We study food Engel curves among the poor population targeted by a conditional cash transfer programme in Colombia. After controlling for the endogeneity of total expenditure and for the (unobserved) variability of prices across villages, the best fit is provided by a log-linear specification. Our estimates imply that an increase in total expenditure by 10% would lead to a decrease of 1% in the share of food. However, quasi-experimental estimates of the impact of the programme on total and food consumption show that the share of food increases, suggesting that the programme has more complex impacts than increasing household income. In particular, our results are not inconsistent with the hypothesis that the programme, targeted to women, could increase their bargaining power and induce a more than proportional increase in food consumption.