An Analytic and Empirical Examination of Zero-Inflated and Hurdle Poisson Specifications

Abstract

This article examines two alternative specifications for estimating event-count models in which the data-generating process results in a larger number of zero counts than would be expected under standard distributional assumptions. The author compares King's hurdle event count model and Greene's zero-inflated Poisson model, using data on congressional responses to Supreme Court decisions from 1979 to 1988. The author shows that each of these models is a special case of a more general dual regime data-generating process that results in extra-Poisson zero counts. Furthermore, because this data-generating process can produce overdispersion in its own right, these models are also shown to be related to variance function negative binomial specifications. The underlying correspondence between these models leads to similar results in estimating and interpreting them in practice.