Teacher Coaching at Scale in Washington, D.C. Public Schools
*Names in bold indicate Presenter
Crafting policy to improve teaching practice at a district or state level is challenging. Teachers have diverse needs based on their own knowledge, skills, and experience as well as the classroom and school environment in which they are situated. Research suggests that professional development (PD) is ineffective, on average, at developing teaching practice (Garet et al., 2008; Garet et al., 2011; Garet et al., 2016). However, recent studies provide some hope that intense and sustained, skill-based, and individualized PD can indeed improve teaching skills (Desimone & Garet, 2015). This kind of PD may be more adaptive to the situated nature of teaching, and therefore more effective. A meta-analysis of causal studies found that coaching programs improve teacher instruction by almost 0.60 SDs and student achievement by about 0.15 SDs (Kraft, Blazar, & Hogan, 2018). However, many studies define treatment as inclusive of workshop-style PD in addition to one-on-one coaching, and most represent small-scale programs. Whether and how coaching can improve outcomes at scale, therefore, is still unanswered.
This study examines features of a district-wide coaching program in Washington, D.C. Public Schools (DCPS) to address the following research questions: 1) descriptively assess teacher selection into one-on-one coaching at scale, and 2) understand the relationship between coaching dosage and outcomes of interest, like teacher satisfaction with school supports, observed teacher performance, student achievement, and teacher retention.
DCPS recently implemented a program called Learning together to Advance our Practice (LEAP), which departs from traditional PD in its use of school-based coaching supports in tandem with district-wide learning modules focused on Common Core State Standards (CCSS) aligned teaching practices. This initiative is unique in its employment of coaches at every school in the district. This study draws on rich administrative data and district-wide survey data from the first and second year of LEAP to examine two research questions.
Methods and Data Sources
We employ administrative data that contain a rich set of demographic and performance information about teachers and coaches, including evaluation scores, years of experience in the district, and student achievement. We describe the selection of teachers into coaching using ordered logistic regression, and then estimate the association between coaching and our outcomes of interest controlling for a variety of teacher, LEAP team and school covariates to isolate the effects of coaching. We also estimate these relationships with a Heckman two-stage estimation model to address potential selection issues.
Preliminary evidence suggests that teacher experience and the job title of a coach systematically influence coaching dosage during the first and second years. For example, assistant principals were less likely to engage in one-on-one coaching frequently, compared to their teacher leader peers. Teachers with more experience were more likely to engage in one-on-one coaching frequently, compared to their less experienced peers.
Some evidence suggests a negative relationship between coaching program size and effects (Kraft et al., 2018). Policymakers hoping to implement coaching at scale need more information about factors that influence the effectiveness of these programs.