This dissertation considers the determinants of countries' real income per capita, controlling for the effects of their physical location. We use meta-analysis to summarize the results of previous research, and then apply a new estimator to control for spatial effects in a system of simultaneous equations. In the meta-analysis we consider a new sample of 86 previously published empirical studies offering 1,267 distinct estimates of annual convergence rates, roughly half of which use some sort of spatial technique. We find that, as predicted by econometric theory, using spatial lags alters estimated convergence rates, whereas using spatial error methods affects only the precision of those estimates. Accounting for spatial lags turns out to significantly reduce estimated rates of convergence, as a large fraction of low-income countries' growth is in fact associated with their location as opposed to their initial income. Our spatial system approach helps analyze how location influences growth, using key variables for 95 countries from 1960 through 2002, pooled into nine five-year averages centered on 1960, 1965, and so on through 2000. We use simultaneous equations to identify alternative channels through which country characteristics might affect income. Each channel is tested with a new generalized moments (GM) estimator for spatial autoregressive coefficients, allowing for spatial error correlation, correlation across equations, and the presence of spatially lagged dependent variables. Results of the spatial system support both "institutionalist" and "technological" determinants of income. A time-varying index of each country's institutional quality has a strong independent effect on that country's current income. But there is also a persistent effect of each country's physical geography, through observable factors such as seasonal frost, malaria transmission, and coastal location that are correlated with income through their links to the endogenous levels of agricultural output, health, urbanization and trade. The implication of these results is that, to overcome geographic barriers countries would need exogenously provided but location-specific technologies, particularly in agriculture and health. More generally, taking account of spatial location in a systems context improves the performance of growth models, offering a promising approach for future work.
