Why closing the gender gap in college admission tests is not a proper safeguard against discrimination

In the media and in public discourse, fairness in college admissions tests is often understood as the absence of mean differences between subgroups, especially when talking about the gender gap. Psychometricians, however, use the model of equal success probabilities to define fairness by also taking academic access into account. Using real-world data, this paper shows how focusing on the absence of a gender gap in test results can hide actual gender discrimination and even lead to incorrect conclusions. The paper then discusses why the model of equal success probabilities can only be as good as the operationalization of academic success, which remains an ongoing challenge.