Democrat-leaning polls have skewed results since 2014

The news media’s Democrat-favoring election polls did not begin with the Donald Trump era, an examination by The Washington Times shows.

The 2014 and 2018 off-year elections were also mistake-filled, as surveyors undercounted the Republican vote in key contested races.

Republicans say it matters because anti-GOP polls can suppress votes, hurt fundraising and lead to news media reporting that the Republican candidate has little chance.

Mr. Trump won the presidency in 2016 by carrying three “blue wall” states — Pennsylvania, Michigan and Wisconsin — in the face of scores of polls that said he had little or no chance. It was an awakening for conservatives’ long suspicions of liberal news outlets and whether they try to suppress the Republican vote.

The trend appears to have started in 2014, when various firms predicted a number of Democrats would win who did not.

Fox News missed several key races that year. Conservatives currently are irate over Fox News’ decision desk for calling Arizona for Joseph R. Biden with more than 600,000 votes still to be counted.

All news media eventually put the state in Mr. Biden’s column by less than 11,000 votes. On Election Night, Fox News’ decision desk also said Democrats would pick up at least five House seats — a call wildly off the mark. Republicans are on their way to gaining 10 or more.

Meanwhile, liberal media polls were so one-sided that pundits such as FiveThirtyEight gave Mr. Trump only a 10% chance of winning. Races in battleground states turned out to be much closer, as they were in 2016.

Some pollsters blamed crummy polling on a failure to locate Trump voters.

But a look at 2014, and another off-year, 2018, show it cannot all be attributed to “Trump-shy” supporters.

Longtime pollster John Zogby told The Washington Times that established firms systemically sample too many Democrats.

“Oversampling Democrats has been going on for years,” he said. “I started to really take notice in 1994.”

Mr. Zogby’s final 2020 poll had Mr. Biden up by 6%. He currently leads the national count by 3.6%.

Mr. Zogby nailed the generic congressional ballot, predicting just a 2% Democratic edge. The current tally is 1.9%. Some polls had that gap at 10-12 percentage points.

The Times found more than a dozen races in 2014 and 2018 that pollsters erred on the side of Democrats. There were no such overshoots for Republicans.

A look at 2014

Kansas. Virtually every big surveyor said the race between Greg Orman and Republican Sen. Pat Roberts was a dead heat. Most gave Mr. Orman a slight lead, as did Fox News, Survey USA and NBC News, according to data posted on RealClearPolitics.com. Mr. Roberts won by 11 points.

Iowa. Polls said Sen. Joni Ernst was engaged in a tight race with Bruce Braley. Quinnipiac University had the Senate race tied, as did Reuters. Fox News said she was up 1 point. She won by 9 points.

North Carolina. Pollsters said now-Sen. Thom Tillis would lose to Democrat Kay Hagan. Seven of the last eight polls had her ahead by 1 to 4 points. Fox News, CNN, Rasmussen and YouGov all foresaw a Hagan victory. Mr. Tillis won by 2 points. He won reelection in 2020 with final polls showing him losing.

Virginia. Pollsters unanimously said Democrat Sen. Mark Warner would easily win reelection in a double-digit landslide. CBS News/New York Times, for example, projected a 10% victory margin. Mr. Warner edged past Republican Ed Gillespie by a mere 0.8 points.

Arkansas. The Real Clear Politics polling average for Sen. Tom Cotton was a 7-point lead. He won by 17.

Maryland. In the 2014 Maryland governor’s race, the news media polling complex gave Republican Larry Hogan no chance of defeating now-Rep. Anthony Brown. CBS/New York Times said Mr. Brown would win by 13 points. The Washington Post said 9.

A look at 2018

The Democrat-skewed polls carried over to the 2018 election, where the name “Trump” was absent from state ballots.

Florida. Ron DeSantis won the governorship despite every mainstream poll saying he would lose to Democrat Andrew Gillum. Quinnipiac placed him 7 points down and CNN at minus 12. Of the last 30 polls, only a TV station and the conservative-leaning Trafalgar Group forecast Mr. DeSantis winning.

In the Senate race, Republican Rick Scott saw 14 of 20 polls predicting a loss to incumbent Bill Nelson. Quinnipiac said he was 7 points down. Mr. Scott won by a scant 0.2 points. Trafalgar put him up by 2.

Indiana. Republican Mike Braun held a long-shot chance to defeat incumbent Joe Donnelly. Fox News placed him 7 points back. NBC/Marist said 2. Harris had Mr. Braun up by 1 point. He won by 6.

Missouri. Polls failed to detect the 6-point win for Republican challenger Josh Hawley against Sen. Claire McCaskill. Fox News had race tied. NBC/Marist put her up by 3 points. Trafalgar, once again the accurate outlier, predicted a Hawley win by 4 points. He won by 6 against a Real Clear Politics average of 0.6 percentage points.

Wisconsin. Gov. Tony Evers beat Republican Scott Walker, but the race at 1.2-point margin of victory was far closer than establishment pollsters prognosticated. NBC/Marist predicted a 10-point win. Emerson College said 5. Hometown Marquette University nailed it as a tie.

Ohio. Five of seven polls said Mike DeWine would fall in the governor’s race to Democrat Richard Corday. The last three polls in October placed him down 3 to 6 points. He won by 4.3.

Georgia. One of few states badly missed by Atlanta-based Trafalgar. It saw Republican Brian Kemp losing the governor’s race by 12 points to Stacy Abrams, who eventually lost by 1.4 points. Other polls placed the contest at 1 or 2 points either way.

The Nov. 3 election was another black eye. Analysts gave Mr. Trump only a 10% chance of beating Mr. Biden, based on bad polls. The president had no shot in Michigan and Wisconsin, the polls said, and a faint opportunity in Pennsylvania.

Republicans contend that constantly telling Trump voters that their candidates have little chance has the effect of discouraging them from going to the polls.

Some polls pegged the GOP participation at 28%. But the actual turnout was 35%, as it was in 2016.

In May 2017, the American Association of Public Opinion Research issued a paper on what went wrong in 2016. The report, done by pollsters, said surveys were generally accurate.

But the paper admitted that firms “under estimated support for Trump.”

“Many polls — especially at the state level — did not adjust their weights to correct for the over-representation of college graduates in their surveys, and the result was over-estimation of support for Clinton,” the report said.

Also, the report said, “some Trump voters who participated in pre-election polls did not reveal themselves as Trump voters until after the election.” The report rejected blaming political bias, saying 2012 polls tended to undercount Democrats.

The findings were produced by 13 pollsters, among them The Washington Post, Gallup, NBC pollster Marist College and the Pew Research Center. The association now has to examine the 2020 election, for which Republicans gave pollsters a failing grade.

Mr. Zogby said he generally “weights” Republican turnout in the mid-30s because that, in reality, is what happens.

He wrote this for Sunday’s Miami Herald: “Many independent pollsters just don’t take voters’ party identification as seriously as they should. In 2020, when 95% of Republicans supported Donald Trump and 94% of Democrats supported Joe Biden, it matters greatly if one party is over-represented in your sample and the other group is not represented enough. Too often pollsters would report results with as many as 44% Democrats and as few as 28% Republicans in their samples. Small wonder their results would reveal a double-digit lead for Biden.”