Conceptual and Statistical Considerations in Estimating College Value-Added to Learning

Session Type: Invited Session

Abstract

While the economic value to college education has been establish, the learning value is contested. The Voluntary System of Accountability in the U.S., for example, seeks to report learning measures and value-added (VA) estimates for participating colleges, but most colleges do not provide this information. This reflects, in part, the fact that estimating learning VA faces multiple conceptual and statistical challenges. Conceptual challenges include the definition of VA: should VA be defined as controlling for context effects (e.g., mean college entrance scores) or should such mean differences be part of the college “treatment?” Statistical challenges include model selection (e.g., random vs. fixed effect), score reliability, bias in estimation, longitudinal vs. cross-sectional designs. The papers in this session address: (a) policy context and assessment-system construction for VA estimation, (b) conceptual issues in modeling VA, and (c) statistical issues in VA estimation.

Sub Unit

• AERA Presidential Session

Chair

• Richard J. Shavelson, SK Partners & Stanford University

Papers

• Modeling College Value-Added in Colombia: Impact of Context Effects, Dropouts, and Nature of Outcome Measure on Empirical Estimates - Benjamin W. Domingue, University of Colorado - Boulder

• The Colombian National Setting for Higher Education Learning Outcomes and Value-Added Assessment - Adriana Molina, Colombian Institute for Educational Assessment; Julian Patricio Mariño von Hildebrand, Universidad de los Andes

• Multidimensional Value-Added Models - Ernesto San Martín, Pontificia Universidad Catolica de Chile; Sébastien Van Bellegem, Université Catholique de Louvain

• Challenges and Best Practices in Estimating Value-Added Using Standardized Assessments of Postsecondary Learning - Jeffrey T. Steedle, Pearson

• Conceptual and Statistical Considerations in Modeling College Value-Added - Edward W. Wiley, SK Partners LLC.

Discussants

• Stephen W. Raudenbush, University of Chicago

• William H. Schmidt, Michigan State University