A comparison of logit and nested logit models with random coefficients

Résumé

We propose a comparison across four types of demand models: a pure logit model, a logit model with random coefficients, a pure nested logit model and a nested logit model with random coefficients. Through the application of these techniques to a recent dataset on the European car market, we show that taking into account the nesting structure is important even when random coefficients on continuous valued product characteristics are allowed. Consequences in terms of elasticities, relevant market definition and merger simulation for each of the four models are analyzed.