This study provides a set of experimental results highlighting a new possible mechanism for racial discrimimation, based on biases in how people recall payoff-relevant information about others, depending on whether they are of the same race or not. Building up on the work by psychologists showing own-race biases in the memory of faces, I show that these biases extend to the recall of payoff-relevant information: Payoffs attached to people are more accurately recalled within race than across races. Consequently, discrimination arises in an environment where there are no differences in the underlying distributions of productivities and productivities have been observed perfectly at first. The experiment involves East-Asian and White subjects who see an equal number of pictures of East-Asian and White faces and each face is mapped to a payoff-relevant value. Incentives are provided to recall faces associated with higher values. I find a clear asymmetry in recall: High value faces are more accurately recalled within race than across races and lead to higher efficiency losses. I contrast these results with a treatment where race is a distinctive attribute for a minority of faces (where there is only a small number of pictures of East-Asians). In that case, East-Asians are recalled more accurately both for White and East-Asian subjects. These results raise new questions on the implications of such cognitive biases for the nature of cross-racial relations, in particular for phenomena relying on re-identification, such as the formation and maintenance of social ties, the establishment of trust and the sustainability of cooperation. |
