Evaluating Statewide School Accountability Systems: Comparison of Growth Models

Abstract

Under the Elementary and Secondary Education Act (ESEA) waivers, and now under the Every Student Succeeds Act (ESSA), states have the opportunity to incorporate elements other than proficiency into their accountability systems. With this flexibility, many states have turned to, or are exploring, the use of growth in their accountability system. Using three years of data from the Minnesota Comprehensive Assessments (MCAs), five growth models which have been used in state accountability systems were explored: transition matrix, trajectory, projection, student growth percentiles, and hierarchical linear models. The relationship between the rank order and school demographics, school size, and school type in each of these models was explored in order to determine which model appeared the least biased. Both the transition matrix and hierarchical linear model appeared to be relatively unbiased, as implemented in this study, but the hierarchical linear model produced results more similar to the other three models explored.