(iii) The 1995 Gemines Study

217. The European Communities explains that it has had access to a study entitled "The possible effects for the pisco industry of a reduction in the tax applied to whisky" carried out in August 1995 by Gemines, a respected firm of consultants, at the request of Chile's pisco industry (hereafter referred to as "the 1995 Gemines study"). ¹³³ This study provides further evidence of significant cross-price elasticity between pisco and whisky.
218. Further, the European Communities explains that the objective of the study was to quantify the effects for the pisco industry, and more generally for the economy of the zona pisquera, in the hypothesis that the ILA was amended so as to equalise the tax rates applied to pisco and whisky. Gemines considered two different scenarios. According to the first scenario, pisco and whisky would be taxed at 35 %. According to the second scenario, they would be taxed at 30 %.
219. According to the European Communities, as a first step, Gemines estimated the cross-price elasticity rate between pisco and whisky on the basis of historical sales and price data covering the period 1985-1992. The estimated rate was 0.26. This would indicate that if, for example, the prices of whisky went up by 10 %, the sales volume of pisco would increase by 2.6 %. ¹³⁴ According to Gemines, that rate of cross-price elasticity is sufficient to conclude that pisco and whisky are "substitutes, albeit to a moderate extent". ¹³⁵ By contrast, on the basis of similar regressions, Gemines reached the conclusion that neither wine nor beer could be considered as substitutable with pisco. ¹³⁶
220. The European Communities states that Gemines then proceeded to estimate the changes in consumption of pisco and whisky that would take place in each of the two scenarios above described. As shown in Tables 25 and 26 below¹³⁷, Gemines concluded that sales of pisco would drop by 10.2 % in the first scenario and by 8.6 % in the second scenario, whereas sales of whisky would increase by 5.8 % and 6.5 %, respectively.

Table 25 ¹³⁸

First scenario

Pisco and Whisky taxed at 35 %

	Pisco	Whisky
Price	+ 7 %	- 18.6 %
Quantity sold	- 10.2 %	+ 5.8 %

Table 26 ¹³⁹

Second Scenario

Pisco and whisky taxed at 30 %

	Pisco	Whisky
Price	+ 3.5 %	- 21.3 %
Quantity sold	- 8.6 %	+ 6.5 %

221. The European Communities comments that these figures may underestimate considerably the increase in whisky consumption. In the first place, Gemines did not take into account the additional demand generated by whisky's cross-price elasticity to increases in the price of pisco, but only and exclusively the elasticity of whisky to changes in its own price. Moreover, unlike in the case of pisco, Gemines did not estimate whisky's actual rate of own-price elasticity. Instead Gemines limited itself to "assume" a rate of 0.31 on the basis of "the characteristics that exhibits this market and alternative studies". ¹⁴⁰ That "assumed" rate, however, seems too low. In fact, that rate would be inconsistent with the cross-price elasticity rate of pisco in response to changes in the prices of whisky previously estimated by Gemines itself. On the basis of the latter rate (which itself appears to be an under-estimate), an increase of 10 % in the price of whisky would generate an increase in sales of pisco which is much larger (2.6 % of approximately 75 % of the spirits market) than the total drop in sales of whisky caused by the same price increase on the basis of Gemines' assumed rate of own-price elasticity for whisky (3.1 % of approximately 4 % of the spirits market).
222. The European Communities further notes that the authors of the study cautioned that in practice the decline in the sales of pisco was likely to be even greater than shown in the above tables due to the fact that in both scenarios the changes in the price of whisky were much bigger than those considered in order to estimate the cross-price elasticity rate between pisco and whisky.
223. In rebuttal, Chile argues that with regard to cross-elasticity of demand, the European Communities has been unable to demonstrate high cross-elasticity of demand. The study cited by the European Communities shows only a "moderate" substitutability between pisco and whisky, based on the 1995 Gemines study. A more thorough study over a longer period, conducted for the Chilean industry, as shown in Table 27 below, demonstrates an even lower degree of cross-elasticity. Such low cross-elasticity hardly demonstrates that the products are, in terms of Article III:2 "directly competitive or substitutable [emphasis supplied]".

Table 27 ¹⁴¹

Cross-price elasticity of pisco with other spirits, wine and beer

Results of the Regression
Statistics of the regression
Multiple coefficient correlation			0,9878
Coefficient of R2 determination			0,9758
Adjusted R2			0,9624
Typical error			0,0640
Observations			15
Analysis of the Variable
			Sum of the squares	Average of the squares	F	Critical Value of F
Regression		5	1,4895	0,2979	72,6767	5,32677E-07
Residual		9	0,0369	0,0041
Total		14	1,5264
Analysis of Coefficients
	Coefficients	Typical error	Statistic t	Probability	Inferior 95%	Superior 95%
Interception	3,5771	3,6554	0,9786	0,3534	-4,6920	11,8461
Variable X 1 (Income)	-0,0072	1,2109	-0,0059	0,9954	-2,7465	2,7321
Variable X 2 (Pisco Price)	-1,3109	0,4574	-2,8661	0,0186	-2,3456	-0,2762
Variable X 3 (Whisky) Price)	0,1248	0,5158	0,2419	0,8143	-1,0421	1,2917
Variable X 4 (Wine Price)	0,5963	0,4030	1,4796	0,1731	-0,3154	1,5079
Variable X 5 (Beer Price)	0,3622	1,2132	0,2985	0,7721	-2,3823	3,1067

224. Chile explains that to estimate the cross-elasticity between pisco and other alcoholic beverages, an econometric methodology analysis was developed on the basis of a time-series of data. It is worth mentioning that econometric models are widely accepted as a useful tool to determine whether two products are strong substitutes or not.
225. Chile goes on to explain that the econometric model reached a highly satisfactory adjustment at global level, and the coefficient values (elasticities) have the sign expected by the economic theory: an increase in per-capita income will lead towards an increase in per-capita pisco consumption; an increase in pisco price will lead towards a diminishing in pisco consumption. However the only variable that, from a statistical standpoint has a significant coefficient (with a confidence level of 95%), was the pisco price, with a value of -1.31. That means that a 10% increase in the price of pisco, will result in a decrease of 13.1% in pisco consumption. According to the results of this mathematical model, the cross elasticity between pisco and whisky is not only very low (0.125), but statistically not significant (with 95% of confidence). An elasticity of 0.125, means that a 10% increase in the price of whisky (for whatever reason), will increase pisco consumption by 1.25%. But this elasticity, being so low that these products cannot be considered close substitutes to each other, is also not significant from a statistical standpoint, that means the elasticity could be 0.
226. According to Chile, in order to deepen this analysis, a second regression was made, but in this case the per capita income variable was eliminated from the equation. It shows an improvement, and the coefficients maintain the signs that the theory indicates. In this case, pisco and wine prices appear as significant variables (with 95% of confidence), with the cross-price elasticity of wine being 0.59. In this case, whisky also presents a low elasticity (0.128), which is not statistically significant (with 95% of confidence).
227. Further, Chile states, that in order to investigate further, the price of beer was eliminated from the second regression analysis, since its elasticity coefficient is not significant (with 95% of confidence), even though it is higher than that for whisky. In this case, the equation of the regression only considers prices for pisco, wine and whisky. With this modification, the quality of the regression improves, and the coefficients maintain the sign theoretically indicated. In this regression, the price of pisco (elasticity = -1.50) and wine (cross elasticity = 0.78) remain as significant variables (with 95% of confidence), while the price of whisky (cross elasticity = 0.07) remains as not significant.
228. Chile then argues that based on the previous commentaries, it can be concluded that, using the econometric analysis (widely accepted at technical level), it is not feasible to demonstrate the existence of a significant cross elasticity of demand between pisco and whisky.
229. Chile further explains that the methodology used in the above regression (demand model with constant elasticity, estimated under the ordinary least squares) is identical to the one employed by Gemines in the 1995 Gemines study. In this case, the results differ from those indicated in the Gemines study because the price series used in one case and in the other are different: in the case of the Gemines study, a quarterly series of prices was used, which covers the period 1985 - 1992 (seven years). In the present case, an annual series of fifteen years was used (from 1983 to 1997). It is necessary to indicate that upon using quarterly series, even though the number of observations increases, a lot of distortions are introduced because of seasonal consumption factors. Some of these seasonal factors could be eliminated upon introducing variables such as a "dummy" (which is the Gemines methodology). However, models constructed with "dummy" variables are generally of a lesser quality than those models that do not require this class of variables, and this does not seem to be the exception.
230. The European Communities alleges that the only piece of evidence put forward by Chile is the regression analysis. The European Communities made several comments on it. First, the regression has serious methodological problems. Second, in spite of those problems, the regression confirms that pisco and whisky compete with each other: pisco consumption goes up when the price of whisky increases and falls when the price of whisky goes down. Finally, contrary to the claims made by Chile, the regression does not show a lower rate of cross-price elasticity than the 1995 Gemines study. Unlike the authors of the 1995 Gemines study, the authors of the Chilean regression analysis have used a simple linear regression. This means that the parameters estimated by them do not represent elasticities and, therefore, cannot be compared to the parameters estimated in the 1995 Gemines study.
231. The European Communities further contests that Chile has failed to rebut the extensive evidence provided by the European Communities showing that pisco and the other distilled spirits are "directly competitive or substitutable" in the Chilean market. The only piece of evidence adduced by Chile is the above regression analysis. That analysis concludes that the Chilean regression analysis is afflicted by fundamental multi-collinearity problems which render the results extremely unstable and deceptive. Further, the paucity of the data set used by Chile is such that it is not possible to address those problems.
232. The European Communities explains that a difficulty inherent to all time-series data is that many valuables are "collinear". This is the term used by econometricians in order to describe the fact that many values move in parallel over time without necessarily implying any causal relationship. An often cited textbook example is that of cumulative rainfall and the consumer price index, which both rise over time. A regression of one on the other would yield a high statistical correlation.
233. In addition, the European Communities recalls that cross-elasticity is not a one way relationship. Pisco prices also affect whisky sales. In light of the respective market shares of the two products, this reverse cross-price elasticity should even be larger than that the effect whisky prices have on pisco sales. In the view of the European Communities, Chile makes no effort to address this aspect.
234. Chile also points out that in the first place, it is for the European Communities to prove that the regression analysis that it has submitted to the Panel meets all the standards required to be considered a sound econometric exercise.
235. Chile further notes that as the European Communities pointed out, there are a number of tests to be carried out in the data employed in a regression, before one could use the results confidently. Chile acknowledges the effort put forward by the European Communities to review the Chilean regression analysis that was made with the annual data covering the 1983 � 1997 period, and agrees with the comments that the data contain some insolvable problems (namely multicolineality).
236. Chile argues that the EC's criticism are ironic, however, because the regression submitted by the European Communities has even more serious problems; it covers a smaller number of years (all of them included in the regression submitted by Chile) and is further distorted by the use of quarterly data. Finally, and even more important, the elasticity coefficient is not significant (at 95% confidence level).
237. Chile again emphasized that in short, neither regression analysis is sufficiently reliable, but the EC's is less reliable than that of Chile, and it is the European Communities, not Chile, that has the burden of the proof.

(iv) The 1996 Gemines Study

238. The European Communities notes that following the submission to Congress of the 1995 Proposal by the Chilean Government, the pisco industry commissioned from Gemines a new study in order to assess the impact of the proposed reduction of the taxes on whisky from 70 % to 50 % (hereafter, the "1996 Gemines study").
239. The European Communities states that it has not been able to obtain a copy of the 1996 Gemines study. Nevertheless, its findings were widely publicised by APICH (the association of pisco producers) in July 1996, following the announcement by the Government that it intended to submit an amendment to the 1995 Proposal providing for an even larger reduction of the tax on whisky. Details of the 1996 Gemines study were also cited by some members of the Chamber of Deputies during the debate of the 1997 proposal. ¹⁴²
240. The European Communities notes that according to press reports¹⁴³, the 1996 Gemines study concluded that the reduction of the tax on whisky to 50 % envisaged by the 1995 Proposal would cause a 47 % drop in the price whisky and, as a consequence, a 17 % reduction in the sales of pisco. The European Communities claims that the accuracy of those reports has not been disputed by Chile.
241. The European Communities notes that the first document it requested was the "1996 Gemines study". The relevance for this dispute of that study is thus unquestionable. Yet, Chile's pisco industry has refused to provide the 1996 Gemines study to the Panel.
242. The European Communities states that according to a letter from APICH (the association of pisco producers) which has been forwarded by the Chilean Government to the European Communities, the reasons for that refusal are twofold. The first reason is that the study contains information "confidential" to Capel and Control. The second reason is that Capel and Control were not "satisfied with the results of the study, which did not achieve the expected results". For those two reasons, the letter concludes, "the study was never made public".
243. The European Communities then claims that to begin with, one may doubt of the "confidentiality" of business information which has been shared by the two main pisco producers without any apparent restriction. If it is true that the study contains confidential business information, then Chile's anti-trust authorities would be well advised to ask for a copy.
244. The European Communities further points out that, in any event, back in 1996 the pisco industry did not treat the study as "confidential". In July of that year APICH convened a press conference at which, brandishing the 1996 Gemines study as irrefutable evidence, APICH warned that if the Government proposal was adopted, sales of pisco would fall by 17 %. This was the same press conference at which the Chairman of APICH recalled that when at the beginning of the 1980s the tax differential between pisco and whisky was reduced to 5 %, "pisco producers were almost chased out of the market".
245. Also, the European Communities states that the press conference of July 1996 was part of a strategy aimed at stopping the Government from submitting an amendment to the proposal then pending before the Parliament that would have provided for a larger reduction of the tax on whisky. With the same purpose, the 1996 Gemines study was also provided by APICH to some members of the Chilean Congress, who quoted it extensively during the subsequent debate of the Government proposal.
246. The European Communities alleges that the inescapable conclusion is that in 1996 the pisco industry was telling to the Chilean Government and to the Chilean Parliament a totally different story from that presented to the Panel.
247. The European Communities argues that, in any event, neither APICH nor the Chilean authorities have explained why it is not possible to provide a non-confidential summary of the 1996 Gemines study. In view of that, the European Communities would urge the Panel to draw appropriate inferences from the attitude of Chile's pisco industry.
248. In rebuttal, Chile comments that the eagerness of the European Communities to request private documents prepared for the pisco industry as the Adimark study and Gemines 1996 study, is absolutely abnormal in this type of process. On one hand, the European Communities, not Chile's industry, has the burden of proving its assertions; on the other, nothing has prevented the European Communities or its industry from conducting all the surveys in Chile that they may wish; the Chilean industry -- like the Scotch Whisky Association -- is under no obligation to provide the information or advice it gets from private consultants.

To continue with The Adimark Survey

¹³³ EC Exhibit 20.

¹³⁴ The European Communities notes that cross-price elasticity measures the relative change in sales of one product as a result of a relative price change in another. However, cross-price elasticity itself is not neutral to the existing market shares of the product involved, and is dominated by the product with the larger share.

¹³⁵ The European Communities notes that the authors of the study cautioned that the estimated rate was likely to be lower than the current rate of cross-price elasticity. See 1995 Gemines Study, p. 57 (EC Exhibit 20).

¹³⁶ 1995 Gemines study, p. 61 and fn. 18 (EC Exhibit 20).

¹³⁷ 1995 Gemines study, p. 64 (EC Exhibit 20).

¹³⁸ EC First Submission, Table 19 (p. 62)

¹³⁹ Ibid., Table 20.

¹⁴⁰ 1995 Gemines study, p. 63 (EC Exhibit 20).

¹⁴¹ Chile First Submission, Annex II.

¹⁴² As an example, the European Communities refers to the intervention by Representative Prokurika, Minutes, p. 44 (p. 47 of the English translation).

¹⁴³ El Diario, 2 July 1996 (EC Exhibit 30).