Standard and Premium Soju

1.214. Korea points to another peculiarity in the treatment of standard and premium soju. Korea observes that in the Dodwell report, the premium soju is offered as a choice side-by-side with western-style spirits in the pairwise choice between white spirits and standard soju and in the pairwise choice between brown spirits and standard soju. Thus, Korea notes, in the Dodwell "brown spirits" test, respondents are offered a choice between scotch, cognac and premium soju on the one hand, and standard soju on the other; and in the white spirits test they are offered a choice between vodka, gin, rum, tequila, liqueur, and premium soju on the one hand, and standard soju on the other.

1.215. Korea argues that grouping soju with western-type spirits rather than with standard soju is eccentric, and suggests a lack of familiarity with these two soju products. According to Korea, standard and premium soju are regarded as close substitutes in Korea. Korea states that premium soju is more expensive than standard soju-the "current price" of premium soju in the Dodwell report is 2,400 won as compared to 1,000 won for standard soju. Korea notes, however, that is hardly a compelling case for putting premium soju in the same group as 20,000 won scotch or 32,000won cognac.¹²⁴

Problems

1.216. Korea states that studies of the Dodwell type must be carefully designed if they are not to fall foul of bias. Korea points out that the Dodwell report clearly has not managed this. Korea discusses two sources of bias and claims that either is sufficient to eliminate the credibility of the Dodwell Report.

Choice of Respondents

1.217. Korea points out that no reason is given for the exclusion of males above the age of 49 or of the rural population when the Dodwell report states "It was decided to select as respondents for the research 500 Korean men in 3 Korean cities aged between 20 and 49 who have purchased soju in the last month."¹²⁵

1.218. Korea states that no information is provided on how and where the same was selected, or on how its members were induced to spend their time answering the survey questions. Korea notes that no information is given on the drinking habits of respondents-for example, where they typically purchase alcohol, or how much they consume(although questions 5-13 deal with these matters). Korea further notes that no information is provided on the actual characteristics of the sample-for example, average age, occupation, income-that would permit a check on whether the sample is even representative of "Korean males aged between 20 and 49 who have purchased soju in the last month."

1.219. Korea argues that the restriction and exclusion of the rural population are not the only ones placed on the sample. Korea indicates that p. 3 of the report announces that "additional criteria for the respondents were whisky purchase in the last 3 months." Korea notes that no reason is given for rejecting persons who have not purchased whisky in the last three months. Korea conjectures that, possibly, the results of the study in the absence of this condition were disappointing. Korea notes that it is certainly true that a good way to obtain a strong response to a fall in the price of whisky and a rise in the price of soju is to select a sample of persons that drink both soju and whisky(and, of course, a whisky purchase in the last three months and a soju purchase in the last month is consistent with a purchase of both in the last week, day or hour).

1.220. Korea states that the effects of selection bias are evident, regardless of the motivation for the additional restriction,. Korea points out that in the Dodwell sample, at current prices, 72 per cent of respondents select soju when asked to choose between brown spirits and soju; and 72 per cent select soju when asked to choose between white spirits and soju (Korea refers to the Dodwell study, p. 6). Korea notes, however, that the actual share of soju in consumption of distilled spirits in Korea, however, is about 95 per cent.

1.221. Korea asserts that the discrepancy between those who prefer soju at current prices in the Dodwell study and the true figure is very large¹²⁶ Korea nevertheless states that the difference between 72 per cent of the sample and 95 per cent that choose soju in reality is likely to seriously understate how unrepresentative the sample is. Korea attributes the discrepancy to the fact that respondents were given a pairwise choice between soju and brown spirits, and then a pairwise choice between soju and white spirits. They were not offered a three-way choice between soju or brown spirits or white spirits.

1.222. Korea acknowledges that some respondents may choose brown spirits rather than soju from that pair, but soju rather than white spirits when offered that pair; while others might select white spirits rather than soju in that pairwise offering, but soju rather than brown spirits. Offered a three-way choice, all of those who reject soju in the pairwise choice will continue to reject soju. Korea, however, emphasizes that those who chose white spirits when offered a choice between soju and white spirits need not be the same as those who choice brown spirits when offered a choice between soju and brown spirits. Korea states that it follows that the figure of 72 per cent choosing soju rather than brown spirits in the Dodwell sample, and the similar percentage selecting soju rather than white spirits, gives the highest possible percentage of those in the sample who would have chosen soju had they been offered a three-way choice.

1.223. Korea points out that the actual percentage choosing soju in a three-way choice would be 72 per cent only if all those who prefer white spirits to soju also prefer brown spirits to soju and vice versa. Korea notes that at the opposite extreme, though, if all of those choosing whisky over soju prefer soju to white spirits; and if all of those choosing white spirits rather than soju prefer brown spirits, the percentage opting for soju in a three-way choice would be 44 per cent-28 per cent would reject soju in favour of brown spirits, and 28 per cent would reject soju for white spirits. Korea states that only 72 per cent of Dodwell respondents choose soju at current prices indicates the unrepresentative nature of the Dodwell sample: that a lower figure is avoided only by the failure to offer a three-way choice; and that in such a choice, the number choosing soju might fall as low as 44 per cent, puts the Dodwell report into a world entirely different from that of Korean reality.

1.224. Korea argues that the Dodwell study is based upon a sample of persons who are strongly biased towards western-type spirits, relative to the Korean population as a whole. Korea asserts that the Dodwell sample is not a credible sample. Korea states that, even leaving aside other grounds for doubt, the response to hypothetical price changes of a group so unrepresentative cannot be taken to reflect anything of the responses of the Korean population as a whole to real price changes.

The Single Choice Drink

1.225. Korea indicates that the Dodwell study respondents are confronted with an either-or choice. Korea notes that according to the Dodwell script, interviewers say: "As you can see, there are five types of spirits and photos of typical brands of these types. Which spirit would you choose at these prices?"

1.226. Korea states that reliable information about how different prices might change drinking habits-that is, whether a person who is a regular soju drinker might switch to becoming a regular whisky drinker were the price of soju higher and the price of whisky lower-might be relevant in this case. However, Korea argues, that is not what the Dodwell interviewers ask about. Korea notes that they ask which bottle a respondent would choose at the different prices. Korea points out in the Lexecon/Hindley Report that respondents might perfectly naturally interpret the Dodwell questions asking: "If you saw these prices the next time you bought a bottle of spirits, which bottle would you choose?" Korea notes that such questions opens the possibility that some respondents interpreted the hypothetical prices as a one-time offer: "If you saw these prices the next time you bought a bottle of spirits, but knew that usual prices would be back in force the time after that, which bottle would you buy? " Korea emphasizes that some Dodwell respondents may simply be saying that they would try a bottle of high-price cognac were it temporarily on offer at such a low price.

1.227. Korea argues that respondents interpreting the question as asking "would you try a bottle of cognac if it was offered at this price?" are almost certainly more likely to answer affirmatively than those interpreting the question as asking "would these prices cause you to change your drinking habits?". Accordingly, Korea concludes, the ambiguity in the question almost certainly increases the number saying they would buy a bottle of brown or white spirits at a lower price.

1.228. Korea emphasizes that there is absolutely no reason to suppose that respondents are speaking about a change in their drinking habits-that their answers imply that if soju rose in price from 1,000 won to 1,200 won, they would switch their regular drink, during meals for instance, from soju to cognac at 32,000 won or scotch at 20,000 won.

1.229. Korea argues that market surveys, carried out specifically for the purposes of legal proceedings at the request of an interested party, are, of course, to be treated with caution. The analysis must be rigorous, and bias must be avoided at all cost. According to Korea, the Dodwell study does not meet these standards. Korea sums up what it perceives to be the most glaring defects of this study:

1.230. Korea further argues that in contrast, more credence can be given to a study, which was not prepared specifically for regulatory purposes, but which tries to explain in objective terms to exporters the situation on the Korean market. Korea refers to a recent report initiated by the European Commission recently, which stated:

1.231. Korea asserts that the constant and independent demand for soju is not the result of any protective government policies. Korea refers to the same EC study which notes that:

(e) Product-by Product Analysis

1.232. Korea submits that as competitive relationships differ from product to product and from market to market, the United States and the European Communities bear the burden of proving for each individual product combination that a 'like' or 'directly competitive and substitutable' relationship exists in the Korean market before they can put the applicable tax rates into question.

1.233. Korea states that, without assuming the complainants' burden of proof, it will demonstrate the failure of the complainants to discharge this burden in the following way:

1.234. Korea, therefore argues that both diluted soju and distilled soju must be compared individually to each of the imported liquors in question. Further, it states that although premium soju is a variation of diluted soju, Premium's price is somewhat higher, though still far below the price of the imported liquors. Premium soju represents only a small volume of diluted soju sales (currently, around 5%). In the discussion below of diluted soju due account is taken, where necessary, of any of premium's special features.¹³⁰

1.235. Korea notes that the only products that the United States and the European Communities have alleged are 'like' are standard soju and distilled soju and vodka. Korea states that it will therefore only make representations about the lack of a 'like product' relationship as far as vodka is concerned. In its view, it goes without saying that Korea does not accept that any 'like product' relationships exist in this case.¹³¹

1.236. Korea also seeks to point out the very considerable price differences that exist between the imported liquors and diluted soju. According to Korea, the complainants recognize these differences at actual market prices.¹³² However, Korea argues that these price differences remain considerable, even when the disputed taxes are eliminated. Korea points out that this is shown by the complainants' own expert study, the Dodwell Study.¹³³ Korea's position is that although it contests the results of the Dodwell Study, the raw price data provided in that study appear to be generally correct. Korea feels that these data are so compelling that it has not felt it necessary to go beyond the data set forth by the complainants. In its view, the one exception is whisky, where, given its importance to this case, Korea has supplemented the Dodwell data with its own figures.

1.237. In short, Korea is seeking to show that an inexpensive local meal drink such as diluted soju is not in direct competition with expensive western-type liquors. In the alternative, Korea is seeking to show why the complainants have not shown that Korea's tax system meets the other criteria of Article III:2, assuming that the Panel would still find a competitive relationship between some products,

1.238. Korea states that contrary to what the complainants are alleging, the so-called soju based cocktails are not soju. According to Korea, these are sweetened mixtures, with a low alcohol percentage (10-15%), that were introduced in 1994. They are not comparable to either standard or distilled soju. Korea adds that to make the distinction, manufacturers never use the word 'soju' in the brand names for these products. They are classified differently, like liqueurs, according to the liquor tax law. Korea points out this classification also covers such imported liqueurs as Bailey's, Grand Marnier, Kahlua, etc. Liqueurs are subject to a tax rate of 50%. It is unclear to Korea what, if any, complaint the European Communities and the United States are formulating in this respect.

1.239. Korea also argues that contrary to EC assertions, sales of soju-based cocktails did not increase by 1250% in 1995. According to Korea, the taxed volume of soju-based cocktails increased by 419% in 1995, from 1,583 kl (1994) to 8,218 kl (1995), and that in 1996 sales decreased by 8% (to 7,562); in 1997 by 22% (to 5,893) kl).¹³⁴

To continue with Diluted Soju

¹²⁴ According to Korea, taxation of premium soju is identical to that of standard soju, and the natural division between soju on the one hand, and western-style spirits on the other would on its face have given clearer results. Rejection of the natural division seems to be a major flaw is the design of the Dodwell study, and it would be interesting to know how it came about and why it was maintained.

Korea claims that, as it is, the outcome of the study is blurred by interactions between standard and premium soju, which account for much of the change reported. The first Dodwell results, for example, are based on holding at current levels the price of brown and white spirits(and premium soju) while the price of standard soju is hypothetically raised from 1000 to 1200 won. The reported result is that standard soju loses 5.2 percentage points to "brown spirits." Three of these points, however, are gained by premium soju. Against white spirits, standard soju loses 5 percentage points, but 4 are gained by premium soju.

According to Korea, if these results are treated seriously, as presumably they were by the authors of the report, they strongly suggest that standard and premium soju are close substitutes in consumption. That finding might have been expected to lead the authors of the Dodwell report to use the natural division. It is curious that it apparently did not have that effect.

¹²⁵ The Dodwell Report at p. 2.

¹²⁶ The figure of 72 per cent is used by the authors of the Dodwell report (p. 6). According to Korea, an alternative would have been to take the percentage of the sample choosing either standard or premium soju at current prices: 84.8 per cent then choose soju and 15.2 per cent brown spirits proper when offered that pair, and 86.2 per cent choose soju and 13.8 per cent western-type white spirits of that pair. This reduces one discrepancy between the Dodwell sample and actual market shares-the Dodwell percentage choosing brown spirits proper fall to only five times actual market share. But it also highlights other discrepancies. In the Dodwell sample 14.85 per cent of those choosing soju opt for premium soju: in fact, premium soju takes about 4 to 5 per cent of total soju sales. In the Dodwell sample, 13.8 per cent choose western-type white spirits, which have an actual market share of 1.8 per cent.

According to Korea, these discrepancies strongly suggest sample bias. The Dodwell report, though, does not provide enough information to translate percentages of Dodwell respondents into a figure directly comparable with actual market shares, so that it does not allow calculation of whether the comparisons above understate or overstate the degree of bias. One problem is that Dodwell respondents are asked to choose a single bottle: in fact, drinkers are likely to consume different drinks at different times. Another problem is that those choosing soju in the Dodwell sample, for example, might on average consume more soju than those choosing whisky consume whisky. The implied market share of soju would then be more than 72 per cent. To approach actual market shares, Dodwell soju choosers, if they drank nothing but soju, would have to consume a volume of soju about 13 times greater than Dodwell whisky drinkers, if they drank nothing but whisky.

¹²⁷ See Attachment 2 of Korea, which is a critique by Lexecon Ltd and Dr. Brian Hindley (London School of Economics) of the Dodwell Study.

¹²⁸ Sofres Report, p.22

¹²⁹ Ibid., p. 12.

¹³⁰ For instance, in giving average prices of standard soju, the higher price of premium is taken into account.

¹³¹ Korea is mindful of the fact that in the most recent Japanese liquor taxes case, a 'like' product relationship was found to exist between one product pair: Japanese shochu and vodka. However, given the differences between Japanese shochu and the Korean sojus, as well as the differences between the Korean and the Japanese markets, this holding is inapplicable to the case at hand, according to Korea.

¹³² Sofres Report, at p. 53 (1997), reproduced in Attachment 3.

¹³³ Pre-tax prices are provided in the Dodwell Study on page 20, in the column marked 'NET'. See Attachment 4.

¹³⁴ Source: National Tax Administration.