Theory and Practice of TFP Estimation : The Control Function Approach Using Stata

Alongside Instrumental Variable (IV) and Fixed Effects (FE), the Control Function (CF) approach is the most widely used in production function estimation. Olley-Pakes, Levinsohn-Petrin, Ackerberg-Caves-Frazer have all contributed to the literature proposing two-steps estimation procedures, while Wooldridge showed how to perform a consistent estimation within a single step GMM framework. In this paper we propose a new estimator, based on Wooldridge's, using dynamic panel instruments à la Blundell-Bond and we evaluate its performance by Monte Carlo simulations. We also present a new Stata module - prodest - for production function estimation, show its main features and key strengths in a comparative analysis with other available user-written commands. Lastly, we provide evidence of the numerical challenges faced when using OP/LP estimators with ACF correction in empirical applications and document how the GMM estimates vary depending on the optimization/starting points used