A systematic approach for incorporating taste variation into a revealed preference framework for heterogeneous consumers is developed. We create a new methodology that enables the recovery of the minimal variation in tastes that are required to rationalise observed choice patterns. This approach is used to examine the extent to which changes in tobacco consumption have been driven by price changes or by taste changes, and whether the significance of these two channels varies across socioeconomic groups. A censored quantile approach is used to allow for unobserved heterogeneity and censoring of consumption. Statistically significant educational differences in the marginal willingness to pay for tobacco are recovered. More highly educated cohorts are found to have experienced a greater shift in their effective tastes away from tobacco.