Chapter 7: Using Blinder-Oaxaca Decomposition to Explore Differential Item Functioning

Detection of differential item functioning (DIF) has historically played an important role in investigations of test fairness. A wide range of statistical methods have been proposed and studied for DIF detection, and an extensive literature now exists to inform the selection among different techniques (see e.g., Penfield & Camilli, 2006; Magis, Beland, Tuerlincx & De Boeck, 2010). Unfortunately, while statistical methodology for DIF is well-established, psychological explanations for DIF are often lacking. Such explanations play an important role in evaluating the effects of DIF on test validity. Test users can be perplexed as to whether the observation of DIF in particular items actually reflects unfairness (i.e., bias), or is instead a reflection of a complex construct for which group differences on certain facets of the construct may be expected.