Tactical Asset Allocation and Model Uncertainty : An Exploratory Study for the Swiss Stock Market

This paper uses statistical model selection criteria and Avramovs (2002) Bayesian model averaging approach to analyze the sample evidence on stock market predictability in the presence of model uncertainty. Based on Swiss stock market data, our posterior analysis finds that neither the cumulative posterior probabilities nor the posterior probabilities of the individual forecasting models are constant through time. We also decompose the variance of predicted future excess returns into three components: uncertainty attributed to forecast errors, parameter uncertainty, and model uncertainty. The empirical results indicate that the respective contributions are highly dependent on the time period under consideration and the initial values of the predictive variables. In contrast to Avramov (2002), model uncertainty is generally not more important than parameter uncertainty. Finally, we also demonstrate the implications of model uncertainty for tactical asset allocation strategies. Our results do not indicate any reliable out-of-sample return predictability. Moreover, in contrast to Avramov (2002), the out-of-sample performance of the Bayesian model averaging approach is not generally superior to the statistical model selection criteria. Consequently, even properly incorporating model uncertainty into a tactical asset allocation model does not seem to help improving the performance of the resulting short-term tactical asset allocation strategies. We use a number of statistical model selection criteria and Avramovs (2002) Bayesian model averaging approach to analyze the sample evidence on stock market predictability in the presence of model uncertainty. The empirical analysis for the Swiss stock market includes a number of predictive variables found important in previous studies of return predictability, as well as their stochastically detrended counterparts. The posterior analysis finds that neither the cumulative posterior probabilities nor the posterior probabilities of the individual forecasting models are constan