Virginia’s governor race is being billed as a test case for 2016, but you can’t read much into polling data this early
Virginia’s 2013 gubernatorial election may not have particularly appealing candidates, but the race has taken on extra meaning because of the state’s swing status in presidential elections. The race pits a so-far unlikable Democrat, Terry McAuliffe (aka T-Mac), against a exceptionally conservative Republican, attorney general Ken Cuccinelli (aka Cuch). Are swing state Virginians willing to elect a person, Cuccinelli, who wants to defund Planned Parenthood and has labeled the Environmental Protection Agency an “agency of mass destruction”? McAuliffe, on the other hand, was a long-time Clinton fundraiser and adviser, and his run is seen as a test of the Clinton name and organization for Hillary’s possible run in 2016.
From two polls produced this week we can make similar conclusions with different results. An NBC/Marist poll has T-Mac up by 2pt among registered voters and down by 3pt among likely voters. A Washington Post survey has Cuch up 5pt among registered voters and 10pt among likely voters. You’ll note that the overall results are different, but that the gap between the registered and likely voters is the same.
Right now, the truth is probably somewhere in the middle. Quinnpiac has polled registered voters three times since January, and each of their results fell in between the registered voter numbers from Marist and the Post. Likewise, a Christopher Newport University survey from January had a similar 4pt gap between registered voters and likely voters. Off-year elections tend to feature more Republican friendly electorates, given that African Americans and young people are a lower percentage of the electorate than in presidential year elections.
But does this current polling actually tell us anything about what will happen on election day? Not really. At this point in 2009, all but one scientific poll had eventual winner and current Republican Governor Bob McDonnell winning by 9pt or less. A few polls even had Democrat Creigh Deeds ahead by mid-June, but most had McDonnell ahead by about 5pt. With a few exceptions – notably from SurveyUSA, which consistently showed double-digit McDonnell leads by early August – most surveys continued to show a single-digit McDonnell lead through September. McDonnell won by 17pt.
You think that’s unusual? Go back to the 2005. Through this point in the campaign, not a single poll had the eventual winner, Democrat Tim Kaine, ahead. A few polls had him down as much as 10pt, while the majority had him behind by mid-to-high single digits. Republican Jerry Kilgore trailed in only one public poll, all the way through early October! Tim Kaine, of course, went on to win by 6pt.
Overall, the polling at this point and through September has been off by about 10-15pt the last two elections. Given that Cuccinelli has a 4pt lead in the HuffPollster aggregate, either candidate could win. Perhaps Cuccinelli will win by a margin greater than 4pt, maybe less, maybe exactly that. Or maybe T-Mac will come out on top.
You might be tempted to look at the candidates’ favorable ratings, but I wouldn’t make too big a deal of those either, as they merely reflect the ballot standing of each man at this point. The important point is that both feature positive net favorables. Given that over 30% of voters have no opinion of Cuch, and over 40% have no opinion of T-Mac, the favorable ratings of these candidates will change. In 2005, Jerry Kilgore had a higher net favorable than Tim Kaine, which could be seen in the ballot test. He ended the election with a lower net favorable rating and lost. In 2009, Bob McDonnell had a net favorable rating equal to Creigh Deeds in early polling, which was reflected by a tight race in the ballot test. McDonnell ended up with a much higher net favorable rating, and indeed, he crushed Deeds.
What about the difference between the likely and registered voter results? The number one polling lesson from 2012 was that when likely and registered voters disagree in elections with high voter turnout, you should go with the registered voter results. Obviously, an off-year affair is not a high turnout election. Still, I was curious to see if over the past two cycles, pollsters have tightened the electorate, keeping in mind that only a few pollsters release results among both registered and likely voters, and most don’t.
Among those who report results from different screens, it’s not really clear that pollsters who use a likely voter screen are getting more accurate results than those that only polled registered voters. A 2009 October Virginia Commonwealth University poll had McDonnell winning by 18pt among likely voters and 16pt among registered voters. Neither was more accurate than the other. A 2009 October Roanoke College poll had McDonnell ahead by 17pt among likely voters and by 19pt among who said they were certain to vote (who we’ll call “definite” for the sake of clarity). In this instance, the looser screen ended up being closer to the final result.
In 2005, tightening the electorate did pollsters no favors either. An October Diageo/Hotline poll found Kaine winning by 2pt among registered voters, 1pt among likely voters, and down by 2pt among definite voters. Kaine, as mentioned, won by 6pt. An October Roanoke College poll had Kaine winning by 8pt among probable voters and 10pt among definite voters. Again the looser screen performed better than the more selective one.
What about earlier in the campaign? In 2009, an August Washington Post poll had McDonnell up by 15pt among likely voters and by 7pt among registered voters. Clearly, the likely voter result was closer to the truth – in this case, at least. In 2005, the opposite occurred. An early September Washington Post poll had Kilgore up by 7pt among likely voters and by 4pt among registered voters. Earlier in the campaign, SurveyUSA showed a dramatic rise in Kilgore’s lead from 4pt to 10pt between May and June when switching between registered and likely voters. Other polling companies showed a consistent race, which means that the far less accurate June result was almost certainly because of a change in the electorate being polled.
Thus, I’m not really sure there’s much to take from the recent polling data on the Virginia gubernatorial race. Yes, Ken Cuccinelli likely has a small lead at the moment, but that edge probably means very little. The difference between the registered and likely voter screen may be instructive, but we can’t yet be sure when to trust it. Tighter voter screens have not proven to necessarily be more accurate in either the 2005 or 2009 Virginia gubernatorial elections.
We can only sit back and allow the campaign to unfold.
• A previous version of this article incorrectly stated the leader of The Washington Post’s survey of the Virginia gubernatorial race and has been corrected accordingly