This paper investigates the persistence of gender stereotyping in the forecasting of risk attitudes. Subjects predict the gamble choice of target subjects in one of two treatments. First, based only on visual clues and then based on visual clues plus two responses by the target from a risk-preference survey. Second in reverse order: first, based only on the two responses, then on the two responses plus visual clues. In isolation the gender stereotype and survey responses both inform predictions about others risk attitudes. In conjunction with one another, however, the stereotype persists and dominates the survey response information.