Using a yardstick to measure a meter: Growth, projection, and value-added models in the context of school accountability

Abstract

As test-based educational accountability has moved to the forefront of national and state educational policy, so has the desire for better measures of school performance. Past debates focused on the potential benefits of using value-added models to measure school performance. These statistical models represent the most sophisticated technique for assessing the unique contribution of schools to the learning gains of their students. However, recent national policy has shifted the discussion. The federally created Growth Model Pilot Program (GMPP) permits states to use projection models in their accountability systems. Most states involved in the pilot program are using projection models to give schools "credit" for those students who have not yet achieved proficiency, but have made sufficient learning gains such that they appear on track to become proficient in the near future. While value-added measures of school performance attempt to compare schools' relative effectiveness, the proposed projection models are used to assess whether schools are effective at getting their students up to a fixed learning target—proficiency.

This dissertation has two components. The first is an empirical comparison of different state projection models, answering the questions: How accurate are states' projection models? Are their goals realistic? The second is a comparison of three approaches to measuring school performance: the status model (used under No Child Left Behind (NCLB)), the projection models (used under the GMPP), and value-added models (used in some state accountability systems).

Findings suggest that the GMPP's projection models are not very accurate. Moreover, even if they were perfectly accurate, the GMPP's measures of school performance are so similar to NCLB's status measure that this new program is unlikely to have significant impact on the way we assess schools' performances. In contrast, using value-added models would dramatically shift the focus of our education accountability system. Highlighted in this work is the inherent tension in the desire to compare schools' relative effectiveness while simultaneously holding schools accountable for bringing all students up to high achievement levels.