Efficiency in Large Dynamic Panel Models with Common Factor

This paper deals with asymptotically efficient estimation in exchangeable nonlinear dynamic panel models with common unobservable factor. These models are especially relevant for applications to large portfolios of credits, corporate bonds, or life insurance contracts, and are recommended in the current regulation in Finance (Basel II and Basel III) and Insurance (Solvency II). The specification accounts for both microand macro-dynamics, induced by the lagged individual observation and the common stochastic factor, respectively. For large cross-sectional and time dimensions n and T , respectively, we derive the efficiency bound and introduce computationally simple efficient estimators for both the microand macroparameters. In particular, we show that the fixed effects estimator of the micro-parameter is asymptotically efficient. The results are based on an asymptotic expansion of the loglikelihood function in powers of 1/n. This expansion is used to investigate the second-order bias properties of the estimators. The results are illustrated with the stochastic migration model for credit risk analysis.