Search

The push for accountability in public schooling has extended to the measurement of teacher performance, accelerated by federal efforts through Race to the Top. Currently, a large number of states and districts across the country are computing measures of teacher performance based on the standardized test scores of their students and using them in combination with other indicators to help categorize teachers as effective or ineffective. With the implementation of the Common Core State Standards in a majority of states and the development of specific assessments that align with them, it is particularly urgent to open up the field right now to discuss and shed light on best practices in computing teacher performance measures. The market for new assessments coupled with derivative products that compute teacher effectiveness measures--each promoting a particular methodology--is becoming increasingly competitive. Policy makers must make informed decisions on which products and procedures they will use. A research team at Michigan State and Indiana Universities has contributed a number of key insights to the body of research on best practices in computing teacher performance measures and is in the process of disseminating its findings. This paper describes the project, its findings, and its preliminary recommendations for policy. Bios of principal investigators are appended.

Accurate indicators of educational effectiveness are needed to advance national policy goals of raising student achievement and closing social/cultural based achievement gaps. If constructed and used appropriately, such indicators for both program evaluation and the evaluation of teacher and school performance could have a transformative effect on the nature and outcomes of teaching and learning. Measures based on value-added models of student achievement (VAMs) are gaining increasing acceptance among policymakers as an improvement over conventional indicators of performance. This paper is the first in a series of papers that aims to resolve controversies surrounding VAMs. The authors' research question is: How well do different estimators perform in estimating teacher effects in a commonly used VAM framework? The authors focus their study on six estimators--most of which are commonly used in the research literature and in policy applications involving teacher effects. They answer their research questions by first outlining the assumptions that must be met for each estimator to have good statistical properties in the context of a fairly common theoretical framework. They then apply the estimators to the task of recovering teacher effects in simulated data. They apply the estimators to different types of data, where each data set is generated to violate or maintain specific assumptions in the attempt to mimic different types of student grouping and teacher assignment scenarios, and compare their performance. They use simulated data and conduct the research at Michigan State University. It is clear from this study that many VAMs hold promise: they may be capable of overcoming obstacles presented by non-random assignment and provide valuable information. Given their context-dependency, however, caution must be applied in interpreting the findings that VAMs, as currently applied in practice and in the research literature, are producing. They may best be viewed as suggestive evidence of effects, at this point. In addition, methods of constructing estimates of teacher effects that they can trust for high-stakes evaluative purposes must be further studied. Clearly, although value-added measures of teacher performance hold promise, before they can be used for policy purposes, more research is needed. There is much left to investigate. (Contains 3 tables.)