Value-Added Models for the Measurement of the Effectiveness of GERES Schools

Abstract

Using data from the GERES longitudinal research program, the study proposes the comparison of statistical models of varying degrees of complexity to determine the effectiveness of elementary schools. The purpose of the comparison is to determine whether a greater degree of complexity is justified in terms of improved accuracy and if there are differences between the models regarding their consistency and stability when portraying school performance. The two simplest models for the contribution of the school to student proficiency, named status models, use as a proxy for prior proficiency either a measure of the school’s average socioeconomic status or a measure of the socioeconomic condition of each student. The other two, called value-added models-VA, incorporate real prior proficiency measures, enabling them to describe the learning gain attributable to the school over the period under study. The results indicate high correlations between the VA models and a low correlation between these and the status models, showing that the gain in precision with the addition of a measure of the socioeconomic condition of each student is small. It is shown that for about 80% of schools the measures of VA across time are stable, suggesting that school effectiveness is a reasonably stable characteristic and that a measure of VA can contribute to the comparison of schools and the definition of interventions, at least during the first years of elementary schooling.

Educational assessment; Value added; Multilevel models