Biased predictions of students' future achievement: An experimental study on pre-service teachers' interpretation of curriculum-based measurement graphs

In an online experiment, a sample of N = 109 pre-service teachers were presented with 14 graphs mimicking graphs used in curriculum-based measurement. Graphs depicted a student's weekly test scores for the first part of a semester, and participants were instructed to use the graphs to predict students' achievement at the end of the semester. Relative to a linear regression model, participants generally tended to underestimate future achievement (i.e., predictions were negatively biased). Predictions were more negatively biased when data variability was low rather than high, when improvement was steep rather than flat, and when the most recent score indicated a performance upturn as opposed to downturn. The results are interpreted in the light of models of judgmental anchoring (Kahneman & Tversky, 1973; Mussweiler & Strack, 1999). Implications for practice are discussed.