In spite of exhaustive research by many analysts on the data generated by the NIT experiments, uncertainty remains over whether work reductions in the experiments should be considered "large" or "small." The authors of this paper argue that this uncertainty arises in part because different...

This paper explores the gains from multiplying the number of sites used in experimental evaluation of the effects of employment and training programs. Using a multilevel (hierarchical) statistical framework, the authors analyze the role of site multiplication in three recent program evaluations....

Of welfare-to-work programs evaluated by random assignment, two stand out as having exceptionally large estimated effects: one in Riverside, California, and the other in Portland, Oregon. The authors use data from 24 evaluations and the tools of meta-analysis to examine why. The findings...