Pollsters Have Adjusted Their Models Since 2016

It’s normal for a candidate to doubt the polls if the polls show them behind. The public can be drawn into that viewpoint if they are surprised by an outcome. Many point to the surprising win of Donald Trump in 2016 as proof that the polls were “wrong.” But, of course, the polls showed a win by Hillary Clinton, and she actually did win the popular vote. Moreover, the final numbers were not even half the margin of error, and we should remember that the polls are always a few days behind public sentiment. In addition, an unusual number of “undecided” voters went with Trump, and third party candidates lost voters on election day.

On the other hand, American elections are not decided by the popular vote. A handful, or even one state, can throw an election. It was the unexpected turn in just a few states that put Trump in the White House, so pollsters looked to see how to do a better job of monitoring swing states. Pollsters took a look at how they were working and have made adjustments.

EDUCATION

FiveThirtyEight asked pollsters what those changes were.

one thing came up again and again in our interviews: Pollsters told us they were now weighting their samples by education, because one key takeaway from 2016 was just how important someone’s level of educational attainment was in predicting their vote. . .

Some pollsters such as Ipsos and the Pew Research Center have taken weighting by education a step further by weighting for educational attainment within racial groups.

POPULATION DENSITY

Another variable is geography. We’ve known for many years that certain states, such as California and New York are sure to vote Democratic, while others, such as Alabama and Montana, will go Republican. Whole areas go that way, with the northeast and west coast going one way, while the midsection of the country goes the other. But now, pollsters are placing more attention on population density.

NBC News/Wall Street Journal polls are even weighted by the share of respondents from urban, suburban and rural areas. “This helps to make sure we are fully representing rural Americans,” said Horwitt, adding that it also “removes another factor which can contribute to poll-to-poll variation.”

REGISTERED VOTERS

In the days when most people answered their phones, a good survey could be achieved by simply dialing phone numbers randomly. The law of averages would give pollsters a fair cross-section of voters. Now, Pew, for example, contacts a prospect by mail first, to “get to know them” before asking their opinion. They choose whom to contact based on party affiliation as a starting point.

CELL PHONES, TEXTING, ONLINE

People have been moving from landlines to cell phones. As of 2019, a full 96% of us have cell phones. That has been a problem, since here are tighter rules for reaching strangers via cell phones. And since older people are more likely to hold onto their landlines, in addition to their cell phones, which phone is used changes the demographics of the calls. Some pollsters are even using texting as a tool to reach younger voters.

Generally, online polling has been suspect, because the contact is not as direct, but some pollsters rely heavily on online polling, and they often appear to be reliable, as well.

THE PANDEMIC, VOTING BY MAIL

This year, we have a unique problem. That is, even if the opinions of voters is perfect, it’s hard to determine who will actually take the chance of going to a crowded voting site to cast their ballots. On the other hand, the act of voting at home may change a person’s mindset. More importantly, as we’ve seen, it’s more likely that mail-in votes will be disqualified by some technicality. So if the pollster is correct, the end result may be off, anyway.

UNDECIDEDS

One of the things that threw off the 2016 results was a large number of undecided voters. It’s hard to figure out what “I don’t know” means. That is not as much the case this year, for better or worse. It comes down to so many people who either love or hate Trump. There’s little in-between, and that means this year will likely be much more predictable than the last election.

SHY TRUMP VOTER

In 2016, we had the phenomenon of the “shy Trump voter.” Many people were afraid to say they were for Trump, afraid of looking racist, or because they didn’t know why they were for Trump or against Hillary, so they didn’t identify as Trump fans, because they were afraid they’d be asked why. That’s just not the case this year. If someone is for Trump, they scream it.

2018

Despite all the difficulties of polling now, the results of the elections in 2018 were reliably forecast by the polls.

Circumstances have not changed much since then, so pollsters are reasonably confident that they have it right this year.

HOWEVER, TRUMP COULD STILL WIN

The polls all show Joe Biden ahead of Donald Trump, by as much as 17%. But there is an argument that they’re still missing something. Last time, the polling of most states was within the margin of error from the final results. But there were a few outliers. Ohio’s results were 8.1% more for Trump than expected. Iowa was 9.4% off.

Insider Monkey thinks Trump will pull another miracle this year. They are banking on upsets. For instance. . .

FLORIDA: 538 shows Biden ahead by 3.6%. Insider Monkey (IM) thinks this will narrow over the next three weeks, with Trump winning by 0.1%.

GEORGIA: 538 gives Biden a slim chance to win by 0.7%. IM thinks Trump will win by 1.1%.

IOWA: This is a bigger swing. 538 says Biden is barely ahead by 1 point; IM thinks Trump will win a by a resounding 7.1%.

MINNESOTA: 538 has Biden ahead by 8.2, but IM thinks it’s more like 1.5—which could easily flip in three weeks.

NORTH CAROLINA: 538 has Biden ahead by 1.9%, but IM says Trump is up 1.8%.

OHIO: 538 has Biden ahead by 0.5%, but IM says Trump is ahead by a full 5.7%.

PENNSYLVANIA: 538 has Biden ahead comfortably, by 6.3%. IM calls it a tossup.

WISCONSIN: 538 has Biden ahead by 6.4%. IM thinks Biden is only ahead by 0.6%–with three full weeks to go.

The bottom line is that IM projects Trump with 258 Electoral Votes, and Biden with 253 at this point. However, if Trump wins Florida and North Carolina, they think he has a 65% chance of winning the election.