A Correlated Random Effects Model for Nonignorable Missing Data in Value-Added Assessment of Teacher Effects

TL;DR

This paper introduces a correlated random effects model that jointly analyzes student achievement and missing data to provide unbiased teacher effect estimates in value-added assessments.

Contribution

It develops a flexible model that accounts for informative missing data by jointly modeling responses and missingness with latent effects.

Findings

Model adjusts teacher scores for missing data bias

Application to university calculus and elementary school data

Improved accuracy of teacher effect estimates

Abstract

Value-added models have been widely used to assess the contributions of individual teachers and schools to students' academic growth based on longitudinal student achievement outcomes. There is concern, however, that ignoring the presence of missing values, which are common in longitudinal studies, can bias teachers' value-added scores. In this article, a flexible correlated random effects model is developed that jointly models the student responses and the student missing data indicators. Both the student responses and the missing data mechanism depend on latent teacher effects as well as latent student effects, and the correlation between the sets of random effects adjusts teachers' value-added scores for informative missing data. The methods are illustrated with data from calculus classes at a large public university and with data from an elementary school district.