Prediction of study success : should selection instruments measure cognitive or non-cognitive factors?

• In a number of countries, both high school GPA and admission tests are used in the selection to higher education. The validity research on such instruments has mainly focused on their predictive validity, and the outcome generally shows that the GPA is a better predictor than the test. One explanation is that the grades also measure non-cognitive factors, which are considered important for school performance. Moreover, many admission tests have been criticised for being too alien to what is being taught in schools, hence lacking in relevance for future education. As a consequence, many admission tests are being revised, to be more similar to the school grades in terms of content. A relevant question is if this is the right way to proceed if the main purpose of the test is to predict future study success? This study compares the subtests of a traditional admissions test (the SweSAT) with high school grades in verbal and quantitative subjects in terms of content construct and predictive strength. Success in higher education is measured by the number of credits achieved by students in economics and business administration programmes at Swedish universities. The purpose is to find out if the difference in predictive strength can be related to cognitive or non-cognitive factors.  The data is analysed by regression analysis and structural equation modelling (SEM). The findings show that there are differences in content as well as construct, but that variations in predictive strength have to do with both cognitive and non-cognitive factors, where quantitative grades and test scores are more similar in construct than verbal grades and test scores, but also better predictors of performance in higher education. The findings are discussed in relation to previous research, and some suggestions for future research are made.

Publikations- och innehållstyp

ref (ämneskategori)

kon (ämneskategori)