In 1999 general assistance recipients in New York City were required to participate in a job training and outplacement assistance program. Initially, recipients were enrolled in ‘waves’ due to capacity constraints. The program’s impact is identified using a quasiexperiment in which selectees are compared to concomitantly eligible non-selectees. Selectees are 15 percentage points more likely to start a job and 10 percentage points more likely to exit welfare than are non-selectees. This methodology is important since random-assignment experiments can be costly and difficult to implement. Further, experiments are not impervious to criticism; this procedure addresses three of five known shortcomings.