The consumer price index (CPI) is designed to track the overall change in prices at a constant level of product quality. It also serves as a measure of cost-of-living changes when it is used for index-linking wages and private contracts. It is therefore important to make sure that the CPI is a good indicator of cost-of-living changes. The authors accordingly seek to test the existence of a divergence between the CPI and cost-of-living changes by means of an intertemporal comparison of Engel curves for food consumption. The curves measure the share of food consumption in household expenditures as a function of purchasing power (real income)-a share that decreases with income. Under the assumption that theoretical Engel curves remain constant over time, the shift in the observed curves reflects a divergence between the change in the price index and the change in the actual cost of living. We borrow this methodology from Costa (2001) and Hamilton (2001) and apply it to French data taken from the CPI series and the “Family Budget” (Budget de Famille) surveys from 1979 to 2006. We perform the analysis by household category using a variety of methods: methods that are robust to the excessive influence of certain observations; instrumentation against measurement errors and endogeneity of total expenditures; and data fitting to adjust for any incorrect reporting. The results differ according to whether we work on data taken directly from the survey or on data fitted to the national-accounts aggregates. Therefore, while our results cannot have a normative value, the intertemporal analysis of Engel curves does not support the claim that the French CPI underestimated the rise in the cost of living between 1979 and 2006.