Charles F Manski*, John Newman** and John V Pepper***
*Department of Economics, Northwestern University (Benjamin Meaker Visiting Professor, University of Bristol)
**The World Bank
***Department of Economics, University of Virginia
Agencies operating social programs often use performance standards to evaluate success in achieving outcomes of interest (e.g. see Cave and Hanney, 1992). Program outcomes are measured and compared with the standard, a threshold deemed to separate acceptable outcomes from unacceptable ones. An evaluation using a performance standard should specify not only the threshold to be used but also the action to be taken if outcomes do not meet the threshold. Discussions of performance standards are often disappointingly vague about this critical matter. However the idea usually seems to be that the threshold should be set equal to an outcome level thought achievable by some alternative, perhaps a change in the management of the program being evaluated or perhaps an entirely different program. Then a possible action is to replace the program being evaluated with the alternative if the program yields an outcome below the threshold.
Evaluation using performance standards is clearly appealing in principle. The hard questions concern implementation. This paper examines two problems of outcome measurement that confront efforts to implement standards. These are the problems of auxiliary outcomes and the problem of counterfactual outcomes.
View the full article (PDF, file 90KB) (PDF, 89kB)
