A problem with quasi-experimental evaluations is that often a particular set of results can be interpreted, but a different set cannot

A problem with quasi-experimental evaluations is that often a particular set of results can be interpreted, but a different set cannot. For example, assume that a service to a disadvantaged group is supposed to lead to higher scores on some variable (see question 3). If the post-program level of this disadvantaged group is higher than that observed in a middle-class comparison group, the interpretation of the evaluation is relatively easy. However, if the postprogram level of the initially disadvantaged group improves ?