Value-added model (VAM)

A value-added model estimates how much a teacher, school, or program contributed to student learning growth, measured by standardized test scores, after statistically adjusting for factors outside the school's control such as students' prior test scores and often demographic characteristics. The output is typically a score showing whether a teacher's students gained more or less than statistically similar students taught by other teachers.

The underlying method is a regression analysis that predicts each student's expected score from prior performance and covariates, then compares actual scores to that prediction; the average gap for a teacher's roster is the "value added." This differs from a raw proficiency or pass rate, which reflects where students end up rather than how much growth occurred, and can penalize teachers of already high-achieving students for a ceiling effect while crediting little real growth elsewhere.

VAM scores have been used in some jurisdictions for teacher evaluation, which is contentious: measurement error and year-to-year instability for any single teacher are well documented, classroom assignment is rarely random, so causal inference about a teacher's true effect is limited, and different model specifications can rank the same teacher differently. Most methodologists recommend using VAM as one input among several, such as classroom observation, rather than a standalone judgment, distinct from student-facing tools like early warning indicators.