We’re in the final stretch of the campaign, with just 15 days to go until the election. Indeed, “the election” is something of a misnomer, since early or mail voting is already underway in most states and around 30 million people have already cast ballots.
This is a period of high anxiety for almost everybody. So here are a few tips for how to process news and polls over the final two weeks. I’m going to keep these fairly short and sweet; we’ll save the more philosophical stuff about election forecasting, etc. for elsewhere.
1. Keep the context of the COVID-19 pandemic in mind.
The most important story in the country is that 210,000 Americans and counting have died from the coronavirus, plus virtually everybody’s lives have been disrupted in some way by the pandemic. And far from the U.S. having turned the corner, cases and hospitalizations are growing again in most parts of the country. President Trump’s approval ratings on COVID-19 are poor, too, considerably worse than his overall approval numbers. The pandemic isn’t a new story, but it’s liable to be more important to voters than, say, whatever is in Hunter Biden’s emails
2. Don’t assume the race is in the bag for Biden.
Although COVID-19 and other issues make Trump’s road to reelection difficult, he still has a 12 percent chance of winning the Electoral College, according to the FiveThirtyEight model as of Sunday afternoon. And if Joe Biden maintains his current lead in the polls, Trump’s chances will fall further — although the forecast thinks it’s more likely that the race will tighten.
But say Trump’s chances do decline further — to 5 percent by Election Day, for example — I’d keep a few things in mind.
First, even a 5 percent chance is something you ought to take seriously if the consequences are very high, something I think both Trump and Biden supporters would say is true of this election. And second, the outcomes in this election aren’t entirely binary. Say Biden wins: His margin of victory will still be heavily scrutinized. Does he win by double digits nationally? Does he win a state like Texas? This could affect both the degree to which Democrats pursue a more aggressive agenda, and the extent to which Republicans regard Trumpism as having been repudiated. Also, many Senate races are competitive, and having control of 50 versus 52 versus 54 or more Senate seats will greatly affect Biden’s first two years in office. Statewide races matter too, especially in states where control of the redistricting process is still in play.
3. But also don’t buy the narrative that “polling is broken.”
Polling is an imperfect instrument, more so in some years than others. However, 2016 — while far from a banner year from the polls — was not quite so bad as some critics assume. The national polls were pretty good, and Trump’s wins in the swing states were not that surprising based on the close margins in those states beforehand. The 2018 midterms was one of the more accurate years for polling on record, meanwhile.
One other thing to keep in mind about polls in an election like this one: They do provide some way to measure public sentiment, however imperfect, independent of election results, which could be important if the election is disputed. It’s not surprising then that Trump frequently disparages polls — which could give Americans more confidence about the results if they closely resemble the polls — when he’s also repeatedly failed to commit to accepting the results of the election.
4. Don’t get too obsessed with comparisons to 2016.
I know it’s fashionable to make comparisons between 2016 and 2020, and to discuss the various ways in which they might or might not be similar (there are fewer undecided voters this year, for instance). But to some extent, I think those comparisons are misguided. You shouldn’t make too much of a sample size of one election. And you should avoid thinking in binaries, i.e., that the polls will either be “wrong” or “right.” Instead, it’s more of a spectrum: The larger Biden’s lead in the tipping-point states, the more that polls could be wrong and he could win anyway. And there’s also no guarantee that a polling error will work in Trump’s favor as it did in 2016. In 2012, polls underestimated then-President Barack Obama and Democrats instead.
At the same time, while there is no particular reason to think that polls will be wrong in exactly the same ways that they were in 2016, there are also precedents for larger polling errors than the one we experienced in 2016. The final Gallup poll in 1948 had Harry Truman 5 points behind Thomas Dewey, for example, but Truman actually won by 4 points, making for a 9-point polling error. A polling error of that magnitude probably would be enough for Trump to win the Electoral College, although not necessarily the popular vote.
5. Don’t pay much attention to individual polls; wait for polling averages to move.
This is perhaps the single piece of advice we give most often at FiveThirtyEight, but it’s especially important in the final couple weeks of a campaign. After a lull this weekend, there are likely to be a lot of polls the rest of the way out. On any given day, it will be possible to take the 2 or 3 best polls for Biden and tell a story of his holding or expanding his lead, or the 2 or 3 best polls for Trump and make a claim that the race is tightening.
Resist buying too much into those narratives. Instead, turn to polling averages like FiveThirtyEight’s that are smart at distinguishing (ahem) the signal from the noise. We do program our averages to be more aggressive in the closing days of the campaign — so if there’s a shift in the race, our average should start to detect it within a few days. But while there is such a thing as underreacting to news developments,One reason that FiveThirtyEight’s model gave Trump a better chance than others in 2016 is because it shifted more following then-FBI Director James’s Comey letter to Congress.
“>1 the more common problem in the last days of a campaign is false positives, with partisans and the media trying to hype big swings in the polls when they actually show a fairly steady race.
6. Beware talk of “October Surprises.” They’re usually overhyped.
Indeed, while the Comey letter really did matter in 2016, contributing to a 3-point shift toward Trump in the waning days of the campaign, it’s more the exception than the rule. On average, in elections since 1972,Before then, polling data is too sparse to calculate this number with any precision.
“>2 national polling averages shifted by an average of 1.8 points and a median of just 1.4 points in the final 15 days of the race.
|Leader in FiveThirtyEight national polling average|
|Year||15 days before ELECTION||Election Day||Change|
Also, keep in mind that relatively few voters are undecided this year, and that many people have already voted, which could dampen the effect of any last-minute news developments.
7. Don’t read too much into the campaigns’ behavior.
These last two points are things I’ve learned by experience. There’s enough conflicting information in the final days of the campaign that it can help to triage, and one category of information I’d generally ignore are reports about how the Biden or Trump campaigns are feeling about the race. Even if reporters have good access into campaigns and are accurately reflecting their thinking, presidential campaigns often do not have a better read on the race than public polls. Campaigns are often just as surprised by unexpected results as anyone else; the Trump campaign’s models gave it a 30 percent chance of winning the Electoral College on the eve of the election in 2016, the same as FiveThirtyEight’s forecast did. And of course, less diligent reporters are subject to being spun by the campaigns, or to publishing information that is designed to deceive or bluff the campaign’s opponents.
8. Don’t get carried away with early voting data.
Democrats have a huge edge in early voting so far … but as I talked about on my weekly segment for ABC’s “This Week,” I’m not sure I’d read too much into it. The early-voting lead for Democrats is largely in line with what polls predicted, and Republicans are likely to draw the race closer with a huge Election Day turnout. Moreover, our experience in past elections is that people tend to read more into early voting data than is warranted and often cherry-pick data in ways that are favorable to their preferred party or candidate.
Also, the huge partisan split in early in-person voting and mail voting is new — historically, it was something that both parties took advantage of — and that makes it hard to put it into context. Maybe it really will turn out to be a bad sign for Republicans that Democrats are banking so many votes. Or maybe Democrats will underperform polls because mail votes have a higher rate of ballot spoilage. On balance I’d rather have a lot of votes locked in than not, but we’re flying pretty blind here. Besides, most polls try to account for early voting — for instance, by asking voters whether they’ve already voted — so to the extent that Democrats are benefiting from it, it should be reflected in the polls already.
It’s going to a long 15 days — and perhaps beyond, since we may not know the winner on Nov. 3. FiveThirtyEight will be providing all the content that you might want, from daily podcasts to near-constant updates of our forecast. But you’ll usually know all that you need to if you’re pacing yourself and only checking in with news coverage of the campaign once a day or a couple of times a week. Stay safe and stay sane, and we’ll enjoy watching the rest of the election with you.