Purpose This research aims to analyse the housing demand in northern France with respect to socio-demographic variables and the distance between the residence and the workplace. Design/methodology/approach Econometrics with discrete choice models are used to study the three main dimensional choices of housing demand: tenure, type and location. A contribution is to use a heteroscedastic logit model where the variance of the error term is allowed to differ over alternatives and to capture in particular the heterogeneity of tastes. As a matter of fact, household characteristics are very likely to influence the magnitude of the scale parameter in the choice of housing alternatives and then influencing the results if it is not taken into account. Applications for housing demand are nearly non-existent. This paper fills this gap. Findings Econometric estimation confirms that residential choices are influenced by age, income and size of the household, as well as by the rent-to-income ratio. An increase in any of these variables decreases the probability of choices of all the alternatives other than the most often chosen alternative (which is for this application house ownership in the suburb). Moreover, the distance to work systematically influences the housing choice for single-parent families and two-earner households. Additionally, preferences are found to significantly differ between local housing markets, specifically between Lille (a large agglomeration and capital city of the North area) and Dunkerque (an industrialised area). The geographical areas are defined based on INSEE employment zones (“zones d’emplois”). Research limitations/implications This research has been performed for the north of France and may not hold for other areas even though the methodology can be replicated and the mechanisms at play are quite similar elsewhere. Practical implications An important conclusion for sustainable development is the importance to improve city centre amenities relative to those of the suburb or to increase the services associated with high-density dwelling because clearly the most desired alternative remains a house in the suburb. The housing market in the Dunkerque area has some special features characterised by a strong industrial landscape (with port and heavy-duty industrial activities). In this context, amenities provided by the city centre offset the strong attraction of a house in the suburb. Social implications This research shows that households with similar characteristics tend to prefer the same type of real estate property. Therefore, to avoid social segregation, it is important that housing supply respond to different household preferences and needs in the different segments of the housing market. Moreover, the housing supply should take into account the specificities of the geographical areas (both in terms of population who may have a different profile) and in terms of amenities. Originality/value This research is one of the very few conducted ones on discrete housing choices in France (with the notable exception of De Palma et al. , 2007 for the choice of location). Three simultaneous choices are considered: tenure (including social housing which is almost always ignored), type of building and location. The authors have shown that it is important to take into account the heterogeneity of the preferences in the econometric model with a heteroscedastic logit model.