| Abstract: | This study investigates the degree to which grades based solely on constructed-response (CR) questions differ from grades based solely on multiple-choice (MC) questions. If CR questions are to justify their higher costs, they should produce different grade outcomes than MC questions. We use a data set composed of thousands of observations on individual students in introductory economics classes at a large public university. We note that the instructors of these classes made conscientious efforts to write CR questions that assessed higher levels of learning (Bloom, 1956 Bloom, B. S. 1956. Taxonomy of educational objectives: The classification of educational goals, handbook 1: Cognitive domain. New York: McKay.. Google Scholar]). Despite this, we find relatively little difference in grade outcomes. Our analysis suggests that switching from an all-CR assessment to an all-MC assessment would produce grade variations that are similar to the differences that are observed for students across different tests. Although other studies have focused on test scores, frequently AP test scores, our study is the first to focus attention on university grades. We hope that our inability to identify substantial benefits to CR questions in terms of grades will stimulate further research to identify substantive benefits from using the more costly CR questions. |
