View article

In this paper, we investigate a dynamic linear panel regression model with measurement error. We consider the panel data estimation whose time dimension (T) is not small and comparable to the cross sectional dimension (N). First, we show that the 2SLS estimator suffers from the bias problem due to many instrumental variables. Using the alternative asymptotics where T 3 N goes to a constant as N, T→∞, we characterize its asymptotic bias due to many IVs. As a bias reduction method, we investigate the JIVE and derive its limiting distribution under the alternative asymptotics.