Explaining the persistent productivity differences between countries has long been a focus of the theoretical and empirical growth literature. While a large body of work on the conditional convergence hypothesis emphasizes the role of factor accumulation, it has been suggested that an explanation of the observed phenomenon is incomplete if not accounting for cross-country differentials in total factor productivity (TFP). (1)
One aspect that has received relatively less attention is an apparent history dependence of countries' productivity rankings. Eaton and Kortum (1997) report that despite large differences in growth rates, over a 40-year period the five leading economies in the world only show two reversals in relative position. (2) Comparing the labor productivity in seven countries to the U.S. level over a period of more than 30 years similarly suggests that economies often retain their relative positions, with countries starting out at lower productivities also converging to lower levels (Figure 1).
This article presents a model that can provide an explanation for persistent cross-country TFP differentials. We make two key assumptions. First, the intellectual property (IP) embodied in new production technologies is subject to worldwide patent protection. The owner of a technology can freely choose where in the world it will be used. Second, we assume that there is complementarity between human capital and the level of technology used in production, that is, the relative productivity advantage of a new technology is bigger when it is used by more skilled workers.
We show that under these conditions, new technologies are first introduced in the most human capital-rich country and subsequently passed on to countries with a less-skilled labor force, before they eventually become obsolete. This mechanism of technology diffusion, which emerges endogenously in our model, is consistent with the observation that less-developed countries typically get access to new technologies later. (3)
Moreover, if there is a positive feedback between the level of technology used in a country and the rate of human capital accumulation, countries' long-term productivity levels become history dependent. Economies that start out with a low human capital stock initially catch up to the leading countries, but never fully converge to the same level.
[FIGURE 1 OMITTED]
The key difference between our approach and existing models of persistent TFP differentials is that we do not impose any ex ante restrictions on the ability to transfer technologies or relocate production processes internationally. In contrast, Parente and Prescott (1994) explicitly model barriers to technology adoption. Howitt (2000) assumes that while researchers have access to all the technological knowledge available worldwide, any technologies that are used in a country must still be developed domestically. As a consequence, countries that invest more in research and development (R&D) will, on average, employ more recent technologies and be more productive. A very general model of technological spillovers and adoption costs that captures many features of the various approaches discussed in the literature is developed in Klenow and Rodriguez-Clare (2005). (4)
The remainder of this article is laid out as follows. Section II discusses the literature relevant to our approach, with a focus on motivating our key assumptions and presenting evidence to support the diffusion mechanism implied by our model. Section III introduces the general model. In Section IV, we derive and discuss the main results. A brief summary is presented in Section V, together with concluding remarks.
II. LITERATURE OVERVIEW
The degree, speed, and nature of convergence have been a focus of interest in the growth literature. While traditional growth models suggest that countries should converge to a steady state that is dictated by their structural properties--a result known as the conditional convergence hypothesis--it is understood that a large number of plausible deviations from the standard neoclassical assumptions generate growth models in which outcomes are history dependent. …