법과 통계의 만남 - 담합으로 인한 손해액의 통계적 추정 -

Many legal professionals are having a hard time figuring out what should be an appropriate amount of damage in a price-fixing case. Estimating damage is difficult since statistics is widely used during the process in order to measure the hypothetical competitive price, which is essential in determining the amount of damage. But judges and lawyers do not have much prior knowledge on statistics. Therefore, they cannot help but relying on expert witnesses who specialize in that field. Most of the time, however, experts also disagree over the adequate amount of damage. Recently there was a court case in South Korea, where expert witnesses of both sides differed in their damage estimates, the gap between them even reaching 100 billion KRW. Thus legal professionals cannot shun from statistical debates if they want to solve this ‘legal’ dispute. Previous literature in the legal studies did not touch upon this issue thoroughly where both statistics and the law are pertinent. This paper tries to fill in the gap, especially with respect to how legal professionals should deal with various methodological issues raised by econometrical inferences. In doing so, the validity of many arguments raised at previous court cases of South Korea and the United States are discussed. For example, the defendants(South Korean oil companies) raised the argument at the recent oil price-fixing case that the coefficient of a variable, estimating the effect of price-fixing in a particular year, could not be used since it was not statistically significant with 95% confidence level. But the Seoul Central District Court(the court of 1st instance) and Seoul High Court of Justice(the court of 2nd instance) ruled against the defendants, holding that the use of that variable is permitted on the ground that the purpose of the statistical analysis of the court-appointed expert was to estimate the amount of damage, rather than to test the hypothesis of whether there was a price-fixing. This paper argues that courts' decisions were reasonable in that 5% significance level is just a rule of thumb among social scientists, so it cannot be that a hypothesis failing to meet 5% significance level is worthless. If a coefficient is statistically significant not with 5% significance level, but with 10% significance level, for instance, it is true that the information is of slightly lower value than that satisfies 5% significance level, but it is still a useful information and the disparity between two are not large.