Classical growth convergence regressions fail to account for various sources

of heterogeneity and nonlinearity. While recent contributions are able to address either

the one or the other, we present a simple two-step method to address both issues. Based

on a slightly augmented version of a recently proposed algorithm to identify convergence

clubs, we formulate a flexible nonlinear framework which allows to analyze convergence

effects on both individual and club level, while alleviating potential misclassification in

the club formation process using simultaneous smoothing over the club structure. The

merits of the method are illustrated for data on different aggregational levels.