Evaluating Training Programmes for the Long-Term Unemployed: An illustration of the matched comparison group methodology

This paper describes an evaluation of a skills training programme for long-term unemployed adults in the UK, for which a random allocation design had been ruled out. After an initial survey of programme leavers, a matched comparison group was constructed from administrative records, based on claim dates, sex, age and locality. The close match achieved was partially undermined by subsequent sample attrition, necessitating a re-matching process. Results showed a high degree of selection into the Training for Work Programme even after matching. Subsequent improvements in administrative data in the UK have increased the viability of this type of design, and where analysis can be carried out entirely with administrative data, the problem of sample attrition is overcome.