Elections matter a lot. At the presidential level this is clear as day — but down ballot, things are much murkier. Before we sit down and watch the results come in Tuesday evening, it helps to take stock of all the information we have and look for clues of where to find more.
What to Expect in Virginia’s Gubernatorial Race
Our first destination is perhaps the most obvious: public opinion polls. The reader familiar with the contest will likely be familiar with polls in Virginia given the somewhat wacky numbers being posted a week or two ago. Though there could be good reasons why the measurements were previously bouncy (POLITICO’s Steven Shepard and The Washington Post’s Scott Clement have you covered here), they have since settled into a relatively clear picture of the race.
That picture, contrary to what the polls may (and I emphasize “may”) have said last weekend, paints a relatively clear 3-4% lead for Democratic candidate Ralph Northam over Republican Ed Gillespie.
I say relatively clear because, for better or for worse, the public polls have seemed to clump around a 3-5% Northam lead since Saturday. In fact, the more recent measures are about 3 times as “focused” as the earlier ones (measured by the difference in the standard deviations of the polls fielded before and after November). That could either be signs of pollsters “herding” their estimate or, perhaps more simply, the effect of more pollsters releasing estimates over the past week. There have been thirteen polls released over the past week, for example, whereas each week prior to October 30th saw less than 4 polls released. More polls in the field would mean less deviation between them — at least theoretically.
Note that I’m not ruling out entirely the possibility that some firms may be adjusting their estimates. We all know that pollsters are under enormous pressure to get elections correct nowadays and often face public ridicule for releasing estimates that seem out-of-line with other polls. That’s what happened when Hampton University faced harsh public backlash for releasing a poll showing Ed Gillespie ahead of Ralph Northam by 8%… when the average put Northam up 5.
So the motivation is there, but is the evidence? As mentioned, the standard deviation of the last week’s polls is a measly 1.6%. Theoretically under average sampling error for polls with 780 respondents (the average sample size of polls of the Virginia Gubernatorial race), we should see a standard deviation closer to 2 or 2.2%. That being said, if the true Northam mean is 3% we would expect to see polls of Gillespie winning at least 5% of the time. In reality, we have seen that happen in roughly 1-in-10 polls. We shouldn’t lose much sleep over this.
Safe to say I’m rather confident that the average of polls this cycle is a sound one. But even the best polling averages can err.
In previous gubernatorial contests, polls have not been the best partner in our attempts to anticipate outcomes. The average error in measures of each candidates performance is a rather large 4.5%, and we can expect there to be roughly 10 points of error for either candidate. The 95% confidence interval on Ralph Northam’s margin of victory, then, is a stout 22%. So while he is ahead by 3% he could lose by as much as 7 percent to Ed Gillespie or win by as large a margin as 15 points.
To be clear, however, an error that large won’t happen more than 5% of the time. Given past error and Northam’s 3% margin in the polls, we can put his win probability somewhere around 65-70%, though I wouldn’t put any money on a higher bet. Northam is in a better position than Jon Ossoff was when he tried to flip Georgia’s sixth congressional district last summer, but does not have quite as strong of prospects that Hillary Clinton had almost a year ago today.
There are other reasons too that may prompt us to hedge our bets on Northam. Aside from the massive uncertainty of our fundamentals-based models for this contest, there is one major factor to consider: turnout.
Voter turnout in Virginia consistently gives Republicans an outside chance of overperforming expectations in midterm cycles. In 2013, for example, the Northern counties of Fairfax and Arlington — Democratic strongholds, to be sure — voted at just about even with the turnout rates of the rest of the state. But the rural central and western counties turned out to vote at up to 5, 7, and 10 points above the statewide average. The story in Virginia (at least so far) is exactly like it is elsewhere in the United States: Off-year elections are better to Republicans than presidential cycles are. The question is whether that will be enough for Ralph Northam to pull ahead. Will it? We’ll have to wait and see.
What to Watch on Election Night
Speaking of waiting and seeing, I have some tips on how to best follow along with the results on election night. Of course, there will be some fantastic tools, commentary, and data streams to witch at the New York Times, FiveThirtyEight, and Decision Desk HQ (respectively) but there are also ways to maximize your viewing experience from your living room.
First and foremost, during the race, you can compare Ralph Northam’s performance in Virginia’s counties with previous democrats relative performance there. Based on this metric, which is calculated by averaging each county’s relative Democratic performance per county per year. For example, since Clinton won Arlington County by 59% in 2016, but she won the entire state by 5 points, Northam’s benchmark there is 54%. Northam needs to beat his overall state margin by 59 points in Arlington to be favored to win the election (unless, of course, he makes up the votes elsewhere).
|County||2016 (Clinton)||2014 (Warner)||2013 (McAuliffe)||Northam's Benchmark|
|Falls Church City||52.6||44.3||45.6||47.0|
|Charles City County||19.5||29.4||26.1||25.3|
|Newport News City||21.3||21.6||21.6||21.6|
|Manassas Park City||22.8||7.5||9.3||12.3|
|Prince William County||15.8||1.7||5.6||7.2|
|Prince Edward County||0.1||6.1||5.3||4.2|
|Virginia Beach City||-8.8||-6.2||-4.6||-6.1|
|King and Queen County||-22.4||-3.8||-6.4||-9.7|
|James City County||-10.4||-11.9||-11.6||-11.3|
|Isle of Wight County||-25.8||-15.7||-16.6||-18.7|
|Prince George County||-22.2||-17.0||-18.0||-18.8|
|Buena Vista City||-35.9||-19.2||-26.6||-27.1|
|King George County||-32.9||-26.1||-28.0||-28.8|
|King William County||-40.4||-27.8||-27.5||-30.8|
|New Kent County||-42.6||-34.2||-33.3||-35.8|
|Colonial Heights City||-44.5||-47.4||-45.8||-45.9|
2. Bellwether precints (with DDHQ)
If you prefer an even tighter look at the data, you can keep your eyes on some precincts that are likely to go as the state goes. These are precincts we call “Bellwether” precincts, and they usually offer a very good (and early) look at how the race is shaping up. If you asked me for my top 3, I’d give you the following list:
- Fairfax 306 - Great Falls
- Virginia Beach City 42 - Brandon
- Chesterfield 404 - Providence
Of course, no three precincts can tell you how the election is going to unfold: that’s why I’ll be tweeting out updates for a set of 30 as election night wears on! You’ll want to follow me and Decision Desk HQ to get updates.
Finally, I’ll be running the live projections of the Virginia governor’s race for Decision Desk on their blog and twitter feed. Here’s a spoiler of what you can expect.
With the evidence available to us, we can’t say confidently that one candidate has a huge lead in the 2017 Virginia gubernatorial race. Rather, Democrat Ralph Northam has shored up enough support to hang on to a 3-4% lead in public opinion polls over the last week or so. But polls have been wrong in the past, and so are some of the normal models we would run that use data about turnout and the “fundamentals” at play in the race. Rather, the contest looks like it’s teetering on a surprise, and we’re just going to have to wait until election night to see how that plays out. Be sure to follow along with us online as we do exaclty that.