Abstract
Long-term regional convergence hypothesis is examined for 32 Mexican states in a regional growth model with poverty traps using a new dataset on regional income inequality for the period 1940-2011. Although zero-growth poverty trap hypothesis is rejected for 28 out of 32 states, the evidence confirms B- convergence and s-convergence for the period 1940-1980 and indicates the breakup of convergence in post-1980 period. The break in the convergence process is attributed to trade liberalization carried out under weak state capacity and clientelistic patronage environment without independent and effective regulators. The widening of regional inequality is characterized by an increase in growth in high-income U.S.-border states and no such increase in poorer states that cannot converge to the frontier under such conditions. When long-term convergence relationship is conditioned on unobserved long-run effects, the speed of convergence for pre-1980 period is around 2% per year and diminishes with the estimation horizon. Sensitivity analysis based on income-specific quantile regressions emphasizes substantial heterogeneity in the speed of convergence across states.
