This article presents propensity score matching as a method to implement randomized conditions to analyze service effects using nonexperimental data. Most social work research is challenged to implement randomized clinical trials, whereas administrative and survey data are often available and can provide valuable information about services received under naturalistic conditions. This article discusses the assumptions of this method and the analytic steps involved; and it presents three examples of the approach, demonstrating that it is possible to approximate the conditions of a randomized controlled trial, and when selection bias is reduced, investigators can have more confidence in their findings. (Contains 1 note.)