IZA DP No. 11603: Testing at Length If It Is Cognitive or Non-Cognitive
substantial revision forthcoming in: Oxford Bulletin of Economics and Statistics
Using Italian data on standardized test scores, we show that the substantial heterogeneity in how performance changes with the position of questions can alter the rank of individuals and classes as the length of the test increases. We examine whether decomposing test scores into initial performance and performance decline allows to separate the influence of cognitive and non-cognitive skills and find that our measure of cognitive skills – the math grade before the test – not only has a dominant influence on initial performance but also affects substantially performance decline.