Discrete Poisson kernel density estimation-with an application to wildcat coal strikes

This paper proposes a nonparametric Poisson kernel density estimation technique for discrete distributions. Economists have been using continuous kernels to approximate discrete distributions. This work introduces a discrete kernel as more appropriate for approximating discrete distributions. Simulation results are presented to compare with standard parametric approaches. We apply our discrete Poisson kernel estimator to approximate the distribution of coal mine wildcat strikes in the United States.