A Variance Decomposition for Swiss Stock Market Returns

This paper analyzes the persistence or serial correlation of expected returns as well as the univariate time-series approach that studies the implied autocorrelation function of realized stock returns, including mean reversion and its conditions. In particular, we critically examine whether increases in expected dividend growth tend to be associated with decreases in future expected returns, a correlation that would amplify the volatility of equity returns. Based on Swiss stock market data and a number of predictive variables found important in previous studies of return predictability, we emphasize that the results are dependent on the particular specification of the information set which agents use to predict equity returns. In contrast to the standard conclusion in the literature, we do not generally find that the variance of news about future returns is greater than the variance of news about future dividends. This result rather depends on the time period under consideration, and, more importantly, whether the dividend-price ratio is included in the analysis. Our conclusions are strongly supported when we additionally introduce Bayesian model and parameter uncertainty into the analysis. The results thus pose a challenge to current conditional asset pricing theory and to the businesscycle related motivation of stock market predictability.