Assessing Program Performance

Establishing the "Activities-Outcomes" Connection: Evaluation Experiments

Quasi-Experimental (Comparison Group) Design

In this design, change is assessed by comparing perceptions or behaviors of program participants with those of non-participants (comparison group). If outcomes for the two groups differ in the expected way (e.g., program participants have lower recidivism rates than non-participants), then the evaluator assumes that the difference was caused by the program.

The assumption here is that the program participants are exactly like the non-participants in every way except that they received the program services, so any differences between the two must be due to the program. In such designs, evaluators often select non-participants who match participants on key factors, such as age, gender, and criminal history.

The trouble with this design, however, is that the evaluator can never be certain that the groups are exactly the same on every factor that might lead to differences in observed outcomes. The evaluator can have more confidence in the results of a quasi-experiment than he or she can in the results of the pre-post design, but still cannot be certain that the program activities caused the observed outcomes.

Learn More...

The Nonequivalent Comparison Group Design (Government Accounting Office)

Non-Random Comparison Group (National Institute of Justice, pp. 4.5-4.6)

| | |
| | |