Using a large field experiment, we show that racial composition of employer neighborhoods predicts employment discrimination patterns in a direction suggesting in-group bias. Our data also show racial disparities in the geographic distribution of job postings. Simulations illustrate how these patterns combine to shape disparities. When jobs are located far from Black neighborhoods, Black applicants are doubly disadvantaged: discrimination patterns disfavor them, and they have fewer nearby opportunities. Finally, building on prior work on Ban-the- Box laws, we show that employers in less Black neighborhoods appear much likelier to stereotype Black applicants as potentially criminal when they lack criminal record information.