The purpose of this article is to compare a set of multinomial logit models derived from revealed choice data and a decompositional choice model derived from experimental data in terms of predictive success in the context of consumer spatial shopping behavior. Data on consumer shopping choice behavior as collected before the opening of a new major clothing store in a shopping center were used to estimate the parameters of the various models. The estimated parameters were then used to predict market shares of the shopping centers after the opening of the new store. Predicted shares were then compared with data on actual behavior collected after the opening of the new store. Results indicate that the two modelling approaches perform almost equally well.