Panel Paper: Are Students' Teacher Ratings Predictable?

Thursday, November 8, 2012 : 1:55 PM
Salon B (Radisson Plaza Lord Baltimore Hotel)

*Names in bold indicate Presenter

Sade Bonilla, Stanford University and Richard Bowman, Albuquerque Public Schools

The Race To The Top (RTTT) grant competition has induced states to change their teacher evaluation systems by tying teacher ratings to student achievement. The Measures of Effective Teaching (MET) Project has encouraged states to adopt evaluation systems that rely on multiple measures to improve accuracy. The push for performance management in the education sector has led several RTTT recipient states to utilize student perception surveys as one measure of teacher performance. However, little is known about how K-12 children will respond to student perception surveys, as much of the research on student survey respondents has focused on college students. Student survey response bias based on student observable characteristics may be problematic if student survey responses are used for high-stakes decisions without controlling for student observable characteristics.

This paper examines the relationship between student characteristics and responses on a student perception survey given in an evaluation and performance based compensation context, and whether these relationships are related to teacher rankings. The study sample includes over 3,000 middle and high school students in a large urban school district. Student perception surveys were gathered for all sections of 76 volunteer classroom teachers at three high schools and one middle school. Student survey responses were matched to district student academic and disciplinary records. Detailed analysis of the effect of including student controls on teacher rankings is ongoing, but preliminary results suggest that students’ survey responses are related to their performance expectations in the class as well as some of their background characteristics, suggesting that rankings of teachers that do not include controls for student characteristics may lead to bias.