This paper analyses the effects that the 2012 VAT reform in Spain had on households’ welfare, focusing on a major expenditure group: food and non‐alcoholic beverages. Households’ demands are modeled as a two‐stage Quadratic Almost Ideal Demand System, which is then estimated by means of a consistent two‐step estimator introduced in Tauchmann () and not previously used in studies of this type. This procedure allows consistent imposition of the traditional parameter restrictions that utility maximisation requires in the context of a censored model. Our results show that the welfare loss and the increment in the tax bill increase with income. We also show that expenditure on food and non‐alcoholic beverages grows less than proportionately with income. Consequently, households with lower income experience a greater welfare loss relative to their income levels. In short, the 2012 VAT reform in Spain, focusing on this expenditure group, can be considered as regressive.