On Specifying and Estimating Economic Growth as a Spatial Process : Convergence, Inequality, and Migration

Sammanfattning: This thesis includes three self-contained papers. The first paper considers the effect of geographically dependent observations on cross-sectional growth convergence and proposes a way of decomposing the level of technology taking into account geographical variation in growth rates. It adopts spatial parametric methods; shows how and why a correlated errors approach may be implemented; and deals with both spatial heterogeneity and spatial autocorrelation. In the context of a cross-sectional test of convergence, geographical models and estimation procedures are then applied to data on per capita income amongst US states. The implication of a spatial approach compared to a non-spatial one in terms of bias and efficiency are discussed.The second paper addresses the question of whether or not there are any systematic geographical patterns in the distribution of income, and if so, how to deal with them. A spatial version of the well-known sigma convergence is derived. It is shown that the effects of geographical dependence on the evolution of inequality do not change the time-path configuration, but it can generate scale effects and possibly also mirror effects. Positive spatial dependence means that regions will form clusters of high and low income levels which become increasingly more differentiated. Statistical versions of the model are defined and procedures for fitting the models are described using spatial statistical methods. This is then demonstrated on per capita income amongst states in the U.S. 1940-1990.The third, and final, paper explores and discusses some aspects of migration in regard to regional economic growth convergence. Observed migration is considered to be the outcome of successful matching in the labour market. Of crucial importance to the matching processes is the speed of filling vacancies, which in turn depend on labour market conditions. Adjustment costs due to spatial mismatch in the labour market are suggested to affect the economic growth process. The relevance of such an approach in a setting of growth convergence is empirically investigated using aggregated data for the U.S. states. In order to capture the spatial aspects of the data, estimation is undertaken by simultaneously specified spatial Gaussian regression models.