In this study the effects of various types of rehabilitation programmes on labour market outcomes are estimated. A main feature of this study is that it jointly evaluates multiple treatments by nonparametric matching estimators. The study is based on a large sample of persons in western Sweden who are long-term sick and could participate in rehabilitation programmes. Our results suggest that workplace training is superior to the other rehabilitation programmes with respect to labour market outcomes, but compared to non-participation no positive effects are found. ©2004 John Wiley & Sons, Ltd.