Value-Added Models

Value-added models are statistical models that try to identify the impact of programs, people, or environments on a specific outcome. In education research, the term typically refers to models that look at the impact of various inputs on student growth in achievement, where achievement is generally measured by standardized test scores. In this sense, these models fit into the broader investigation of the education production function, which describes the process of producing learning as a formula that relates various inputs to the output of learning. So, within this context, for example, educators and [Page 1784]researchers can use a value-added model to ask whether a particular program or intervention helps raise achievement—to see “what works” in improving student learning. Such an investigation would fall under the heading of program evaluation. Alternatively, educators and researchers could also use value-added models to estimate how effective individual teachers or schools tend to be at raising their students’ achievement. These types of analyses are generally used for accountability purposes—to see who is performing well.

In all cases—whether value-added models are oriented toward program evaluation or accountability systems—the models estimate how the features, institutions, or individuals of interest improve student achievement during a specific time period. To help restrict their focus to a particular time period, the models always adjust the estimates of impact in some way for a student’s prior achievement to determine the “value added” of the input of interest after that point. It is this feature of the model—the adjustment for prior achievement—that gives it its name.

This entry begins by explaining how to compute a value-added model. It then discusses concerns with these models and how those concerns are commonly addressed. Next, the entry presents debate and controversy and concludes by reviewing research findings and summarizing the current and future outlook.