One of the most appropriate quasiexperimental approaches to compensatory education is the regression-discontinuity design. However, it remains underutilized, in part because of the need to clarify the link between the mathematical model and administrative decision-making. This paper explains the derivation of a program efficiency index congruent with the regression-discontinuity design. The efficiency index is based on a confidence interval calculated for expected mean posttest performance. If the observed posttest mean corresponds exactly to the upper limit of the confidence interval, the efficiency index is +1 (optimal efficiency). If the observed posttest mean falls at the lower confidence boundary, the index value is zero (minimal efficiency). Data from a remedial mathematics program (grades 2, 3, 7, and 8) illustrate the possibilities for: (1) net growth (index above 1); (2) breakdown (index below 0); and (3) gradations of maintenance (index between 0 and 1). As conceived, the efficiency index is comparable to eta-square, the correlation ratio. Thus, its analytic context differs from that of the effect size coefficient commonly associated with classical control group design. A model is presented showing how variations in the size of the efficiency index may lead to different decision making options. (Author/LPG)