Underreporting in count data models: some new results and an application to labor mobility

Abstract

Abstract Winkelmann and Zimmermann (1993), although available only as a working paper, has initiated a small but persistent literature on underreporting in count data models. Their Poisson-logistic (Pogit) model was initially developed to estimate the determinants of the number of job offers during a given time interval, when the researcher only observes the number of accepted offers. Later Pogit applications can be found in epidemiology, engineering, public health, and criminology, among others. I review some of the recent developments in the literature. I also consider an extension of the original model where the reporting process can be estimated separately from the event generating process due to the availability of a validation dataset.