The Bradley Effect

I was 11 in 1948 when the U.S. went to bed having elected Tom Dewey president, and awoke to find Harry Truman had been surprisingly reelected after all. I don't remember much about my early years, but I remember when World War II ended (I was 8) and I discovered that my parents had been frightened--had been looking up for German bombers--my whole life, and they weren't any more. And I remember that strange November night in 1948 when Truman defeated Dewey.

Public opinion polling has greatly improved since then. For one thing, the egg they got on their faces that night made them significantly more humble--they ran scared for decades. And social science and statistical mathematics have improved greatly. In addition, computers and ubiquitous telephones have changed the polling landscape. We simply will never see again the kind of polling and reporting errors that characterized that election in 1948.

However, there is the "Bradley Effect." In 1982 Tom Bradley was comfortably in the lead over the Republican candidate, George Deukmejian, in the California governor's race. But, all polling (including exit-polling) notwithstanding, he lost. The myth is that both white and non-white voters are inclined to tell pollsters they favor a non-white candidate over a white one, and then, in the moment of truth, to pull the white candidate's lever. However, in the case of Bradley's defeat this was not the real story. Bradley was probably defeated by two other relatively silent factors. One was an unpopular gun control initiative; the other, an aggressive Republican absentee ballot program that generated hundreds of thousands of Republican votes no pollster anticipated.

The so-called "Bradley effect," unpollable racism, may have reared its head again the Virginia gubernatorial race in 1989 (and is occasionally therefore called the "Wilder Effect") and the election of David Dinkin for mayor of New York City over Rudi Guliani in 1989 (the "Dinkin Effect').

Will it affect Barack Obama's victory over John McCain next month? Probably not significantly.