The paper introduces a neat regression setup for over- and under-dispersed count data, which generalized Poisson regression and logistic regression (yes, it can be applied to binary data!). The model is GLM-based and has lots of neat properties. Estimation is fast too.
You can find the paper at http://ssrn.com/abstract=1127359 and R code here. This is joint work with Kimberly Sellers.