The time series properties of per-capita income in U.S. regions are te
sted for consistency with the neoclassical growth model's prediction o
f per-capita income convergence. Two conditions are required for conve
rgence. Shocks to relative regional per-capita incomes should be tempo
rary (stochastic convergence), and initially poor regions should catch
up to rich regions (beta-convergence). We find evidence for stochasti
c convergence across U.S. regions during the 1929-90 period after allo
wing for a trend break in 1946. We also find that U.S. regions have ac
hieved beta-convergence. These findings support the neoclassical model
's prediction of conditional convergence.