"By systematically analyzing the universe of black, Latino, Asian American Senate and Gubernatorial candidates, a representative sample of female candidates, and a large sample of white male comparison cases, we demonstrate that there is a Bradley Effect for black candidates that continues to the present day. We also show that women are susceptible to a reverse Bradley Effect. In addition to this contribution, the article also introduces a model of preference falsification which fills a gap in the survey response bias literature and is consistent with our overall findings, though to our knowledge no data exists to test it directly."Source: Center for the Study of Democracy. U.C. Irvine Paper 08-08.
Download full pdf publication | Link to online abstract