Likelihood-Based Estimation of Dynamic Panels With Predetermined Regressors

This article discusses the likelihood-based estimation of panel data models with individual-specific effects and both lagged dependent variable regressors and additional predetermined explanatory variables. The resulting new estimator, labeled as subsystem limited information maximum likelihood (ssLIML), is asymptotically equivalent to standard panel generalized method of moment as N → ∞ for fixed T but tends to present smaller biases in finite...