Recent tests for the convergence hypothesis derive from regressing average growth rates on initial levels: a negative initial level coefficient is interpreted as convergence. These tests turn out to be plagued by Francis Galton's classical fallacy of regression towards the mean. Using a dynamic...