A new diagnostic test for cross-section independence in nonlinear panel data models

Abstract

In this paper, we propose a new diagnostic test for residual cross section in dependence in a nonparametric panel data model. The proposed test is a nonparametric counterpart of an existing test proposed in Pesaren (2004) for the parametric case. First of all, we establish an asymptotic distribution of the proposed test statistic under the null hypothesis. As shown in the parametric case, the asymptotic distribution is a standard normality. We then analyze the power function of the proposed test statistic under an alternative hypothesis that involves a nonlinear multi-factor model. In order to compute both the sizes and the power values, we select a simulated critical value in each case based on a simple bootstrap simulation scheme in the context of nonparametric panel data models. We finally provide several numerical examples and an empirical analysis of a set of CPI data in Australian capital cities to illustrate the finite sample performance of the proposed test statistic.