EVALUATING ACADEMIC PROGRESS WITHOUT A VERTICAL SCALE

Alternatives to vertical scales are compared for measuring longitudinal academic growth and for producing school‐level growth measures. The alternatives examined were empirical cross‐grade regression, ordinary least squares and logistic regression, and multilevel models. The student data used for the comparisons were Arabic Grades 4 to 10 in Qatar, and results were examined in the scale score and performance level metrics. It is found that vertical scales and cross‐grade regressions can show different results at the individual student level, but at the school level, the different measures of growth were strongly correlated, particularly in the scale score metric. Differences between the methods appear more likely in the performance level metric than the scale score metric and for grade pairs with more extreme performance.