Association for Public Policy Analysis & Management

One of the criticisms of accountability policies is that high-stakes tests distort instruction, causing those tests to lose validity, especially for students at risk of missing benchmarks. Using a longitudinal dataset from North Carolina, I compare the usefulness of high-stakes test scores in predicting college grades in honors track classes to standard track classes to test whether test scores lose predictive validity in the lower-tracks, where teaching to the test is more likely to occur. According to the results, while test scores for standard track students are indeed less useful predictors of later academic achievement, high school GPA remains just as strong a predictor for standard as for honors classes. Interviews with teachers suggest that this pattern results from the limitations of the exams in contrast to the decisions by teachers to shape instruction and grading to the needs and characteristics of their students.