Convergence Speeds and Transitional Dynamics in Non-Scale Growth Models

Abstract

Neoclassical and endogenous growth models yield strikingly different predictions regarding the determinants of long-run growth rates and their implications for long-run cross-country convergence characteristics. On the one hand, Mankiw, Romer, and Weil (1992) and Barro and Salai-Martin (1995) have shown that countries converge to identical growth rates, but to distinct income levels. Since these empirical findings run counter to the predictions of endogenous growth models, they have cast doubt on the relevance of such models to explain long-run cross-country convergence and transition paths. On the other hand, while the empirical evidence confirms the implications of the traditional neoclassical in terms of cross-country convergence, calibrations show that the neoclassical model’s implied convergence speed of about 7 percent, greatly exceeds the empirical estimates of approximately 2 percent. This excessive speed of convergence is also accompanied by implausibly high rates of return (in the standard model) or by implausible rates of investment (in models with human capital). In addition, Bernard and Jones (1996a) maintain that the neoclassical convergence approach overemphasizes capital accumulation at the expense of technological change. They document that, at least since the 1970s, there exists little evidence for cross-country convergence of manufacturing technologies within the OECD. In this paper we seek to reconcile these empirical findings by using a two-sector model of capital accumulation that incorporates endogenous technological change (knowledge). To do so, we examine the transition dynamics and convergence characteristics of a new class of non-scale growth models. In many respects these models are a hybrid of endogenous and neoclassical models, and indeed the traditional Solow-Swan model is a special example. Technology is endogenous as in Romer (1990), and emerges as the outcome of agents’ optimizing behavior, while the dynamic These analyses controlled for parametric differences across countries such as savings rates. The introduction of adjustment costs can slow the speed of convergence, while factor mobility increases it, see King and Rebelo (1993), Ortigueira and Santos (1997), and Barro and Sala-i-Martin (1995). Non-scale refers to the characteristic that variations in the size or scale of the economy do not permanently alter its long-run equilibrium growth rate. For example, RD see Jones (1995b) characteristics are similar to those of the neoclassical model. But in contrast to the latter, as our calibration exercises highlight, the two-sector non-scale model generates remarkably plausible convergence speeds, without having to introduce adjustment costs as in Ortigueira and Santos (1997).