There is an ongoing debate on whether the recent surge in unemployment in the US is due to cyclical or structural factors. The distinction is important because the policies required to reduce each type of unemployment are very dierent. We use a novel approach to empirically quantify the extent of structural unemployment and to shed light on its sources. We propose a simple model of a segmented labor market. Within each segment, search frictions generate unemployment. In addition, there is structural unemployment due to heterogeneity across segments. Four different types of adjustment costs give rise to structural unemployment: worker mobility costs, job mobility costs, wage setting frictions, and heterogeneity in matching efficiency. We construct data on job and worker surplus and job and worker finding rates and use them to estimate these adjustment costs and assess the contribution of each to unemployment. We find that, across US states, worker mobility costs are larger than job mobility costs. Across industries, mobility costs are very high for both workers and jobs. We then construct a time series for structural unemployment and its components. This helps us to understand the extent, to which unemployment is driven by structural factors in the current and past recessions.