President Trump in the Oval Office. (Evan Vucci/AP)

For the third time in two weeks, President Trump on Sunday touted poll numbers that showed half the country approving of the job he’s doing as president.

“Just hit 50% in the Rasmussen Poll, much higher than President Obama at same point,” Trump wrote, with his characteristically idiosyncratic capitalization. “With all of the phony stories and Fake News, it’s hard to believe! Thank you America, we are doing Great Things.”

That tweet provides a great opportunity to review some of the ways in which Trump uses poll numbers to present a rosier case for his presidency than is warranted. So let’s do so.

1. Rasmussen Reports polls are consistently friendlier to Trump (and were more unfavorable to Obama) than most polls.

The pollster uses an unusual pool for its polls, talking only to “likely voters” — a nebulous descriptor for a subset of the population that doesn’t try to approximate the views of all Americans. Often, likely voter pools lean more heavily Republican because, well, Republicans are often more likely to vote. (Even in this year, when Democrats consistently say they’re more motivated to vote, the end result is that Democrats are about as likely to vote as Republicans.) Rasmussen calls only landline phones, supplementing that pool with an online survey. It’s an unorthodox methodology that has produced mixed results.

But those results have a consistent pattern. From the day of each president’s inauguration through April 14 of his second year in office, Rasmussen released hundreds of poll results assessing Obama and Trump’s national approval.

  • In 93.7 percent of the polls taken from Jan. 20, 2009, to April 14, 2010, Rasmussen’s Obama approval number was lower than the RealClearPolitics average of polls on the same day. (On average, the Rasmussen polls were 3.9 points lower than the RealClearPolitics number.)
  • In 99 percent of the polls taken from Jan. 20, 2017, to April 14, 2018, Rasmussen’s Trump approval number was higher than the RCP average of polls, by an average of 4.9 points.

This is what that looked like earlier this month.

In early 2010, FiveThirtyEight Editor Nate Silver attributed Rasmussen’s Republican lean to a “house effect”: How it conducted its polls led to a Republican advantage.

“If you’re running a news organization and you tend to cite Rasmussen’s polls disproportionately,” he wrote, “it probably means that you are biased — it does not necessarily mean that Rasmussen is biased.”

After the November 2010 election, he changed his position: Rasmussen was in fact biased. But his point about citing Rasmussen stands — and we’ll come back to it.

2. Most polls have consistently had Trump’s favorability rating well under 50 percent.

So if Rasmussen’s 50 percent figure is high, what’s correct?

“Correct” in polling is hard to say with certainty. Polls have acknowledged, built-in errors that mean their findings are always within a range. That’s why most of those who are tracking polls look at poll averages, like RCP’s: Averages tend to diminish the effects of those margins of error.

Here, then, is how RCP’s averages for Trump and Obama compared through mid-April of their second years in office.

Obama’s average has always been higher — although the gap between the two has narrowed. More on that in a second, too.

Trump is cherry-picking when he looks only at that Rasmussen poll. If we cherry-pick a poll that’s closer to our hearts — a new poll from The Washington Post and our partners at ABC News — Trump’s approval rating is 40 percent. In mid-April 2010, Obama was at 54 percent in our poll.

3. Trump’s poll numbers are unusually low.

It’s important to reiterate, too, that these are not just two regular old polling averages for two presidents, one of which is better than the other. Pollsters keep determining that Trump’s unpopularity is historically low. His approval has improved recently, but it’s still very bad by historic standards.

So even if Trump were to correctly tout an uptick in a poll that wasn’t an outlier like Rasmussen, it’s simply climbing up a bit further from the bottom of the barrel.

4. Trump’s poll numbers are unusually static.

But Trump’s approval ratings haven’t really moved much at all. We can express that visually:

The range of approval ratings for Trump is much more narrow than it was for prior presidents. Only Obama had a range that was nearly as small.

Why? Because the past seven years of Obama’s presidency set a pattern that Trump continues. His party loved him, the other party hated him, and independents moved around. Republicans were open to Obama for the first year of his presidency before they calcified into opposition; the Democrats never gave Trump that grace period. Independents made up much of the movement in Obama’s polls — and independents view Trump more negatively than they did Obama.

That pattern emerged in Obama’s second year, which is why it looks as though those two lines have converged. By April 2010, Obama was in the upper 40s; he’d eventually settle in the mid-40s for much of his time in office. For about half of his time in office, Trump has been in the 30s.

Trump’s poll numbers are low and stay low — and it’s not clear what, if anything, might break that pattern. Movement in the polls, then, is largely minor (or insignificant).

5. Polls don’t keep rising to the same percentage over and over.

As we said, Trump has tweeted three times about polls this month — all three polls from Rasmussen.

These are the three polls. Notice anything?

He’d “just hit” 50 percent on April 3, was “still rising” to 51 percent on the 4th — and “just hit” 50 percent again on the 14th. What’s implied there is that at some point the polls had gone back down. After all, if you just hit 50 percent, went up to 51 percent and then hit 50 percent again, the second time hitting 50 percent seems like it’s headed in the other direction.

But this, too, is just Trump cherry-picking. That 51 percent, for example, wasn’t the day after Trump had hit 50 percent. He was at 50 percent on April 1, 49 percent on April 2, 51 percent on the 3rd, and 47 percent on the 4th and 5th. Rasmussen is like other pollsters in that it has margins of error, and, as such, even similar poll numbers will naturally bounce around a bit.

Of the Rasmussen polls this month, Trump has been at or above 50 percent three times — and each time he tweeted about it. No other polls — even from Rasmussen — made the cut. And that he has “just hit” 50 percent twice makes that clear.

Now please remember: This article is not meant to be a response to Trump’s Sunday tweet. Instead, we’re confident that it can stand the test of time, holding important lessons whenever Trump tweets about his poll numbers. Because the common pattern in such tweets matches that of Sunday’s: seizing on an outlier and presenting it as fact.