Estimating Firm Size Elasticities of Product and Process R&D

This paper provides an empirical test of the particular product life-cycle hypothesis which postulates that the firm size elasticity of process R&D exceeds the firm size elasticity of product R&D. Panel data on German manufacturing firms is used which is affected by attrition and sample selection. An inverse probability weighted generalized empirical likelihood (GEL) estimator is proposed, which corrects for the selectivity bias under the identifying assumption of conditionally independent selection and benefits from the superior small sample bias properties of GEL compared to generalized method of moments (GMM). The product life-cycle hypothesis is clearly rejected in all specifications.