How Does District Principal Evaluation Affect Learning-Centered Principal Leadership? Evidence From Michigan School Districts

Abstract

This study used Hierarchical Multivariate Linear models to investigate relationships between principals' behaviors and district principal evaluation purpose, focus, and assessed leadership activities in 13 school districts in Michigan. The study found that principals were more likely to engage in learning-centered leadership behaviors when the purposes of evaluation included principal professional development, school restructuring, and accountability; when the focus of evaluation was related to instructional leadership; and when evaluation addressed leadership in school goal setting, curriculum design, teacher professional development and evaluation, and monitoring student learning. The findings from this study have implications for improving district evaluation policies and practices.

The authors would like to thank the Michigan State University Educational Policy Center for financial support and Wang Jun Kim for assistance with survey administration.

Notes

1. The district's new principal evaluation system was based on its teacher evaluation system, which, in turn, was based on Danielson's Framework for Teaching (1996); the new system was characterized by clear performance standards (dimensions) and rubrics differentiating performance on the dimensions. Depending on a given principal's years of experience and prior evaluations, he or she was evaluated with regard to their performance on two or more of the dimensions each year. Under the new evaluation system, most of the data were collected by district administrators through observations and conferences. The district's old evaluation system, like those in many other districts, consisted of a checklist of personal traits and behaviors, and it did not feature a rubric differentiating levels of performance (Kimball, Milanowski, & McKinney, 2007).

2. In the study reported here, we did not include measures of professional community or relational trust in our models or analyses. This was due to the fact that our measures of these constructs were based on teacher surveys, which were not a focal part of this study. In other work, we investigate possible links between principal leadership behaviors and teachers' perceptions of professional community and relational trust in their schools.

3. The response rates were calculated by dividing the number of valid returned surveys by the total number of eligible participants (i.e., the total number of people who were initially contacted minus the number of respondents who were ineligible). With regard to the district administrator surveys, ineligible respondents were those who had not been responsible for principal evaluation in 2006–2007 in the same district (where they were working in 2007–2008). In terms of the principal surveys, ineligible respondents were those who had not worked as principal in 2006–2007 in the same school (where they were working in 2007–2008). With regard to the teacher surveys, ineligible respondents were those who had not worked as teachers in 2006–2007 in the same school (where they were working in 2007–2008).

4. The theory in this study supports the assumption that constructs (factors) are not independent. Therefore, we used oblique rotation to run the factor analysis, which relaxes the assumption of independent factors (constructs). The following aggregation used the same procedure.

5. Hierarchical Linear Modeling (HLM) is based on and makes use of nested data. Nested data can include, for example, repeated observations nested within persons (e.g., four dependent variables in this study can be considered as repeated observations of principal's leadership behaviors), persons nested within organizations (e.g., principals' work within school districts), and the organizational units themselves nested within larger entities (e.g., districts are located in states). The application of HLM addressed three general research purposes: (a) to improve estimation of effects within individual units (e.g., developing an improved estimate of a regression model for an individual principal by acknowledging that similar estimates exist for other principals within the district); (b) to improve the modeling of crosslevel effects (e.g., how principals' leadership influenced by district evaluation policy); and (c) to partition variance-covariance components (e.g., to decompose the variation among a set of individual-level variables into within- and between-school or district components). Please refer to Raudenbush and Bryk (2002) for further details.