The polls for Republicans in 2020 were awful. Awfully wrong.
So awful that the normally sedate Pew Research Center unleashed a post-election scolding to the polling triumvirate of media/colleges/consultants who consistently undercounted GOP chances.
“It’s clear that national and many state estimates were not just off, but off in the same direction: They favored the Democratic candidate,” Pew said in a post-Nov. 3 analysis.
The error gap is important, Republicans say because bloated Democratic poll numbers can tamp down fundraising and influence some voters to give up — before even voting. It is what Democrats would call “voter suppression.”
Seven weeks away from the Nov. 8 midterms, conservative backbiting has surfaced as they see the “generic congressional” poll in which voters are asked which party they will back tilt away from Republicans.
YouGov just posted numbers that show the Democrats are up 6 points. But Republicans would point out that in 2020, the same firm had the Democrats up a whopping 10 points near Election Day. Democrats actually lost 14 House seats and its post-election spread was three, not 10, or just about a tie, according to data posted by RealClearPolitics.com.
YouGov’s final 2020 presidential poll had Joseph R. Biden winning the popular vote by 10, a giant landslide. He beat then-President Donald Trump by 4.5%.
Afterward, noted GOP pollster Frank Luntz called on colleagues to quit, referring to mistakes in 2016 too.
“I think what is happening is accountability in action,” Mr. Luntz said on Fox News’s “Media Buzz.” “And if you got it wrong this time, you got it wrong twice in a row, you shouldn’t be working in the business. There are other things you can do. You can sell real estate. You can sell stocks.”
Pollsters gave Mr. Trump virtually no chance of gaining the magic 270 electoral votes in 2016. He garnered 306, winning in the “blue wall” states of Pennsylvania, Michigan and Wisconsin.
I’ve spoken to conservative-aligned polling companies with generally better records than the big guys. They say the major mistake is not to “weight” the curated sample of, say, 900 voters. Weighting means to increase or decrease the share of certain demographics, such as white or Black if the final sample is out of whack with recent election history.
Here is some history: In 2020, Quinnipiac University’s final polls had Mr. Trump losing Florida, 47-42. He won by 3.3%. Quinnipiac foresaw Mr. Trump beaten in Ohio 47-43. He won Ohio by 8.2 points. Nationally, it said Mr. Biden would win by 11 points.
CNBC polls had Mr. Biden winning Florida and North Carolina, where he lost, and saw an 8-point win for him in Wisconsin, where he squeaked a win by 0.7. One major poll said Mr. Biden would win Wisconsin by 17 points.
None of 14 polls found Maine Sen. Susan Collins in the lead. Some predicted a substantial defeat. She had a substantial win — by 9 points.
“After this election, I think the polling industry needs to take a hard look at what it does,” Ms. Collins told Fox News.
The GOP undercount goes back further than 2016.
In 2014, pollsters said Iowa Sen. Joni Ernst was tied with Bruce Braley. She won by 9 points.
Kansas Sen. Pat Roberts faced a deficit against Greg Roman. Mr. Roberts won by 9.
Two-term Maryland Gov. Larry Hogan was hopelessly behind, according to polls by the New York Times and Washington Post.
In 2018, Florida Gov. Ron DeSantis saw every mainstream media poll foretell his defeat — except Trafalgar. He beat Andrew Gillum in a half-point squeaker.
After Nov. 3, 2020 voting, Pew blamed “systematic” flaws.
“The fact that the polling errors were not random, and that they almost uniformly involved underestimates of Republican rather than Democratic performance, points to a systematic cause or set of causes,” Pew said.
Mr. Trump, of course, is not on the ballot this year in person, but his presence is everywhere. He is actively campaigning and endorsing candidates. His voters need to turn out if the GOP is going to win.
I asked Quinnipiac if it has changed any methodology since 2020.
Polling Director Doug Schwartz told me: “We found that our polls were generally accurate in measuring the Biden percentage of the vote, but we had a notably high incidence of respondents not providing a response to the election matchup question. In an effort to reduce the number of respondents not providing an answer to our election matchup question, we have modified the way we probe those respondents who do not initially provide an answer to this question.”
I asked YouGov America the same question and spokesman Allen Houston. He said: “What is different for 2022 vs. 2020? We have a new frame (used for selection and weighting of samples) based upon imputing the 2020 vote to the voter file and aggregating it to state-level sampling units. This attempts to correct for the 2020 miss, so that the sample is composed at least of the right fraction of 2020 Biden and Trump voters. We are weighting the sample to 2022 primary participation and vote choice, in addition to demographics and 2020 vote. We are using a new turnout model, combining past turnout from the voter file and self-reported likelihood of voting from the survey.”
There it is. Technical, yes. But it does show that pollsters listened to the Pew Center’s criticism and have a commitment to correcting bad polls.
Tom Bevan, co-founder and president of RealClearPolitics.com, wrote in early September about generic ballot history. An important hint at the GOP congressional landslide in 2010, he said, was the Virginia governor’s race the previous year. Republican Bob McDonnell won. Last year, Republican Glenn Youngkin defied Virginia’s blue-ness and captured the governorship.
Mr. Bevan also examined other general ballot years when congressional Republicans did well, such as 2014, and when Democrats prevailed, as in 2018.
His conclusion: “This year appears to be following the pattern from previous good Republican years — especially 2014. Perhaps Democrats will be able to defy recent trends, but if the pattern holds, expect Republicans to gain in the Generic Congressional Ballot post-Labor Day.”
On Sunday, new generic polls seemed to back Mr. Bevan’s prediction.
An ABC News/Washington Post poll had Republicans up 5 points. Rasmussen has them up 2. And CBS News puts them 1 ahead.
These are much better GOP numbers compared with August’s, when, for example, The New York Times wrote, “Growing Evidence Against a Republican Wave.”
• Rowan Scarborough is a columnist with The Washington Times.