Center Affiliates Advance Value-Added Research

NCPI affiliates continue to make advancements in value-added research, resulting in several recent journal publications and a presentation at the Association for Public Policy Analysis and Management’s (APPAM) fall 2010 conference. Two recent journal articles were featured inEducation Finance and Policy and The Stata Journal.

These articles stemmed from research conducted by several NCPI affiliates at the RAND Corporation along with Tim Sass of Florida State University. In 2009, Dan McCaffrey, J.R. Lockwood, Tim Sass, and Kata Mihaly’s article, “The intertemporal variability of teacher effect estimates”, was published in Education Finance and Policy, and examined the year-to-year variability in value-added measures for elementary and middle school math teachers to determine the utility of value-added estimates for distinguishing between high- and low-productivity teachers and predicting future teacher performance. Their research found year-to-year correlations in value-added measures ranging from 0.2 to 0.5 among elementary teachers and 0.3 to 0.7 for middle school teachers. Predicting teachers’ future performance was greatly improved when averaging value-added estimates from two years.

A 2010 article in The Stata Journal resulted from additional value-added work by these scholars. “Centering and reference groups for estimates of fixed effects: Modifications to felsdvreg” highlights advances made to software that improve researchers’ ability to provide more useful value-added estimates. The article recognizes the increased availability of large, multilevel longitudinal databases in various fields including education research. This has heightened the use of software for estimating fixed effect models. Unfortunately, many of the existing software routines are not ideal for such research objectives. That is, existing programs using regression models with teacher fixed effects and possibly student fixed effects often report estimates of teacher effect relative to an arbitrary hold-out, rather than the more desirable approach of estimating teacher effect relative to the average of some comparative group. The software advances highlighted in this 2010 article explain these improvements for providing estimates on a more useful scale.

Jennifer Steele of the RAND Corporation presented at the fall 2010 APPAM conference on the topic “The distribution and mobility of effective teachers: Evidence from a large, urban school district.” The presentation highlighted results from a recent study examining how well the distribution of teacher effectiveness – measured as teachers’ value-added estimates – mirrors the distribution of teacher qualifications such as experience and academic background in the district. The study also looks at the extent to which patterns of teacher mobility within and out of the district appear to contribute to differences in students’ access to effective teachers. Overall, the study found that discrepancies in teachers’ value-added effectiveness resemble discrepancies in teachers’ observed qualifications. Additionally, teachers leaving their schools were found to be less effective, on average, than their counterparts who stay. And, among teachers changing schools within the district, there was little evidence of a relationship between a teacher’s value-added effect and the characteristics of a destination school relative to the prior school.