According to recent and largely untested theories, unemployment benefits (UBs) could improve the extent and quality of job reallocation even at the cost of increasing unemployment. Using yearly panel data from a large number of countries, we evaluate empirically the relationship between unemployment benefits and structural change. Unlike previous work assessing the effects of UBs on labor market stocks, we focus on flows and rely on policy "experiments", notably the introduction from scratch of unemployment benefits in many countries. We exploit the longitudinal nature of our data to lessen the potentially important selection, endogeneity and omitted variables problems. We find a positive, sizable and significant effect of the introduction of UBs on job reallocation, arising mainly from the job destruction margin although this effect fades away over time. UBs are also found to induce more sectoral shifts from agriculture to services. These findings appear to be robust to changes in the countries in the sample, control variables or estimation methods. We discuss to which extent our results are consistent with equilibrium matching models with or without endogenous sorting of workers into jobs providing entitlement to UBs and stochastic job matching.