Heterogeneous Convergence

We use U.S. county-level data containing 2,696 cross-sectional observations and 34 conditioning variables to study heterogeneity in convergence rates across 32 individual U.S. states. Using a 3SLS-IV estimation method, we find significant heterogeneity in the state-level convergence rates. Across the 32 states studied, the point estimates range from 3.8 percent (California) to 15.6 percent (Louisiana), with an average of 8.1 percent. We find that the convergence rates are negatively correlated with the initial income.