Bradley Effect

The Bradley Effect is named after Tom Bradley, the Black mayor of Los Angeles who lost the 1982 California gubernatorial race to Republican George Deukmejian despite leading in most pre-election and exit polls. Pollsters and political scientists subsequently hypothesized that some white respondents told interviewers they would support the Black candidate to avoid appearing prejudiced, then voted otherwise in the privacy of the booth — a form of social-desirability bias.

The concept was invoked to explain similar gaps in other races involving Black candidates, including Douglas Wilder's narrow 1989 Virginia gubernatorial win (he had led by larger margins in polling) and David Dinkins's 1989 New York City mayoral race. The related "Wilder Effect" is sometimes used interchangeably.

Evidence for a persistent Bradley Effect has weakened over time. Daniel Hopkins, then at Harvard, published a 2009 study in the Journal of Politics finding that the gap between polls and results in races with Black candidates largely disappeared after the mid-1990s. Barack Obama's 2008 primary and general election performances were closely watched as a test case: while he underperformed some polls in the New Hampshire primary, his national general-election result tracked final polling averages closely, and most analysts concluded no significant Bradley Effect occurred.

Methodologically, the effect is one of several explanations for poll-to-result divergence, alongside differential turnout, late-deciding voters, undercounted demographics, and nonresponse bias. Debate continues over whether observed gaps reflect racial dissimulation specifically or broader survey error. Some researchers have extended the framework to other identity categories, examining whether women or LGBTQ candidates face analogous effects, though findings are mixed.

For MUN and IR researchers, the Bradley Effect is useful primarily as a case study in survey methodology, identity politics, and the limits of polling as a predictor in racially salient contests. It should be cited carefully: it is a contested hypothesis, not an established empirical regularity.