Random Walk oder Mean Reversion? Eine statistische Analyse des Kurs/Gewinn-Verhältnisses für den deutschen Aktienmarkt

The present contribution considers the question whether the random walk model or an AR(1)-process (“mean reversion”) is a better representation for the development of the price/earnings ratio of the German blue-chip index DAX. Empirical evidence for one of these alternative model hypotheses is crucial to the predictability of the underlying variable, i.e. the P/E ratio. While the random walk hypothesis implies the non-existence of a long-run “fair” value for the variable of interest, an AR(1) process, in contrast, possesses a long-run mean and exhibits mean reverting behaviour in that it fluctuates around this constant long-run value. Both an exploratory data analysis and a set of formal statistical tests equally lead to the conclusion that the hypothe-sis of an AR(1) process, in a statistical sense, better represents the investigated time series data than the random walk model. The consequences of this key result are not only discussed with respect to the predictability of the P/E ratio of the German stock market index, but also with regard to forecasts for the development of the DAX itself.