Measuring knowledge spillover effects via conditional nonparametric analysis

The issue of knowledge spillover is central in modern theories of innovation and growth. There is a large gap, however, between the recognition of the role of spillover in several theories and the empirical appreciation. This is done mainly through indirect estimation, surveys, direct estimation of cross-effects, and spatial dependence models. We review these methods and discuss their potential and limitations. In this paper we try to explore a new approach to the measurement of spillovers, which radically departs from the approaches followed in the literature. The approach is based on the exploitation of a recently developed family of techniques in nonparametric efficiency analysis (robust conditional efficiency), which allow the estimation of the impact of external factors on the technical efficiency of productive units. While in the tradition of these techniques a number of conceptual and practical issues severely limited their use, the new developments solve for these issues and deliver a flexible and powerful set of tools. We advocate the use of these tools and give a demonstration of their potential, using data at territorial level for Italy.