Matthew Notowidigdo


Consumer-Financed Fiscal Stimulus: Evidence from Digital Coupons in China


We study a new form of fiscal stimulus undertaken by municipalities across China starting in 2020: government-issued digital coupons designed to encourage spending in certain categories such as restaurants, groceries, and entertainment. Using unique account-level and transaction-level data from a large online shopping platform that distributed many of the government coupons, we estimate the effect of the coupons on spending for many different types of coupons with different spending thresholds (e.g., “Spend at least ¥X, get ¥Y off”). We identify the effects of the coupons on spending using a bunching estimator that uses the transaction-level spending distribution in the weeks before each coupon is distributed as the counterfactual, and we validate the bunching estimates using the random assignment of a subset of the coupons in our data. We find that the coupons cause large and persistent increases in spending in the targeted spending categories, and we do not find evidence of any substitution away from spending on the platform in non-targeted spending categories. We estimate that the consumer spending increases by 2.7-2.8 yuan per yuan spent by the government. As a result, we conclude that the digital coupons increased spending substantially in the targeted spending categories at very low fiscal cost. We show that a standard consumption model can generate these results since the coupons’ spending thresholds create “notches” that lead to large spending responses from consumers. We calibrate the consumption model to match our empirical results, and we find that the coupons generate about half of the increase in consumer welfare as an equivalent amount of fiscal stimulus distributed as cash. We then use the calibrated model to simulate alternative coupon designs, and we find that lower coupon thresholds and higher coupon discounts would be less cost-effective but would deliver greater aggregate stimulus.

Book a meeting with Matthew

- 1:1 (four slots)

- PhD (15:00-15:45)