This paper evaluates a hiring subsidy for lower-educated youths in Flanders (Belgium) that reduced labour costs by approximately 13% for a period of two years, starting in 2016. Using a donut Regression Discontinuity Design, we find no evidence that the subsidy improved the job finding rate of eligible job seekers in 2016-19, a period marked by a tight labour market. We then investigate the role of temporary work agencies, which disproportionately employ the target group and obtain 25% to 34% of the subsidies. Using Difference-in-Differences regressions, we demonstrate that agencies did not raise wages of eligible agency workers in response to the policy. Remarkably, despite a 3.3% labour cost reduction, full-time equivalent employment of eligible workers in these agencies decreased by 9.2% over the three years following the reform. Our findings highlight how an active labour market policy affects agency employment.