This paper uses an empirical connection between real stock market indices of Germany and the USA for forecasting corresponding returns. We are starting from the random walk as the traditional forecasting model in stock market applications, extending it by co-integration. Since the cointegrating relation considers information about a systematic link between the stock market indices, containing a common stochastic trend of both, differences from the random walk occur particularly in the long run. Thus, the estimation period shows that with increasing forecasting horizon predictability of simple real returns of the German stock market gets more accurate than reflected traditionally.
Co-integration of international stock markets random walk discretely and continuously compounded returns impulse responses