The Relationship between Principals and Teacher Development: Evidence from the District of Columbia Public Schools

ABSTRACT

Principals’ influence on teachers is an important route through which principals may affect student outcomes. The advent of teacher evaluation systems allows estimations of principal contributions to teacher effectiveness. Employing data from the District of Columbia Public Schools, this paper examines how principals vary in their contributions to improvements in teacher effectiveness, and the relationship between this measure of principal quality and the measures DCPS employs to assess principals’ contribution to teacher development. Results indicate that principals are associated with meaningful changes in teacher development. These estimated contributions are modestly correlated with DCPS’s measures of principal quality.

Acknowledgments

I am grateful to the Bankard Fund for Political Economy’s Pre-doctoral Fellowship in support of the pursuit of my dissertation. This work would not have been possible without the District of Columbia Public Schools. To that end, I would like to thank the district for supplying the data employed in this research and to Scott Thompson, Luke Hostetter, Lauren Norton, Liz Kim, and Chris Lewis for answering my many questions. I extend my gratitude to The New Teacher Project for allowing me to the use the Insight survey data in this work. Finally, I am thankful to Jim Wyckoff, Dan Player, Michelle Young, and Daphna Bassok for their invaluable feedback.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1. Principals’ ability to affect important school outcomes is likely also a function of principals’ training and education. The author acknowledges this (see, Figure 1) but does not focus on it in this paper.

2. The number of schools and principals are not congruent because schools may have multiple principals or principals may serve at multiple schools. The number of schools also varies over time due to school openings and closures.

3. Five percent of schools in these data have multiple principals, and a handful of principals serve at multiple schools. In these situations, it is unclear how to attribute school outcomes to leaders, and so the paper drops them from the primary analyses. To explore the robustness of results to this choice of an analytic sample, models are estimated on all schools as well.

4. The rating process was simplified and score calculations were made more transparent.

5. Principals scoring the highest rating are eligible for a $20,000 bonus, with an additional $10,000 if they are at the lowest performing and poorest schools.

6. Beginning in 2016-17, teachers are only evaluated by their principals. Starting in this year, DCPS changed their observational rubric as well.

7. MEs attend a six-week summer institute where they analyze the TLF rubric, video-norm to establish inter-rater reliability, and practice observations and post-observation conferences (The New Teacher Project, 2011).

8. This sample may still be capturing the effects of principals at successfully retaining certain teachers for the duration of the panel. Such a restriction is, however, the best attempt at isolating principal effects on improving teacher quality.

9. Results, available upon request, are similar to those from the main model.

10. The author omits the first couple of years of SL-IMPACT scores because the evaluation system was inconsistent in its methods of measuring and recording principal quality prior to 2013-14.

11. Tabachnick and Fidell (2007) suggest 0.32 as the minimum loading of an item, and denote loadings higher than 0.71 as “excellent”.

12. A mixed-effects model often presents a noise-bias tradeoff. Specifically, when modeling principal and school effects, the model reduces noise in the principal estimates. When estimating principal-by-school and school fixed effects, however, the model introduces more bias in principal effects than it reduces noise (Bartanen & Husain, 2022). When estimating only principal effects this issue doesn’t cause concern. Because this paper primarily relies on models with only principal effects, mixed-effects models are employed for all relevant analyses.

13. The author examines if principals make greater improvements for teachers who have lower initial TLF scores. To do this, the analytic sample is divided into quartiles based on 2010-11 TLF scores, and equation 1(1) $\mathit{Ob}{s}_{\mathit{ipst}}={\beta }_0+{\beta }_1\mathit{Ob}{s}_{\mathit{it}-1}+{\beta }_2{T}_{\mathit{it}}+{\gamma }_t+P_p+{\epsilon }_{\mathit{ipst}}$(1) is run separately for each subgroup. Across groups, principals contribute between 0.10 and 0.19% of a standard deviation (0.04 to 0.08 TLF points), but they do not systematically contribute more to the improvements of those teachers who begin with lower TLF scores.

14. Under IMPACT, as teachers gain experience and receive consecutively high ratings, they are no longer evaluated by the ME and only receive observations by their principal, explaining the relatively larger sample size when examining principal-assigned observation scores.

15. Available upon request.

Additional information

Funding

The research is supported by the Institute of Education Sciences, U.S. Department of Education, through Grant [R305B100009] to the University of Virginia.