In order to measure the effect of tax credits on private R&D investment, researchers confront the difficult problem of finding an exogenous measure of tax policy that exhibits sufficient variation to support robust identification. This paper takes a new approach based on exploiting differences in the average capital-labor ratio of R&D investment across industries and variation in the tax treatment of different expenditure types across countries and over time. The estimated short-run elasticity is 0.50 which is somewhat more than double previous estimates derived from cross-country analysis.