Count-Based Research in Management

We review 11 years (2001-2011) of management research using count-based dependent variables in 10 leading management journals. We find that approximately one out of four papers use the most basic Poisson regression model in their studies. However, due to potential concerns of overdispersion, alternative regression models may have been more appropriate. Furthermore, in many of these papers the overdispersion may have been caused by excess zeros in the data, suggesting that an alternative zero-inflated model may have been a better fit for the data. To illustrate the potential differences among the model specifications, we provide a comparison of the different models using previously published data. Additionally, we simulate data using different parameters. Finally, we offer a simplified decision tree guideline to improve future count-based research.