In the thesis, we investigate the relationship between productivity change and R&D in the presence of international R&D spillovers by trade and technological distance. The study covers total manufacturing industries in OECD countries. We assume the main channels of international R&D spillovers are trade and technological distance. We use the non-parametric frontier and parametric frontier approaches to examine the influence of domestic R&D and international spillovers on TFP change, technical change, and efficiency change. We expect domestic R&D and international spillovers to affect these changes in different ways.
Most empirical studies use the traditional total factor productivity index, which assumes that all units of decision making are technically efficient. In order to consider inefficiency and decompose productivity into the components mentioned above, we introduce the concept of frontier. Using the concept of frontier, we also decompose the change of total factor productivity into efficiency change and technological progress. Our empirical results indicate that for all countries as a whole, both domestic R&D and disembodied spillovers of R&D have statistically significant and positive effects on TFP growth, and especially on technical change.
We further investigate R&D and spillover effects by dividing countries into two groups: a high-R&D-intensive group and a low-R&D-intensive group. For the high-R&D-intensive group we select the top five countries that exhibit relatively large ratios of R&D expenditures to value added, and for the low intensive group we select the bottom five countries. For high-R&D-intensive countries, the growth of TFP is mainly attributable to the increase in domestic R&D. For low-R&D-intensive countries, however, the international spillovers are the main factors for the TFP growth. Technical innovation is attributable to disembodied international spillovers, and efficiency improvement to embodied spillovers.
For the comparison of these results, we use productivity changes using the non-parametric approach and stochastic frontier approach. We obtain clearer results when using the Malmquist index than when using stochastic frontier analysis. These results may be caused by the methods of productivity measure: the parametric approach allows random noises, and the non-parametric approach assumes no specific production function. It is either due to misspecification in the models or high levels of random noises.
