Recent evaluations of active labor market policies are not very optimistic about their effectiveness to bring unemployed back to work. An important reason is that unemployed get locked-in, that is they reduce their effort to find a regular job. This paper uses an administrative dataset from the Slovak Republic on durations of individual unemployment spells. The focus of the analysis is temporary subsidized jobs. By exploiting the variation in the duration of these jobs it is possible to investigate whether or not the locking-in effect is important. It turns out that it is.