Can Autoregressive Betas Account for the Statistical Properties of Stock Returns?

This paper analyzes the implications of autoregressive betas in single factor models for the statistical properties of stock returns. It is demonstrated that this assumption alone is sufficient to account for the most important stylized facts of stock returns, namely conditional heteroscedasticity, leptokurtosis, weak serial correlation over short horizons, asymptotic independence and aggregational Gaussianity. This means that the autoregressive assumption of beta is sufficient to reproduce the observed non-linear dynamics of stock returns, even in the most extreme case in which both the factor and the non-systematic component follow Gaussian independent processes