Do Consumption-Based Asset Pricing Models Explain Serial Dependence in Stock Returns?
We show that the Bansal-Yaron, Campbell-Cochrane and Cecchetti-Lam-Mark models of asset prices cannot explain the serial correlation structure of stock returns. We show this by estimating these models and deriving expected returns from them and then testing whether the difference between observed and expected returns is a martingale difference sequence. We use variance ratio and rescaled range tests which we modify to account for the expected returns being functions of estimated parameters. We also use a weighted quantilogram test based on a bootstrap procedure robust to this estimation. The evidence against the BansalYaron and Campbell-Cochrane models is significant. While the evidence against the Cecchetti-Lam-Mark model is not in general significant, our point estimates strongly suggest its residuals are not a martingale difference sequence. Furthermore, a semi-parametric maximal predictability test suggests there is some evidence that the three models’ state variables struggle to explain the degree of predictability observed in the market return. A timing strategy designed to exploit predictability in the market can significantly outperform the market in certainty equivalent terms under the Bansal-Yaron model. The timing strategy may underperform the market by less than it ought to under the Campbell-Cochrane model.