Nonlinearities and economic growth new evidence.

摘要

This study examines the implications of the assumption of linear coefficients on models of economic growth. Two variants of the Solow Model of economic growth are studied using both linear estimation (ordinary least squares) and nonlinear estimation (generalized additive models (GAM) and multiple adaptive regression splines (MARS)). Results from earlier studies are replicated with linear estimation and extended with the nonlinear techniques, and earlier empirical models are tested against newly available data extending through the year 2003. The effects of population growth rate, physical capital accumulation rate, and human capital accumulation rate on per worker income are examined. The effects of these variables along with initial income level (at the beginning of the study period) on the rate of income growth are also examined.;Using five cross-section data sets, this study argues that while income per worker is correlated positively with investment in capital (both physical and human) and negatively with population growth, a simple linear empirical model oversimplifies these relationships. Specifically, the effect of human capital accumulation (or level) on income is not linear but increasing in the accumulation rate (or level) of human capital.;Further, nonlinearity is discovered in the effect of initial income on the rate of growth. The conventional model suggests a negative relationship between initial income and growth rate; this indicates convergence. In this paper, it is demonstrated that there is a stronger negative relationship between initial income and growth rate for richer countries. The interpretation is that richer countries grow faster than poor countries. Though not statistically strong, this effect is observed across multiple periods and data sets. In some cases, investment in human capital appears to overcome the negative effect of low initial income.