We investigate the employment effects of job access within cities. To overcome the endogeneity of residential location choices, we exploit the quasi-random allocation of applicants across affordable housing units in five German cities and propose a novel IV strategy. Our job access measure combines proximity to jobs in both geographical and skill space, taking into account potential competition from other workers. We compile a unique administrative dataset comprising the complete labor market biographies of individuals who moved into affordable housing. A one-standard-deviation increase in job access reduces the unemployment hazard of unskilled workers by 3.4 percentage points. We do not find similar effects for skilled workers, who are more likely to own a car and are therefore less geographically constrained. Simulations show that full segregation, with all unskilled allocated to central neighborhoods, would reduce the city-wide unemployment rate of the largest city in our sample, Munich, by 1.3 percentage points.