Professional development quality and instructional effectiveness: Testing the mediating role of teacher self-efficacy beliefs

ABSTRACTThe purpose of this study was to understand the mechanism linking the quality of professional development (PD) to instructional effectiveness by focusing on the mediating role of teachers’ efficacy beliefs. Based on social cognitive theory and using the Teaching and Learning International Survey (TALIS) 2018 dataset, we confirmed that teacher self-efficacy beliefs significantly mediate the relation between PD quality and three indicators of instructional effectiveness – clarity of instruction, cognitive activation, and classroom management. Specifically, partial mediation was found for clarity of instruction and cognitive activation, while full mediation was found for classroom management. Our findings suggest that the relationship between PD quality and specific types of instructional practice is highly nuanced and mediated by teacher’ self-efficacy beliefs. We discuss the implications of our findings for theory, future research, and practice.KEYWORDS: Professional development qualityinstructional effectivenessteacher self-efficacystructural equation modellingmediation Disclosure statementThere are no relevant financial or non-financial competing interests to report.Notes1. This sample does not include teachers who are marked as ‘logically not applicable’ (e.g. those who did not participate in any PD activities during the past academic year) and ‘not administered’ (e.g. those who were not asked to complete this portion of the survey).2. While Cochran-Smith and Lytle (Citation1999) observed that the informal patterns of interaction and social discourse in which teachers engage within their communities of practice can also constitute a form of PD, we were only able to reflect formal PD activities of teachers given the measures available in the TALIS dataset.3. Because Mplus 8.1 cannot automatically conduct bootstrapping procedures with multiple imputed datasets simultaneously, we conducted the 1,000 bootstrapping procedures for each of the 20 imputed datasets separately and reported their average confidence intervals.