Nate Silver (AP photo)

The data journalists seemed so confident, so calming. "Why Donald Trump Isn't a Real Candidate, In One Chart," read the headline in a June 16 post at FiveThirtyEight. On July 18, even after the tycoon-cum-candidate surged in presidential polls, New York Times Upshot reporter Nate Cohn wrote that Trump had reached a "turning point" and would crumble under the weight of Republican attacks.

At the time, the RealClearPolitics average of national polls put Trump at 15 percent. Today, he is at 23.8 percent, now far ahead of the field. If there was a turning point, Trump hooked it and left his haters choking on the exhaust.

In a rough year for polling analysis, the Trump surge stands out. The first-time candidate whom so many people wrote off has done for 2016 what Isaac Asimov's Mule did for the psycho-historians of Foundation -- a conquest from out of nowhere, unpredicted by any of the calculations, turning his enemies' blasted palaces into new (and classy) throne rooms.

Two Fix-ers, Chris Cillizza and Philip Bump, have already conceded that they blew the Trump call. Bragging disclosure: I didn't, and took a few Twitter swipes at #datajournalism for it. Still, I was curious as to why the people who explain elections through hard numbers had been wrapped around the axel by Trump. It was one thing for people who covered politics from the cattle calls to over-rate the damage from Trump's joke about John McCain. What about the data journalists?

First up: FiveThirtyEight. I asked FiveThirtyEight's Harry Enten if he had been surprised by Trump's surging favorable numbers (his low favorables among Republicans had been cited by Cillizza and others as a reason he could never take off), whether his surge was all due to media coverage and whether we had learned something from the affair.

FiveThirtyEight founder and editor-in-chief Nate Silver responded, saying I was asking the wrong questions.

"To be honest, dude, it seems like you’re being pretty straw-manny and cherry-picky," said Silver. "I’d encourage you to read a cross-section of Trump coverage at FiveThirtyEight and the 537 other 'datajournalism' sites. It’s pretty nuanced, and offers perspective that is sorely lacking in the mainstream media’s hyperventilated coverage of him."

His throat duly cleared, Silver made three arguments:

  1. Trump's horse race numbers (e.g “Trump is at 25 percent in the new XYZ poll”) are inflated, for several reasons. First, they depend heavily on name recognition and on being in the news a lot. Second, they ignore that his unfavorable ratings are still quite high even within the Republican Party, which limits his room to expand his current share of the vote. Third, these are often polls of adults or registered voters, not the very small number of people who will actually vote in Iowa or NH. Fourth, most voters have lives and have not started thinking about the race seriously yet.
  2.  The GOP establishment is vehemently opposed to Trump. That matters a lot and makes it extraordinarily difficult for him to win the nomination, even though he can play off that resentment in the near term.
  3.  Despite the mainstream media’s implications to the contrary, what’s happening with Trump has plenty of precedent. It occurred for Michele Bachmann, Herman Cain, Newt Gingrich (twice) and other GOP candidates in 2011/12. And his case is similar to bygone fringe or factional candidates like Pat Buchanan and Pat Robertson. Sometimes these candidates hold 20 or 30 percent of the vote through Iowa or New Hampshire, and sometimes (as was the case for most of the candidates in 2012) their bubble pops sooner. But, they tend to have a ceiling on their support, which manifests itself eventually.

Furthermore, Silver pointed out that FiveThirtyEight followed the Trump-skeptical coverage (including a piece titled "Two Good Reasons Not To Take The Donald Trump ‘Surge’ Seriously" on July 16) with a piece about Trump's trolling benefiting him with a certain type of voter. "None of them will necessarily be deterred from declaring their support for him because of his comments about McCain," wrote Silver. "Some of them might even be encouraged." Indeed.

Next up: The Upshot's Cohn. He had followed up his "turning point" piece by arguing that the polls showing Trump ahead were "fooling a lot of people" who did not check the small print. Trump's McCain joke, made on July 16, had not been fully factored in. "There are just two national polls conducted entirely after Mr. Trump’s remarks," Cohn noted.

Every poll conducted since the remarks has found Trump in the lead. Cohn left room for that possibility ("entirely possible that Mr. Trump’s poll numbers won’t take a beating in the short run"), but it wasn't his top theory.

"I'm not surprised that he hasn't collapsed," Cohn explained in an e-mail. "I think I wrote that scrutiny would 'erode' his numbers. But, as such, I am surprised that there isn't any evidence of erosion at all."

Initially, Cohn wrote that Trump's McCain gaffe would allow party elites to bring him down like a wounded rhino. Why had that not worked? And why had Trump not taken on damage for his old, well-promoted, flip-flops from liberal positions?

"I think it takes traditional, sustained criticism from party elites and journalists, but I also think it's difficult for policy-oriented critiques to get attention when Mr. Trump provides so much news himself," Cohn said. "Most voters aren't paying much attention -- a third of Republicans still don't know enough to formulate an opinion of Scott Walker, and this is among the politically interested voters who take public opinion polls -- and oddly, I think he's benefited, at least so far, from running such a substance-free campaign. He hasn't provided a 'news hook' for a policy discussion."

So: Was Cohn wrong? It was "too early to say," he argued. Trump had proven sturdier than a few recent fringe candidates. Perhaps that had to do with his built-in celebrity; perhaps it was something else.

"If he does shrug off the critics, I would interpret that to mean that he was quickly able to build a relatively strong base of support," Cohn said. "I think that would be really interesting: His poll numbers depict an ideologically diverse coalition, not the sort of coherent base that I would expect to be most resilient."

That might mean that the days of writing off Trump are over. His fate is not predetermined by polls or by prior right-wing insurgencies. The press may actually have to cover a campaign.

"[Trump] has a plurality in polls of Republican-leaning adults in August when nobody’s paying attention to the campaign," Silver told me in a follow-up e-mail. "That’s not all that meaningful, and it's not sufficient to make him a contender. It’d be like declaring a baseball team a contender based on their spring training stats."