Gruppskillnader i resultat på högskoleprovet
2004 (Swedish)Report (Other academic)
The Swedish Scholastic Aptitude Test (SweSAT) has been used for selection to higher education in Sweden since 1977. The most important demand on the test is that it should rank test-takers as fairly as possible with regard to their expected study success. Another demand is for unbiased admission, which means that, no group should be discriminated against because of e.g. gender or social class. Several studies have been performed on gender differences in test results, but differences between social classes have been scarcely studied. The main aim of this study was to examine differences between social classes on item level. Since results from the gender studies have had some impact on item construction, a secondary aim was to compare the results of social classes with the results of females and males. For group of 16 354 test-takers in spring 1992, who had finished upper secondary school the same year, differential item functioning (DIF) analyses were performed by means of the Mantel-Haenszel method and by comparing item characteristic curves (ICC). The DIF-items were classified into categories A (negligible), B (intermediate) and C (large) according to the severity of DIF. The test in 1992 consisted of 148 multiple-choice items, divided into six subtests. The outcome of the Mantel-Haenszel analysis for social groups was that 30 items turned out to have DIF, one of which in the B-category, and 29 in the A-category. For gender 113 items turned out to have DIF, 16 in category C, 20 in category B, and 77 in category A. Out of the 30 social group DIF-items, six were common to social group I and females, six to social group III and females, eight were common to social group I and males, and two to social group III and males. Four six out of the common DIF-items (four of which were in favour of social group I and females, and two in favour of social group I and males) the ICCs supported the results from the Mantel-Haenszel analyses. The conclusion was that there are very few items which function differently for social groups. Therefore, in order to find any pattern among DIF-items favouring social group I, an enormous amount of items would have to be analysed.
Place, publisher, year, edition, pages
Enheten för pedagogiska mätningar, Umeå , 2004. , 41 p.
, PM, ISSN 1100-696X ; 192
IdentifiersURN: urn:nbn:se:umu:diva-14180OAI: oai:DiVA.org:umu-14180DiVA: diva2:153851