A few years ago, I published an article on using Poisson, negative binomial, and zero inflated models in analyzing count data (see Pick Your Poisson). The abstract of the article indicates:
School violence research is often concerned with infrequently occurring events such as counts of the number of bullying incidents or fights a student may experience. Analyzing count data using ordinary least squares regression may produce improbable predicted values, and as a result of regression assumption violations, result in higher Type I errors.