Convergence of nations' wealth: an empirical approach.

Summary This thesis focuses on testing the hypothesis of convergence of GDP per capita across nations. In Chapter I, we summarize the theoretical debates and present the main approaches followed to test the convergence hypothesis of GDP per capita. Our definition of convergence corresponds to the hypothesis of stationary GDP per capita differentials of zero mean. In chapter II, we characterize one by one the GDP per capita series of 16 industrialized countries between 1900 and 1989 and of 23 OECD countries between 1950 and 1988 and show that they contain a stochastic trend. In chapter III, we test the convergence hypothesis itself. We characterize the GDP gaps separately and then simultaneously. The rejection of the stationarity of the gaps leads us to look for cointegration relationships between GDP per capita in order to determine whether there are no common stochastic trends. In chapter IV, we apply unit root tests with panel data. These tests reject the stationarity of the differences for 100 countries between 1960 and 1985. We accept stationarity of deviations with a non-zero mean for OECD countries between 1950 and 1988. In chapter V, we consider a less strict definition of convergence: the lower the initial GDP per capita, the higher the growth rate per capita if the countries have common levels for some variables. We then estimate a heterogeneous dynamic panel model.