Social insurance programs, including Unemployment Insurance (UI), have been adopted in many countries where informal employment is prevalent and monitoring of eligibility imperfect. Whether social insurance programs can increase welfare in such a context often remains an open question. To address this issue, we study the Brazilian UI program. Social insurance trades off welfare gains from providing income support and efficiency costs from distorting behaviors. Imperfect monitoring may both exacerbate behavioral responses and, with the possibility to work informally, reduce the need for insurance. Using matched employee–employer data, and two complementary empirical strategies, we estimate the impacts of UI extensions on program and labor market outcomes and their efficiency costs. We find large percentage reductions in hazard rates of formal reemployment, in particular around benefit exhaustion. However, because hazard rates are very low, UI has little impact on formal reemployment and efficiency costs amount to only 5%–11% of the cost of extending benefits. Using survey data, we further estimate that 35% of our sample of job–losers is actually unemployed after 5 months of UI, a figure comparable to the US. In our normative framework, these results imply that even a low social value of insurance is consistent with an increase in welfare from extending UI. We obtain that welfare effects from the existing UI program are likely to be positive and may be sizeable.