The authors discuss the use of propensity scoring in experimental program evaluations to estimate impacts for subgroups defined by program features and praticipants' program experiences.  They examine both the traditional propensity score matching method (which sorts the control group on predicted subgroup membership and compares subsets to treatment group members actually in the subgroup of interest) and the cutoff method (which applies symmetric predictive sorting to both the treatment and control groups).  Their research demonstrates the similarity of findings from the two approaches in the particular application used--a large-scale experimental evaluation of Job Corps, the nation's largest federal education and training program for disadvantaged youths--and demonstrates the effectivenss of both approaches in producing credible subgroup findings when predictive accuracy is high.