Item-Specific Efficacy Judgments in Mathematical Problem Solving: The Downside of Standing Too Close to Trees in a...

Educational researchers assess self-efficacy by asking students to rate their capability of succeeding at specific target tasks (e.g., math test items) and then testing their performance to actually solve similar test items. Pajares and colleagues (Pajares & Kranzler, 1995; Pajares & Miller, 1994, 1995, in press) argued for the use of identical items to assess self-efficacy and performance in order to maximize self-efficacy's predictive power. In two studies, structural equation models (SEM) demonstrated that this variation led to positively biased estimates of paths from self-efficacy to performance and negatively biased estimates of paths from self-concept to performance. Whereas corrections for this bias did not substantially alter the size of effects or substantive interpretations, results from both studies were consistent with a priori predictions about the nature of this bias. Researchers are encouraged to use similar but not identical items to assess self-efficacy and performance, a construct validity approach to interrogate their interpretations, more diverse outcome measures, and SEM approaches like those demonstrated here.