A change of pace for you from all the polling doom and gloom lately. Most readers know the name “Trafalgar” at this point, but for the few who don’t: They’re the firm that called Trump’s wins in Michigan and Pennsylvania four years ago when nearly every other pollster was predicting Clinton victories, some by blowout margins. (Trafalgar was the only outfit to find Trump ahead in Michigan in any survey taken that year.) Two years later, with the rest of the field polling field expecting an Andrew Gillum victory in Florida, Trafalgar’s final poll found Ron DeSantis ahead. Again, they were right, everyone else was wrong.

Their secret is that they try to adjust for “social desirability bias,” the reluctance of some people to tell pollsters the truth about their preferences for fear of being judged disapprovingly. Robert Cahaly, Trafalgar’s lead pollster, is of the belief that that effect is especially prevalent among Trump voters. If in fact there are a ton of “shy Trumpers” still out there who are lying to pollsters — which many experts doubt — then the polls this year are way off.

Except Trafalgar’s, perhaps. If in fact Trump performs much better on Election Day than the competition expects, and especially if he does well enough to win another term, Cahaly will instantly be the most renowned pollster in America. The rest of the polling industry will never recover its reputation. That’s what’s on the line on November 3, in addition to control of the White House and Senate.

All of that is an introduction for you to the conversation below between Cahaly and Mike Pesca of Slate. If you can spare 18 minutes to listen to the whole thing, please do. Cahaly takes the listener through his approach to polling and explains the importance of social desirability bias, which he believes is even *more* pronounced this year than it was when Trump first ran for president. That’s borne out in some of his numbers over the past month, which reliably show Trump scoring five points higher — or more — than the polling average. Skim through the latest numbers from Michigan, the site of Cahaly’s most celebrated success in 2016. Of the last 11 surveys there, Biden leads by eight points or better in nine of them. In the 10th poll he leads by six.

In the 11th, from Trafalgar, he … trails Trump by one. That’s an eight-point difference between Cahaly’s result and the current average of 7.2 points. But Cahaly isn’t cowed. He tells Pesca in their chat that he thinks Michigan is the likeliest of the Rust Belt swing states to stay red this year, which completely contradicts the conventional wisdom that Michigan is actually Trump’s toughest “hold” in the region. In fact, Trump had been off the air in the state from late July until a few days ago, when he began running ads again. I’d been wondering lately if he was going to give up on the state altogether and focus his efforts on Pennsylvania, treating MI as a lost cause. On the contrary, says Cahaly. Expect another win in Michigan.

It’s hard to overstate just how far off the beaten path his numbers are relative to everyone else’s, including pollsters that tend to be friendlier to Biden. Trafalgar isn’t just expecting “shy Trump voters” to be a factor. They’re expecting them to be a massive factor:

What's also interesting about Trafalgar is that they're always biased toward Trump by about the same amount, unlike Quinnipiac where they do have a handful of polls that are better than the average for Trump. https://t.co/KxJa7WMQJv

— Nate Silver (@NateSilver538) October 15, 2020

Via RCP, here are the latest Trafalgar numbers from battlegrounds (although bear in mind that some of these are close to a month old):

Florida: Trump by two
Pennsylvania: Biden by two
Wisconsin: Biden by three
North Carolina: Trump by two
Michigan: Trump by one
Ohio: Trump by four
Arizona: Trump by four

If all of those results held on Election Day, assuming no unlikely Biden victories elsewhere, Trump would win 276-262. That’s right in line with Cahaly’s forecast for the election, which he delivers to Pesca at 26:00 of the clip below — Trump in the mid-270 range. Which is the most encouraging polling news Team MAGA has had in months.

But don’t get too cocky. Look back again at the numbers above and notice how tight the margins are. A polling miss of two or three points would be perfectly ordinary for any pollster, yet Cahaly’s prediction depends upon him missing nowhere. If Biden fares just two points better in Michigan than Trafalgar expects, he’ll be president. If he’s three points better in Florida and North Carolina, he’ll be president comfortably. And as impressive as Cahaly’s 2016 polling was, he *did* overshoot the mark on Trump in Michigan, predicting a two-point victory in a state the president won by just three-tenths of a point. If Trafalgar’s current Michigan numbers are off by the same margin, Biden would win the White House by the skin of his teeth.

There’s no margin for error. The interview with Pesca starts at 10:30 below.