Working Paper: NBER ID: w16969
Authors: Liran Einav; Amy Finkelstein; Stephen P. Ryan; Paul Schrimpf; Mark R. Cullen
Abstract: In this paper we explore the possibility that individuals may select insurance coverage in part based on their anticipated behavioral response to the insurance contract. Such "selection on moral hazard" can have important implications for attempts to combat either selection or moral hazard. We explore these issues using individual-level panel data from a single firm, which contain information about health insurance options, choices, and subsequent claims. To identify the behavioral response to health insurance coverage and the heterogeneity in it, we take advantage of a change in the health insurance options offered to some, but not all of the firm's employees. We begin with descriptive evidence that is suggestive of both heterogeneous moral hazard as well as selection on it, with individuals who select more coverage also appearing to exhibit greater behavioral response to that coverage. To formalize this analysis and explore its implications, we develop and estimate a model of plan choice and medical utilization. The results from the modeling exercise echo the descriptive evidence, and allow for further explorations of the interaction between selection and moral hazard. For example, one implication of our estimates is that abstracting from selection on moral hazard could lead one to substantially over-estimate the spending reduction associated with introducing a high deductible health insurance option.
Keywords: health insurance; moral hazard; selection
JEL Codes: D12; D82; G22
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| introduction of new health insurance options (G52) | reduction in total medical spending (H51) |
| selection on moral hazard (D82) | overestimation of spending reduction (H68) |
| not accounting for selection on moral hazard (C24) | estimated spending reduction of $350 per person (G52) |
| accounting for selection effects (C34) | estimated spending reduction of $130 per person (H51) |
| comprehensive coverage (G52) | greater behavioral response to changes in coverage (E71) |