In this paper, I use a unique empirical setting that allows me to compare the performance of teachers hired in a discretionary process led by the teachers’ union in Mexico with the performance of those hired on the basis of a screening rule (test scores on a standardized exam). My results show that the discretionary hires perform considerably worse than the rulebased hires (as measured by value added to student achievement). The evidence presented here shows the impact of personnel selection mechanisms on the quality of public service delivery.