The Washington PostDemocracy Dies in Darkness

This is Trump’s most spectacularly wonderfully terrible tweet ever

The president who makes up his own poll numbers has thoughts on polling methodology.

President Trump delivers remarks as he holds a Cabinet meeting in the East Room of the White House in Washington on May 19. (Kevin Dietsch/Pool/EPA-EFE/Shutterstock)

I mean, where do I start?

I’m tempted to start by noting the staggering irony of President Trump insisting on Twitter that a new poll showing him losing by a wide margin to former vice president Joe Biden relies on the same bad math that convinced a slew of Republicans in 2012 that Mitt Romney, then the party’s nominee, was likely to easily defeat incumbent President Barack Obama. Romney did not defeat Obama, as you may recall, because the bad math was bad. But that Trump is using the idea that the polls are skewed against Republicans today — as his press secretary bashes Romney’s performance in 2012 because Trump now hates the guy — is remarkable.

Except I’m also tempted to start by pointing out that the president who now fancies himself an expert on polling began the day by sharing a poll number that is by now pretty obviously made up. Trump keeps talking about how his approval rating among Republicans is 96 percent, without ever citing a poll that includes that figure. He used to repeatedly claim his approval was 95 percent. Before that he repeatedly claimed that it was 94 percent. It’s like his sense of his own net worth, which he once said depended to some extent on how he felt on a given day. Today Trump feels like 24 out of every 25 Republicans approve of the job he’s doing, so that’s what he’s going to tweet.

But where I’m actually going to start — 250 words in and without having actually introduced the subject of this article — is by talking about how good a pollster John McLaughlin is. Or, more accurately, how not good he is.

Remember when congressman Eric Cantor of Virginia lost his reelection bid in the 2014 Republican primary? It was stunning, given his senior leadership position within the caucus and the presumption that he was going to win easily, as senior members of a caucus usually do. But Cantor lost by double digits.

Cantor’s pollster was McLaughlin, and he had assured the world that Cantor would win the primary by 3.4 points. Oops, sorry. The decimal place was in the wrong place. By 34 points. The poll was off by 44 points.

That’s not “the race shifted” territory. That’s “the pollster got some very big things very wrong” territory.

It wasn’t McLaughlin’s only error, though it was his most notorious. After Cantor, the Republican Party’s congressional campaign arm recommended its candidates not use his shop. Some did anyway in 2018, such as Rep. Rob Woodall (R-Ga.), who commissioned a poll in October of that year showing him with a 27-point lead. He won by 2 points. Oops! Another decimal point error. Zero-point-two points.

McLaughlin also did polling for then-Rep. Barbara Comstock (R-Va.), showing her running close in her reelection bid — and prompting the party to keep spending in the race. When a Post-Schar School poll showed her losing by 12 points, McLaughlin disparaged it as a “partisan Democratic poll release.” Comstock lost by 12 points.

There are plenty of other examples. But why do we care about John McLaughlin’s abilities as a pollster? Because Trump does.

After CNN released that poll showing Trump down 14 points to Biden, Trump shared a memo from McLaughlin on his Twitter account. McLaughlin claimed that the methodology of the poll, conducted by SSRS, was flawed, and Trump used that to argue that this was therefore a “SUPPRESSION POLL,” undertaken to dampen enthusiasm.

First of all, that’s a goofy argument. If you want to suppress enthusiasm — something which Trump’s own campaign in 2016 admitted doing — you generally try to do so when people are actually voting. Trying to suppress votes now is like trying to suppress your appetite for Thanksgiving dinner on the Fourth of July. Lots of time for it to return.

What’s goofier, though, are McLaughlin’s claims about the partisan splits used by SSRS.

“Instead of the 33% Republican turnout which actually happened in 2020,” the memo reads, “they are reporting polls on only 26%, 25% or even 24% Republicans.”

In fact, 25 percent of respondents in the CNN poll were Republicans. Of course, according to Gallup, 28 percent of Americans identify as Republican. It’s not the case that everyone in America who isn’t Republican is a Democrat; most of them are independents. Forty-four percent of the respondents in CNN’s poll were independents — more than the 37 percent of the electorate who identify that way, according to Gallup.

McLaughlin has a typo there, too. He means that 33 percent of the electorate in 2016 was Republican, according to exit polls. This, he suggests, is the baseline that should be used in polling, instead of weighting to demographics of the actual population of the country and measuring their current party identification. That latter approach is something done, he says, because news networks “have Democrat operatives like Chuck Todd, George Stephanopoulos and other Democrats in their news operations.”

You know who paid for the exit polls? Those same news networks.

As the election gets closer, polls will switch from registered voters to likely voters — once we have a sense of who those likely voters are. That’s very hard to predict five months out, yet this is what McLaughlin demands.

He also takes issue with the timing of the poll and the questions, complaining that it was conducted mostly before last Friday’s jobs numbers, as though CNN and SSRS should have known those numbers would be better than expected and then waited to poll until they were out.

But those were secondary concerns. His real concern was that the sample of respondents was skewing the results.

“The refusal to screen for actual likely voters is creating an under-polling of Republicans and therefore Trump voters,” McLaughlin wrote in the memo. “It seems intentional. It’s exactly what the media did in 2016. Let’s prove them wrong again.”

It is intentional — because CNN wanted a poll which reflected the voter pool, not a poll which would “prove wrong” Trump’s opponents. If that’s what Comstock was paying for in 2018, no wonder McLaughlin’s poll results turned out the way they did.

That word, though! The “unskewing!” Those who were around eight years ago for the presidential race will remember the broad effort to “unskew” the national presidential polls that year to better capture Mitt Romney’s support. The idea promoted by people such as Dean Chambers, who set up a whole website dedicated to unskewing the results of polls showing Romney trailing, was that rejiggering poll results to include more Republicans would yield a more accurate result.

It didn’t. Romney lost — something which Trump now lords over him repeatedly to disparage his political chops.

Oh, speaking of the effort to unskew the 2012 polls in Romney’s favor? Here’s how one prominent poll-understander revealed his expertise on the process.

If that argument seems familiar, it’s because of stuff you read about 60 seconds ago.

As I wrote earlier Monday, the CNN poll may well be an outlier, though Biden’s lead over Trump has grown recently. There will always be polls that show a candidate doing particularly well or particularly poorly, polls which are often later shown to be outside the broader trend of what’s happening. Sometimes, though, they’re at the forefront of how a race is shifting. Time will tell.

What hasn’t been shown to be the case is that pollsters such as CNN and SSRS end up wildly off the mark in national polling because of how they select their respondents and weight their responses. Put another way: If there’s a pollster whose results are suspect in this particular tiff, it’s not the pollster Trump is criticizing. It’s the pollster he has asked to offer the criticism.

White House Press Secretary Kayleigh McEnany said on June 8 that Trump took "great offense" at Sen. Mitt Romney's 2012 comments on race. He hasn't always. (Video: The Washington Post)