Endogeneity in pharmaceutical knowledge generation: An instrument-free copula approach for Poisson frontier models

Abstract

This study provides an assessment of the R&D–patent relation of European pharmaceutical firms that are not flawed by endogeneity biases. Firms invest in R&D and generate latent knowledge which then manifests in observable patent outcomes through a Poisson model. The process of turning R&D into knowledge is described by a production process subject to inefficiency and endogeneity. To estimate a Poisson stochastic frontier model, the suggested novel copula-based approach directly accounts for the dependence between the endogenous regressors and the inefficiency component. Hence, its implementation does not require any instrumental variables. Simulation results underline that the proposed estimator outperforms conventional instrumental variable estimators. Neglecting endogeneity leads to a substantial underestimation of the R&D elasticity of patents generated in the European pharmaceutical industry.