2

I keep reading that almost every single poll known to man shows that Biden is leading Trump. The only exceptions are polls from a few Republican states (like Kansas) where Trump easily won in 2016 but now is only winning by a small margin.

My question: at this point in time in 2016 (October & November) what did these same polls look like between Clinton and Trump? Was Trump actually losing by so much in the polls in 2016? I know it was a surprise that he won in 2016, but I don't know how big of an underdog he was.

I ask because every single article I've read (mainly from NY Times, Washington Post, and CNN) mentions that Biden will be winning these elections by a landslide, yet I don't feel so certain. I know these news outlets (specially Washington Post) may be a bit biased against Trump, but I suppose they're the most "neutral" in terms of news. Yet I still haven't read an article saying that there's a poll where Trump is winning.

7
  • 1
    This is a very open ended question. Much of the news coverage you refer to goes specifically into 2020 vs 2016 analysis. Google "trump biden poll 2016" and you'll have much to look at. Maybe you could then ask something more specific? – Italian Philosophers 4 Monica Oct 22 '20 at 21:40
  • 2
    Confounding the question is that the polls themselves have changed significantly. A systemic failure to cover certain demographics (poor and not college educated, I think) was cited at least in part for the 2016 polling "failure", and pollsters have changed their methodologies to deal with this. It is expected that the polls are more accurate now, though anyone on the wrong side of 2016 is probably apprehensive at best about the polls right now. – zibadawa timmy Oct 23 '20 at 9:29
  • @ItalianPhilosophers4Monica: I didn't follow the 2016 elections, so I'm just trying to understand if Trump was such an underdog in 2016. – fdkgfosfskjdlsjdlkfsf Oct 23 '20 at 15:18
  • I didn't downvote, merely remarked that there is a lot of coverage of precisely that will give you professional comparisons, probably better than here. If you have a particular, specific, question, then yes, ask away. Joe C made a good comment that a rabbit dropped out of a hat for Trump at the last minute then and that could still happen now. – Italian Philosophers 4 Monica Oct 23 '20 at 15:34
  • "may be a bit biased towards Trump" Do you mean "biased against"? – Acccumulation Oct 23 '20 at 22:53
9

The main issue was that there was a disconnect between the media and what the polls said. Overall the polls predicted a close race within their margins of error. As it happened several key states flipped within that margin of error between poll numbers and actual results.

Usually that margin of error is stipulated at 3% by most (decent) polls.

The media essentially did away with that inherent uncertainty and claimed the polls were wrong.

However with accounting the uncertainty of 3% polls declare themselves in, only Wisconsin was actually wrong aka beyond a 3% jump, I believe.

So Trump was not winning in the polls, but the polls did measure a race within their margins of errors aka too close to call with any certainty and ultimately only missed their mark in one state.

Contributing factors in 2016 not present in 2020:

  • significant third party candidates sapping Democratic votes
  • a higher percentage of undecided voters going into the last two weeks
  • the Comey press conference reopening the entire E-Mail affair for the nth time in the news cycle which in itself could have swung voters on the basis of "if there is so much smoke so close to an election..."
  • a polling bias where most pollsters did not get enough data from certain demographics (mostly non college white men) that then swung the vote

But because of 2016 the pollsters stress far more at every opportunity how propabilities work and they obviously do not replace the actual election.

However in 2020 factors contributing to uncertainty:

  • Covid
  • Mail ballots and early voting on unprecedented levels
  • Election law uncertainty in various states if votes will be counted
  • Calls for voter suppression/intimidation (which is illegal so don't try it)
  • Calls for double voting (which is illegal so don't try it)
  • Preemptively declaring the election invalid on the basis of voter fraud that has not happened, yet, and there is no basis of it having happened in any election before...

Yeah, I am biased on the last few, sue me. The last four points can manifest and might mean the vote is irrelevant because it will be ultimately the courts that decide.

6
  • 2
    Good answer overall, but note that if several states flip in the same direction from the margin of error, it seems statistically likely that the flip resulted from a systemic flaw in the analysis. – Italian Philosophers 4 Monica Oct 23 '20 at 15:37
  • 6
    @ItalianPhilosophers4Monica A good pollster knows that their errors can include systematic components (and they adjust for these as best they can all the time), which is why FiveThirtyEight gave Trump about a 30% chance according to their modeling in 2016. That was based on the assumption that true results deviating from the predictions would be more likely to deviate systematically rather than randomly. A bunch of random deviations have little impact on the actual outcome, but systematic ones do. – Bryan Krause Oct 23 '20 at 16:17
  • 1
    @BryanKrause We're talking about 2 different things here: 30% chance for Trump win in '16 merely means that the 30% branch happened, no more, no less. Talking about uncertainty in polls and then stating that a number of them happening all in the same direction retroactively explains the outcome is another. Possible, certainly. Also possible that the polls at the time didn't fully take into account the dynamics of Trump's rather unprecedented candidacy. Essentially, Rumsfeld's "unknown unknowns" (much as I don't like Rumsfeld, I think that, oft-denigrated, statement is quite insightful). – Italian Philosophers 4 Monica Oct 23 '20 at 16:26
  • 3
    @ItalianPhilosophers4Monica Not sure I see the difference, except for a difference in interpretation or maybe you misunderstand me. People who understand polling know that polling errors are not likely to be independent when polls using the same methodology are applied to different populations. The contrast is between the 30% chance that aggregators like FiveThirtyEight gave, and the ~1% chance that other outlets looking at the same data gave. The difference was in the understanding that polling errors are unlikely to be independent. "30-70" is an underdog but not an extreme one. – Bryan Krause Oct 23 '20 at 16:30
  • @ItalianPhilosophers4Monica BTW, the term for "unknown unknowns" isKnightian Uncertainty – Acccumulation Oct 23 '20 at 22:57
6

At this stage in the 2016 cycle, the RealClearPolitics polling average gave Clinton a lead of 5.4 points over Trump nationally. This compares to a Biden lead of 7.9 today.

You would be right to say that it is not a certainty that Biden would win. Modelling done by FiveThirtyEight currently give Biden an 87% chance of winning the election, and Trump a 12% chance.

1
  • Comments are not for extended discussion; this conversation has been moved to chat. – JJJ Oct 24 '20 at 19:45
1

Similar to an above answer, there are some polls that didn't really show Trump as an underdog, but someone who was close in the 2016 election. On 270towin, the National Polls on average showed the chance of a Clinton victory to be 47.2%, Trump 43.6%, and 9% for another option (i.e. not reaching 270). The margin of error is about 3.1% to 3.6% for most of the polls, so while Donald Trump was shown as generally having a disadvantage, it wasn't enough to be considered a underdog battle. A early poll on 10/27/2016 from IBD/TIPP showed Hillary having a 44% chance, Trump a 42% chance, and 14% of other. Meanwhile, the polls for 2020 on 270towin currently show Biden winning by a 6.4% margin with a 51% chance for Biden and a 45% chance for Trump.

1

Trump was always an "underdog." The fact of the matter is that he lost the popular vote nationally, and the polls (properly) reflected that.

Trump won at least three states (Michigan, Pennsylvania, and Wisconsin) by razor-thin margins within the "margin of error." Because of this fact, polls are not nearly as good at picking up electoral college trends regarding individual states, even though they may do fine on a "national" basis. Therefore, it would be hard to tell that Trump was winning in the electoral college in 2016, at the same time he was losing the popular vote.

There may be other confounding factors like the following: Even in 2020, Trump may have won the election "on the ground" in the above three states. What hurt him was absentee votes. Let's say that a bunch of old "snowbirds" take off for Florida in October of each year. In 2016, they registered and voted in Florida. After observing the 2016 results, they decided that their votes in Florida (their second home) were "wasted," so they re-registered in their "home" (northern) states, and provided the margin of victory for Biden in the electoral college in 2020. In this example, you have the same people, different vote locales, different outcome.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .