Comparing the incomparable: a predictive validity analysis based on matching methods
2008 (English)In: The impact of testing on people and society, 2008Conference paper (Refereed)
This paper evaluates the selection instruments for higher education in Sweden, by comparing how students selected on either grades from upper secondary school (GPA) or an admissions test (the SweSAT) perform in higher education. The study appreciates that evaluating multiple instruments is not without problems, as applicants in different admission groups are not necessarily comparable with respect to personal traits, and different instruments may benefit individuals differently. The study investigates if combining regression models with matching methods will give additional or better information about the different selection instruments, compared to traditional methods in predictive validity studies. The data consists of register data from students born 1972-1974, who have been admitted to a Business Administration or Edconomics programme in a Swedish university between the years 1993 to 1996. The number of credits taken in higher education serve as criterion for academic success. The results from the regression analyses support previous findings in validation studies, showing that the GPA admitted students are more productive in the economics programme than students admitted from the SweSAT group, and even a negative correlation between test scores and number of credits among those admitted. However, the results also show that the SweSAT group are more productive than the GPA group overall, when the total number of university credits serve as criterion.
Place, publisher, year, edition, pages
selection, admission, prediction, validity
Research subject didactics of educational measurement
IdentifiersURN: urn:nbn:se:umu:diva-37797OAI: oai:DiVA.org:umu-37797DiVA: diva2:370054
6th ITC International Conference ; July 14-16, 2008, Liverpool UK