Linear and nonlinear growth models for value-added assessment: an application to Spanish primary and secondary schools’ progress in reading comprehension

Abstract

Value-added models are considered one of the best alternatives not only for accountability purposes but also to improve the school system itself. The estimates provided by these models measure the contribution of schools to students’ academic progress, once the effect of other factors outside school control are eliminated. The functional form for the growth of students’ performance is a key aspect to be considered when calculating value-added estimations, and, in this regard, there are several reasons that might suggest a nonlinear growth trajectory over time. This aspect is especially important when a variable such as reading comprehension is analyzed, since this is a skill in which nonlinear progression can be expected. This paper focuses on this issue and proposes to estimate schools’ value-added under nonlinear growth models where changes in performance follows a quadratic trajectory, analyzing differences in results with respect to those provided by linear growth models. To illustrate this point, the value-added in reading comprehension was estimated for three parallel cohorts that collect data from 153 primary and secondary schools in Madrid (Spain) and 6,755 students who were assessed at four different times during the academic years 2005–2006 and 2006–2007. Different models have been calculated using hierarchical linear models. The results show that nonlinear growth models fit better. Additionally, the inclusion of the students’ individual and family characteristics in the model provides more accurate VA measures for schools. Finally, it could be noted that the value-added is a relative magnitude that depends on the reference set.