We evaluate an experimental program in which the French public employment service anonymized resumes for firms that were hiring. Firms were free to participate or not; participating firms were then randomly assigned to receive either anonymous resumes or name-bearing ones. We find that participating firms become less likely to interview and hire minority candidates when receiving anonymous resumes. We show how these unexpected results can be explained by the self-selection of firms into the program and by the fact that anonymization prevents the attenuation of negative signals when the candidate belongs to a minority.