George,

It is time once again to put to rest the notion that many white respondents in polls give “politically correct” or socially acceptable responses when an African-American candidate is one of the choices.

This phenomenon has once again been cited in the news lately. The Doug Wilder and Tom Bradley examples are often cited.

The argument goes something like this: A white respondent receives a call from some distant stranger from a polling organization. The respondent really doesn’t like the idea of having an African-American as governor or president, but wants to be “politically correct” in the eyes of the interviewer (even though the interviewer can’t see the respondent). So the respondent says he is going to vote for the African-American candidate even if he/she has no intention to do so.

This is nonsense. People just do not go through this kind of physiological contortion when answering surveys.

In citing ONLY the Wilder and Bradley examples, political pundits are engaging in false analysis. It is called cherry picking data – picking only those examples that prove your case. In 2006, there were three gubernatorial contests in which an African-American was one of the candidates – Deval Patrick in Massachusetts, Ken Blackwell in Ohio and Lynn Swann in Pennsylvania. In Ohio, the polls were right on – Blackwell received exactly the same vote as predicted by the polls. In Massachusetts and Pennsylvania, the African-American candidates received MORE votes than predicted. If we add the Wilder contest in Virginia and the Bradley contest in California to the three contests cited above, we have five contests. In two of them, the African-American got more votes than the polls predicted, in two the African-American got fewer votes than expected, and in one the polls were right on. In short, there is no systematic evidence of the Bradley effect.

We keep forgetting that polls, including Exit Polls, are very crude instruments. Their predictions are often quite different than actual vote. Many times the polls show that a candidate is far ahead only to have that candidate win by only a small margin, or even lose. There are many reasons why ANY poll is off the mark. To pick only two examples – the Wilder and the Bradley elections – and say that they were off because of some contorted way in which supposedly biased respondents gave their vote preference is nonsense. To single out the Wilder and Bradley contests and invent some racial reason for them being inaccurate (when so many polls are inaccurate for all sorts of reasons) is singling out black candidates for unusual treatment. That is being unduly racially conscious on the part of the analyst.

David RePass (Electronic mail, October 9, 2008)