The Value of Unemployment Insurance

In the absence of unemployment insurance (UI) choices, the standard approach to estimating the value of UI is to infer it from the observed consumption response to job loss in combination with some assumption on preferences. Exploiting the unique data and policy context in Sweden, we propose two alternative approaches, which we implement and compare to the standard consumption-based approach on the exact same sample of workers. We nd that the drop in consumption expenditures upon job loss is relatively small ( 13 percent), but the marginal propensity to consume (MPC), estimated using variation in local government transfers, is around 25 percent higher when unemployed than when employed. We show that this wedge in MPCs, the focus of our rst approach, reveals a high relative price of smoothing consumption, con rmed by direct evidence on the limited consumption smoothing means available during unemployment. The estimated relative price provides a lower-bound on the value of UI, which turns out to be substantially higher than the consumption-based estimate under standard preference assumptions. Exploiting the UI choices embedded in the Swedish UI system, we also propose a Revealed-Preference approach, which con rms that the average value of UI is large in our setting, but also reveals substantial dispersion in the value of UI, above and beyond the variation in consumption drops.