Dynamic Measurement Modeling: Using Nonlinear Growth Models to Estimate Student Learning Capacity

Single-timepoint educational measurement practices are capable of assessing student ability at the time of testing but are not designed to be informative of student capacity for developing in any particular academic domain, despite commonly being used in such a manner. For this reason, such measurement practice systematically underestimates the potential of students from nondominant socioeconomic or ethnic groups, who may not have had adequate opportunity to develop various academic skills but can nonetheless do so in the future. One long-standing approach to the partial rectification of this issue is dynamic assessment (DA), a technique that features multiple testing occasions integrated with learning opportunities. However, DA is extremely resource intensive to incorporate into educational assessment practice and cannot be applied to extant large-scale data sets. In this article, the authors describe a recently developed statistical technique, dynamic measurement modeling (DMM), which is capable of estimating quantities associated with DA—including student capacity for learning a particular skill—from existing large-scale longitudinal assessment data, allowing the core concepts of DA to be scaled up for use with secondary data sets such as those collected by Statewide Longitudinal Data Systems in the United States. The authors show that by considering several assessments over time, student capacity can be reliably estimated, and these capacity estimates are much less affected by student race/ethnicity, gender, and socioeconomic status than are single-timepoint assessment scores, thereby improving the consequential validity of measurement.