Essays on Empirical Finance: Out-of-sample Methods and Applications

Financial research faces a replication crisis as many empirical studies of long-short investment strategies (originally developed in the U.S. equity market) cannot be replicated or are susceptible to p-hacking, data snooping, HARKing (i.e. hypothesising after the results are known) and are therefore sample specific. This is a highly debated topic, given the large degrees of freedom researchers have in defining an investment strategy and the number of strategies that have been published in recent years. This cumulative thesis approaches this discussion in some ways. How likely is it that one set of rules will fit all empirical questions? The problem is exacerbated when international data are used for replication. Thus, how likely is it that a set of rules defined for the U.S. will apply to all international stock markets?

The major part of this dissertation focuses on international scientific replications using similar (but not same) data and applications of the existing U.S. results. The results show that the straightforward application of a long-short strategy originally developed in the U.S. can be highly misleading and that for any long-short strategy, one needs to reconsider what an appropriate test in international markets might look like from a theoretical and pragmatic perspective. This is an interesting contrast to the current trend towards large comparative studies of sometimes hundreds of anomalies, in which the theoretical and conceptual specificities of individual strategies are lost. One major reason that long-short strategies cannot simply be applied from the U.S. to international equity markets is because these markets are much smaller than the U.S.. Adjustments are therefore inevitable.