In this paper we propose and implement a mechanism of modeling the price indices of food purchases by income groups of households. These indices could be interpreted as differentiated by income food inflation. This approach is based on the differences in prices of purchases for the income groups within each year. We provide the calculations of these indices for the RLMS data and Households Budget Survey conducted by Rosstat (HBS). We discuss possible modifications of the proposed procedure for goals of forecasting of inflation differentiated by income groups. In the result of the comparison with direct calculation of inflation separately for each income group we conclude that the proposed in the paper approach has several advantages, including lower requirements of amount of incoming information.