A model demonstrating how performance incentives may encourage cream skimming (selecting participants with a high likelihood of success) was applied to data from 16 Job Training Partnership Act training centers. Short-term performance measures were weakly and perversely related to long-term impacts, but efficiency gains/losses from skimming were…

Analysis of Tennessee Job Training Partnership Act data shows that creaming (nonrandom selection of participants, especially nonselection of disadvantaged persons) takes place but not as greatly as critics suggest. Selection of economically disadvantaged persons would lower the placement rate only 8%; targeting high school dropouts would reduce…

The first three steps to developing a program that is both effective and efficient are participant selection, recruitment, and assignment. They require thinking about the populations to be targeted, considering how to open the door to the population and how to get the population in the door, and determining the services that are needed. The Job…