Compound distribution models of stock returns: an empirical comparison

The study provides empirical tests and comparison of mixed diffusion‐jump processes and finite mixtures of normal processes as models of stock price behavior. For weekly returns, both specifications have significantly higher descriptive validity than a stationary normal distribution, and, in most cases, mixed diffusion‐jump processes are empirically superior to finite normal mixtures. The distribution of monthly returns, however, can be safely assumed to be approximately normal.