Transformed Estimation for Panel Interactive Effects Models

We propose a transformed estimator for the slope coefficients of panel models with interactive effects. The estimation method does not require prior knowledge dimension factor structure. It is consistent and asymptotically normally distributed under fairly general conditions when N fixed T→∞ or T N→∞, both are large NT→a≠0<∞. Moreover, because transformation equivalent to aggregating cross-sectional units time before implementing least-square over across units, it can bypass issues arising from heteroscedasticity serial correlations in idiosyncratic errors. Furthermore, case that errors independent time, there no asymptotic bias even explanatory variables contain lagged dependent NT→a<∞ as T→∞. Extensive Monte Carlo simulations also conducted examine finite sample performance method.