It may not seem like it, but this tweet is actually an improvement.

Normally, President Trump doesn’t try to rationalize why the only approval rating he cites comes from Rasmussen Reports. He just plucks that juicy polling cherry off the polling tree and tweets out the results. We’ve repeatedly cautioned against taking Trump’s insistence on Rasmussen at face value (occasionally annoying the team at Rasmussen). After all, Rasmussen’s approval numbers are consistently outliers relative to other polls — a function, in part, of their focus on likely voters and not Americans on the whole.

This tweet, at least, tries to explain why Trump uses only Rasmussen’s numbers. The correct reason, of course, is that Rasmussen’s numbers are almost always higher than the polling average. (They have been 99 percent of the time since Trump was inaugurated.) But Trump tries to argue here that he uses Rasmussen because the pollster is more accurate than other polls — including our own.

Before we assess the claim in the tweet — that Rasmussen outperformed other pollsters in 2016 — we should note that predicting election results is different from assessing views of politicians. Rasmussen’s figures include only likely voters, a pool of people that’s proprietary to the firm and could be presidential or midterm voters. But it’s clearly only people who are likely to vote, not Americans on the whole. When Trump highlights Rasmussen’s approval number, he is highlighting that he is approved of by, say, half of the less than 40 percent of the country that votes in midterm elections.


But we must also defend our honor on 2016 polling, given both Trump’s tweet and Rasmussen’s frustrated defense of its own polling. (“The Liars Say We’re Outliers — Again,” the pollsters wrote Tuesday.)

This is what Rasmussen wants to focus on: Its final poll had the margin between Trump and Hillary Clinton at two points — just under the 2.09-point margin in the actual results. The Post (and our partners at ABC News) had the margin at four points. CNN’s final poll, conducted with ORC in late October, had the head-to-head margin at five points.

Rasmussen did better!

We were about 1.9 points off. CNN (two weeks before the election) was 2.9 points off. Those are margin-of-error-level misses, but Rasmussen was still closer in predicting the margin.

“Who got it right? The three daily tracking polls — Investor’s Business Daily, the Los Angeles Times and Rasmussen Reports,” Rasmussen’s Fran Coombs wrote Tuesday. “We’re the ones who were taking the temperature of the electorate every single day, not dropping in for a handful of days like the others — usually after a controversy — for a snapshot of popular opinion.”

Even assuming that the argument about tracking polls being somehow more accurate than other polls is correct (which is not a fair assumption), it’s not a good description of what The Post and ABC did. For the last few weeks of the 2016 election, we also ran a tracking poll.

But we hasten to note that Rasmussen’s articulation of “who got it right” is itself very wrong. The IBD and Los Angeles Times polls were correct in the sense that they thought Trump was going to win the election — but they thought he was going to do so while winning the popular vote. That’s what this figure measures: the national vote split between the candidates. Both IDB and the Times had Trump winning the popular vote by a few points. Because it was Clinton who won by a few points, those pollsters were actually much worse than The Post and CNN. Having Trump up two in a race he lost by two is a four-point miss.

The Times and IBD polls were a bit like someone saying God was sad because the New York Knicks didn’t make the playoffs and then claiming credit when it starts to rain. They got the effect right, but the reason for it was quite a miss.

Here, we’re also measuring only the margin, which is misleading. For example, a pollster could have predicted that the 2016 results would have been Clinton 14, Trump 12 and 74 percent for Tori Spelling. They would have gotten the Trump-Clinton margin right! But.

So let’s look at how the pollsters predicted the results of the four-way contest between Trump, Clinton, libertarian Gary Johnson and the Green Party’s Jill Stein. Here are the results from the polls Trump cites in his tweet.

Notice that the dark blue-green Post-ABC dots are as far from the results as Rasmussen — or closer. For all four candidates.

If we add up how far each pollster was from each of the four candidates’ actual results, the Post-ABC poll was off by only four points in total. Rasmussen was off by more than six. Even CNN-ORC’s poll from October was closer to the final mark than Rasmussen’s final tracking poll.

During the primary, Trump’s assessment of polling was more nuanced — because those polls showed him winning, as he did. He called me once during the primaries to discuss an article I’d written about polling (I don’t remember which) and made a point about online-vs.-live-caller polling that demonstrated a good sense of how polling worked. Trump understands that the point of polling is to explain something about the world. As a politician, he has consistently preferred that those explanations be ones that cast him in a favorable light, not necessarily explanations that are entirely accurate.

We pick on Rasmussen only because Trump insists on citing them to the exclusion of other polls. He does so because they tell the story that he wants America to hear: He’s popular, and the media is lying to you when it says he’s not.