Detecting DIF in cognitive item scores on the PISA financial literacy questionnaire

Differential Item Functioning (DIF) is used to detect bias within assessments. In this project, uniform DIF and nonuniform DIF were modeled and observed to assess the fairness of the cognitive items on the Financial Literacy Questionnaire (FL). The FL is provided to countries to assess a variety of 15-year-old’s aptitudes by the Programme for International Student Assessment; including financial literacy. This project uses, logistic regression (LR), a hierarchical logistic regression with a random-intercept and a random-coefficient (HLR-RI/RC) to assess uniform and nonuniform DIF to provide further support of HLR in DIF assessment. Results indicate that both types of DIF were present in the first collection of models; which gender bias was examined for male and female participants. The second collection of models detected and nonuniform DIF, indicating item bias toward participants that were native and nonnative to their countries.