Negative growth and convergence of the Mexican states -: A panel data analysis

This paper attempts to characterize the convergence process among the Mexican states during the period 1970-1995. The proposed methodology uses dynamic panel data models without exogenous regressors, under the stationarity assumption. It is argued that the Breusch-Pagan (1980) LM-test and the F-test for individual effects in the panel, can be used jointly to discriminate between the hypothesis of absolute and conditional convergence. Given that the temporal dimension of the panel is small we use Monte Carlo simulations to asses the magnitude of the biases of the AR parameter and we correct the estimates using median-unbiased estimation. It is found that the dynamics of quinquennial per capita GDP of the Mexican states during this period is consistent with conditional convergence. Also it is found that the long run growth rates implied by the model are negative, which is consistent with the continuous slow down of the quinquennial growth process observed in the data for thus period.