Does equating matter in value-added models?

The purpose of this study was to examine the effect of equated and non-equated data on value-added assessment analyses. Several models have been proposed in the literature to apply the value-added assessment approach. This study compared two different value-added models: the unadjusted hierarchical linear model and the generalized persistence model. The former model assumes equated tests while the latter one relaxes this assumption. Two different data sets (equated and non-equated) were analyzed with both models. Value-added estimates for both models based on a statewide examination (equated) and a countrywide examination (non-equated) data were generally consistent. School rankings showed differences between the two models. The practical implication of this study is that although there were small differences in school rankings, a model requiring an equating assumption can be applied to a non-equated data set in a case when equating between test forms is not possible