California pollsters earn victory laps while national forecasters take lumps
Voters in key battleground states fooled pollsters again this presidential election, but in California they did just what the polls suggested they would.
California’s election forecasts were almost smack on the final numbers, a rare bright spot for the muchcriticized industry. It hasn’t always been that way in California, where Democrats have built a comfortable margin in the past three decades — making the state much easier to predict.
California pollsters say their detailed knowledge of the state helped. National pollsters often misstep as they forecast the outcome in states they don’t know well.
“We did pretty well,” said Mark DiCamillo, pollster for UC Berkeley’s Institute of Governmental Studies.
Nationally, America’s pollsters took an embarrassing beating for the second presidential election in a row. After confidently miscalling Hillary Clinton as the preelection winner in 2016, this time many of the same prognosticators predicted a Joe Biden landslide that never came.
Some of the state polls were even worse. In Ohio, for example, a survey taken two days before the election by Quinnipiac University of Connecticut found Biden leading President Trump by four percentage points. After all the votes were counted, Trump won Ohio, 53% 45%.
DiCamillo’s final presidential poll of likely voters, which was completed Oct. 21, found Biden leading 65% to 29%, with 6% undecided.
California’s unofficial final result? Biden won, 64% 34%.
Mark Baldassare, president and CEO of the nonpartisan Public Policy Institute of California, has been doing the organization’s polling since 1998. His last preelection poll of likely voters, finished Oct. 18, found Biden at 58% and Trump at 32%, with the rest either undecided or supporting minor party candidates.
State polls even caught the surprising rightward shift that already has cost California Democrats two congressional seats, with a pair of races still undecided.
In that same October poll, Baldassare found that while voters statewide would support a generic Democrat over an unnamed Republican by a 58% 37% margin, Republicans had a two percentage point edge in the state’s eight most competitive districts, as determined by the nonpartisan Cook Political Report.
In the two congressional districts that flipped, both in Orange County, the Republicans each won, 51% 49%. In the two contests that haven’t been called, Republicans hold narrow leads.
But the fallout from the failure of many of this year’s state and national polls is going to have an effect on the survey industry, even in states like California, which saw few problems.
“We were pretty accurate this cycle, but it has an impact on everyone in the industry if someone else is doing badly,” said Baldassare. “If it’s going badly, people don’t believe polls.”
This year, California had a number of polling advantages over many of the election’s battleground states, he added. The state’s long history of mail voting meant that there weren’t the type of partisan battles over early voting that roiled states like Pennsylvania, Nevada and Texas. Even without the threat of the coronavirus, Californians were used to casting ballots without going to a polling place.
“The way people went about voting in California was different,” Baldassare said. “No one waited until the last minute to go to the polls, so we didn’t have to do a lot of guessing about who was going to vote and who wouldn’t.”
The same voter enthusiasm — and strong partisan divide — that resulted in a record number of ballots being cast this fall also meant there were few undecided voters in the days leading up to the presidential election.
“So many people had already made up their minds” and weren’t going to be swayed by anything they heard at the last minute, which makes the pollster’s job easier, Baldassare said.
With mail ballots going out in the first days of October, many of those polled late in
California had already sent in their ballots, making them the likeliest of likely voters.
But the biggest advantage for California pollsters is that they know the state they’re working in and have plenty of experience with the politics and the people in every part of the sprawling state. DiCamillo, for example, started working for Mervin Field’s California Poll in 1978, before it was renamed the Field Poll. Baldassare did his first survey in 1982, when he was a professor at UC Irvine.
That’s not always the way it is in other states, especially those that become election battlefields. Pollsters from the national newspapers, television networks and universities flood those states, publishing surveys that don’t always reflect the idiosyncrasies that each of those places might have.
“The national organizations have the best pollsters in the country. I know those people and they’re really good,” DiCamillo said. “But they’re asked to go into battleground states and they don’t have experience in all of those states.”
Over the past year, the New York Times worked with Siena College to produce a series of state polls on the presidential race. On the Sunday before the election, the paper released its final polls from four battleground states: Arizona, Florida,
Pennsylvania and Wisconsin.
Biden “holds a clear advantage over President Trump across four of the most important presidential swing states,” the paper stated.
In Arizona, the Times found Biden with a sixpercentagepoint lead. The former vice president hung on to win by 0.3%. The misreads were similar in Pennsylvania and Wisconsin.
In Florida, it was Biden winning, 47% 44%. On election day, it was Trump, 51% 48%.
“You want to look for polls by people who have a lot of experience in those specific states,” Baldassare said.
Iowa showed the advantage of a state poll. The Iowa Poll, done by Ann Selzer for the Des Moines Register, found that Trump had surged into a sevenpercentagepoint lead the weekend before the election when national pollsters had Trump up by one or two points and, in the case of Public Policy Polling, a Democratic firm based in Raleigh, N. C., Biden leading by one percentage point.
Trump won the state, 53% to 45%, almost the exact margin seen in the Iowa Poll.
The other advantage California pollsters had in this year’s presidential race is that it was never close. Trump, who is massively unpopular in this heavily Democratic state, was far behind Biden from the start and never narrowed the gap.
It’s different when the race is close.
Even though pollsters are careful to warn that every survey is just a snapshot in time and that there’s a margin of error of at least a couple of percentage points in either direction, “if you’re on the wrong side, it’s a killer,” DiCamillo said. “If you’re on the right side, people will ignore it if you’re a couple of points off, but they remember if you got the winner wrong.”
And even when a race is too close to call, a public pollster has to call it.
In the 1986 Senate race, the final Field Poll had Democratic Sen. Alan Cranston leading Los Altos Rep. Ed Zschau by just one point, well within the margin of error.
“We said the race was a virtual tie,” DiCamillo said. “But we also said that if there had to be a favorite, it would be Cranston.”
Cranston won reelection by that single percentage point, 49% to 48%.
“How lucky can we be?” he said. “Sometimes you just have to cross your fingers when it’s that close.”