Testing for distributional features in varying coefficient panel data models

Abstract

This paper provides several tests for skewness and kurtosis for the error terms in a one-way fixed-effects varying coefficient panel data model. To obtain these tests, estimators of higher-order moments of both error components are obtained as solutions of estimating equations. Additionally, to obtain the nonparametric residuals, a local constant estimator based on a pairwise differencing transformation is proposed. The asymptotic properties of these estimators and tests are established. The proposed estimators and test statistics are augmented by simulation studies, and they are also illustrated in an empirical analysis regarding the technical efficiency of European Union companies. [1]

Nonparametric multidimensional fixed effects panel data models

“…j=1 Y ijt is the cross-sectional average, to remove the time fixed effects (i.e., λ t ). Second, inspired by Stromberg et al (2000), Honoré and Powell (2005) and Soberon et al (2020) (in a different context), a pairwise differencing transformation is proposed to remove the cross-sectional heterogeneities (i.e., µ i and γ j ), simultaneously. [7] Hence, subtracting from time t of (Y ijt − Y t ), time s, for s = t, we get…”