We start from an aggregate random coefficients nested logit (RCNL) model to provide a systematic comparison between the tractable logit and nested logit (NL) models with the computationally more complex random coefficients logit (RC) model. We first use simulated data to assess possible parameter biases when the true model is a RCNL model. We then use data on the automobile market to estimate the different models, and as an illustration assess what they imply for competition policy analysis. As expected, the simple logit model is rejected against the NL and RC model, but both of these models are in turn rejected against the more general RCNL model. While the NL and RC models result in quite different substitution patterns, they give robust policy conclusions on the predicted price effects from mergers. In contrast, the conclusions for market definition are not robust across different demand models. In general, our findings suggest that it is important to account for sources of market segmentation that are not captured by continuous characteristics in the RC model.