The Leaning Tower of Pisa
Decomposing achievement test scores into cognitive and noncognitive components
Lex Borghans, Maastricht University
Trudie Schils, Maastricht University
Draft version July 11, 2012
Test scores on achievement tests depend on both cognitive and noncognitive skills. The predictive power of achievement tests is therefore potentially due to all components. The question of this paper is whether it is possible to disentangle cognitive and noncognitive factors from the performance on the test. Using data from the international achievement test PISA, we decompose the test scores into two factors. We investigate the development of the performance of students during the test, utilizing the (randomized) difference in the order of the test questions in the various test booklets. We document that performance substantially drops during the test and that this performance drop differs between types of students and countries. The estimated size of the drop is very stable over the years, while correlation between this drop and the test scores is small. This suggests that the decline in test scores during the test picks up something else than just cognition. The size of the decline in test scores during the test is related to personality traits, mainly to agreeableness, and to motivational attitudes towards learning. It also predicts outcomes in later life such as income and smoking in addition to the pure test score. The motivation effect can explain 19 percent of the variation in the average test scores between countries.
I saw this paper at the NBER summer institute last week and thought it was pretty cool.