An Empiricist’s Guide to Nonparametric Analysis in Accounting

Abstract

Recent advancements in statistical packages and computing power have made various forms of nonparametric estimation accessible to empirical researchers. This study explores several of these nonparametric estimation techniques, focusing on kernel density estimates and locally weighted regression for tractability. We provide a discussion of these research design choices, including their statistical properties, limitations, and key inputs over which researchers have discretion. Then, we provide examples of these techniques, analyzing effective tax rates in the financial service industry using kernel density estimates and the relation between audit fees and size using nonparametric regression analysis. Additionally, we discuss how nonparametric techniques may be used in univariate estimates, to complement ordinary least squares regression, and in other areas in accounting.