Accounting for persistence in panel count data models. An application to the number of patents awarded

We propose a Poisson regression model that controls for three potential sources of persistence in panel count data; dynamics, latent heterogeneity and serial correlation in the idiosyncratic errors. We also account for the initial conditions problem. For model estimation, we develop a Markov Chain Monte Carlo algorithm. The proposed methodology is illustrated by a real example on the number of patents granted.