Articles and Analysis


More New Polls in Virginia

Topics: 2006 , The 2006 Race

Our most recent update to the Slate Election Scorecard focuses on a new poll in the New Jersey Senate race, but I would like to say a few words about Virginia. We updated the race on Thursday, but two new polls have been released there since. Taken together, the new surveys show George Allen maintaining a narrow lead over Democratic challenger Jim Webb despite a swarm of new allegations regarding his past use of racial slurs. Those allegations may be having an effect, however, in blunting the impact of Allen's massive spending advantage in September.

First, a brief timeline: SurveyUSA fielded its first Virginia poll this week Sunday through Tuesday night, releasing data on Wednesday showing Allen leading among likely voters 49% to 44%, slightly (but not significantly) more than their previous poll conducted two weeks earlier. However, a new story broke in Virginia as the poll was in the field and the SurveyUSA release speculated about "volatile" race they saw in "day-to-day data:"

On Sunday 9/24, after Allen had been accused of using racial slurs in college, he led by 7 in SurveyUSA Sunday-only data . On Monday 9/25, after Allen strongly denied the accusations, he led by 11 in SurveyUSA Monday-only data. On Tuesday 9/26, after more people corroborated the accusations, Allen trailed Webb by 3 points, in Tuesday-only data. The 5-point Allen advantage shown here, when the 3 days of data are combined and averaged, cannot be considered stable [emphasis in original]

Presumably wanting to check a potential break to Webb, SurveyUSA continued to conduct interviews Wednesday and Thursday nights. They released new data yesterday based on a "rolling average" of surveys conducted Tuesday through Thursday. The result: Allen leads in the second wave by six points (50% to 44%). The four SurveyUSA polls released since mid-august indicate a small but consistent Allen lead (see their chart below - click it to see a full size image). However, the slightly growing Allen lead cannot be considered statistically significant given the reported sampling error of four or more percentage points of each survey.


Several readers wondered why the Slate summary on Thursday made no mention of the day-to-day results mentioned in the fine print of the first SurveyUSA release. The reason was partly the space limitation of the Slate summary, but mostly because - as Gerry Daly on Crosstabs correctly noted - "volatility" is a built in feature of small nightly samples. SurveyUSA interviewed roughly 200 to 210 likely voters each night this week, not enough to make the day-to-day shifts they reported statistically significant. The additional interviews conducted Wednesday and Thursday nights confirm that Allen's lead was undiminished by the end of the week.

Let's give SurveyUSA credit for sharing their suspicions of a mid-week trend to Webb and for conducting additional interviews to help sort things out. But let's also remember a rule of thumb I have learned over the years: When the larger sample indicates consistency but smaller subgroups look unstable, the "volatility" is usually in the survey, not the voters.

But wait, there is one more survey in the mix, this one released late yesterday by Mason-Dixon. They show Allen and Webb tied at 43% each among 625 likely voters** interviewed Saturday through Wednesday last week. Three weeks earlier, they had Allen ahead by 4 points (46% to 42%). While the difference between a four point lead and a tie obviously has great psychological significance, the two results are within sampling error of each other.

Are the trends in these polls hopelessly conflicted? Not really. Neither trend is statistically significant.  Consider that chart below that Charles Franklin created for us this morning. It plots the Allen margin (Allen's vote minus Webb's vote). While the chart shows the random variation from poll to poll, the grey "local trend" regression line (based on the telephone polls only) that fits the points well and shows the narrowing of the race that occurred over the summer mostly flattening out in September.


[Obviously, the Zogby Internet surveys are an exception. While the most recent release snaps back into line with other polls, four out of six Zogby polls have shows results much more Democratic result, and the huge Allen rebound the show in September is inconsistent with what the other surveys have shown].

The recent general stability in the telephone surveys is remarkable if you consider how massively the Allen has been outspending Webb over the same period. The controversies surrounding Allen may be dominating the press coverage of the race, but the air war is a totally different story. In June, Allen reported $6.6 million in cash-on-hand to Webb's $424,000, and that advantage is now playing out. The Hotline (subscription only) reported on Thursday that Allen has already spent over $3.5 million on electronic advertising and his campaign has purchased another $650 thousand in television time next week, and much of that attacks Webb.

Webb's advertising is very light by comparison. The Associated Press reported on Wednesday that Webb started airing his second television ad on Tuesday, but only in "the Norfolk and Roanoke broadcast markets and cable in the Washington suburbs at a total cost of $65,000 through Oct. 2." So in the voter rich Richmond and Washington DC markets, voters have seen plenty of this advertisement attacking Webb, but have seen next to nothing of the spots introducing Webb.

So the controversies surrounding Allen are having an effect. For now, at least, they appear to be blunting the impact of Allen's attack ads, and that alone is remarkable.

**The language of the Mason-Dixon release is a little confusing, but they are reporting results among "likely voters" and not all registered voters: "A total of 625 registered Virginia voters were interviewed statewide by telephone. All stated they were likely to vote in the November general election" (emphasis added).


Gary Kilbride:

Someone on another site made a good point yesterday, that the SurveyUSA poll on Allen/Webb is undersampling blacks. It includes 14% blacks while the typical percentage in Virginia is low 20s. Also, given the recent allegations and historical trends it's hard to believe Allen will receive 27% of that black vote, which is what SurveyUSA gives him.

And once again I'm baffled that SurveyUSA is finding Hispanics who overwhelmingly prefer the GOP. In that Virginia poll it's 17 Hispanics for Allen and 6 for Webb. That has been a consistent pattern in SurveyUSA polls this year -- and granted it's tiny raw numbers -- but the GOP with more than half of the Hispanic vote, often 70+%. Meanwhile, Democrats figure to receive roughly 60-65% of the Hispanic vote nationwide. I'm wondering if SurveyUSA makes calls in English to Hispanic households, and somehow a disproportionate percentage of those who complete the automated phone survey in English lean Republican. That's pure guesswork but I can't come up with anything else.

BTW, to get to the raw numbers on the SurveyUSA polls, at the crosstabs page go to the dark column at far left and change the pulltab from "Show Percentages" to "Show Counts (Frequencies)"

On progressive sites many posters are baffled that Webb isn't zooming into a large lead. I think some of the numbers from the Mason-Dixon poll tell the story. Allen maintains a net positive rating of 41-32 with 27% neutral. That's what I suspected, that the old charges would have much more impact on a political newcomer than someone Virginians have known and formed an opinion of for decades. The new Virginians in the northern counties may be influenced but based on demographic trends they were probably less likely to support Allen to begin with.

Note the finding from the Virginia Mason-Dixon poll, that 64% say their vote will be "for candidate" to 35% "against candidate." That's been my frustration for years, that Democrats seem to believe that percentage is reversed and want to focus heavily on negativity. It wasn't negative campaigning that won in 2002 and 2004 for Republicans, it was women switching in small but critical percentage to the GOP due to a perceived positive, that the Republicans were better suited to protect them on national security concerns.

IMO, the Swift Boat campaign was the most overhyped and irrelevant aspect of 2004 or any recent cycle. Kerry was not considered charismatic or particularly likable long before the Swift Boaters, and there were indications all year the womens vote was drifting to Bush. I remember this specific section from a NY Times article, which I bookmarked at the time: http://www.nytimes.com/2004/09/22/politics/campaign/22women.html?ei=5090&en=5375fdd1ed0d73e7&ex=1253505600&partner=rssuserland&pagewanted=print&position=

"Andrew Kohut, director of the Pew Research Center for the People and the Press, said that women over all were "cross-pressured'' this year because they had more confidence in Mr. Bush on terrorism but were skeptical that he would handle the economy better.

"There's a tension among middle-class women who are attracted by Bush's perceived strength on terrorism and concerns about his poor performance on the domestic agenda,'' Mr. Kohut said. "They will be the story of this election: the way women make this choice.''

Mr. Kohut said that in 2000, men voted for Mr. Bush at the same rate that women voted for Mr. Gore. "Now,'' he said, "Bush has more solid strength among men than Kerry has among women, and if Kerry can't match that margin among women, he will lose.'"

This year I'm convinced Bush has forfeited that strength among women due to relentless incompetence, leading to a restored gender gap, but Democrats are not doing nearly enough in terms of emphasizing economy/jobs to insure young women show up at the polls.

Anyway, Webb is never going to take full advantage of Allen's woes, particularly among the womens vote, when he has 8% who don't know who he is and 41% with a neutral opinion. The old comments about women in the military may be more of a factor than many believe. I've seen anecdotal evidence on many progressive forums that Allen's ads on the matter are effective, with posters from Virginia saying their wives and daughters weren't thrilled with Webb after viewing those commercials. IMO, Webb needs to get on the air with some warm and fuzzy spots to boost his own favorables.


Post a comment

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.