Establishing Benchmarks for Outcome Indicators

Abstract

Indicators of program performance are used as evaluative measures in a variety of fields. A particularly vexing problem for evaluation is the development of empirically based performance expectations. Should program sponsors be satisfied with average performance? How should evaluators account for client differences? This article presents a statistical technique for developing performance standards based on benchmark groups. The benchmark groups are selected using a multivariate technique that relies on a squared Euclidean distance method. For each observation unit, in this case a school district, a unique comparison group is selected. The performance of the district is compared to the performance of its benchmark group. Then the credibility, predictability, and equity of the method are tested. The approach meets or exceeds these test criteria and appears to be a viable, albeit controversial, approach for developing comparative performance standards.