Fairer testing

The use of standardized achievement tests is increasing in society, and scores obtained may have a decisive impact on people’s life choices, so it is essential to ensure they are also fair and reliable.

The purpose of the project is to develop innovative statistical methods of analyzing standardized achievement tests to ensure reliable, valid and fair test scores over time for all candidates.

Standardized tests such as the Swedish university entrance exam, national school tests or the driving test theory exam are used for purposes such as higher education selection, to set grades or for certification.

Achievement tests are given to different groups of people at different times. Several versions of a test are used to ensure the questions are not known in advance. In the interests of fair testing, it is important to examine the nature of the tasks to be performed, for instance how difficult they are, before a test is composed, and that there are methods of comparing candidates’ results from different test versions used on different occasions.

The project focuses on methodological basic research and is being conducted in three stages. Stage one is to create analytical statistical models and methods. Stage two is to examine the methods under controlled conditions using simulated data, allowing the conditions to be varied, e.g. length of the tests, ability of the candidates, nature of the tasks. In stage three the methods are applied to standardized achievement tests from three countries.

The research findings and all software codes will be made freely available so they can be used by test developers, public agencies and researchers who are responsible for designing, analyzing and evaluating achievement tests.