On Saturday I told you about an analysis of polling in presidential battleground states by Mason-Dixon Polling & Research telling what might happen on Election Day based on past experiences. Here is that analyst:
MASON-DIXON BATTLEGROUND STATE SUMMARY
STATE OBAMA MCCAIN OTHER UND DATES
Colorado 49% 44% 3% 4% 10/28-29
Florida 47% 45% 1% 7% 10/29-30
Missouri 46% 47% 2% 5% 10/29-30
Nevada 47% 43% 2% 8% 10/28-29
N Carolina 46% 49% – 5% 10/29-30
Ohio 45% 47% 2% 6% 10/29-30
Pennsylvania 47% 43% 1% 9% 10/27-28
Virginia 47% 44% – 9% 10/29-30
*** Missouri, North Carolina, Ohio, and Pennsylvania courtesy of NBC News.
*** Colorado courtesy of The Denver Post.
*** Florida courtesy of The Miami Herald, Orlando Sentinel, Tampa Tribune, Florida Gannett Newspapers and Florida NBC Television affiliates.
*** Nevada courtesy of the Las Vegas Review-Journal and reviewjournal.com.
*** Virginia courtesy of The Richmond Times-Dispatch, Norfolk Virginian-Pilot, Roanoke Times, Lynchburg News & Advance, Charlottesville Daily-Progress, Bristol Herald-Courier, Martinsville Bulletin and Danville Register & Bee.
625 were sampled in each state.
HIDDEN RACIAL VOTING COULD SAVE MCCAIN
J. Bradford Coker
Managing Director, Mason-Dixon Polling & Research, Inc.
The 2008 presidential election features the first African-American nominee of a major political party, and this historic event will provide a valid testing of public opinion polling as it relates to stated voting preferences and race.
There has been considerable discussion of the so-called “Bradley Effect” on election polls, which is a reference to the 1982 California governor’s race between Democrat Tom Bradley, the African-American mayor of Los Angeles and Republican George Deukmejian, the state’s Attorney General. Deukmejian won a narrow victory after trailing Bradley in all of the pre-election polls. Some academics have recently written opinion pieces challenging the actual validity of the “Bradley effect” — defined as the tendency of white voters to tell pollsters they were “undecided” rather than admit they were voting against the black candidate – while others say it was real, but no longer valid.
In 1982, I was a market research analyst for a national chain of bowling centers, and while politically active and avid an election student, I had not yet founded Mason-Dixon. As such, I have no personal or professional connection to that 1982 race, have not reviewed all of the relevant polls and do not feel that I can offer an opinion specifically about that election.
I do, however, have first hand experience with several other elections that had very similar circumstances. These are the 1989 Virginia governor’s race between Republican Marshall Coleman and Democrat Doug Wilder (an African-American), the 1990 North Carolina Senate race between Republican incumbent Jesse Helms and Democrat Harvey Gantt (an African-American), and the 1990 Louisiana Senate race between incumbent Democrat J. Bennett Johnston and Republican David Duke (a former leader of the KKK). In each of these three races, the final pre-election polls showed a high percentage of “undecided” white voters, almost all of whom ended up voting for the white candidate, or in the case of Louisiana, the candidate with a clearly racist background.
In Virginia, our final Mason-Dixon poll showed Wilder leading Coleman 48%-44%, with 8% undecided. 96% of the “undecided” voters were white. On Election Day, Wilder won by the very slim margin of 50.2% to 49.8%. In other words, the “undecided” whites broke almost all in favor of Coleman.
In North Carolina, our final Mason-Dixon poll showed Gantt leading Helms 48%-44%, with 8% undecided. 92% of the “undecided” voters were white. On Election Day, Helms won 53% to 47%. Again, the “undecided” white voters broke to the white candidate.
In Louisiana, our final Mason-Dixon poll showed Johnston leading Duke 53%-26%, with 8% for a more mainstream Republican candidate and 13% undecided. 90% of the “undecided” voters were white. On Election Day, Johnston won with 54%, but Duke’s support jumped 18-points and he finished with a strong 44% of the vote. This surge was from both “undecided” whites and some white voters that stated they were voting for the other Republican ended up pulling the lever for Duke.
Of note regarding Wilder and Gantt, their historic candidacies drew a tremendous amount of attention from the national news media. The history-making impact of electing a black candidate to major office for the first time in the South was not lost upon the voters. In the case of Wilder, our poll showing him only up 4-points was considered “questionable” by some in the national press. A Washington Post poll that came out at the same time gave Wilder a 15-point lead (52%-37%) and their headline stated “Wilder Taking Command in VA Race, Poll Shows; Coleman Slips as Campaign Nears End.” The paper’s lead paragraph stated “Virginia Lt. Governor L. Douglas Wilder, riding a wave of support among women, younger voters and suburbanites has opened up a significant lead over Republican J. Marshall Coleman.” If you substitute “Barack Obama” for “Doug Wilder” and “John McCain” for “Marshall Coleman” this sounds very similar to what is being written in some quarters lately.
So, the million dollar question is will there be a “Wilder/Gantt/Duke Effect” in the 2008 presidential race? No matter what anyone theorizes, the answer today is that no one knows for sure. The three races I cited were all almost 20 years ago, and changes in the country may have mitigated many factors regarding race and voting. But, since this is the first time an African-American has been this close to the winning the presidency, it would be foolish to completely rule it out.
IF (and I emphasize the word “if”) there is still a racial effect on voter behavior as it relates to polling, the circumstances today have a number of statistical similarities. In all of the battleground states, the “undecided” vote is overwhelmingly white and generally higher than in recent presidential elections.
– In Missouri, 96% of the “undecided” voters are white
– In Ohio, 95% of the “undecided” voters are white
– In Virginia, 93% of the “undecided” voters are white
– In Pennsylvania, 93% of the “undecided” voters are white
– In Nevada, 86% of the “undecided” voters are white
– In Florida, 84% of the “undecided” voters are white
– In Colorado, 84% of the “undecided” voters are white
– In North Carolina, 81% of the “undecided” voters are white
The potential effect if all of these “undecided” white voters break for McCain would clearly alter the outcome in the battleground states.
– In North Carolina, McCain’s support would increase to 53%
– In Ohio, McCain’s support would increase to 53%
– In Missouri, McCain’s support would increase to 52%
– In Virginia, McCain’s support would increase to 52%
– In Florida, McCain’s support would increase to 51%
– In Pennsylvania, McCain’s support would increase to 51%
– In Nevada, McCain’s support would increase to 50%
– In Colorado, McCain’s support would increase to 47%
I must emphasize that this is not a prediction. It is simply to lay out the statistical scenario that could occur if there is a “Wilder/Gantt/Duke Effect” in the 2008 presidential election. If McCain should come back to win in the key battleground states, this phenomenon has to be considered as one of the factors in such an outcome.