In this paper we use nonparametric and semiparametnrc estimation procedures to analyse whether the restart programme in the UK has had any effect on reducing unemployment duration. The restart programme consists of an interview of the long-term unemployed to counsel them on effective job search. The statistical results utilize experimental data in which a control group does not receive the restart interview. The results show that the programme has had a significant effect (for the treatment group) of reducing the duration of unemployment. However, if we distinguish between exits to 'any job' and to full-time jobs lasting at least 3 months we find that the treatment group is no different from the control group in exits to 'stable jobs'.