Nate Silver, pollsters flunk election forecasts in 2016 sequel

America’s premier pollsters flunked Tuesday’s presidential election just as they did in 2016 by habitually overestimating the Democratic vote in battleground states.

Pollsters also gave up on Republicans in key Senate races only to see them win.

A Washington Times examination did not find similar poll bias for Republicans. No surveys predicted a Republican victory.

Republican Sen. Susan M. Collins in Maine won reelection, but you would never know it by the polls. All of them — 14 — had her losing, by a lot.

Angry Republicans said news media and their research firms are misleading the American public into thinking President Trump and other Republicans can’t win.

“Look at the polling,” former Arkansas Gov. Mike Huckabee said on Fox News. “It’s obsolete. It’s an 8-track tape. And it’s as worthless as a milk bucket under a bull. Let’s never hear about polling anymore. It’s worthless. And all it did was cause some people to think, ’Oh well, there is no point in going to vote.’ I think we’ve got to get to a place where we don’t depend upon every unscientific methods of polling that have proven to be horribly wrong both in 2016 and 2020.”

Said a victorious Sen. Lindsey Graham, South Carolina Republican — who won reelection by 11 points in what was supposed to be a tight race — “To all the pollsters out there: You have no idea what you’re doing.”

Why?

Longtime pollster John Zogby told The Washington Times that polling firms are trapped in the past, relying too much on landlines instead of less-intrusive online mechanisms.

“I think that the rigid rules of yesteryear that require the landline telephones and the cellphones are such that Democrats continue to be more likely to answer the phone than Republicans do,” Mr. Zogby said. “I think for starters, Republicans may be intimidated by what are on occasion effete voices. I know The New York Times always used out-of-work actors and artists on the telephone.”

Then there is the sample of voters who are questioned. Some polls this year included self-identified Republicans at less than 30% while the turnout was 35%, just like in 2016.

“Ridiculous,” Mr. Zogby said. “It’s been like that. I can tell you in my modeling … I never get less than 34% or 35% Republicans. Even when I need to weight a little bit, I never get so much out of whack with an oversample of Democrats and an undersample of Republicans.”

Pollsters often defend their work by pointing to national polls that came close to predicting Hillary Clinton’s 2% edge in the popular vote.

But Americans elect their leaders by states, not by a national vote.

As in 2016, pollsters this time gave Mr. Trump virtually no chance in Wisconsin and Michigan and portrayed Pennsylvania as barely reachable.

On Wednesday, the balloting was so close that votes were still being counted in states that could propel either Joseph R. Biden or Mr. Trump to victory.

Voters who relied on polling guru Nate Silver and his political statistics website FiveThirtyEight would not have thought that the election would still be undecided Wednesday evening. He put Mr. Biden’s chances of winning at 90%. Mr. Trump, in other words, had no chance.

Four years ago, Stuart Rothenberg wrote in The Washington Post that Mr. Trump had absolutely no path to 270 electoral votes.

This year, he wrote in Roll Call on Oct. 19 that Republicans had given up on any chance that Mr. Trump would be reelected.

“Veteran Republicans have all but given up on Donald Trump’s reelection and are now focused on minimizing the damage down-ballot,” he wrote. “They criticize Trump’s poor messaging and salute the Democrats’ fundraising numbers, which reflect Democratic enthusiasm and foretell Republican problems down the stretch.

“While Republicans still have a chance to hang on to the Senate, Democratic gains are inevitable, and a net Democratic gain of at least three seats is likely.”

That loss in the Senate did not happen, according to initial returns.

Why the sureness of a Trump defeat? The answer is battleground polling.

President

⦁ Wisconsin: Mr. Trump eked out a win four years ago. He lost by less than 1% this year.

Voters would not have expected the tight margin by reading polls on Real Clear Politics. All had Mr. Trump losing and most by significant amounts. Topping the list of perhaps 2020’s most inaccurate poll is ABC News/Washington Post. It reported with a week to go that Mr. Trump trailed by a whopping 17 points.

⦁ Michigan: The second must-win state Mr. Trump narrowly won in 2016. Like last time, pollsters gave him no shot this year, either.

CNN said the president would lose by 12 percentage points, Reuters by 10, CNBC, Emerson College by 7. Only conservative-leaning polls gave him a chance.

On Wednesday, when media outlets called the state for Mr. Biden, Mr. Trump was behind by 1 point.

⦁ Pennsylvania: A 2016 win for Mr. Trump in another state where most surveys gave him little chance, although the “blue” numbers were single digits.

With 63% of precincts reporting Wednesday, Mr. Trump held a 550,000-vote lead, but state officials said many Democratic-leaning jurisdictions had uncounted votes.

⦁ Florida: Mr. Trump earned a healthy win this year by nearly 400,000 votes and a 3-point edge, yet the average of polls pointed to a Biden 1 percentage point win.

Quinnipiac University predicted a 5-point win, and CNBC saw a 3 percentage point Biden victory. Only conservative-leaning polls pegged the Trump victory.

⦁ Iowa: A tough-to-read state where pollsters were generally on the mark, but again undercounted the Republican turnout. Quinnipiac saw a 1 percentage point win for Mr. Trump, who routed Mr. Biden by an 8-point margin, 53% to 45%.

⦁ Ohio: Mr. Trump won a decisive victory, but Ohioans would not know it was coming by reading established pollsters. Quinnipiac conducted two late polls that showed Mr. Biden ahead by 4 to 5 points.

Again, the conservative polling firms, Trafalgar and Rasmussen this time, saw the victory of 53% to 45%.

The Times asked Quinnipiac poll director Doug Schwartz why so many polls overestimate the Democrat turnout.

“A full examination of what went wrong with polls this year is going to take a while,” he said. “At the moment, I still need to see the final election results and final exit poll results, and without those, I’m not able to make even preliminary hypotheses about what exactly the issues are.

“After the 2016 election, it took six months for the American Association of Public Opinion Research to release their findings of polling errors. I would expect a full evaluation of 2020 to take at least as long, though we might have some idea of the situation before then.”

Senate

Many predictions had Democrats winning the Senate, with at least six Republican incumbents seemingly vulnerable.

Larry J. Sabato and his Center for Politics at the University of Virginia predicted a 52-Democratic Senate advantage. In other words, a loss of five Republican seats. He also said Democrats would make a net gain of 10 House seats. Instead, Democrats lost seven seats, according to The Associated Press.

By Wednesday afternoon, four of the six Republican senators survived and Republicans picked up a seat as expected in Alabama, on a path to keep its 53-seat majority and blunt Mr. Biden’s promise to become the most progressive president since Franklin D. Roosevelt.

If the pollsters had been right, all six may have lost.

⦁ Michigan: Republican challenger John James, a former Army helicopter pilot, did not see one poll predict a win, or even a close race in his Senate run against. Sen. Gary Peters.

CNBC, Emerson College, Reuters and The New York Times all said he would lose by as much as 7 points. CNN said the gap was 12 points.

As of Wednesday, Mr. James trailed by 0.1 percentage point with 99% of votes counted.

⦁ North Carolina: Republican Thom Tillis holds a 49% to 47% lead, or about 100,000 votes, over Democrat Cal Cunningham.

But the last week of polls all showed Mr. Tillis losing by 1 to 4 points. NBC said he would lose by 10 points.

Trafalgar aced the number at 2% for Mr. Tillis.

⦁ Maine: Ms. Collins should not be returning to the Senate. Not one of 14 polls on Real Clear Politics had her tied or leading.

Quinnipiac had her down by 12 in September. Emerson College pegged her deficit at 6 points on Oct. 31.

On Wednesday, AP called the race for Ms. Collins with a big 51% to 42% lead over Democrat Sara Gideon, who conceded.

⦁ South Carolina: Mr. Graham was so irate at bogus polls that he attacked them in his victory speech. All three of Quinnipiac’s polls had Mr. Graham tied. He won by 55% to 44%.

He later tweeted at the media: “Biggest losers of 2020 election cycle. Mainstream media polling which was incompetent. Mainstream media coverage of the Trump presidency which was deplorable.”

⦁ Georgia: Sen. David Perdue trailed or was tied in four of the last five polls. With 94% of ballots counted, Mr. Perdue held a 167,000-vote lead, 51% to 47%.

Solid polls

Trafalgar Group in Atlanta burst onto the political scene in 2016. It was the only firm that predicted Mr. Trump would win Pennsylvania and Michigan, which he did to claim the presidency.

This election, chief strategist Robert Cahaly extended his range, polling in 10 contested states.

A review at this point shows he was correct in seven.

Based on incomplete returns, he correctly judged that Mr. Trump would win North Carolina, Florida, Georgia, Ohio and Pennsylvania and Mr. Biden would win Minnesota and Wisconsin. He spot-on put Mr. Biden’s Wisconsin margin at 1%.

Mr. Cahaly seems to have been wrong about Arizona and a Michigan victory for Mr. Trump. But Trafalgar surmised that, with a 2% lead in Michigan, Mr. Trump was far more competitive than establishment polls showed.

The news media called Arizona for Mr. Biden. Trafalgar has Mr. Trump winning Nevada, where he is slightly behind as the tabulating continues.

Mr. Trafalgar said he can identify Trump voters, who have disdain for liberal media, by conducting short surveys of a couple of minutes and relying on emails as well as phone calls.