Many countries provide extra resources to schools serving disadvantaged pupils. We exploit a discontinuity in the assignment of such personnel subsidies in Flanders to estimate the impact on cognitive outcomes via a regression discontinuity (RD) design. Because bias can be substantial in RD designs, we include a bias correction in the specification of the control function. Overall, we find positive effects for mathematics, reading and spelling, but the impact is significant for spelling only. The effects are larger for disadvantaged pupils defined on the basis of family background, smaller – or less reliable – for low initial performers, and again larger at schools that used the resources to foster socio-emotional development.