Identifying effective and ineffective schools for accountability purposes: a comparison of four generic types of accountability models

The stakes associated with student performance have been raised to new highs under the No Child Left Behind Act (NCLB, 2001). Many people are concerned with the adequacy and appropriateness of the statistical models used in identifying low-performing schools for educational accountability. The purpose of this study was to compare four generic types of accountability models (i.e., status models, improvement models, growth models, and value-added models) and see if they reach consistent/inconsistent conclusions regarding the effectiveness of the same set of schools. Further, the four models were also compared in terms of “fairness”. A fair model is defined as one that produces school effectiveness indices that have low correlations with various student background variables (Webster, Mendro, Orsak, & Weerasinghe, 1998). The sample included this study consisted of all 297 K-5 schools in Louisiana. The results indicate that (1) the school effectiveness indices produced by the status model, the improvement model, and the growth model diverged significantly from those produced by the value-added model but converged highly among themselves; and (2) the school effectiveness indices produced by the value-added model had the lowest correlation with various student background variables. The methodological and policy implications of these findings were discussed