo and Jinro which account for more than 90% of sales in Chonbuk-42.5 and 50.6% in 2004, respectively ( Source: Companies' annual report. The margins of Hitejujo were lower than those of other soju producers, which was due to large expenses of marketing and administration. But the margins of the combined company were close to the average of soju industry since Hitejujo was very small in comparison with Jinro. Critical losses, CL = (X/(X + M)), for M = 29.4% and X = 5%, 10%, 20%, 30%, and 40% are as follows (Table 18): Critical loss 14.5% 25.5% 40.5% 50.5% Table 18. Critical losses for the combined company of Hitejujo and Jinro.
We consider here high levels of SSNIP such as 20 and 30% in order to see whether there are a considerable number of consumers who show strong loyalty to local products as in the Moohak-Jinro case.

Estimation of actual sales loss and geographic market definition
Gallup Korea conducted a survey in 2005 with consumers in Chonbuk area for the analysis of their choice of soju products. The sample was selected to represent the average behavior of soju consumption in terms of sex, age, and regions, and the final 810 consumers were screened who responded that they usually consumed "Hite 2" (Hitejujo's brand name) or "Chamisle" (Jinro's brand name). 24 The results of critical loss analyses are (Tables 19 and 20 These results show that Chonbuk consumers would switch their purchase of soju from their current main favorite brands to others to the extent that the relevant geographic market should not be confined to the Chonbuk region.
It still remains the issue of how far the geographic market should be enlarged, i.e., whether it is the country as a whole or some regions are excluded. To make a definite conclusion on this issue, we have to conduct further critical loss analyses. Instead, we made a tentative suggestion that the geographic market might be nationwide, excluding some southern regions such as Busan, Kyungnam, Kyungbuk, Chonnam, and Jeju. The main reason is that each of those regions has a dominant local producer and consumers with strong loyalty to a local product. Moreover, their current soju sales in Chonbuk are negligible. The KFTC also defined the geographic market relevant to the merger between Hitejujo and Jinro as "the country except Busan, Kyungnam, Kyungbuk, Chonnam, and Jeju." It is to be noted that the market definition in an antitrust case is case-specific. In other words, it depends on a starting point of analysis, which is a product or a region for which competitive concerns are raised. The previous market definition in Moohak-Daesun was confined to Busan and Kyungnam since consumers in the region did not switch much their purchase from local products to products in other regions, e.g., Hitejujo and Jinro, in response to local price increases. On the other hand, there would be nothing wrong if the geographic market relevant to Hitejujo-Jinro had been extended to the whole country even including Busan and Kyungnam, even though it did not actually go that far. This does not involve any inconsistency, since consumers in two regions could have different preferences.
Besides conducting critical loss analyses, we observe two facts that differentiate Hitejujo-Jinro from Moohak-Daesun. First, LIFO index in Chonbuk was only 42.5%, while that in Busan-Kyungnam was 94.7%. Compared to the conventional standard of 75-90% of LIFO-LOFI test, 40% must be too low. Second, there is a common pricing constraint in soju industry in that producers should apply the same wholesale price in the country. This suggests another basis of the national market in this case. One of the merging companies, Jinro, is a national producer which cannot raise soju price in Chonbuk without risking sales losses in other regions. On the other hand, both Moohak and Daesun are local producers which had virtually no other regions to consider in setting prices.
It is worthwhile to elaborate a bit more on why opposite conclusions were arrived at with respect to expanding the geographic market in the two cases of Moohak-Daesun and Hitejujo-Jinro. Notice that we tried to apply the same methods of critical loss analysis: e.g., the same classification of variable vs. fixed costs in deriving margins and the same construction of questionnaire in conducting surveys. Hence, the opposite results in two cases were tied to the idiosyncratic nature of the two geographic regions rather than the application of the test. People in Busan-Kyungnam have strong loyalty to their local brands, which have been rooted in the past history of local purchase requirement regulation and the political atmosphere aforementioned. On the other hand, people in Chonbuk are not so loyal to their local soju firms which are not indigenous anymore. Hitejujo and Jinro are hardly regarded as Chonbuk-based since Hite, a national beer company, acquired the former indigenous Chungbuk soju and Jinro is a national soju company. Furthermore, political regionalism in Chonbuk has not been as keen as Busan and Kyungnam since Chonbuk has never been a power center in recent Korean political history.

Implications for anticompetitive effects of Hite-Jinro merger
Since beer and soju are separate products, Hite-Jinro merger is a conglomerate one. KFTC considered four possible anticompetitive effects of the conglomerate merger in this case: (i) excluding competitors, (ii) strengthening entry barriers, (iii) limiting potential competitors, and (iv) raising prices. To alleviate concerns, the merger was given conditional approval. The imposed conditions of corrective measures included a price cap of RPI + 5% on Hite and Jinro's beer and soju, division of marketing workforce and organization of Hite and Jinro for five years, and some self-arrangement of not committing exclusionary practices in the future.
The first two concerns stem from the fact that beer and soju producers share the same distribution channels of liquor wholesale: Hite and Jinro accounted for 34.5% and 22.1% of liquor wholesalers' sales in 2004. It was alleged that the combination of two dominant companies in beer and soju could enhance its bargaining power against wholesalers and press them to influence final demands in favor of its brands. Moreover, strongholds of the two companies were different; Hite was dominant in southern provinces while Jinro was in Seoul and its adjacent regions. So the leverage effect of expanding monopoly powers across regions through tying or bundling was worried about by competitors of Hite and Jinro in beer and soju markets. 25 On the other hand, the defending party argued that pushing wholesalers or leveraging cannot be effective in the long run and that the ultimate determinant of final demands is consumer preference. 26 Jeon et al. conducted an econometric study which rejected the existence of leverage effects in previous mergers in beer and soju market (see [2]). The third concern of eliminating potential competitors is an often-claimed anticompetitive effect of conglomerate mergers. Lastly, the concern about raising prices is peculiar for conglomerate mergers. KFTC considered beer and soju as exerting some competitive constraints, even though it defined the two as different products. This sounds contradictory since market definition is nothing but a consideration of competitive constraints. A more practical basis for this concern was that the acquirer expended too much-over $3 billion dollars for the deal-and that it could not but increase product prices in order to resolve financial difficulties. Given the sunken nature of financial costs for the merger deal, the last concern does not seem to be warranted.
On the other hand, the KFTC did not consider the horizontal merger between Hitejujo and Jinro as restraining competition materially in the soju product market. Given the geographic market definition of the country as a whole except for five regions (Busan, Kyungnam, Kyungbuk, Chonnam, and Jeju), the market share of Hitejujo was merely 2.5%. The KFTC's merger guideline stipulates that anticompetitive concerns are insignificant if the gain of market share after the merger is less than 5% in the relevant market.

Recommendation
Several issues have been raised in the course of applying critical loss analysis in merger cases of Moohak-Daesun and Hite-Jinro. First of all, estimation of margins using accounting data involves some degree of discretionary or ad hoc classification of fixed and variable costs. The parties who defend an enlarged market definition argue for more fixed costs, and vice versa; it is because the larger the portion of fixed costs, the higher the rate of margins, and hence the lower the level of critical loss. The most controversial and significant components are labor costs and advertising expenses. The arguments of those who regard them as fixed are as follows: the time horizon in the SSNIP test is one year, and hiring regular workers and expensing advertising budgets do not vary in accordance of output changes within one year. But Seoul High Court and the KFTC interpreted the time horizon appropriate for SSNIP test as a significant period over which monopoly power can be exercised, which is not be confined to only one year. Given that, labor and advertising costs are not necessarily fixed.
Second, we used survey data in estimating price elasticities of consumer demand and actual losses corresponding to various levels of price increase in the SSNIP test. Economists usually prefer using actual historical data in estimating demands rather than resorting to survey data. However, in many cases we have data problems; data with the necessary degree of desegregation, time span, and representativeness are not available. Recent spread of bar code scanning 25 See [14] for discussions of portfolio effects. 26 Recent development in soju market seems to confirm this argument. Doosan recently emerged as a strong rival to Jinro with more than 10% market share in Seoul and its vicinities. system in retailing shops, and consequent availability of POS (Point of Sale) data, is a promising development that may resolve data problems in the future. 27 But there are still many cases, such as Moohak-Daesun and Hite-Jinro, where POS data have a serious problem of sample selection bias in that they constitute a small portion of population data, and sample characteristics do not reflect those of the entire population. In such cases, a well executed survey can be a useful alternative source of data. Survey data are often discredited because questions tend to predetermine answers and that respondents tend to overstate (or understate) their responses. But questions in our survey are so straightforward that they do not risk predetermining answers; the questions simply ask how consumers will change purchases in cases of price increases. There still remains a problem of overstatement (or understatement) tendency. We have to take it into an appropriate account in drawing conclusions. Our results had considerable margins so that the conclusions were robust even after accounting such a tendency.
Third, we considered high price increases of 15 and 30% in the SSNIP test as well as conventional 5 and 10% in the Moohak-Daesun case. It was due to the fact that there were two groups of consumers in the regions-a price-sensitive one and another with loyalty to local brands. In such case, there is a concern that a local monopolist can exploit captured brand-loyal consumers by a high price strategy even though he cannot make profits with low prices. Of course, the usual criterion of 5-10% in the SSNIP test will be sufficient in most of cases where there are not many brand-loyal consumers.
Fourth, there is a subtle issue of how we should account income effect in price effects in hypothetical monopoly test. If the spirit is to include close demand substitutes in a relevant market, then we may have to focus substitution effect in price effects and to give less weight to actual losses due to income effect. On the other hand, if the spirit is to identify a set of products and regions for which a monopolist can exercise market power, then we have to consider both substitution and income effects equally. US and EU guidelines are not clear about which is the right perspective, while the current version of the KFTC merger guideline seems to be based on the former.
The last remark is on case specificity and possible asymmetry of market definition. That is, market definition in antitrust cases is case-specific; it depends on the starting point of analysis, from which we check whether the relevant market should be enlarged further. There would be nothing wrong if the geographic market relevant Moohak-Daesun was confined to Busan and Kyungnam, while the market relevant to Hitejujo and Jinro had been extended to the whole country including Busan and Kyungnam. Similarly, there would be nothing wrong if a critical loss analysis starting from soju product implied that soju was a separate market while an analysis starting from beer product had implied that beer and soju were in the same market, even though it did not turn out that way. Such possibilities do not involve any inconsistency, since consumers in two regions, or of two products, could have different preferences, and their consumption behaviors could result in asymmetric cross-elasticities.

Further study
Critical loss analysis is a convenient tool for practical market definition, and it proved to be very useful and successful in the two cases examined in the paper. Unfortunately, however, there are not many industries for which the analysis can be actually implemented. Soju and beer industries in Korea are rather special in that companies in these industries focus on basically one liquor business, which enable us to calculate the ratio of margins with public accounting data. Also there are not many different products in the industries, for which we can survey consumers' purchasing behavior. On the other hand, it is practically impossible to conduct critical loss analysis for industries with too many businesses and products, for example, retailing or banking industry. For those industries, the SSNIP test will remain just as a conceptual framework without a quantitative support of critical loss analysis.