Session: Selecting Insurance
Room: Upson 117
Time: Wed 10:15-11:45

Presenter: Jeremy Barofsky (Harvard School of Public Health. Health Economics, Global Health and Population)

Discussant: Marisa DominoUniversity of North Carolina, Chapel Hill

Abstract

RATIONALE: Individuals often express long-run preferences that conflict with their short-run behavior. We would like to exercise, eat well, and save more, but regularly ignore our well-laid plans. In addition, we frequently express interest in health insurance coverage when sick, yet evidence from Mexico’s free health insurance program suggests that we often fail to enroll when healthy. Indeed, estimates of take-up rates for various social programs range from 40% to 80% (Dahan and Nisan, 2006), meaning that low-income families often "leave money on the table" that could improve their lives. As Currie, 2006 explains, the literature generally provides three reasons: 1) stigma, 2) lack of knowledge, and 3) transaction costs.

OBJECTIVES: This study advocates a new explanation for low uptake in social insurance programs: inconsistency between short- and long-run preferences due to procrastination. Specifically, we model time inconsistency using the quasi-hyperbolic discounting model from Laibson, 1994 where β < 1 refers to time inconsistent preferences and δ represents the discount rate. We borrow heavily from the framework laid out in Carroll et al., 2005 which describes an agent's decision to opt-out of a 401(k) default contribution rate. This work extends Carroll’s model to an agent that weighs the immediate cost of affiliation against the expected discounted future benefits in avoided medical costs. The theoretical model predicts which types of households will affiliate and we test these predictions using data from Mexico’s recent health insurance expansion called Seguro Popular (SP). Even though nearly none of the affiliates paid an insurance premium, the data shows that over 35% of families did not affiliate after 10 months of intensive enrollment efforts. The data allows tests of the knowledge and transaction-cost hypotheses and permits estimates of expected household benefits to SP using the baseline survey.

METHODS: First, we test for evidence that both information and transaction costs produce significant changes in affiliation rates using logistic regression. Then, we implement a calibration by estimating each household’s expected medical costs (based on last period’s spending) and affiliation costs from a qualitative enrollment study. Plugging these estimates into the theoretical model produces an enrollment probability for each household which, when aggregated, generates a predicted percentage of affiliated households that varies by the β’s and δ’s chosen. We produce an envelope of β-δ discount rates that are consistent with the affiliation data, using actual minus predicted percent affiliated as our measure of model fit.

RESULTS: We find an association between household size and enrollment suggesting a role for transaction costs in uptake, but no evidence that knowledge raises affiliation. When we assume a standard δ of 0.03, we find that the β’s which minimize model error are between 0.8 and 0.9, indicating the presence of time-inconsistent behavior.

CONCLUSIONS: If transaction costs constitute the main reason for lack of take-up, then low enrollment reflects a rational decision and our policy response should be minimal. However, if low uptake occurs because individuals with time-inconsistent preferences plan on acting in the future and never do so, then policy interventions to encourage affiliation are welfare improving and governments must implement further initiatives to encourage health-care coverage.

Key Terms

Health insurance, program take-up, hyperbolic discounting

Authors:

Jeremy Barofsky (Harvard University, School of Public Health. Health Economics, Global Health and Population)