Performance in pass-fail assessments

Abstract Pass-fail tests are meant not only to assess but also to elicit performance of individuals who trade the cost of improving it off a higher probability that an imprecise test’s performance estimate exceeds the pass threshold. Each individual may perform strongly and fear failure because of negative estimation errors, or perform weakly and hope to pass because of positive errors. This paper shows how that choice depends on the test’s precision and threshold, which the tester may choose to elicit strong performance, and on the individual’s pass reward and performance cost, which the tester must take as given and may not know exactly. When it is costly for the tester to increase precision and individuals are heterogeneous it is optimal for some to fear failing, for others to hope to pass. This is empirically true in the results of exams that conform to the model’s assumptions. The model is also applicable to editorial and research funding processes.