Daniel Koretz

More than two decades of research has documented that when tests are used for accountability, scores may become inflated, sometimes by a very large amount. A growing body of research is finding that this bias is not uniform and that it can vary systematically with variables that we often use as substantive predictors in research. Nonetheless, researchers often use the scores from high-stakes tests as outcomes, with no consideration of the potential bias from differential score inflation. In this presentation, I will present results from several studies to demonstrate that the results of studies using high-stakes test scores as outcomes may be misleading.