Date of Award
Spring 2023
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Economics
First Advisor
Abaluck, Jason
Abstract
This dissertation includes two chapters on public economics and health economics, centered around designing public policies when agents make behavioral mistakes. In the first chapter, I study the nono-fungibility of SNAP and its normative implications. The second chapter studeis the optimal design of the health insurance when when the insurers can use bureaucratic cost as a policy instrument.The first chapter studies the non-fungibility of SNAP and its normative implications using novel consumer panel data of shopping history across multiple grocery retailers and restaurants. Exploiting a recent state-level policy change in the amount of SNAP benefits, I estimate that the propensity to consume groceries out of one dollar amount of SNAP benefits is 44 cents, which is smaller than the estimate of Hastings and Shapiro (2018) from a single retailer, which is 59 cents. I show that an estimate from a single retailer can be substantially biased when shoppers differentially select which payment method to use across stores, and I document substantial selection in practice. SNAP participants may want to use cash or credit cards in stores that attract more affluent consumers and to use SNAP benefits in stores with fewer affluent stores. This finding suggests that researchers should be careful in the selection of payment methods when partnering with a grocery retailer for SNAP-relevant research. Next, I assess the implications of non-fungibility from the perspective of two distinct normative philosophies. Following Chetty (2009) and Whitmore (2002), I derive sufficient statistics for calculating efficiency loss. The formula requires estimating the MPC of groceries out of SNAP and the price elasticity of grocery baskets. Using a shift-share design that exploits the geographical distribution of national retailers, I estimate the price elasticity of groceries to be 0.4 and the deadweight loss of providing an in-kind transfer instead of a cash transfer to be 12 cents out of each dollar of the benefit. Annually, the efficiency cost is $12 billion in total. To understand SNAP’s nutritional impact, I estimate SNAP’s effect on food-away-home expenditure and find that a 100% increase in SNAP decreases food-away-from-home spending by 13%. Linking the granular product-level consumption data of both food-at-home and food- away-from-home to a nutrition database, I find that providing in-kind rather than cash leads to a large increase in calories per day in food purchases without providing a gain in dietary quality. The second chapter studies optimal insurance when the insurers can use bureau- cratic costs as a policy instrument. Economists conventionally think of insurance as trading off risk protection and moral hazard. Cost-sharing mitigates overconsumption but reduces risk protection. In the last decade, prior authorization has become an increasingly important alternative means for insurers to control costs. Prior authoriza- tion imposes a bureaucratic burden on physicians for prescribing some drugs, especially high-cost ones. This paper studies optimal insurance when the insurers can use both cost sharing and prior authorization, and applies our results to the Medicare Part D prescription drug insurance market. First, we consider a model where patients and doctors both correctly understand the value of each treatment. In this model, the key optimality condition is that the cost to risk protection of the coinsurance increase necessary to avert a dollar of treatment must equal the “paperwork costs” that must be imposed on physicians to avert a dollar of treatment. Taking this condition to the data requires estimates of conventional elasticities of utilization with respect to coinsurance rates as well as estimates of elasticities of utilization with respect to prior authorization. We estimate both, finding that the latter are considerably larger; our model implies that in the status quo, prior authorization is much less costly than cost-sharing on the margin. We expand our model that departs from the assumption about the patient’s rational decision and re-investigate the optimal design of insurance. This is guided by the fact that doctors are more informed about the value of medical treatment than patients, and prior authorization directly affects’ the doctor’s prescription decision, while cost-sharing affects the patient’s decision as well. Our model shows that there is an additional gain to consider in this new model. Lowering the coinsurance rate while increasing prior authorization costs can efficiently target the medical expenditure for utilizing high-value rather than low-value drugs. We derive a prediction of this behavioral hazard model, if there is behavioral friction on the patient side, the marginal rate of substitution (MRS) between cost-sharing and prior authorization (e.g., a change of previous authorization cost needed to offset the effect of a coinsurance rate change on medical expenditure) should be the same regardless of the value of drugs. We show that MRS is much higher for high-value drugs, consistent with our model. Our findings suggest that we consider this targeting gain for the optimal coinsurance rate and bureaucratic cost, and we conclude the paper with discussions about its empirical implementations.
Recommended Citation
Song, Jintaek, "Essays in Public Economics" (2023). Yale Graduate School of Arts and Sciences Dissertations. 1039.
https://elischolar.library.yale.edu/gsas_dissertations/1039
