Yesterday, I determined that the safest bet of this election was that the House of Representatives stays in Democratic hands. In contrast, I wouldn’t feel confident wagering on the the presidential and Senate results. Of the two, I feel more confident — or, that is, less unconfident — in the presidential race, so today it is on that race I’ll focus. Later this week, I’ll look at the scramble for the Senate.
I should start off by saying that if you read my One Month Out piece — “Where Have You Gone Ken Bone? (Our Nation Turns Its Lonely Eyes to You)” — then you, unlike Ken Bones’s pants, are in great shape heading into today’s analysis. Not much has changed. In summary, I tried to show that predicting this election should come down to a surprisingly simple conclusion based on three premises:
- Data shows that nearly every decided voter is committed to their candidate.
- Data shows that unlike 2016, there are very few undecided voters.
- Data shows that Joe Biden’s lead is generally big enough to withstand President Trump’s current support plus winning undecided voters at the surprising margins he won them in 2016, even though as the incumbent he’s less likely to do so.
- 1-3 should mean a Biden victory.
With Biden above 50% in both averages, we know that Trump + everyone else polled cannot catch him, and it’s exceedingly unrealistic Trump adds everyone else. Biden’s probably looking at 52% of the vote, maybe 53. No one who has hit 51% or higher has ever lost.
The President’s numbers in the above average closely mirror his approval rating, which has lived within a stone’s throw of 43% for nearly his entire presidency. Whenever he drifts too far from 43, he always returns to it. To win, his numbers would need to do something they’ve almost never done: consistently climb, and fast.
Of course, with all this reliance on data, on all of your minds — either as a result of right-leaning confidence or left-leaning anxiety — is a single question:
Can’t the polls be wrong again?
In 2016, the polls also suggested that Hillary Clinton would win, and prognosticators like me agreed with those polls as we careened toward our collective embarrassment. If that is your prevailing feeling in this moment, I encourage you to read the Ken Bone piece to best understand why 2020 is not 2016.
You should also consider that an average of 2016 national polls were only off only by a point. They projected Clinton to win the national popular vote by 3 and she won by 2. If today’s national polls are just as wrong, that means Biden wins by about 6 points. The chances that the President’s support is so well-placed in the battlegrounds that he loses the national popular vote by six but nevertheless wins the Electoral College are incredibly remote, as determined by the statisticians at FiveThirtyEight:
It’d be by some measure the most wildly deviated difference we’ve seen from the handful of national popular vote losers who won the Electoral College in a two-way race.
That said, it’s worth saying there are no guarantees in politics. When people ask me who’s going to win, I say Biden. Then they ask if I’m sure. I say no. That’s not how this works. No one knows who’s going to win. If you’re looking for certainty, go crack open a holy book. Politics can’t help you. All we can do is make our educated guess.
Keep in mind, however, that just because a statistical surprise happens once doesn’t make likely to repeat. It’s reasonable to give Trump about a 1-in-6 chance of winning, or the probability of a die turning up on the number 6. That happens often, but just because it happens once doesn’t mean it’s more likely to happen again, and just because someone correctly predicted they would roll a 6 doesn’t make them particularly insightful on the art of rolling dice.
Nevertheless, Democrats’ bedwetting is understandable, since, to them, a Trump re-election borders on apocalyptic. If rolling a six meant your house burned down, you’d be pretty terrified about that dice roll even if there were a five-in-six chance your house would emerge unscathed.
Of course, the case against such a passionless mathematical analysis is that politics is not a true roll of the dice. Indeed, political science is no science at all. Democrats fear and Republicans hope that there’s something systemically wrong about polling, and that this error revealed itself in 2016 and has not been corrected since. Leading theories of the systemic problem include the Shy Trump Voter, who is too embarrassed to out themselves to a pollster; the Shenanigans Theory (my terminology), which proposes Trump supporters are purposefully misleading pollsters because it would be so sweet to prove everyone wrong again; and the Bad Samples Theory (my terminology), which posits that most pollsters aren’t correctly sampling Republican and/or conservative and/or non-college educated white voters.
I’ll leave you to your own internet research, which never ends badly for anyone, but for what it’s worth, I don’t think they hold up:
- Shy Trump Voter: Anecdotally, Trump supporters are pretty loud and proud. Data, meanwhile, should suggest that if the Shy Trump Voter phenomenon were real and widespread, other Republicans should perform much better than Trump in the polls. Consider that the generic House ballot in 2016 had the country on a razor’s edge (Democrats +0.6), with Trump just a couple points behind that. In 2018 and now in 2020, the Democratic generic ballot lead is about 7 points, again only barely better than Trump’s deficit. Meanwhile, we have surprisingly close Senate races in Republican-controlled Iowa, North Carolina, Georgia (X2), Arizona, South Carolina, and Montana. If it were truly just a “Shy Trump Voter” phenomenon, we wouldn’t see Republicans struggle in the House nationally and in red states their Senators had previously won.
- Shenanigans Theory: I don’t buy this one either. Recall earlier how I pointed out Trump’s polling numbers aligned with his approval rating during most of his presidency, so it’s not as if they’re just now pretending they don’t support him. Would a Shenanigans Trump voter have told voters they disapprove of his presidency this whole time? And if they did, do they realize that low popularity undermines what a president tries to accomplish? Do they really want to Own The Libs so much that hampering Trump’s presidency is worth it? Are they that stupid? I say no.
- Bad Samples Theory: I mean, maybe? Those that endorse this theory are mostly speculating. Pollsters claim to have improved their sampling approach, but they might again underweight a certain demographic like they did non-college educated white voters last time. Still, it’s possible pollsters have over-corrected as they “fight the last war” instead of focusing on reading the 2020 electorate. In sum, it feels like a flip of the coin to me whether Biden or Trump come out better than their polling.
Speaking of polls, a couple of pollsters — Rasmussen and Trafalgar — are doing their part to keep Trump within reach or even portray him as the favorite. Rasmussen is the only pollster to have Trump within 5 (they have it at 2). Trafalgar’s chief pollster, in addition to having Trump tied or the leader in nearly every swing state when nearly all other non-Rasmussen pollsters have Biden ahead, is officially predicting the President’s re-election.
Liberals are quick to dismiss these ostensibly conservative pollsters because of their tendency to favor Republicans while the main pack of pollsters does not. However, you might recall Rasmussen as tremendously accurate in 2016 (which, as I frequently see on my feed, their Twitter handle loves to remind us).
They again have Trump down just 2 points, and if Trump loses by just 2 then he probably wins the Electoral College. Trafalgar also correctly predicted Trump’s victory.
Conservatives like to point to these two pollsters as evidence that the liberal media pack is either bad at their job or purposefully goosing the numbers. Since Rasmussen and Trafalgar were right in 2016, shouldn’t we trust them again now?
I’m not so sure. Though Rasmussen loves to show that Real Clear Politics graphic from 2016, you won’t find them relaying information from a more recent election — that of 2018. No highlighting is necessary here, as Rasmussen finds its own way to stick out:
Just two years ago, when considering the makeup of the national electorate, they were nearly ten points off the mark. In fact, it was their favoring of Republicans that disrupted what would have been an otherwise incredibly accurate RCP generic ballot average.
As for Trafalgar, their surveys seem to be bending over backwards to make Trump look more competitive, somehow producing samples that suggest Trump will win 29% of the Democratic vote in Michigan and Trump will win independents by a 2:1 ratio. They have Trump winning 18-24 year-olds in Florida by 18 points. Suffice it to say, nothing about our political landscape suggests those things will happen in this election. Oh, and once this surprise sampling got attention, they took down the links to the sampling. They don’t appear to be on the up-and-up.
In every presidential election, just like in every supposedly one-sided sporting event, there are a few prognosticators who go against the grain, and once in a while they’re right and look like geniuses. Unlike sporting events, however, in presidential elections they have four years to bolster their reputation and brag. We’ll see if it was luck or skill in just six days.
How Trump Wins
So I really don’t see the bulk of polls being so wrong that Trump would win if the election were held today. Still, how might the President pull off rolling that 6 by Tuesday? It likely requires three things to be true over the next six days:
- He makes up a point or two in the polls. That’s totally doable. However, again, there are far fewer undecided voters this time around, so we shouldn’t expect that this alone is enough to make up major polling deficits. Just a point or two.
- The polls are off by a couple points — and in his favor. Polls have margins of error for a reason: they are imperfect. Though national polls were, on average, only off by one point in Hillary Clinton’s popular vote win, they could reasonably be off by 3 points or so. Now, it’s worth noting that this polling error might skew toward Biden, giving him an even larger victory, but they might skew toward Trump, too, particularly if there really is a systemic problem with sampling. However, polling was generally good in 2018, so such an error really would have to be Trump-specific.
- 1 and 2 are concentrated in the right states to help give him another win in the Electoral College. If Trump gains a point or two in the next six days and polls are off by most of their margin of error and this error exclusively underestimates Trump’s chances, that gets him to about 2 or 3 points back. Trump then has to hope the above gains and errors are more pronounced in battleground states. In other words, perhaps Biden wins California and New York by a few more points than expected but loses Pennsylvania and Wisconsin in the process, the amalgam of which wouldn’t be reflected as a national polling error.
It’s certainly possible.
Examining the Electoral College
I’ve waited too long to get into how one of these candidates will actually win — by getting to 270 electoral votes. Though Biden leads nearly every relevant swing state, these states aren’t isolated from each other in any way but geography, and sometimes not even by that much geography. In other words, a rising Trump tide could life all boats.
On Monday, NPR’s Domenico Montanaro tweeted a concise description showing Trump’s difficult but not impossible situation:
In other words, if Trump runs the table on ALL swing states Biden leads by less than he leads Pennsylvania, then he would just need to win Pennsylvania to win the election. So if Trump receives sufficiently broad momentum, we could be in for an Election Day surprise. Again, it’s possible.
But it’s unlikely. I choose to trust polling data that suggest that Biden is indeed running ahead of Hillary Clinton’s numbers from four years ago. With six days to go before the 2016 election, Clinton’s RCP average lead was just 1.9 points. Biden’s is 7.2. Trump won Pennsylvania, Michigan, and Wisconsin by less than a point each to win the election.
You do the math.
I’ll leave you with the quickest of Electoral Math breakdowns and what I’m thinking for an electoral map six days before election night.
I love what 270toWin.com allows you to do with its interactive electoral map, so I encourage you to toy with it. The map currently does most of the work for us. It combines forecasts from nine leading organizations to give us a color-coded map of each party’s “Safe,” “Likely,” and “Lean” states, in addition to total “Toss-up” states. You as the user can then cycle through the colors for any given state, likely modifying the Lean and Toss-up states to come to a prediction.
Because 2016 taught us that “Lean” states can’t quite be trusted, I’m going to start with a map that converts all of them to “Toss-ups.” Such a map looks like this:
As we can see, Biden has a clear lead in Safe + Likely at 212 to 125, but, with many electoral votes on the table, he’s far from a sure thing.
I think we should pencil in Democrats’ perpetual pipe dream, the Great State of Texas, for President Trump. No Democrat has won a statewide race there since 1994. It has long been Democrats’ state of the future, and it will remain so at least another election cycle. Florida, too, should end up in Trump’s column with some active momentum there. Those two massive states combine for 67 electoral votes, helping Trump leap to 192.
I also think the states of Ohio, Georgia, North Carolina, and Iowa, all of which are incredibly tight in polling, will tilt toward Trump. Republicans have closed well in tight statewide races, not just in the 2016 presidential but also in the tight 2018 Senate races, where they won seats in Texas, Florida, Indiana, Missouri, and Tennessee despite close polling. Giving Trump these states adds up to more 55 electoral votes, bringing him to 247, or 23 short of victory.
We can go through a similar process with Biden. His polling looks almost insurmountably strong in Michigan and New Hampshire (and I again must emphasize the lack of undecided voters compared to 2016, when Trump beat the polls in Michigan by winning late-deciding voters), while both 2016 and 2018 revealed Minnesota and Nevada as blue states in Trump’s America. Those four states give Biden another 36 electoral votes, bringing him to 248, 22 short of a win. Our new map:
That’s right: 248 to 247, with just Pennsylvania (20 electoral votes), Arizona (11), Wisconsin (10), and Nebraska and Maine’s second Congressional districts (who each get an electoral vote) left in play. That’s 43 electoral votes. Biden currently looks like the favorite in all but Maine’s CD2, so he should still be considered the strong favorite to win the election. The fact that red states like Iowa and North Carolina are even close suggests a national political climate where purple areas should tilt blue.
But just suppose Biden wins Pennsylvania and Nebraska’s CD2 for 21 electoral votes, and Trump wins Arizona, Wisconsin, and Maine CD2 for 22. The result would be…
THEN what happens? I’ll tell you tomorrow.