The Production of Knowledge: A Meta-Regression Analysis

The production of knowledge has been subjected to quantitative analysis following the development of the R&D based endogenous growth theory. There is a spirited discussion about the empirical validity of the semi-endogenous or endogenous (Shumpeterian) growth theories, with clear policy relevance. While the first theory points out ineffective policies in the long run, the latter allows for important effects of fiscal policies, namely subsidies or taxes to R&D. We survey the empirical literature on this topic and implement a meta-analytic regression for the spillover effect, which crucially determines the validity of those theories. We discover that the average spillover effect is lower than one but with the upper bound of the confidence interval above one. We also find that the spillover effect tends to be higher once patents are used to measure knowledge and the estimation of knowledge production accounts for foreign inputs and time-effects; it tends to be lower when the sample includes more countries, only rich economies and industrial data, and when estimations use instrumental variables. Although most recent contributions argue in favor of the Shumpeterian growth theory, we show that when looking at a more complete picture, semi-endogenous theory cannot be neglected.