This study investigates the question of convergence among 15 OECD countries, where convergence is defined in several ways, for example as joint stationarity of differences of log per capita real output. No convergence is found for the entire set of OECD countries, so a clustering algorithm is used to identify sets of countries for which convergence occurs. The sensitivity of these 'convergence clubs' is investigated to the definition of convergence, the data used, and various details of the econometric procedure.