Election Polling When There’s No Election Day
Late-breaking news may not matter that much anymore.
Via memeorandum, I see various right-leaning outlets touting a new Fox News poll showing Republican Glenn Youngkin ahead of Democrat Terry McCauliffe in the Virginia governor’s race:
Republican Glenn Youngkin has moved ahead of Democrat Terry McAuliffe in the Virginia governor’s race, less than a week before the election.
McAuliffe receives 45 percent to Youngkin’s 53 percent in a new Fox News survey of Virginia likely voters. Youngkin’s eight-point advantage is outside the poll’s margin of sampling error.
That’s a big shift from two weeks ago, when McAuliffe was ahead by five, 51-46 percent.
Likely voters are a subgroup of registered voters, identified mainly based on self-reported vote intention and interest in the election.
Among the larger pool of registered voters, it’s a one-point race: McAuliffe 47 percent vs. Youngkin 48 percent. Two weeks ago, McAuliffe led among registered voters by 11 points, 52-41 percent.
Under the way we used to think about elections in this country, this would be more interesting news. The election is this upcoming Tuesday and there seems to be a shift. Assuming it’s reflected in other polls and not simply an error in the way this one particular poll was constructed and/or conducted, it would indicate a possible upset is brewing.
And, it seems, the pollsters being interviewed for the story are buying into it:
What changed? GOP enthusiasm. The race is largely focused on education and this has energized Republicans, as 79 percent of Youngkin supporters are “extremely” interested in the election compared to 69 percent of McAuliffe supporters.
Democratic pollster Chris Anderson says if Democrats sense defeat, enthusiasm could shift in the final days of the campaign.
“With the race essentially tied among the full registered voter universe, McAuliffe could still pull this off,” says Anderson, who conducts the Fox News survey along with Republican Daron Shaw. “But it would take something big to ignite enthusiasm for McAuliffe’s candidacy and a massively effective get out the vote effort.”
Again, this is quite plausible. While both candidates’ ads (which I thankfully have to endure only during live sporting events) are sleazy and distorting, Youngkin’s campaign has, among other things, leveraged a statement by McAuliffe that parents shouldn’t be deciding what their kids learn in school to good effect.
But here’s the thing: I’m not sure any of this matters all that much. While Election Day is November 2, voting has been ongoing since September 17. Virginia allows and encourages voting starting 45 days before the election, not only via absentee ballot but in person at various facilities. In the past, voting early requiring attesting that one would be out of state or otherwise unable to making it to a polling station during normal hours on Election Day but that requirement was rescinded in April. Now, Virginians can vote for 45 days on a “no excuse” basis.
Whether replacing a communal Election Day with a 45-day voting period is a good thing is something I can preach either way. On the one hand, making it easy for people to vote is a good thing. On the other, it means that the debates and other late-breaking news are irrelevant in the case of those who’ve already voted. And, as of two days ago, nearly 870,000 Virginians had already voted.
If the trends shown in this poll are genuine, the disparity should redound to the benefit of McAuliffe, the candidate I unenthusiastically prefer. But I think we’re going to need to rethink how we think about polling when the horse race doesn’t have a single finish line.
There’s always an element of self-fulfilling prophecy whenever a state is deemed solidly blue or red. A lot of Democrats in Alabama don’t bother to vote because they “know” the Republican will invariably win; the party apparatus doesn’t put much effort into fielding a strong candidate, and it doesn’t invest much in campaigning. But every now and then when a race looks surprisingly competitive, it’s like voters from the party that doesn’t usually win awaken from their slumber. This is what happened in 2017 with Doug Jones in Alabama, and in 2010 with Scott Brown in Massachusetts.
Virginia is a bit complicated since it’s traditionally thought of as more of a swing state, and all its gubernatorial races in recent memory have been competitive, including the last one in 2017. But after Northam won by a surprisingly wide margin, after Dems flipped the legislature, after Biden took the state by over 10 points–it’s produced a narrative that the state has pulled away from the GOP and can now be regarded as a safe blue state.
Despite the source, Fox News polls don’t have a bad track record; they aren’t garbage right-wing propaganda polls like Trafalgar or Rasmussen. That doesn’t mean this one is necessarily accurate; even the best pollsters miss, and odd-number-year races have long been hard to poll accurately. (The 2017 race was off from the RCP avg by nearly 6 points.) Still, my concern is that we’re seeing the same kind of bandwagon effect we saw in 2010 with MA and 2017 with AL, where Republican voters who previously felt the race wasn’t worth bothering about have become newly engaged.
Whether early substantial early voting impacts polling is a great question. I don’t think we have the metrics to really know (yet) but if I had to guess, I will guess that an earlier voter is simply more likely than an Election Day voter to waive off a poll.
@Raoul:
That’s a theory I was hearing in 2020, which would have suggested Dems would outperform their polling, which obviously did not happen.
Consider the question, “how likely are you to vote on Nov. 2nd?”
When there was one voting day for most people, the answer was rather clear. Either a binary yes/no, or a continuum ranging from definitely will not vote to definitely will vote.
With early voting and more absentee ballots, how will people who’ve already voted answer the question? They could say “I’ve already voted,” which raises the further question of how the pollster classifies this. They may also answer “I won’t vote on Nov 2nd,” without really being a non-voter.
I’d look at the states where mail in voting is the norm, and most people therefore vote early while few vote on election day.
@Kylopod: It is interesting to note that the national polls had a bigger miss 2020 than 2016 (4% vs 1.5%- 538 stats), yet Biden still won by 7 million votes. The real problem with polling comes at a state level, it appears the pollsters miss a lot of data points. For a national poll, the state errors would tend to even out, so the number may not be that off even though the subsets of states may be way off. I guess the point is that we shall find out about polling without a lot of noise from other parts in the VA election.
@Raoul:
I read somewhere (maybe 538) that 2020 had the most inaccurate polling in a presidential race since 1980.
My personal opinion–and this has been surprisingly overlooked in most of the analyses I’ve come across–is that a lot of it had to do with the unusual conditions created by the pandemic. To start with, the massive shift to vote-by-mail totally screwed up traditional turnout models which try to predict which voters will vote based on past turnout patterns. I saw an interview with Ann Selzer, whose Iowa polls were among the more accurate in that cycle, and she said her polls simply asked respondents if they intended to vote or not. That hasn’t always been the most reliable method of predicting turnout, but apparently it worked in 2020.
Then there’s the fact that Republican campaigns were likelier to ignore the pandemic restrictions, which may not have been smart from a health perspective but enabled them to engage in more on-the-ground campaigning than Democrats did–again, spiking turnout in a way that the polls wouldn’t have picked up. Traditional door-to-door canvassing was almost entirely absent from Democratic campaigns that year–a lesson they later corrected in the Georgia runoffs.
Much of the commentary I’ve seen suggests that elections in which Trump is on the ballot have a tendency to underestimate Republicans. (After all, the 2018 polls were pretty good overall. There were a few big misses, but it was a wash between the parties.) The most popular theory along these lines is the shy-Trump-voter hypothesis, the notion that Trump voters lie to pollsters by claiming they aren’t voting for Trump. I believe this theory is flawed and doesn’t hold up to scrutiny. But there’s a related hypothesis that does have some plausibility, the idea that Trump in both 2016 and 2020 inspired certain people who don’t normally vote to come out to the polls, and that’s the type of thing that escapes the detection of pollsters. Also, Trump voters may be less likely to respond to pollsters who approach them, thus creating an imbalance.
That last theory has actually been a conservative criticism of election polling for decades–the idea that Republicans, being more skeptical of polls and the media in general, are less likely to participate in polls than Democrats. And I think it’s the consensus position by analysts over what went wrong in the 2004 exit polls (even though the pre-election polls that year were fairly accurate). The problem I have is that, while this theory isn’t unreasonable, the fact is that contrary to the claims of conservatives, there’s no general tendency for election polls to underestimate Republicans and overestimate Dems; if you examine polls over the years without cherry-picking, they do the opposite just about as often.
Well, pollsters and the media are whistling by the graveyard with these long voting periods. They make their money off focusing people on the ‘big event’ and little off an election that has no chance of being upended at the last minute. Really, Hollywood isn’t even going to make an election movie that doesn’t have a “big day” with a last minute comeback available.
What’s surprising is that even the political operatives seem to be ignoring that elections are ground games with no real chance of picking up a fumble to run into the end zone.