Anthony Salvanto is using a tracker poll to measure voting patterns, which shows people’s views over time.
Anthony Salvanto was imprisoned in a windowless office on Nov. 8, 2016. His phone was confiscated; guards escorted him to the bathroom. And all he had to look at were numbers, numbers and more numbers.
It wasn’t some sadistic punishment. It was Election Day, and Salvanto, head pollster for CBS News, was one of the few experts allowed see the exit-poll data as it rolled in.
Hillary Clinton held a lead all day, as most polls had predicted. But Salvanto saw a radical shift brewing. Clinton’s advantage was narrow, and many of the voters who hadn’t yet cast ballots were not on her side.
“This is a contested race,” he announced to the network’s anchors and producers once he returned to the newsroom. “I told them to get set for a very late night.”
Donald Trump’s 2016 presidential victory was one of the most stunning upsets in American electoral history. In the campaign’s final days, polls had given Clinton up to a 7-point edge.
In “Where Did You Get That Number?” (Simon & Schuster), out Tuesday, Salvanto explains the lessons pollsters learned that year and are putting into practice now, ahead of this November’s crucial midterm elections.
“In 2016, a lot of us assumed we knew what would happen in Michigan and Wisconsin,” Salvanto told The Post. “It was a great lesson for us pollsters: Even if you think you know what will happen, poll it if you can.”
Blame it on the Blue Wall. Since 1992, in six consecutive presidential elections, a solid block of 18 states had voted for the Democrat every single time. They would have given Clinton 239 electoral votes — 89 percent of the way to victory — if she had held on to them all.
Few bothered to ask those states’ voters what they thought, though. “Pre-election polling in the Midwest . . . just wasn’t there,” Salvanto said. When Trump knocked out Michigan and Wisconsin, two of Clinton’s critical Rust Belt supports, her Blue Wall crumbled.
Late deciders made the difference. CBS’s final 2016 poll gave Trump a substantial lead among voters who made up their minds at the last minute. In another Blue Wall state that went red in 2016, 10 percent of Pennsylvanians made their choice in the campaign’s closing week, netting Trump 97,000 votes — more than twice his margin of victory there.
But that vital indicator was buried under the headline news from the same CBS poll that Clinton held a 4-point lead over Trump nationally.
“The horse-race analogy is compelling but it’s wrong,” Salvanto said. In racing, he pointed out, the distance already run gets your horse closer to the finish line, but “in a campaign, everything can change tomorrow.”
If voters see the election as a dynamic instead of a race, we can use polls the way campaign managers do: not as a crystal ball, but as a tool. “That means asking where a candidate or party is weak, where they’re strong, and which groups may be moving in what direction,” Salvanto said.
That’s why Salvanto is relying more on CBS’s ongoing tracking poll and less on random-sample telephone polling, ahead of this year’s midterm elections.
“That decision is very much informed by 2016,” Salvanto said.
A tracking poll sets up a panel of thousands of voters and returns to them repeatedly over months. That lets pollsters discern the factors driving voter decisions — and gauge how attitudes change as the campaign wears on. In 2016, only two major surveys’ final predictions foresaw a Trump victory. Both of them — from the LA Times/USC and IBD/TIPP — were tracking polls.
“It gives us a great advantage in trying to explain the meaning of the poll results,” Salvanto said.
This year, the CBS Battleground Tracker “is concentrating on the districts that we think will make a difference,” he said. “Remember, in a midterm, you have to watch each congressional seat — don’t pay attention to national numbers.”
Just as a presidential race is not a national contest but a collection of 51 separate elections (one for each state and for the District of Columbia), November’s midterm involves elections for 435 House districts and 35 Senate seats. The collective result will determine which party gains enough seats to control each congressional chamber.
“For pollsters, midterms are the most interesting and difficult challenge we face,” Salvanto said.
Not only must they consider 470 unique races, “there’s never more than a third or at best 40 percent turnout in midterm elections,” he noted. “So we’re looking for a subsample of a subsample” of voters.
For 2018, the CBS News Battleground Tracker has gathered a panel of nearly 5,700 registered voters. Almost all of them live in the 50 to 60 districts that might switch from Republican to Democrat, or vice versa, in November — the only races that matter, when it comes to control of Congress.
The CBS Battleground Tracker keeps tabs on voters in congressional districts that have the potential to “flip.”Source for district ratings: The Cook Political Report
Salvanto’s polling currently indicates that few House seats will change hands in November — and that the GOP could very well hold its majority in the House. “In this era, a district’s voting patterns from the past tend to stay that way,” Salvanto said. “Not as many partisans today are willing to cross party lines.” Of the nation’s 435 House districts, fully 85 percent will almost certainly stick with its current party affiliation come November, Salvanto projects.
Choosing the 15 percent of districts that could flip is as much an art as a science, Salvanto said, but “there’s already a consensus forming among pollsters on what districts they are.”
The potential swing districts are scattered from coast to coast. One of Salvanto’s battlegrounds is right here in New York City — NY-11, in Staten Island and south Brooklyn, currently held by Republican Dan Donovan and being challenged by Democrat Max Rose. Others are in states from Texas to Maine to California to, yes, previously underpolled Michigan.
Some are picked for their demographic profile. “For example, we know Democrats have been trying to win suburban voters with college degrees, so we’re looking at the districts with many of those voters,” Salvanto said.
“We also look at where the campaigns are spending money,” he noted — a sign that the parties’ own internal data sees a potential pick-up or a possible loss.
“Overall, the districts in play tend to be more suburban and more affluent than the rest of the country,” he said. “But no single district will tell the national story.
“Right now I think this election looks like a toss-up,” Salvanto said. “We see a Democrat pickup in the House of Representatives in the 20-odd seat range, but Republicans could certainly hold on to the House.” The GOP holds a slim 43-seat House majority, with six vacancies.
“Even though Republicans have not fared well in special elections so far this cycle, it does look like they will be turning out for the midterms,” Salvanto said. “So far we do not see a large number of Republicans saying they will flip and vote for a Democrat.”
GOP voters in the past have been much more likely than Democrats to turn up and cast ballots in midterm elections, regardless of each party’s enthusiasm level ahead of Election Day.
So Democrats are literally betting the House on their ability to capture large numbers of voters who don’t normally vote in midterm elections. “They have to bring new voters in,” Salvanto said.
Democratic anger against President Trump gives them a shot at doing just that. “A large number of Democrats are contesting districts they have not contested before,” Salvanto pointed out. “That has new people coming into politics — younger, more women, more ethnicities — making for some interesting dynamics.”
But, according to his tracking poll, “voters say the Democrats need to do more than just oppose Trump,” he said. “They’re asking, ‘What are they arguing we’ll get if they take the majority?’ ”
A key indicator to watch is whether voters are thinking nationally or locally when they choose their congressional representative.
“I think we are picking up a shift of more voters looking to national issues,” Salvanto said, a startling departure from the old truism that all politics is local.
That’s another change that can be chalked up to Trump.
“Views of the president are a major factor,” he said. “The more intensely you feel about him, the more important you think the midterms are” — a correlation that’s equally strong on both sides of the aisle.
“The most motivated Democrats say this is very personal to them,” Salvanto said. “They don’t feel President Trump respects people like them — even if they say the Trump economy has been good for them.”
But despite all the White House drama since Trump’s inauguration, his overall approval ratings — currently at 43.3 percent, according to the RealClearPolitics polling average — have barely budged.
“We looked at how presidential approval ratings vary over their first year, and it turns out President Trump’s have been the steadiest ever measured,” Salvanto said. “That’s because his approval numbers among Republicans have been very steady and very strong.”
Negative media coverage actually has the potential to drive GOP turnout to higher-than-average levels in November. “Among Republicans, Trump’s biggest backers feel driven to come to his defense,” Salvanto’s poll finds.
Moreover, 75 percent of Republicans want their congressional candidates to fall in line with Trump’s agenda, he said, and are not looking for a representative who bucks the president or his policies.
“But one number will not tell the whole story,” Salvanto said. Instead, “look at the arguments deployed by both sides, and use polls to understand what is resonating or not resonating.”
“The best pollsters will tell you what could change, whose opinions are in flux and who could be persuaded,” he said.
“Maybe you use that information to change someone’s mind or maybe just to have a better conversation. Either way, it’s never been more important for us to try to understand each other and to give people a voice.”