Regression to the Mean: Using School Accountability Test Score Changes Across Years to Illustrate Deleterious Effects

• In Event: 65.065 - Perspectives in Conducting and Disseminating Research

Abstract

The purpose of this paper is to analyze school-level test data for regression artifacts as an effective demonstration of how regression occurs and how it could lead to inaccurate and sometimes very detrimental conclusions about school effectiveness. Using data across three years and middle school scores in mathematics and social science from a large state as the unit of analysis, the results show overall negative correlations among adjacent year test score changes. Outliers identified in scatterplots one year show regression to the mean the next year, explaining some of the dramatic gains shown by some schools.

Authors

• James H. McMillan, Virginia Commonwealth University

• Heather Carlson-Jaquez, Virginia Commonwealth University

• Shawn Sthreshley, Chesterfield County Public Schools

• Falcon Rankins, Virginia Commonwealth University