R & D-based growth models with transitional dynamics: Evidence from OECD countries

Abstract

Our study quantifies the impact of research and development (R&D) spending on technological progress, and hence on economic growth following the hypothesis that the second-generation growth theory best describes the data generation process of knowledge accumulation in the economy. Understanding the relationship between R&D and productivity is important for policy design geared toward enhancing economic growth. In order to estimate the long-run relationship between total factor productivity (TFP) and R&D activity, this study estimates an error correction model to test for transitional dynamics and long-run dynamics among productivity, investment in R&D and spillovers. To achieve this, we apply the econometric methods proposed by Pesaran and Smith (1995), Pesaran (1997) and Pesaran et al. (1999), the pooled mean group (PMG) estimators, to estimate the relationship among TFP investment in R&D and spillovers using an annual dataset for 21 OECD countries from 1965 to 2015. The advantage of using this estimator is that it uses the information in the data to estimate both the long-run and the short-run dynamics simultaneously. As independent variables, we have constructed stocks for R&D expenditure as a percentage of GDP, also known as R&D intensity, which is a measure defined as a firm’s R&D expenditure divided by its sales. R&D intensity varies according to a firm's industry, product knowledge, manufacturing, and technology, and is a metric that can be used to gauge the level of a firm's investment to enhance productivity. The aim of R&D expenditures, ultimately, is to increase productivity. R&D intensity at the country level is defined as its R&D expenditure as a percentage of gross domestic product (GDP). This is an indicator that reflects the level and structure of the efforts undertaken by countries in the field of science and technology. In the second-generation growth models R&D intensity is also referred to as a measure that captures R&D adjusted by product proliferation (Young, 1998; Madsen, 2008). We find that R&D has a large effect on a country’s TFP in the long-run. The elasticity of TFP with respect to own-R&D adjusted by product proliferation at the country level ranges from 0.571 to 1.510, implying marginal returns to country TFP from R&D spending adjusted by product proliferation that range from 19% to 462%. There are also positive R&D spillovers; the elasticity of TFP with respect to R&D adjusted by product proliferation performed outside the country ranges from 0.194 to 0.533, implying marginal return spillovers that range from 8.2% to 25.6% with a maximum of 0.43% of the total returns accruing to spillovers. Regarding the short-run dynamics, the error correction estimates are statistically significant and do not differ significantly among different specifications. The range for the speed of adjustment ranges between -0.0881 and -0.0695, which implies that half of the time that would take for a 1% deviation of TFP from the long-run relationship to close will be approximately of 5 years.We find empirical evidence that there is a structural break in the long-run relationship between TFP and R&D in 2002. We can infer that the use of a common currency in the Eurozone improved the mechanisms whereby knowledge was transmitted among the countries in our sample. The productivity increase after 2002 can be explained by the conclusion of Coe and Helpman (1995) that trade increases productivity by the mechanism of international technological spillovers; therefore, an increase in trade would enhance productivity among trading partners. Frankel (2008) conducted a survey on the impact of the European Monetary Union (EMU) on bilateral trade and found a positive effect on bilateral trade where the positive effects of a common currency range between 10% and 15% for the first years of the entry in force of the EMU.The three robustness checks that we have conducted validate our baseline estimates. The estimates of the robustness checks in addition to being statistically significant show similar features. First, the higher the depreciation rate used to compute the R&D stocks adjusted by product proliferation, the larger the long-run coefficients. Second, the higher the depreciation rate used to estimate the R&D stocks adjusted by product proliferation, the lower the error correction term in absolute terms. Finally, the spillover ratio and the spillover fraction vary little with respect to the depreciation rate used to compute R&D stocks adjusted by product proliferation. From the forecasts performed based on the baseline and VAR estimations we arrive to the following conclusion. Since the forecasts for the growth rates of the own-R&D stock are constant for the representative country composed by all the countries in our sample, all the variation of total factor productivity has to be explained from the behavior of the R&D performed by other countries. On one hand, for the representative country composed by all the countries in our sample, the decreasing forecasted TFP growth rates are accompanied by an increase in the forecasted R&D performed by other countries growth rates. This would imply that growth for the representative country conformed by the 21 countries in our sample would be explained in part by spillovers. On the other hand, for the representative country composed by the G7 countries, the sizeable increase in forecasted growth rates in TFP goes hand with hand with relatively constant forecasted growth rates of R&D performed by other countries and an increasing trend in within-country R&D investment. These results would suggest that G7 countries’ R&D activity spills over the rest of the countries in the sample. Moreover, the G7 countries have greater incentives to keep performing R&D activity than the rest of the countries in the sample; and the benefits in productivity are greater for G7 countries than for rest of the countries in our sample.