Estimation and Inference in Large Heterogeneous Panels with Cross Section Dependence

Abstract

This paper presents a new approach to estimation and inference in panel data models with unobserved common factors possibly correlated with exogenously given individual-specific regressors and/or the observed common effects. The basic idea behind the proposed estimation procedure is to filter the individual-specific regressors by means of (weighted) cross-section aggregates such that, asymptotically as the cross-section dimension (N) tends to infinity, the differential of unobserved common factors are eliminated. The estimation procedure has the advantage that it can be computed by OLS applied to an auxiliary regression where the observed regressors are augmented by cross sectional averages of the dependent variable and the individual specific regressors. It is shown that the proposed correlated common effects (CCE) estimators for the individual-specific regressors (and its pooled counterpart) are asymptotically unbiased as N ? 8, both when T (the time-series dimension) is fixed, and when N and T tend to infinity jointly. Further, the CCE estimators are asymptotically normal for T fixed as N ? 8, and when (N,T) ? 8, jointly provided vT/N ? 0 as (N,T) ? 8. A generalisation of these results to multi-factor structures is also provided.