Final Report on the Effects of Sample Attrition on Estimates of Channeling's Impacts

This report results from an investigation of the extent to which differential attrition from the research sample in the Channeling Demonstration might have led to biased estimates of program impact. Two analytical approaches were adopted--a heuristic approach and a statistical modeling approach. The results from these approaches lead to the conclusion that, in spite of observed differences in attrition rates between treatment group and control group members, there is very little evidence that such attrition resulted in biased estimates of Channeling's impacts. The occasional bits of evidence to the contrary were scattered and inconsistent across time, Channeling model, and outcome variables. [107 PDF pages]