We investigate unemployment due to mismatch in the US over the past three decades. We propose an accounting framework that allows us to estimate the overall amount of mismatch unemployment as well as the contribution of the frictions that caused the mismatch. Mismatch is quantitatively important for unemployment and the cyclical behavior of mismatch unemployment is very similar to that of the overall unemployment rate. Geographic mismatch is driven primarily by wage frictions. Mismatch across industries is driven by wage frictions as well as barriers to job mobility. We find virtually no role for worker mobility frictions.