Which Polls are right? Thoughts on the state of the race
Due to conflicts with work, I have avoided commenting on US politics for most of the last four years. I am not blind to the impact that has had on the profile of this website. During one of the most momentous periods in American politics, I wrote next to nothing about it. I am still, out of respect and loyalty, inclined to avoid judgements, but that does not extend to an effort to provide insight on where I think the election is today.
In doing so I am entering a crowded field. A number of forecasters have made careers out of trying to statistically model precisely that, finding ways of integrating polls, fundamentals, and historical data points to produce numbers. My former employer Nate Silver at 538 was an innovator in the field, and most recently has placed the odds on Joe Biden winning the election at over 80% for the first time. The 20% modeling in the uncertainty that the polls are wrong.
Nate Silver received a lot of criticism before the 2016 election for providing a similar "uncertainty buffer" which led him to predict that Donald Trump had a 23% chance of victory when other projections had Hillary Clinton at 99%. I think what happened is a vindication of the decision to include a margin of error, but it is nevertheless a crude instrument. The wider problem is with the usage of polling. Polling in the commercial field has many uses, among those tracking the responses of an audience to events and messages. It is very good at that. What it is less good at, however, is magically measuring the views of an undefined audience. Here is where political polling has always erred. The holy grail of a "representative sample" of the elector has eluded pollsters ever since Readers' Digest ran a poll in 1936 which showed Roosevelt losing in a landslide among subscribers. Since that time, numerous methods have been developed to try and remedy the defect and produce a representative sample. This included weighting for gender, party id, ideology, age, race and other demographics at one end, or blind calling on the other. The problem of course was what to weight for? Ultimately, because pollsters do not know what the electorate will look like until it turns out, all weighting is towards an "educated guess". This furthermore is complicated by the behavior of subsamples. Lower turnout among say Hispanic voters does not only reduce their proportion of the electorate. Republican voting Hispanics are far more likely to turn out, so lower turnout among Hispanic and Black voters also tends to raise the Republican voting percentage of both groups. By treating these variables as independent, pollsters tended to see wider misses in wave years.
There has long been a split as well between national and state polling. In 2012, national polls showed a close race or even a Romney lead. State polling showed a comfortable Obama victory. In the end state polls proved overwelmingly correct. The inverse was true in 2016, when state polls continued show Hillary Clinton comfortably ahead in Wisconsin, Michigan and Pennsylvania even as her national lead was underwhelming. Her team, and much of the media, chose to believe in a repeat of their 2012 experience and behaved as if they were up 8-9 points rather than 3-4. In the end, that helped cost Clinton the election.
In 2020 we again have a split between state polls. Most swing state polls show close races, often with a Biden lead, but nevertheless, with the President in striking distance in all the states he won in 2016. It is true some have shown larger margins in recent days, especially Sienna and Quinnipiac, but in general they seem to indicate a race which looks similar to 2012 when Obama won by 4. By contrast, national polls have shown Biden leading by between 10-16 points. Again, a minority show a closer 2-4 point race, but a distinct minority.
Which are correct? For Republicans, the vast margins for Biden in the national(and some state) polls are a deliberate effort to demoralize them and dampen turnout. For Democrats, there is a conviction Republicans have become "unskewers" but a fear they may be right. I think any reality needs to consider the fluidity of this race. For all the talk of polarization, it is unclear the polarization is absolute when it comes to a vote for Biden or Trump, as opposed to blocks which will or will not for one candidate or the other. As such, while I would posit there is little movement between the following blocks, when they move they tend to all swing together for the same reasons.
Voters who loathed Trump from day one and are relatively mainstream liberals
Voters who are diehard Democrats but are not invested in Trump hatred, often non-white voters(white supremacy really seems to be a talking point for whites)
Leftwingers who have no use for Biden but dislike Trump
Romney/Republican voters who voted Clinton in 2016
Romney/Republican voters who voted third party
Voters who enjoy the Trump Administration but have doubts about competence
Voters who generally support the Administration but have reservations about tone
And many more. My theory is that far from being untouchable, the difference between a blowout is the behavior of the voting block who generally like what the Trump Administration has done, do not like the Democratic party or what it stands for, reject virtually all of its ideological charges whether about racism, culture war issues, misuse of government resources, etc and see the Russia and other investigations as partisan bickering. And yet....this group has doubts about the competency of the Administration, and the state of mind of the President. For this group, Biden and the Democrats have offered next to nothing. The only thing they offered abstractly was Biden's nomination, but the convention and campaign was a long list of demands for actions this block is dubious about(lockdowns), issues they see as extremist(BLM symbolism) and things they don't see as a problem even if they have guilt about the process(ACB's Supreme Court nomination). Democrats have done nothing to win them. As they might have been expected to do nothing as trying to do so would have alienated their own supporters.
Nonetheless, these voters may defect anyway. Because while they might not object to the policy and overall record of the Trump Administration as a whole, the President's entourage and personal eccentricities have always been a source of conflicted feelings. The last two weeks have brought them to the fore. it was not only the debate, which I believe did far less damage than the media assumed. The President's illness also did little. It was rather the President's behavior after his illness. Not the recklessness per se, but the sheer randomness of it. It is speculated that the President's behavior is being influenced by the steroid treatment he is on. If so, what recent events have highlighted is a central truth of the Administration. For good or ill, the Trump Administration is a one man show. The President will do whatever he feels like, and no one, not his doctors, not the cabinet, not even his chief of staff or family have any influence. In effect, for these voters, they have just witnessed the President on mind altering substances, being allowed to behave in a manner self-destructive to himself and the country without any intervention.
Many of the voters in this group enjoyed the President's eccentricities and humiliation of figures and institutions they did not respect. But now the President is humiliating himself, and what is revealed is a yawing vacuum at the heart of the entire operation. Many in the left-leaning media have speculated that there is nothing which can shake the President's base. For a good number of voters that is true. But what is happening is almost certainly able to shake a group that nothing else has yet reached. You cannot have a drugged President. The failure to set up any sort of transitional authority or delegate it raises questions as to whether such delegation would even be possible. What if the President was delirious with fever? Is there any point while he breathes that any authority would ever be delegated?
If this seems a late moment for voters to make up their minds it is worth considering the history of losing incumbents. It is rare for incumbents to lose, and rarer for them not to lose in blowouts. Jimmy Carter managed to maintain the fiction of a close race with Ronald Reagan throughout most of 1980. Disregarding Democratic conspiracy theories about deals to keep hostages in Iran, what is generally believed to have cinched it was the final debate where Reagan asked voters "whether they were better off now than four years ago." This is usually seen as a general indictment of the Carter Administration's record or the condition of the country but it really was of Carter. It is worth noting that by 1980 Carter had so utterly abandoned his previous dovishness in foreign policy that he was boycotting the Olympics, aiding the Mujahedeen in Afghanistan and already blockading the Sandinista regime in Nicaragua. Nor was it domestic where Carter had veered between tax raised to tax cuts. Rather it was an indictment of Carter. Carter from the start had seemed to blame voters for the problems of the world. His "malaise" speech had blamed them for the energy crisis. He had seemed to blame them for educational shortfalls, and his switch to the metric system was perceived in a similar way. Carter did not listen. And what Reagan was pointing out in response to Carter ranting about his policies working with Iran and the Soviets was that Carter was never going to change. For Carter, things were always going to be someone else's fault. Even had the hostages been returned the day before the election, it would have likely backfired. Carter would have treated it as a vindication of everything he had done, and in consequence, proof that everyone who had ever disagreed, including the voters prepared to evict him from the White House, were wrong.
The result was for the race to be blown wide open. A 2% race the week before the election ended up in a 10% rout. The voters who did not trust Reagan, thought Carter was smart, wanted to believe he had learned his lessons and that the 2nd term would be different had maintained hope until the debate he might. When they abandoned that hope they abandoned him.
Similarly there is a constituency which would be fine with another four years of the 2017-2019 Trump Administration. They would vote for that over Biden. Attacks on Russian Collusion, Norms, "Children in cages", conservative judicial nominees, will not move them. What will is being persuaded that 2020 is the new normal. Trump 2017-2019 is likely only behind by 3-4 points and has a chance of sneaking in through the electoral college. The October 2020 version of Donald Trump is at serious risk of losing in the most lopsided election since 1984.
These voters have not decided what to do. And quite likely they are sending mixed messages. They "might" tell a pollster they are considering voting for Biden, but given their past support for Trump, their hostility to liberals and Democrats, and the fact that they really want to be able to vote for the President they are quite likely to come off as Trump supporters in "neighbor polls". Recent events haven't changed their desire to vote for Donald Trump. Just made them increasingly believe that it may not be something they will be able to do. Because the Donald Trump they liked and supported is no longer available as option. Rather, what is available is whatever was on display for the last week or so.