Evaluating innovation policy: a structural treatment effect model of R&D subsidies

This paper studies the welfare effects of R&D subsidies. We develop a model of continuous optimal treatment with outcome heterogeneity where the treatment outcome depends on applicant investment. The model takes into account heterogeneous application costs and identifies the treatment effect on the public agency running the programme. Under the assumption of a welfare-maximizing agency, we identify general equilibrium treatment effects. Applyiing our model to R&D project-level data we find substantial treatment effect heterogeneity. Agency-specific treatment effects are smaller than private treatment effects. We find that the rate of return on subsidies for the agency is 30-50%.