With matching this doesn't necessarily cause any problems since the goal is to estimate the average treatment effect (or the average treatment effect on the treated or something like that). The general approach allows for that variability. The classical regression approach, in contrast, often assumes constant treatment effects, and so heterogeneity of effects can cause bias or incorrectly estimated standard errors. Gary -- *Gary King* - Albert J. Weatherhead III University Professor - Director, IQSS - Harvard University GKing.Harvard.edu <http://gking.harvard.edu/> - King(a)Harvard.edu - @kinggary<http://twitter.com/kinggary>- 617-500-7570 - Asst 495-9271 - Fax 812-8581 On Thu, Mar 24, 2011 at 2:55 PM, georgebaah &lt;gtg498n(a)mail.gatech.edu&gt; wrote: