Costs, Values and Uses of Performance Assessment

K-12 Testing

Researchers Brian M. Stecher and Stephen P. Klein of the RAND Corporation have scrutinized the likely costs of hands-on science performance assessments. They conclude that such science tests will be about 20 times more expensive than multiple-choice tests for a class period of testing that will yield scores with a reliability of at least .80 for individuals. They note that science is unusual in its reliance on equipment, so costs should be lower in other subjects. Still, performance exams are likely to be three or more times as expensive as multiple-choice tests.


The question, Stecher and Klein ask, is whether performance assessments are worth the price. They concur with those who claim that they can tap important learning areas not assessed by multiple-choice items, can positively affect curriculum and instruction, and can provide valuable staff development opportunities. They add, however, that far more research on these points is needed to demonstrate the benefits more conclusively.


The authors also note that it might make more sense to use performance tasks on a matrix sampling basis, rather than test every student. This would produce school-level data for far less cost. Further, they explain, "If the classroom or school is the unit of analysis, then performance and multiple-choice tests produce almost equally reliable scores for a given amount of testing time," unlike what happens with individual scores. That is, obtaining reliable data with performance exams at the classroom school level is quite feasible.


(Reliability, a statistical measure of the consistency of test results is typically over .90 for individuals on standardized, norm-referenced achievement tests, but lower on writing samples and constructed-response tasks. Reliability of .80 or higher is considered acceptable for most purposes.)


FairTest, among others, has advocated using state- or district-mandated performance exams on a sampling basis as part of accountability and has argued that classroom uses of performance tasks and portfolios should be the basis for providing data on individual students to parents (Examiner, Summer 1996). This would keep costs down and, more importantly, keep the attention where it most belongs, on improving education in the classroom.


-- "The Cost of Science Performance Assessments in Large-Scale Testing Programs," Educational Evaluation and Policy Analysis, Spring 1997.