Abstract
A central puzzle in labour economics is not simply whether discrimination exists, but why. Decades of audit studies and field experiments show persistent bias in hiring (Bertrand and Mullainathan 2004; Oreopoulos 2011), yet the underlying mechanisms remain contested. Do employers discriminate because they dislike certain groups (Becker 1957)? Because they rationally infer productivity from noisy signals (Phelps 1972)? Or because they conform to perceived organizational norms (Arrow 1973; Rivera 2012)?
These mechanisms are not observationally equivalent, but they are rarely distinguished in empirical research. Existing audit studies and résumé experiments typically conflate them, because they present evaluators with résumés containing perfectly observed signals, thereby ruling out statistical discrimination from the outset. Social-norm influences are also difficult to isolate, as field settings often entangle peer effects with monetary or institutional incentives.
This study offers the first experimental design capable of jointly disentangling all three mechanisms within a single unified framework. We achieve this by integrating noisy productivity signals, norm-based treatments into a carefully controlled résumé-screening task. To our knowledge, no prior experiment has combined these elements, nor has any been conducted in the distinctive institutional setting of Saudi Arabia, a high-context, authoritarian labour market where religion, nationality, and cultural fit are particularly salient. This dual novelty, theoretical innovation and contextual originality, ensures the contribution is both globally relevant and locally significant.