Third, I structurally estimate the theoretical model parameters using Simulated Method of Moments on Compustat firms. Finally, I use these estimated parameters to conduct a macro policy experiment to evaluate the e¤ects of a size-dependent R&D subsidy on different sized firms. In conclusion of this analysis, the optimal size-dependent R&D subsidy policy does considerably better than optimal uniform (size-independent) policy. More interestingly, the optimal (welfare-maximizing) policy provides higher subsidies to smaller firms.