Root-n Consistent Kernel Density Estimation in Practice

Abstract

Abstract This paper details implementation of the recently proposed root-n kernel density estimator of (Escanciano, J. C., and D. T. Jacho-Chávez. 2012. “ n $\sqrt n $ -uniformly consistent density estimation in nonparametric regression models.” Journal of Econometrics 167: 305–316.) that circumvents the slow rate of convergence of traditional nonparametric kernel density estimators. We discuss implementation issues such as bandwidth selection and controlling for heteroskedasticity. Two empirical examples are provided; we re-examine the classic study of the emerging multimodality of the cross-country distribution of income per capita, finding more local structure with this new method, and we study the distribution of lean body mass across gender, where we demonstrate robustness of the new methods to alternative bandwidth selection mechanisms.