The Rise and Fall of Value-Added Models for Teacher Evaluations

• Activity Number: 71
• Type: Topic Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistical Education
• Author(s): J.R. Lockwood*+ and Derek Briggs*+ and Jesse Rothstein*+ and Catherine Durso*+
• Companies: RAND Corporation and University of Colorado and University of California at Berkeley and University of Denver
• Keywords: Value Added Models (VAM) ; Longitudinal Student Test Scores ; LA Times Teacher Ranking ; Noise in VAM ; Missing Data ; Measurment Error in Test Scores

Abstract

Value added assessments rely upon linear mixed effect modeling of longitudinal student test score data to isolate the effects that schools and teachers may have on student learning. Despite alarming shortcomings of Value Added Models (VAM), the political pressure in holding schools accountable for student performances has contributed to the adoption of these models from the district level to the state level. On August 2010, the Los Angeles Times published the results of value added modeling of student test data to measure the effectiveness of elementary schools and teachers in the Los Angeles Unified School District (LAUSD). The LAUSD also publicly released its own ratings of the school scores. In this panel, while presenting alternative paradigms to VAM, we discuss the issues associated with the estimation bias in teachers' effects, the problems of missing data and the measurement error in test scores, along with looking into whether VAM-based performance evaluation can appreciably change the effectiveness composition of the teacher workforce, given the noise in VAM estimates. We also provide recommendations from independent analyses of the LA Times data.