Research

Job Market Paper

I show that spending responses to cost sharing in health insurance depend not only on contractual incentives but also on household myopia and price perception. Using administrative claims data from a large employer plan, I study a natural experiment that introduced a deductible and raised the out-of-pocket maximum. Households on average reduce spending modestly, but the reduction rises steeply with baseline health risk and is concentrated among the highest-risk households. Because high-risk households should readily anticipate reaching the out-of-pocket maximum under either contract, a structurally estimated forward-looking model predicts that they respond least, contrary to the observed risk gradient. To resolve this puzzle, I develop and estimate the first structural model of moral hazard that jointly incorporates household myopia and imperfect spot-price perception. In the model, myopic households respond to spot prices inferred from medical bills, with two frictions biasing perceived spot prices upward: (1) bills arrive with delays, causing price beliefs to lag behind actual prices, and (2) households may infer the price of subsequent care from the average price on a bill rather than the marginal price. I provide evidence for both mechanisms using administrative data and an incentivized survey experiment. The estimated model matches the modest aggregate response and captures the steep risk gradient that the forward-looking model cannot. More broadly, the model provides a framework for understanding spending responses under nonlinear health insurance contracts when households learn their spot prices gradually and imperfectly.

Status: Under review.

This paper shows that standard models of moral hazard predict a generosity paradox: less generous insurance can increase rather than decrease medical spending. Higher cost sharing makes it easier to reach the out-of-pocket maximum, where the marginal price is zero. Forward-looking consumers near that threshold therefore optimally increase their spending to reach the maximum. Using existing empirical estimates of the health need distribution and moral hazard responsiveness, I find that in many realistic scenarios, decreases in generosity lead to aggregate increases in spending and welfare losses. I discuss the practical implications of this for plan designers.

Suboptimal health insurance choices impose substantial welfare costs on households, with enrollment patterns frequently violating financial dominance despite stakes exceeding thousands of dollars annually. We use a randomized field experiment with public university employees during open enrollment to evaluate whether decision aids that clarify these financial consequences affect enrollment patterns. The setting features a financially dominant high-deductible plan that saves money for all workers regardless of health spending, with typical savings around $2,000 annually, and a requirement to make an active choice confirming plan selection. We find that decision aids improve cost recognition by 22 percentage points. Yet they increase intended enrollment in the high-deductible plan by 6 percentage points and actual enrollment by only 2 percentage points, revealing substantial attenuation from understanding to behavior. Survey responses reveal that concerns about managing a health savings account, aversion to out-of-pocket costs, and reluctance to change from familiar plans limit the translation into enrollment. Treatment effects are largest among workers with limited prior plan engagement and vary substantially by liquidity constraints.

There is a sizable academic literature studying the demand for life insurance products. Little is known, however, about how surviving spouses utilize life insurance payouts. We study individuals 50 and older and ask how they adjust their savings, spending, and bequest behavior after life insurance payouts are received. We show that, compared to widow(er)s not receiving payouts, widow(er)s receiving life insurance payouts do not experience shocks to consumption. We also show substantial heterogeneity in other household finance responses to payouts along the wealth distribution. The wealthiest survivors tend to save their payouts, spend down slowly, and experience little change in their bequest plans. The least wealthy, on the other hand, experience enhanced bequest motives for a short period before quickly spending their payouts. In the long run, these individuals even end up more likely to be receiving government support than widow(er)s without life insurance payouts.