<p>Evidence that welfare-to-work programmes in the USA succeed in boosting welfare recipients' earnings at modest cost has helped shape policy in Britain since 1997. So too has the belief that programmes that prioritise moving people into work quickly are more effective than ones that seek to enhance human capital. However, there is little evidence on how long the beneficial effects of 64 US welfare-to work programmes that have all been evaluated using random assignment. It concludes that, on average, these programmes have a positive effect on participants' earnings for five to six years. The effects of 'work first' interventions are most marked early on the decline more rapidly than those of programmes emphasising human capital. Nevertheless, work first interventions typically increase earnings received over six years by more than two-and-a-half times that achieved by human capitl approaches.</p>