This paper examines two limitations of the ASPES methodology and introduces design innovations to address each issue.  One concerns the tension between predicting subgroup membership using sample members' background charactertistics and the assumption of equal impacts for predicted subgroups.   Here, the author proposes selecting predictors that are instrumental variables for impact magnitude.  The second limitation of ASPES involves its focus on one endogenous factor at a time, when multliple mediators of impact may be at work.  The author broadens the conceptaul foundation of the original Peck formulation to address this more complex--and possibly more common--case.