We study the dynamics of the cross-section distribution of patents per capita for the 48 continental US states from 1930 to 2000 using a discrete-state Markov chain. We test for and find evidence in favor of the (knowledge) convergence hypothesis. The distribution of patents is converging to a limiting distribution that is significantly more concentrated than its initial distribution. States in the extreme are more mobile than states in the middle of the cross-sectional distribution and are likely to move to the middle. However, the rate of convergence to the limiting distribution is slow.