Productivity Spillovers between OECD Industries

Productivity spillovers are an important element in recent economic models of new growth and trade theory. The importance of spillover effects is also underlined by several empirical studies using regional or country level data. There are hardly studies at the industry level, however. In particular, there is no study that has attempted to provide a comprehensive assessment of intra- and inter-industry spillovers between developed countries using industry level data. It is this gap in the literature, which the present research project targets using a spatial econometric approach and data from OECD manufacturing industries. The first contribution is to construct measures of the intensity of interactions between industries. In studies using geographic units, some function of distance is a natural choice. Such a natural measure is missing for industry data, and this may partly explain the lack of studies in this field. Thus, the first step of this project is to construct measures of the intensity of interactions between industries at the national and international level. In the present setting, which aims to investigate spillovers between industries, a proper measure of distance (economic interactions) is trade between industries. Productivity spillovers are supposed to take place mainly between firms. Since a large share of trade between firms is trade in intermediaries, input-output data appear to be a suitable (and feasible) choice to measure the extent and intensity of interactions between different industries. These measures are constructed from OECD input-output data and will be summarized in a so called weights matrix where each element measures the interactions between industry i of country j with industry k of country l. This matrix is an important building block of the empirical model in the present study but it will also be of interest in other applications investigating interactions between industries. In a next step, to test for inter- and intra-industry spillovers between and within countries, it takes a spatial econometric approach, which nicely matches the economic hypothesis that are being tested. Having determined the proper empirical modelling of spillover effects using a sequence of specification tests, we go on to estimate the model using the recently developed estimator by Kelejian and Prucha (2006) that allows for heteroscedastic disturbances in the spatial regressive process. We will then extend the empirical model, testing for differences in inter- vs. intra-industry spillovers, for the relevance of border effects, and whether there a significant differences between countries with respect to the strength of (within country) productivity spillovers. The estimation of the extended empirical model will also require a methodological extension of the estimator by Kelejian and Prucha (2006) to allow for parameter heterogeneity in the spatial regressive process. While this motivated by the present research project, the generalized estimator to be developed in this extension should be of interest more generally in other applications as well.