During his very brief(!) press conference on the Mexican border on Thursday, Donald Trump threw out a statistic that he's been using a lot. "I think I'll win the Hispanic vote," he said. "They just did a big poll in Nevada, the state of Nevada. I'm way ahead -- and more importantly as fas as I'm concerned, I'm way way ahead with the Hispanics. Into the thirties."
Trump's obvious goal is to, first, reinforce his ongoing claim that he would win the Hispanic vote if he were the Republican nominee. (In the most recent Post/ABC poll, 84 percent of Hispanics said they would "definitely" not vote for Trump.) Second, he wants to dismiss the suggestion that by calling Mexican immigrants criminals and rapists in his announcement speech, Hispanics view him unfavorably.
So Trump relies on this random poll from Nevada to make his case. The problem is, it doesn't.
Here's the press release for the survey, conducted by Gravis Marketing on behalf of "One America News Network," a media organization founded by the Herring family of Southern California as an attempt to compete with Fox News.
Regardless of who paid for the poll, the survey itself has some question marks. Gravis determined that Trump led Wisconsin Gov. Scott Walker by 12.7 points, 27.7 to 15 in the Silver State. (More than a fifth of respondents said they were "unsure.") Among Hispanics, 31.4 percent chose Trump, versus 11.4 who chose Walker.
First, note that the figure is Republican Hispanics, not all Hispanic voters. That's a critical caveat.
To get those numbers, the company spoke to 623 Republicans. How many of those Republicans were Hispanic isn't clear; the smaller the number, the less accurate the breakdown. (I reached out to Gravis to fill in some of these blanks. The person from the firm who could answer the question was not available, but I'll update this when we do speak.) There's reason to suspect that the number of Hispanic Republicans who were contacted is low. Scott Clement of the Post's polling team notes that Gravis has historically only called landline phones. In 2014, 59 percent of Hispanics in the U.S. used only cell phones. In Nevada in 2013, 43.1 percent of Hispanics were cell-only. If cell phones were excluded from the poll, that makes it much harder to get a large sample of Hispanics.
What's more, only five percent of those surveyed in a 2012 Republican caucus entrance poll were Hispanic, suggesting a very low density of Hispanics in the party. If that same percentage holds for the sample Gravis used, we're talking about 31 people. That means there's a pretty large margin of error even if we're just talking about the Republican vote -- much less Trump's extrapolation to the rest of the state and country.
Update: A reader writes in to make a mathematical point. The percentages for Hispanic Republicans suggest that the number of respondents was a multiple of 35. That's because 31.4 percent of 35 is 11; 11.4 percent is 4. It's possible that there were 70 Hispanic Republicans surveyed, 11.2 percent of the total. But even so, that's 22 votes for Trump (11 times two) -- which probably doesn't translate to the entire population of Hispanics in the country.
Why does Trump keep bringing up this number? Because, as our national poll showed, he is viewed very unfavorably by Hispanics, no doubt thanks in part to his comments about immigrants. Trump doesn't want to admit he's wrong, so he seizes upon and inflates whatever's at hand.
As he's done with some regularity over the course of his campaign.