R&D and Productivity: A Two Way Avenue? CLAUDIO BRAVO-ORTEGA Intelis, Department of Economics, Universidad de Chile, Santiago, Chile and A ´ LVARO GARCI ´ A MARI ´ N * Anderson School of Management, UCLA, USA Department of Economics, Universidad de Chile, Santiago, Chile Summary. — Most of the empirical studies assessing the R&D–productivity relationship at the country level fail to consider the possible simultaneity of these variables. Using a 65-country panel for the period between 1965 and 2005, this paper studies the relationship be- tween R&D and productivity using several R&D indicators. We establish that per capita R&D expenditure is strongly exogenous to productivity. This result allows us to develop a further argument that demonstrates the high social returns to R&D spending. Our esti- mates also indicate that a 10% increase in R&D per capita generates an average increase of about 1.6% in the long-run TFP. Ó 2011 Elsevier Ltd. All rights reserved. Key words — innovation, productivity, endogeneity, economic growth, country panel 1. INTRODUCTION The relationship between productivity and research and development (R&D) expenditure has been a topic of inquiry since the early works of Schultz (1953) and Griliches (1958). Since then, this area of research has produced a significant amount of empirical and theoretical work. While Zvi Griliches posed and approached most of the empirical questions, recent theoretical work credits a substantial role to R&D as a driver of productivity and, hence, of economic growth. 1 In these theoretical models, the connection between economic growth and R&D is generally established through an equilibrium equation that determines the resources allocated to this sector that spur total factor productivity (TFP) growth. The theoretical causality of R&D on productivity is thus estab- lished. Notwithstanding these research efforts, the question of whether productivity might also affect the levels of R&D has been, unfortunately less explored. Accordingly, the nature of the relationship between productivity and R&D expendi- ture at the country level remains partially unanswered. At the country level, there is no clear-cut evidence determin- ing whether there is a causal relationship—in either direc- tion—between productivity and R&D investment or if both events occur at the same time. Answering this question has crucial relevance for developing countries, since alternative an- swers lead to very different sets of policy recommendations regarding innovation and technology policies. In particular, if R&D expenditure produces significant increases in produc- tivity, it would be an appropriate strategy to invest in R&D as a mechanism to boost per capita income. However, if more productive countries invest more in R&D but the investment does not have clear impact on productivity, it would be unrea- sonable for developing countries to focus on increasing R&D at the expense of other types of investments. This paper ad- dresses these questions using a panel that includes 65 devel- oped and developing countries for the period 1965–2005. Theoretically, R&D expenditure can increase productivity through several different channels. First, it makes it possible to produce new goods and services that bring with them a more effective use of existing resources. Second, it makes it easier to reap the benefits of adapting other countries’ techno- logical progress. Third, the gains derived from foreign R&D activities can increase domestic productivity directly through the learning of new technologies and productive processes and indirectly through the imports of goods and services that have new technology incorporated (Coe & Helpman, 1995). The empirical literature generally confirms the enormous benefits of developing the R&D sector in terms of total factor productivity (TFP). However, a significant number of studies overlook the potential problems of simultaneity and reverse causality between R&D and TFP. 2 On one hand, more re- sources should make technological change or R&D more likely, which in turn would influence productivity. However, given the strong relationship between both variables and income, 3 it is likely that both R&D spending and productivity could respond in a similar way to demand shocks without the two necessarily having to be related. The causal relationship could even move in the opposite direction, if R&D spending were to respond positively to expected changes in demand (Frantzen, 2003). As ZviGriliches points out, [...]If research and development is chosen on the basis of economic incentives, it is unlikely to be fully independent of the shocks and errors which affect the production relations we are trying to estimate. 4 The available evidence for statistical causality between R&D spending and productivity comes from several firm- and * For valuable comments and suggestions, we thank Roberto A ´ lvarez, Martin Bell, Jose ´ Miguel Benavente, Mauricio Calani, Ro ´ mulo Chuma- cero and Gustavo Leyva, as well as the participants at the internal Central Bank of Chile seminar and the SECHI 2007 meeting of economists. We are especially grateful to two anonymous referees whose comments greatly improved the final version of our paper, and to Daniel Lederman who generously shared his data with us. Claudio Bravo-Ortega thanks financial support from Fondecyt Grants Numbers 1061137 and 1085027. The usual disclaimer applies. Final revision accepted: November 15, 2010. World Development Vol. 39, No. 7, pp. 1090–1107, 2011 Ó 2011 Elsevier Ltd. All rights reserved 0305-750X/$ - see front matter www.elsevier.com/locate/worlddev doi:10.1016/j.worlddev.2010.11.006 1090