Does screening applicants using exams help or hurt the chances of lower-SES candidates? Because individuals from lower socioeconomic backgrounds fare, on average, worse than those from richer backgrounds in standardized tests, a common concern with this “meritocratic” approach is that it might have a negative impact on the opportunities of lower-SES individuals. However, an alternative view is that, even if such applicants underperformed on exams, other (potentially more discretionary and less impersonal) selection criteria might put them at an even worse disadvantage. We investigate this question using evidence from the 1883 Pendleton Act, a landmark reform in American history which introduced competitive exams to select certain federal employees. Using newly assembled data on the socioeconomic backgrounds of government employees and a difference-in-differences strategy, we find that, although the reform increased the representation of “educated outsiders” (individuals with high education but limited connections), it reduced the share of lower-SES individuals. This decline was driven by a higher representation of the middle class, with little change in the representation of upper- class applicants. The drop in the representation of lower-SES workers was stronger among applicants from states with more unequal access to schooling as well as in offices that relied more heavily on connections prior to the reform. These findings suggest that, although using exams could help select more qualified candidates, these improvements can come with the cost of increased elitism.