Competing models of global dynamics: evidence from panel models with spatially correlated error components

This paper considers two competing theories with differing implications for the long run dynamics of the global economy, neoclassical theory, implying conditional convergence, and New Economic Geography, implying multiple equilibria. These rival non-nested hypotheses are combined within an artificial nesting, panel data, model, embodying a spatial autoregressive error process allowing for unmodelled inter-country heterogeneity and spatially dependent errors, and tested using a panel of 77 countries over the years 1970, 1980, 1990 and 2000. The New Economic Geography theory is shown to have superior explanatory power, but it also introduces some significant problems of estimation and measurement.