The estimation of preference heterogeneity is of key importance in the choice modelling literature. We propose a semi-parametric estimation approach to analyze observed heterogeneity. Under the assumption that more similar people in terms of socio-economic characteristics have more similar preferences, we show how to estimate the nonparametric distribution of preferences and show how to perform statistical tests. The model is applied to a stated choice experiment intended to measure the willingness to pay for travel time savings and arriving at the preferred arrival time at work. We find that there is substantial observed heterogeneity.