Food Subsidies at the Base-of-the-Pyramid: Take-up, Substitution Effects and Nutrition

This paper studies food purchasing patterns among low-income individuals in India using objective scanner data with an aim to estimate the effects of food subsidies and improve their design. We digitally tracked the purchase histories of over 20,000 individuals by installing scanners in 39 food-selling vendors including groceries, butchers, and street food sellers, in an underserved urban settlement in India. We document heavy consumption of packaged junk foods. In groceries, the share of calories coming from the purchase of packaged snacks (e.g., chips or candy) exceeds that from the purchase of rice and wheat combined. To study how food subsidies impact shopping behavior and in turn nutrient purchases, we opened a store mimicking the Indian government’s food subsidy stores, and randomly assigned 1,258 individuals to a weekly rice and wheat subsidy treatment arm or to a no-subsidy control arm, for a 6-week period. Subsidy take up was 82%. We find that our in-kind subsidy induces both substitution and complementarity effects on beneficiaries’ grocery transactions. The purchase of packaged snacks decreases whereas that of complementary foods – such as spices and accompaniments – increases – with no detectable change in total grocery spending or spending at other tracked food vendors. These patterns indicate a substitution from snacking out towards home-cooked meals. These effects are most prominent for working parents. We find no evidence of a negative spillover on the purchase of calories, carbohydrates, protein, fat, and several micronutrients, suggesting that nutrients delivered through our in-kind subsidy program are likely to have increased beneficiaries’ nutrient intake. Finally, we explore the design of alternative, customized in-kind programs where we exogenously vary the subsidized foods. Based on estimates of the take-up rates, we uncover a tradeoff between the nutrient richness of different staples and their attractiveness to customers. We find that subsidizing lentils may increase the amount of proteins delivered through the program.