The Keane and Runkle estimator for panel-data models with serial correlation and instruments that are not strictly exogenous

Abstract

In this article, we introduce the new command xtkr, which implements the Keane and Runkle (1992a, Journal of Business and Economic Statistics 10: 1– 9) approach for fitting linear panel-data models when the available instruments are predetermined but not strictly exogenous. This is a common case that includes dynamic panel-data models as a leading example. Monte Carlo simulations show that, in certain situations, this approach offers an improvement over the popular difference generalized method of moments and system generalized method of moments estimators in terms of bias and root mean squared error. An empirical application to cigarette demand also demonstrates its usefulness for applied researchers.