Abstract
Using new consumer-panel data across multiple grocery retailers, I estimate that an additional SNAP dollar raises food-at-home spending by 48 cents (s.e. 6 cents). This multi-retailer estimate is broadly in line with the 59-cent estimate in Hastings and Shapiro (2018), reinforcing evidence that SNAP benefits are non-fungible. Linking food-at-home and food-away-from-home purchases to nutritional data, I find that SNAP substantially increases calorie purchases but has little effect on nutritional composition per calorie. A behavioral welfare calibration implies an efficiency cost of in-kind provision of 12–14 cents per SNAP dollar; because this calculation is highly sensitive to the MPC, pinning down the parameter is quantitatively important.