How The Pollsters Got It Wrong. Again.

By Brian Kirwin

Americans have lost confidence in pollsters.

People have more faith in the local weather forecast than in political polls, especially in polls done by and for the media.

Campaign experts know the difference between media polls and internal ones. When campaigns commission polls they insist pollsters structure the surveys properly and get accurate results so they can craft a game plan to win.  

Pollsters know how to do that. But the spin polls that show up in the media have a different purpose: they’re often designed to push a narrative. 

The pollsters know the results they want before they start.

Media polls are dangerous. They can depress the turnout of one side, squash fundraising for a “doomed” candidate and dampen the enthusiasm of volunteers.

Trouble is, spin polls look legitimate when published.

In the past, spin polls were created by slanting questions or stacking one party with an advantage.

They were easy to debunk.  

Today’s pollsters are more sophisticated. Their techniques are harder to detect, and the public never notices anything is amiss until Election Day when a poll is off by 15 points.

The 2020 election season was teeming with putrid polls.  In Maine, polls showed GOP Sen. Susan Collins losing by five points. She won by nine. For weeks, polls showed South Carolina’s Lindsey Graham tied with his opponent. He won by more than 10 points.

Six days before the election, a Washington Post/ABC poll showed Biden up by 17 points in Wisconsin. That race is so close it’s headed for a recount with Biden up by .7 percent.

These aren’t mistakes.  You don’t get so much so wrong so often without doing so on purpose.

How do they do it?

Selection of subjects: 

With the data that is available, it’s easy to find voters who actually vote.  If someone isn’t in the voter file, they don’t get polled.  Right?

Wrong. Pollsters know the propensity of non-voters to vote for certain candidates, so if they want to pad the numbers for one side, they will contact voters who registered, but who haven’t been near a voting booth in years. They ask how likely they are to vote in the upcoming election.

People always tell pollsters they are going to vote. 

Any poll that says they contacted “registered voters” or “adults” isn’t worth publishing or reading.

Exclusion of Subjects:

Polling likely voters is more complicated.

If pollsters want to exclude people who only vote in presidential years, they filter out respondents who voted in fewer than two of the last four elections.  

If they want to exclude people who only voted for Obama in 2008 and never showed up again until 2012 they use the “2 out of 4” rule.  

If they wanted to screen out Trump voters who only showed up for Trump, the “2 out of 4” rule will work again.

Pollsters know that this tactic eliminates many voters who will be voting in the presidential election. So why would they do that?  Because they don’t want them in the poll. They want a slanted sample.

Another exclusionary tactic is to use long surveys. Most people will not answer a 20-minute poll. They’ll hang up after 5 minutes or so.

Who talks to pollsters for 20 minutes? Those are almost always highly educated liberals.  

Want to debunk a poll?  Look at the percentage of college graduates in it. Roughly a third of American adults have a college degree according to most sources. I’ve seen many polls with 55, 65, even 70% of those surveyed with a college degree. It’s dreamland.

One poll I saw had Biden leading Trump 64-30 among college-educated respondents. Trump won non-college graduates 54-34. How did experts tilt a poll in favor of Biden? They concocted surveys so lengthy that highly educated voters spent half an hour talking politics, while working-class voters said, “I have to go.”

It’s bias without showing bias.

Finding the right party split

Pollsters have gotten too smart to be caught by the usual “Republican vs Democrat” split in a poll.  If they show +13 Democrats in a poll demographic, they know they’ll be skewered.  So they include the kind of Republicans they want: lots of highly educated Republicans who will check the box for party ID but who only support people like John Kasich or Joe Biden.

Never a Donald Trump.

Bingo! A defensible poll that still veers off into Liberalville.

Another way to weight a poll is with the income question. 

You’ll never know how rich your area is until you see the income percentages in a liberal poll.  This goes hand in hand with the college bias: When pollsters bulk up on college-educated in their surveys, they get a wealthy sample group.

Problem is, the people in line to vote rarely match the six figure incomes who made up the polling sample.

People Lie

Pollsters know people lie. Many people who vote for the same party year after year will tell a pollster they are “independent.” 

Some of the most liberal voters in the world will say they’re independent. People who vote Republican for decades will tell the stranger on the phone, I vote for the person, not the party.

Why does this matter?  Because in this age of data, pollsters can have the voting history of subjects in a snap.  If someone votes in Democratic primaries year after year, a pollster could know that without asking. Most don’t check.

Why?  Because then their samples would actually be accurate.

Pollsters want an outcome, and they can only get there if lifelong Democrats describe themselves as independents and if conservative Republicans claim to be moderates.

If all else fails, pollsters can weight the categories to get the results they want. These weighted models are never publicized and never scrutinized, but they are used to make sure their sample is “representative.”  If the model is wrong, the poll is wrong.

But they’ll never tell you. They’ll just say they need to do better next time.

Meanwhile, how many people saw a skewed poll this year and decided not to contribute to a candidate or even vote because they believed the race was over before Election Day?

It’s media-fueled voter suppression.

And it’s legal.

Brian Kirwin is a political consultant based in Virginia Beach, Virginia with more than 60 campaign victories.