Did the polls in 2016 GET IT WRONG?

I am interested in cases where almost everyone seems to believe something that is actually NOT true.

An interesting example from the 2016 election is that most people seem to believe that: THE POLLS WERE WRONG! I have heard Trump supporters say this and I have heard Trump opponents say this. It is accepted by so many people on both sides of the political divide that it might seem crazy to even challenge it now.

But maybe these are the beliefs - the ones most widely and unquestioningly accepted - that we most need to challenge.

So is this belief true? Were the polls really wrong in 2016?

Let's consider what the polls were saying just before the election. Here is an article that came out the day before the election that discusses the latest poll results. It is called “Presidential Election Polls for November 7, 2016” and appeared in Newsweek.

According to this article “Democratic presidential nominee Hillary Clinton leads her Republican rival Donald Trump by 2.5 points, according to the Real Clear Politics average of most state and national polls. Clinton has 46.8 percent of voter support compared to Trump's 44.3 percent.”

The article also says that “Forecasts still show Clinton winning the election. “FiveThirtyEight” shows Clinton with a 65.5 percent chance of winning the election, while Trump has a 34.5 percent chance of victory.”

***

[So why did Hillary lose in 2016? Read this new book to find out.]

[If you want to support "Anything Smart" just click on book links like the one below to buy your books. "Anything Smart" will receive a commission. Thanks!]

***

There is additional information in this article but I will focus on the two extracts above to evaluate the claim that “the polls were wrong.”

Before we look at the results of the election I would like to consider four points about the information above: what polls do, what “chance” means, the danger of lazy language, and what did these polls actually measure?

1. What do polls really do?

First, what do polls really do? Well, for one thing, polls do NOT predict the future. Polls tell you what answers people gave at a certain point in time. When you take polls at different points in time you get different results. If people who say “the polls were wrong in 2016” mean that the polls predicted Hillary would win but then Hillary lost, then that just means these people don't understand what polls do.

According to the article we quote from above Hillary had 46.8 percent of voter support at the time the polls reported on were taken. That does not PREDICT she will get 46.8 percent of the vote at some later date. If the election were still several weeks away many things might change, and many voters might change their minds, between the poll and the election. On the other hand, it is probably logical to assume that, if nothing significant has changed between the polls and the election, then Hillary will probably get about as much support as she had in the poll. But this is an ASSUMPTION based partly on the poll and partly on the belief that nothing has changed since the poll. The poll ITSELF does not predict anything about the future.

2. What does "chance" mean?

Second, what does “chance” mean? The article we are studying here says that Hillary has a 65.5 percent chance of winning the election. This does not mean that Hillary will win! It means exactly what it says, Hillary has a 65.5 percent chance of winning.

Let's illustrate this with an example. Suppose I give you two coins and a cup. I tell you to shake up the coins in the cup and then toss them out onto a table. I tell you “there is a 75 percent chance that there will be at least one head showing when you toss these coins onto the table.” You toss the coins and we see there is no head showing. If you then say, “Ha ha. Your prediction was wrong” you would be mistaken. I did not predict there would be a head showing, I just said there was a 75 percent chance that there would be a head showing and that statement is absolutely true even if, on any particular toss, no head is showing.

This is the same thing that happened to forecasters in the 2016 election. The polls themselves did not predict anything. The forecasts, based partly on polls and partly on other information, did not predict anything either. They did try to calculate the “chance” or “probability” that Hillary would win, and the fact that Hillary lost does not prove those calculations were wrong any more than getting no head when tossing two coins proves that there is NOT a 75 percent chance of seeing at least one head.

[Note: One big difference between tossing coins and having elections is that we can toss the coins many times to see if the calculated 75 percent chance of seeing at least one head really works out over many tries, while we cannot repeat an election many times to see if the calculated probability was correct. Still, the principle is the same. If the probability of something happening, like Hillary winning, is calculated as 65.5 percent, the mere fact that she did not win does not prove that the calculated probability was incorrect.]

3. The danger of lazy language

Third, the danger of lazy language. One of the statements from the article we quoted above is “Forecasts still show Clinton winning the election.” This certainly looks like a prediction that Hillary will win. Notice first, that this statement is not saying that polls show Clinton winning, but rather that forecasts show Clinton winning. But is even that LITERALLY true?

The statement above is immediately followed by another statement that explains what the author means. “FiveThirtyEight shows Clinton with a 65.5 percent chance of winning the election, while Trump has a 34.5 percent chance of victory.” In other words, saying that forecasts show Clinton winning just means that forecasters have calculated that Clinton has a higher probability of winning. As we showed above, even if Hillary loses, which she did, that does not prove that the calculation of her probability of winning was incorrect.

The problem here is just that people sometimes save time by using lazy language. Instead of saying the more accurate “Forecasts calculate that Hillary has a 65.5 percent chance of winning the election” sometimes people take a shortcut and say the less accurate “Forecasts show Hillary winning the election.” We have to be on the lookout for this kind of lazy language and it should usually be fairly obvious from the context of what we are reading.

4. What did the 2016 polls actually measure?

Fourth, what does the poll actually measure? The poll results quoted above, from just before the election, are talking about popular vote and not Electoral vote. It is natural to assume that whoever wins the popular vote will also win the Electoral vote because that is what usually happens. But it does not ALWAYS happen and 2016 was one of those unusual years when the winner of the popular vote did not also win the Electoral vote.

So here again, the fact that Hillary lost the Electoral vote on election day does not mean that a poll measuring popular vote a few days before the election, was wrong.

With all of these technical preliminaries out of the way we are finally ready to look at what actually happened in the election. According to the American Presidency Project Hillary ended up with 48.2 percent of the vote and Trump got 46.1 percent of the vote. What did the last polls say just before the election? According to the article we are discussing the average of poll results was 46.8 percent for Hillary and 44.3 percent for Trump.

This is pretty close agreement between the polls and the election, isn't it?

• The polls showed Hillary at 46.8 percent and she actually got 48.2 percent. A difference of just 3%.
• The polls showed Trump at 44.3 percent and he actually got 46.1%. A difference of just 4%.
• The polls showed Hillary ahead by 2.5 percentage points and at the time of the election she led by 2.1 percentage points.

Anyone who says the polls in 2016 “got it wrong” should take a close look at these numbers. The polls got it right! What caused the surprise was an incorrect assumption that whoever wins the popular vote will also win the Electoral vote.

***

[So why did Hillary lose in 2016? Read this new book to find out what her explanation is.]

[If you want to support "Anything Smart" just click on book links like the one below to buy your books. "Anything Smart" will receive a commission. Thanks!]

***

Copyright © 2018 by Joseph Wayne Gadway