When regional disparities follow a cyclical short-run pattern, convergence analysis results can be sizably distorted. To tackle this issue, we propose a method based on the extraction of the trend from regional income time series that eschews misleading results when the nature of the cyclical pattern changes over time. Using real per capita personal income data for forty-eight conterminous US states and the distribution dynamics approach, we identify the following three distinct consecutive phases: strong convergence (1930–1970), substantial persistence (1971–1980), and divergence (1981–2010).