Do people blame refugees for negative events? We propose a novel experimental paradigm to measure discrimination in responsibility attribution towards Arabic refugees. Participants in the laboratory experience a positive or negative income shock, which is with equal probability caused by a random draw or another participant's performance in a real effort task. Responsibility attribution is measured by beliefs about whether the shock is due to the other participant's performance or the random draw. We find evidence for reverse discrimination: Natives attribute responsibility more favorably to refugees than to other natives. In particular, refugees are less often held responsible for negative income shocks. Moreover, natives with negative implicit associations towards Arabic names attribute responsibility less favorably to refugees than natives with positive associations. Since neither actual performance differences nor beliefs about natives' and refugees' performance can explain our finding of reverse discrimination, we rule out statistical discrimination as the driving force. We discuss explanations based on theories of self-image and identity concerns.