The paper discusses the problem of pooling cross-section and time-series data in the framework of likelihood functions and the associated posterior distributions and shows how this analysis can be useful in deciding whether or not to pool.