Differential Item Functioning

Differential item functioning (DIF) is formally defined as a lack of equality between two group’s [Page 517]conditional probability functions that relate a trait of measurement to an item’s response data. DIF indicates that examinees who are equal on the trait of measurement, but differ according to some external variable, show differential performance on a test item. For example, DIF is said to be present if groups of examinees who are matched in quantitative aptitude (trait of measurement), but vary in country of origin (external variable), show differential performance on a test item that is intended to measure quantitative aptitude. The presence of DIF is viewed as problematic because it implies that some factor external to the trait being measured is influencing responses to test items in different ways for different groups of examinees, providing an advantage to one or more groups. DIF is a commonly explored phenomenon in educational measurement data because it assists in the statistical process of gauging fairness at the item level, which contributes to the overarching evaluation of measurement validity. Educational research on DIF includes developing methods to estimate DIF; comparing and contrasting the wide variety of DIF methods available in the literature; connecting DIF to other psychometric phenomena; conducting DIF analysis on particular test items and/or across particular groups of examinees; and overcoming challenges to estimating, interpreting, and addressing DIF in practice. The remainder of this entry introduces the concept of DIF in relation to other educational measurement concepts, defines various manifestations of DIF in data, describes some select methods for evaluating DIF, and reviews some of the challenges to evaluating DIF in practice.