We study the anatomy of responses to a major activation programme targeted at unemployed youth, introduced in Sweden in 2007. We use a regression discontinuity design to analyse individual reactions to the programme. We find that individuals who have a relatively high predicted probability of finding a job respond to the threat of activation, whereas there is no significant effect for individuals in a weaker labour market position. This is consistent with activation programmes working as a screening device between those who are able to find work on their own vs. those who are not. In addition to examining traditional predictors of poor labour market outcomes (e.g. education and school dropout status), we find a strong concentration of health problems among individuals with poor labour market prospects. We use register data covering the entire Swedish population, including very detailed information on health.