Testing Two Models for Estimating Teacher Effectiveness

logo-act-printEdgar Sanchez and Yan Zhou, writing for the ACT policy brief, have examined the effects of two different models on estimates of teacher effectiveness. When using value-added models to estimate teacher effectiveness, one key consideration is whether or not to include school effects such as available resources into the model. The authors examined value-added models that did and did not include such school-level information and compared the results.

Key findings in report include:

  • Teacher effectiveness estimates from both kinds of models were highly correlated.
  • When school-level effects were accounted for, fewer teachers were found to be significantly different from the average teacher. The model that does not account for school-level effects identified a greater number of higher- and lower-performing teachers.
  • There was a high agreement rate on the identification of teachers who were not significantly different from the average teacher whether or not school-level effects were taken into account.
  • School-level variance tended to be larger than teacher-level variance. That said, student-level characteristics accounted for considerably more of the variance in student achievement scores.

The report concludes, “If there is a strong theoretical basis for believing that school differences have an important influence on student achievement, it would be appropriate to use a three-level VAM [i.e., a model that accounts for school-level effects]. This type of model will estimate teacher impacts on student achievement while accounting for important school-level factors such as resource allocation.

To read the report, visit www.act.org/research/policymakers/pdf/Model-Choice-Report.pdf