题目：Efficient estimation of heterogenous coefficients in panel data models with interactive effects (joint with Jushan Bai)

主讲：李鲲鹏

时间：4月3日（周三）上午10：30-12：00

地点：博学楼1120教室

主办：国际经济贸易学院

摘要：This paper considers efficient estimation of heterogenous coefficients in panel data models with interactive effects , which have received much attention by recent theoretical and empirical literature. The existing estimation methods, such as the CCE method in Pesaran (2006, Econometrica), are neither efficient nor  satisfactory for finite sample properties. We propose a two-step method to estimate the heterogenous cross-sectional coefficients, in which the ML method is conducted in the first step to estimate the loadings and variances of the idiosyncratic errors, and the heterogenous coefficients are estimated in the second step by making use of the structural relations implied by the models. To analyze the proposed estimator, we first extend the statistical theory of factor models from the diagonal covariance of the idiosyncratic errors to the block-diagonal case. Based on the new theory,  we then establish the asymptotic theory of our estimator. The proposed estimator is asymptotically efficient in the sense that it has the same limiting distribution with the infeasible GLS estimator. In addition, the estimator is also consistent under fixed $N$ if some regular conditions are satisfied. Monte-carlo investigation shows that our estimators have very goodfinite sample properties.