The Effect of Teacher Coaching on Instruction and Achievement: A Meta-Analysis of the Causal Evidence

• In Event: 34.047 - Mentoring Matters, but How? Conceptualizing Mentoring Effectiveness From Preservice to In-Service Teaching

Abstract

Objectives & Connection to the Literature
Research documenting the failure of most traditional professional development (PD) programing to improve instruction and achievement has motivated scholars to study specific conditions under which PD programs might produce more favorable results. These efforts have led to a growing consensus that effective PD programs share several “critical features” (Desimone, 2009) including job-embedded practice, intense and sustained durations, a focus on discrete skill sets, and active-learning (Darling-Hammond et al., 2009; Desimone & Garrett, 2015; Garet et al., 2001; Hill, 2007; Wayne et al., 2008). Teacher coaching programs have emerged in recent years as a promising new form of PD that integrates these features identified as best-practices.
Existing handbooks and reviews of the teacher coaching literature have focused on describing the theory of action, creating typologies of different coaching models, cataloging best implementation practices, and “vote-counting” positive and negative findings (Cornett & Knight, 2009; Devine, Meyers & Houssemand, 2013; Fletcher & Mullen, 2012; Kretlow & Bartholomew, 2010; Obara, 2010; Schachter, 2015; Stormont et al. 2015). This study is the first meta-analysis of causal research studies evaluating the effect of coaching programs on teachers’ instructional practices and students’ achievement on standardized tests.
Methods and Data Sources
We conducted a systematic review of the literature through a three-phase process where we, 1) searched via academic databases, 2) reviewed references in relevant papers, and 3) reached out to leading scholars in the field. Past studies have found that coaching is rarely implemented in isolation. Thus, we included PD programs that focus exclusively on one-on-one coaching as well as those that combine coaching with workshops and instructional materials or use coaching as intensive implementation supports for new curricula or instructional programs. We restricted our sample of studies to only those that used research designs capable of supporting causal inferences which included randomized control trials as well as quasi-experimental methods such as regression discontinuity, difference-in-differences and robust fixed-effects analyses. Our search yielded a total of 38 studies that we coded and classified in two broad categories, content-specific coaching (n=28) and coaching for general pedagogy (n=10). Among the content-specific studies, nearly all of the available evidence pertains to programs focused on reading and literacy (n=25 for reading, 2 for math and 1 for science).
Findings and Significance
Across all coaching models, we find large positive effects of almost half a standard deviation on teachers’ instructional practices as measured by scores on classroom observation instruments assigned by outside observers. Estimates for standardized student achievement assessments in reading show a small positive effect of .04 standard deviations, while data for other subjects is insufficient for providing precise estimates. Funnel plots, which demonstrate a clear pattern of more modest results for larger-scale studies, point to the challenge of taking high-quality coaching programs to scale. Exploratory analyses suggest that pairing coaching with group workshops and training sessions as well as supplemental instructional resources enhances their effectiveness. We conclude by providing suggestions for improving the quality of causal research studies on teacher coaching.

Authors

• Matthew A. Kraft, Brown University

• David Blazar, Harvard University

• Dylan Hogan