Heterogeneity in Convergence Rates and Income Determination across U.S. States: Evidence from County-Level Data*

We utilize county-level data to explore growth determination in the U.S. and possible heterogeneity in growth determination across individual states. The data includes over 3,000 cross-sectional observations and 39 demographic control variables. We use a consistent two stage least squares estimation procedure. (We report OLS estimates as well.) The estimated convergence rate across the U.S. is about 7 percent per year – higher than the 2 percent normally found with OLS in cross-country, U.S. state, and European region samples. Estimated convergence rates for 32 individual states are above 2 percent with an average of 8.1 percent. For 29 states the convergence rate is above 2 percent with 95 percent confidence. For seven states the convergence rate can be rejected as identical to at least one other state’s convergence rate with 95 percent confidence. In examining the determinants of balanced growth path heights, we find that government at all levels of decentralization is negatively correlated with economic growth. Educational attainment of a population has a non-linear relationship with economic growth according to our estimates: growth is positively related to high-school degree attainment, seemingly unrelated to obtaining some college education, and then positively related to four-year degree or more attainment. Also, finance, insurance and real estate industry and entertainment industry are positively correlated with growth, while education industry is negatively correlated with growth. Heterogeneity in the effects of balanced growth path determinants across individual states is much harder to detect (or dismiss) than in convergence rates.