The charging of school user fees is a much-debated policy issue in developing countries. In this paper, I evaluate the impact of a South African fee elimination program that was targeted at the poorest two quintiles of schools based on a community poverty score. Fixed effects estimates find that the program increased enrollment by almost 2 percentage points in treated secondary schools, an increase concentrated in earlier secondary grades. There is substantial heterogeneity in the estimated secondary school effect: it is driven entirely by an increase of around 3.5 percentage points in the poorer of the two treated quintiles. Regression discontinuity estimates confirm that the relatively wealthy schools near the treatment cutoff did not experience any effects on enrollment. Overall, the abolition of fees seems to have been reasonably effective in increasing secondary school enrollment in particularly poor communities. This is despite the fact that the eliminated fees were relatively low, comprising only around 1.5% of annual household income (per child) in these communities.