Faculty of Economics and Business Administration Publications Database

Zur Diskrepanz zwischen gerichtlichen Beweisfragen in Kartellschadenersatzverfahren und den Ergebnissen des ökonomischen „Standardansatzes“ bei statistischen Analysen

Frank, Niels
Oldehaver, Gunnar
Volume: 17
Number: 1
Link External Source: Online Version
Year: 2019

In this article we ask whether the econometric results that are typically produced in expert reports, notably in follow-on cases, can indeed answer the questions that are usually of interest to courts. Of particular relevance are therefore questions that relate to probabilities, such as the probability that an infringement resulted in damages at all or the probability of damages of a certain amount. The typically used (so-called frequentist) approach does however not generate direct answers to these questions. We explain this by following the well-known approach of re-interpreting these results within the framework of testing a given hypothesis, in this case the hypothesis of “no effect”. The statistical evidence that is typically generated by expert reports then asks solely whether the hypothesis of “no effect” can be rejected with certain confidence, the so-called type I error. The type II error, that of not rejecting the hypothesis despite a positive effect, is typically not discussed at all. The reported significance levels or confidence intervals also do not provide direct information about the likelihood with which the true effect takes on a particular value or falls into a given interval, and, in particular, commonly reported results do not allow an inference as to whether a particular effect, or any effect at all, is more likely than another effect or no effect at all. While we do not suggest at all to discard such econometric evidence, the actual meaning of the econometric results presented in expert reports should be well understood by lawyers and courts, and courts, in particular, should take the limitations of the statistical analysis into account when posing particular questions to court-appointed experts. We conclude with a short reference to an alternative approach, that of the Bayesian statistics.