Distorting value added: The use of longitudinal, vertically scaled I student achievement data for growth-based, value-added accountability

摘要

The passage of the No Child Left Behind Act (N.cLB, 2002) legislates that individual states implement accountability systems based on student test scores to track Adequate Yearly Progress (A YP) and the closure of achievement gaps. V AA (value-added assessment) holds educators accountable only for certain types of gains that students make during the time those educators taught their students. Longitudinal, student performance-based V A accountability models have become popular and such models require student data to be vertically scaled across wide grade and development ranges. Many assessment companies provide such vertical scales and claim that those scales are adequate for longitudinal V A modeling. But scales spanning wide grade/developmental ranges also span wide content ranges and the scores are not exchangeable along the portions of the scale. This shift in the constructs measured from grade to grade jeopardizes the validity of inferences made from longitudinal V A models. As the accountability use of growth-based V AA models has extraordinarily high stakes, it is important to understand the potential impact of violating the assumption of measurement invariance that exists with vertical scales. (The term construct shift is coined to describe this violation). The author suggests that high stakes use of V AA systems based on vertically scaled student achievement data is to be discouraged or eliminated. (46 refs.)