DO SPILLOVERS MATTER WHEN ESTIMATING PRIVATE RETURNS TO R&D?

A large body of literature on the estimation of private returns to R&D adopts the Griliches knowledge production framework, ignoring the impact omitted spillover effects may have on consistent estimation. A separate body of literature is primarily interested in the presence and magnitude of spillovers but imposes a rigid ad hoc structure on the channels these can take, e.g. within-industry, within-country or determined by industry input-output matrices. In this paper we adopt a common factor approach which accounts for R&D spillovers without imposing any arbitrary structure on their nature and channels. At the same time we can account for other unobserved common processes which may affect countries or sectors differentially, e.g. economic shocks or business cycles, as well as heterogeneous evolution of TFP over time. Panel data from 12 industrial sectors of 12 OECD and EU countries (1980-2005) is used to arrive at unbiased estimates of private returns to R&D. Our results indicate the presence of substantial cross-sectional dependence in the residuals of the Griliches knowledge production function, pointing to the presence of knowledge spillovers. Further, our estimations suggest that when ignoring the presence of spillovers, R&D produces positive returns. However, when cross-sectional dependence is accounted for, we do not and any convincing evidence for positive private returns to R&D. These results suggest that spillovers may not be additively separable from own-R&D and need to be accounted for in the estimation even when the exclusive interest lies in obtaining estimates for private returns to R&D.