A quartet of independent polls in Virginia find wildly differing results in Democrat Terry McAuliffe's lead over Republican Ken Cuccinelli in the state's governor's race, with one indicating a tight race one week before Election Day and two suggesting we are headed for a landslide.
The smallest lead comes from a Quinnipiac University poll released Wednesday showing McAuliffe up 45 to 41 percent; Hampton University finds McAuliffe up six, the Washington Post and Abt-SRBI show a 12-point lead and Roanoke College finds McAuliffe up 15. Here's a table showing each of the polls.
The divergent results are surprising given both the consistency of McAuliffe's lead in polls over recent weeks as well as similarities in how the new polls were conducted - within the past nine days using live interviewers who called both land-line and cellular phones.
So, how can four polls -- all conducted via similar methods -- show such different results?
Here's why: The polls differ greatly in a) the share of Democrats and Republicans in the pool of likely voters and b) whether the likely electorate is becoming more Republican as Election Day approaches.
Take the Quinnipiac and Post polls, for instance. The Quinnipiac poll (McAuliffe +4) finds 31 percent of likely voters identifying as Republicans and 29 percent as Democrats (39 percent identify as independent or another party). This marks a sharp break from the firm's recent polling in the state. Only a week ago Quinnipiac found an eight-point Democratic advantage in party identification.
The Post poll, by contrast, finds hardly any change in the partisan makeup of likely voters or registered voters overall over the past month; Democrats hold an eight-point advantage in the latest survey, similar to seven points in the previous WaPo poll.
The polls' differences on party ID are not a function of pollsters "choosing" or "oversampling" one party's identifiers since both surveys draw random samples of Virginia adults, weight the results to Census demographics like age and sex, and determine the likely electorate by asking about voter registration, intention to vote and interest in the campaign.
Other demographics in the two polls also fail to explain the difference. Likely voters in the Post and Quinnipiac polls have similar shares of men and women, and the two surveys' age composition is similar. Likely voters in the Post poll are seven points more apt to be white than Quinnipiac (78 vs. 71 percent), but this difference would actually make the Quinnipiac sample appear friendlier to McAuliffe since he is running far ahead of Cuccinelli among minorities rather than showing a tighter race.
So, what's the answer? Put simply: The polls' samples of likely voters appear to have different political leanings.
The eventual partisan split of voters will not be determined until Election Day. Exit polling from the 2009 governor's race showed Republicans outnumbering Democrats by four points as Republican Bob McDonnell cruised to a double-digit victory. But Democrats held six and seven-point advantages in party identification in 2008 and 2012, respectively, when the Commonwealth voted for President Obama.
The big question is whether this Election Day will more closely resemble 2009 or 2012 in terms of the electorate. At the moment the atmosphere is quite bad for the GOP. Republicans took on more blame than Democrats for the highly unpopular federal government shutdown this month, after which negative ratings of the party hit a record high in national Post-ABC polling. The impact of the shutdown was strong in Virginia, with 35 percent of registered voters saying they were inconvenienced by it.
In Virginia, McDonnell has been mired in a scandal over large gifts he received, with his approval down sharply among voters this fall. That drop blunted much of the benefit Cuccinelli could have received from associating with a governor who was once overwhelmingly popular. Cuccinelli has also struggled to keep up with McAuliffe on the airwaves, being outspent by as much as 11 to 1 in closing weeks.