During a speech at the NRA convention in Texas on Friday, President Trump cited an incorrect statistic.

“Kanye West must have some power, because you probably saw I doubled my African American poll numbers,” Trump said, referring to the supportive comments made by West on Twitter. “We went from 11 to 22 in one week. Thank you, Kanye. Thank you.”

CNN’s Brian Stelter tracked that number back to a Daily Caller report that picked out a figure from a Reuters-Ipsos poll. The number wasn’t about black Americans on the whole but black men — and even then it wasn’t a reliable number. In addition to methodological questions about the poll (The Washington Post doesn’t generally cite Ipsos polling because of how the polls are conducted), the Daily Caller report ignores the vagueness in the figure.

The report (and Trump) looked at the poll like this.


From 11 percent to 22 percent. Boom, as they say.

But Reuters’s polling editor noted that the “credibility interval” among black male respondents was a giant nine percentage points. This isn’t the same as a margin of error given Ipsos’ methodology; you can read about the difference if you want. But the idea is the same: The fewer people who are included in the sample for a poll, the more uncertain the final number. And when you start breaking down a poll into smaller groups (black Americans, black male Americans) the sample size drops — and that uncertainty goes up.

Meaning that the “real” figure (if you will) for the week of April 22 could be anywhere within that blue box, according to Reuters. The “real” value for the week of April 29 would be in the green box.


So as Reuters told CNN, it’s possible that the “real” figures indicate that Trump’s support with the group could have dropped.

Again, though, Stelter did a good job of breaking down the problem with Trump’s comment. We’d like to focus instead on another habit of Trump’s: repeating claims about dramatic gains in the polling — that he had already claimed as gains in the past.

Take Trump’s support among black voters. We went over this in January, when Trump claimed basically the same thing as he did on Friday, that his support from black Americans had doubled. At that point, he made the claim on Twitter.

And, at that point, he was referring to an iffy Breitbart article, not an iffy Daily Caller one. We broke down the various egregious statistical mistakes in the Breitbart piece at the time, but a quick review is in order.

In 2016, Trump got about 8 percent of the vote from black Americans in exit polling. SurveyMonkey polling over the course of 2017 showed that 23 percent of black men approved of his job performance and 11 percent of black women did. Average those figures, and you get 17 percent approval — double that vote share! It’s just that you can’t average those two numbers, you can’t compare exit polls with approval ratings, exit polls are notoriously fluky, and actual apples-to-apples data show that his approval rating is essentially flat since January of last year.

Gah, got distracted again. The point is: Trump’s approval rating can’t keep doubling to about the same value without some dramatic reversals. If a business put out a news release touting that its stock price doubled from $40 to $80 a share over the year and then, four months later, put out another news release celebrating a doubling from $45 to $90 in a week, you would be justified in wondering either what was going on — and/or why their stock tanked so much over those four months.

Trump does this with his overall approval rating, too.

“By the way,” he said in that NRA speech, “you just saw the recent poll? It came out, the Rasmussen, 51 or 52. It’s the highest level I’ve ever been at. How does that happen when you only get bad publicity? How does that happen?”

The short answer to that question is that Rasmussen is consistently much friendlier to Trump than most other pollsters. Rasmussen almost always recorded a lower approval for President Barack Obama than the RealClearPolitics polling average over Obama’s first 16 months in office; it has almost always — as in 99 percent of the time — recorded higher approval for Trump than that average.

But also, how many times is Trump going to celebrate getting more than 50 percent in Rasmussen’s poll? (The poll he was citing had him at 51, not 52.)

What has Rasmussen shown over the past month? As Trump’s constant tweeting about being at 50 or 51 percent suggests, the pollster has had him bouncing around at about 50 percent. On average, a little below, 48.7 percent.


At least this, unlike that support-from-black-men number, is based on statistically valid analysis.

What Trump’s doing is like a guy bragging about the big boost to his bank account after every payday and not telling anyone about how much he spends at the racetrack. It’s as if Trump’s standing on the shore telling everyone how much water is coming in with each wave and then staying quiet when the waves slip back out. Maybe the tide’s coming in for him; his approval ratings from other pollsters have ticked upward. But repeatedly trumpeting that he hit 51 percent in the same poll doesn’t give the impression of an inexorable rise in the public’s estimation.

I mean, Trump celebrated hitting 51 percent in Rasmussen’s poll exactly a month before he did so in Texas.

An impressive bit of politician-speak to repeatedly celebrate the same achievement every time.

This article was corrected to properly refer to the uncertainty in the Ipsos poll.