Wednesday, October 22, 2008

POLLS UNDERESTIMATE RACIAL BIAS FACTOR

THE BRADLEY EFFECT, RACIAL BIAS UNDERESTIMATED

With the elections less than 2 weeks away, everyone is talking about the likely fact that polls have not reached Democratic voters that harbor racial prejudice.  A fraction of these people may vote for Sen. McCain or not vote at all.

Are the polls capturing all that bias?

One voter model attempts to factor in hidden racial bias that was published this month suggests it might drag down the numbers for Obama by 6%. 

The hidden factor is commonly called the “Bradley effect’ dating back to 1982 when Tom Bradley, an African American lost the California governors race despite being in the polls on Election Day.  A similar pattern in Virginias governors race by Douglas wilder, an African American had a 15% lead but won by a half percent.

Many African Americans have won mayoral, state, and congressional elections since.  Racial attitudes have certainly changed.  Dan Hopkins, a Harvard political scientist examined every gubernatorial and senate race since 1989 which had either an African American or female candidate, on rationale that a woman could also experience voter prejudice.  Of 133 elections only 18 had an African American.  There was no Bradley effect in any of them since 1996.  He does not expect a Bradley effect.

Kohut, president of Pew Research Center, which performs public opinion polls does not think people polled are dishonest when asked if they will vote for Obama.  But he thinks there are reluctant responders that refuse to be polled.  They have a less favorable attitude toward racial minorities and the polls may miss their intent.

Nine groups published their models this month in PS: Political Science & Politics and only one, Lewis-Beck from the Univ. of Iowa adjusted his model to take race into account.  He fine-tuned his model on voting patterns in the primaries and on how honest people are in saying they will support an African-American presidential candidate.

Beck's prediction:  Obama will get 50.1% of the popular vote but lose the election in the Electoral College.  When he did not take race into account Obama would get 56% of the vote.  It’s a feel good thing: people don’t want to admit race is a factor.  In the past he has been 1.5% off in his predictions.

It’s hard to detect hidden bias, since polling is done on the phone and involves a one on one interaction.  You really will not tell a stranger you have racial prejudice.

A study by Krysan in 1994 asked Detroit residents if they would vote for an African American.  Face to face 11% said no, while in a mail survey 22% said no, privately.

In 2007, sociology students at New York University asked three groups related race questions that were hidden and statistically the groups were all similar.  They found 14% of the people lied.  Democrats were more likely to be dishonest than Republicans, as were those with less education.

Racial bias is clearly visible, and it has shown in preference polls.

The bottom line is Mc Cain still has a shot at the White House.

Sources: Science Oct 10, 2008,Political Science & Politics Oct, 2008

Visit www.drneedles .com for more blogging on controversial subjects.

 

 

 

 

 

No comments:

Labels