With just a few days to go before Tuesday’s election, this is the time when both sides tend to start panicking about the outcome. While the Biden supporters may take comfort from polls showing that he has a comfortable lead, we all know that there are problems with polls. It is not that the polls were that wildly off in 2016. The national vote at the end was consistent with predictions of Hillary Clinton winning. But the state polls were close enough that relatively small deviations were enough to give Trump the electoral college win. We should always bear in mind that polls are always statistical and there is always a margin of error. Black swan events can occur and grey swan events have even greater chance of occurring. (I don’t know if there are grey swans but you get the idea behind the metaphor.)
So what about this time? While Biden seems to have a bigger lead than Clinton did, it may be that the polls may have even larger margins of uncertainty than before because of systematic factors. One reason is that it is getting harder and harder to get people to respond to polls. Two people who predicted that Trump would win in 2016 think that this will happen this year too because polls are undercounting the large number of what they call ‘shy’ Trump voters, people who are reluctant to tell pollsters that they are Trump supporters because of their feeling that the pollsters are likely Biden supporters. There is greater feeling of being judged negatively if you say that you are a Trump supporter than if you say you are a Biden supporter and so shy Trump voters will outnumber shy Biden voters.
To compensate for this, these two pollsters asked people not whom they were voting but for whom they thought their neighbors will vote for because this was more revealing of their true intent. They say that when asked this way, Biden’s lead shrinks considerably. They also use a different method of finding and asking people, not the telephone which they think is useless these days.
We have an Internet panel, but it’s a little different from most others. We recruit our respondents by sending them letters. We buy addresses from the post office—or from a vendor—draw randomly from addresses in the United States, invite people to participate in our studies and we pay them really well. We pay them join, and then $20 for a 30-minute interview. We have a relationship of trust with them.
I agree that telephone polling in the traditional way, as far as I can tell, is pretty close to death. You get extremely low response rates, and there is this issue: Who is still answering the phone?
…We live in a country where people will lie to their accountant, they’ll lie to their doctor, they’ll lie to their priest. And we’re supposed to believe they shed all of that when they get on the telephone with a stranger and become Honest Abe? I cannot accept that.
Meanwhile, Michael Moore who also predicted that Trump would win in 2016 fears that it is going to happen again this year because he sees huge enthusiasm among Trump supporters that is not matched by the Biden supporters. This is not surprising because, after all, the Trump movement has a certain cult-like quality. The question is whether that enthusiasm gap translates into greater rates of voting by trump supporters. Another problem are the efforts to disenfranchise Democratic constituencies in Republican controlled states.
We have also seen signs that the lead that the Democrats built in early voting has started to get reduced due to a late surge in Republican efforts and that Republican voter registration drives have picked up steam and that their door-to-door canvassing and get-out-the-vote efforts have greatly exceeded that of Democrats who did not do this until recently because of pandemic fears.
In addition to the public polls, each campaign also conducts internal polling that they do not reveal (unless it is very good news for them) and that information is used by them to decided where they will campaign and spend ad money in the waning days of the campaign. So where candidates spend time and money in the final days is a good indicator of where they think things are close. That is not a perfect indicator as we saw in 2016 when Clinton did not visit Michigan thinking it was safe and ended up losing there.
So what is the bottom line? It is true that Biden has a greater lead nationally than Clinton did at this time (7.9% versus 2.3% with five days to go) and is likely to win a majority of the national votes and also seems to be on track to win the electoral college, at least as far as the polls go. William Saletan looks at all these factors and explains why he thinks Biden is still in good shape to win.
But I am not so sanguine. Against Biden’s polling leads we need to set the intangibles such as the undercounting of shy Trump voters, the enthusiasm gap in favor of Trump, the Democratic deficit in efforts to register voters and canvass door-to-door, and the structural impediments created in those Republican controlled states to disenfranchise potential Democratic voters. To my mind, this makes the election too close to call and I do not feel comfortable making a prediction. This does not matter since my election predictions tend to be terrible anyway.
So in one sense we have returned to an earlier time before polling became a significant factor in predicting election outcomes, in having to wait until the votes are actually counted to know the result.
https://youtu.be/IdnHfYbr1cQ
I think 538 does a pretty good job at sorting through all these weeds, and they’ve had a few things to say about the recent polling/predictions.
First, they say that at this point polling doesn’t really matter. People have made their decisions and we’ll know what those decisions are when the votes are counted. Polling can be useful for tracking changing opinions about a candidate, but at this point in any race by the time polls are completed the voting will be over. So stop being concerned about polls.
Second, a lot of the predictions are made based on what is known about mail-in ballots. There are voter lists of people who are registered to which party (in most states), and most of what we know about the mail-in ballots are based on how many were sent and returned from people who are registered from each party. For example, out of 100 ballots sent and returned by mail, 40 of them are known to have been returned by registered democrats, 40 of them are known to have been returned by registered republicans, and 20 have been returned by people not registered in either party. (The numbers are really in the range of 50, 30,20, but that doesn’t really change the point I’m leading to.) What we don’t know is the contents of those ballots, and we won’t until the ballots are counted. Of the 40 returned by registered democrats, 5 of them may have voted for Trump. Of the 40 ballots returned by registered republicans, 10 of them may have voted for Biden. We just don’t know, so the predictions based on returned ballot counts by party registration is a huge unknown.
That being said, there is some data suggesting that Trump is getting about 90% of the republican vote and Biden is getting 98% of the democrat vote, so there is some math which can be done to predict how the mail-in vote will go. But we’re adding uncertainty onto uncertainty, so the margin of error in this case is probably greater than the possible polling errors.
What the polls do suggest is that the independent voters have broken strongly for Biden, and what the mail-in votes suggest is that people are motivated to vote. For all the enthusiasm Trump has generated at his rallies, that’s only a small fraction of the voters. So what will motivate a non-rally going republican to vote for Trump? Very little. Trump has shattered so many of the “norms” which republicans hold dear that we may be seeing a retention rate of less than 90% for republicans for Trump. The reason I say that is because the mail-in ballots compared against the voter registration lists are not accurate. Someone may have registered as a republican in 2012, but no longer considered themselves republican today, but the voter registration lists will still show them as republican.
Personally, I think the republican base has shrunk significantly, without anyone really being able to tell because the data they are using is outdated. One bit of evidence I see for this is the significant drop in small donations to the republican party. I’m not suggesting they have all become democrats, but they will probably vote against Trump (if they vote at all). This isn’t to suggest that they won’t return to the republican party if the party finds a way back to Reagan republicanism.
So if the republican base has shrunk, and even the self-identified republicans are only 90% going for Trump, I think the results will be a somewhat worse for the republicans than predicted.
Of course, I also said I thought the republicans would wait to confirm Barrett until after the election and in the lame-duck session, and I was wrong about that. I am not an oracle, even if I try to peep with security into futurity.
The title of 538’s “model talk” podcast yesterday:
There Just Isn’t Good Evidence That ‘Shy’ Trump Voters Exist (48:59, youtube). Nate Silver has pretty much lost whatever patience he had with those spurious arguments. They cover several other topics too.
My biggest concern isn’t with the polls but with votes not being counted for assorted reasons due to court interference.
My neighborhood and the nearby areas are full of Biden/Harris signs that have been getting more varied and more witty recently. There are only 2 yards in my walking area that have signs supporting right-wing candidates (for state offices) and zero Trump signs. I bet those 2 households have some ‘shy Trump supporters’.
From what I hear on the radio -- a young woman in West Virginia said she was supporting Trump for how he improved the local economy. While distributing free meals to children who were at home due to the pandemic.
I fear that at this point the fate of the election rests more on the willingness of the courts and the state governments to go along with a Trump ‘coup’ that the actual ballot count (not really a coup because it will be ‘legal’ as it goes through some not illegal process). The supreme court has already strongly indicated that it will go along with rejecting ballots on pretty weak premises and some state elected officials have been talking about rigging the electoral college voters.
anat, #5:
How much are you willing to bet?
“They don’t have a Trump yard sign, so they probably lie to pollsters when asked if they’ll vote for Trump.”
That’s what you’re betting on. Maybe two whole dollars? Two-fifty?
Also, how many people do you think there are, who don’t have Biden yard signs because they’re “shy” (i.e., lying to pollsters)? If you were in a very conservative state, you might not want to the extra attention from a bunch of gun-toting thugs — sounds reasonable enough to me, so sure, that might be a thing. Maybe not a big thing, but perhaps a thing. That still doesn’t mean it’s important to you to lie to some random pollster. Is that a thing? Why think that it is?
Speculate about what some anecdotes might mean all you like. Go for it. But we still have the fact that some pollsters were not weighting their polls for education level in 2016. Although that’s been recognized now and will be less of an issue this time, those 2016 polls were simply underrepresenting a fairly large class of Trump voters in their samples (because on this basis and others like it, they’re not deemed “likely” to vote).
Doesn’t that sound like a more important factor for explaining the polling discrepancies we saw, compared to what you think you can make out your two neighbors’ yard signs? Did you not know about it, or did you really think your story is giving a better explanation? Personally, I’ll take the real data over some sloppy reasoning and weak evidence, which only gets weaker if you don’t even live in a swing state.
My prediction: Trump loses, by a large margin. As for what happens next, you might expect that he’d surrender power, but that is something I really can’t predict.
Joe Biden knows in his heart that he is the only one who can truly defeat Ronald Reagan this November……Oh sorry, I meant Joe Biden knows in his heart that he is the only one who can truly defeat Joe Biden this November.
The first female US President…. Is just Biden her time.
Sima, wrong both times. Biden wasn’t running against himself.