The theory about cartel pricing and descriptive price statistics suggests that the price path over a cartel life cycle can be subject to gradual, non-linear transitions where the price path moves from (to) the non-collusive to (from) the maximum collusive equilibrium. Ignoring such transitions can lead to biased estimates of the cartel and damage effects. Smooth transition regression (STR) models are a class of models well suited to capture such transitions, also under realistic conditions when the transition start and end dates are uncertain, and when the two transitions are asymmetric. We evaluate the international uranium cartel during the 1970s, using both the mainstream approach based on a linear specification with a dummy variable to capture the cartel, and an STR model. We are the first to use STR models in the evaluation of a cartel/damage effect. Using the STR model, we find that the damage effect is about 18 times higher as compared to the mainstream model.