Theory and methods of panel data models with interactive effects (Q2448726): Difference between revisions

Panel data have found a wide range of applications in economics and other social sciences. In certain cases the explanatory variables are correlated with the unobserved effects. This case creates a series of problems in the standard estimators and hence it needs special effort. The current paper treats this case, namely panel data models with interactive effects. This paper considers the maximum likelihood (ML) estimation for such models. The usual pooled least squares estimator or even the within-group estimator is inconsistent for the regression coefficients. Existing methods for consistent estimation like GMM are either designed for panel data with short time periods or are less efficient. Hence the authors consider ML estimation and in fact the pseudo-Gaussian maximum likelihood method under large number of individuals and large time series. The theory does not depend on normality. The approach has both practical and theoretical interest. The maximum likelihood estimator has desirable properties and is easy to implement, as illustrated by the Monte Carlo simulations. This paper develops the inferential theory for the maximum likelihood estimator, including consistency, rate of convergence and the limiting distributions. It is also extended for the case of models that include time-invariant regressors and common regressors (cross-section invariant). The regression coefficients for the time-invariant regressors are time-varying, and the coefficients for the common regressors are cross-sectionally varying. The paper also has an electronic supplement. This supplement provides detailed technical proofs and inferential theory for the estimated coefficients of time invariant and common regressors. The EM solutions are shown to have local optimality property.