Panel Data Analysis with Heterogeneous Dynamics

摘要: This paper proposes the analysis of panel data whose dynamic structure is heterogeneous across individuals. Our proposed method easy to implement and does not rely on any specific model for dynamics. We first compute sample mean, autocovariances, and/or autocorrelations each individual, then estimate parameter interest based empirical distributions estimated autocorrelations. illustrate usefulness our procedures by applying them study earnings productivity dynamics find that both exhibit substantial heterogeneity. investigate asymptotic properties estimators using double asymptotics under which cross-sectional size length time series tend infinity. prove functional central limit theorem distribution estimator. Further, if we can write as expectation a smooth function individual mean reduce bias split-panel jackknife bias-correction. also develop an inference procedure bootstrap. The results Monte Carlo simulations confirm in finite samples show are informative regarding finite-sample properties.