Inverse Probability Weighted Generalised Empirical Likelihood Estimators : Firm Size and R&D Revisited

The inverse probability weighted Generalised Empirical Likelihood (IPW-GEL) estimator is proposed for the estimation of the parameters of a vector of possibly non-linear unconditional moment functions in the presence of conditionally independent sample selection or attrition.The estimator is applied to the estimation of the firm size elasticity of product and process R&D expenditures using a panel of German manufacturing firms, which is affected by attrition and selection into R&D activities.IPW-GEL and IPW-GMM estimators are compared in this application as well as identification assumptions based on independent and conditionally independent sample selection.The results are similar in all specifications.