Did James Comey’s final-fortnight reopening of the probe into Hillary Clinton’s e-mail scandal cost her the election? Some polling analysts believe that the data shows evidence that it did, but an investigation by a polling trade group discounts the effects. That lede gets buried somewhat in Politico’s coverage of the conclusions to an American Association for Public Opinion Research study, whose primary purpose was to answer how the polling turned out to be so wrong in 2016:

“The evidence for a meaningful effect on the election from the FBI letter is mixed at best,” the report states, citing polls that showed Clinton’s support beginning to drop in the days leading up to the letter. “October 28th falls at roughly the midpoint (not the start) of the slide in Clinton’s support.”

In fact, while the Comey letter “had an immediate, negative impact for Clinton on the order of 2 percentage points,” the report finds that Clinton’s support recovered “in the days just prior to the election.”

That would have been about the time Comey announced that the probe had been concluded again with no further action taken. Comey explained his reasons for these decisions earlier this week in hearings on Capitol Hill, saying the thought of impacting the election made him “mildly nauseous.” Perhaps this will soothe his stomach, but David Axelrod put it best even before Comey testified to Congress:

“Jim Comey didn’t tell her not to campaign in Wisconsin after the convention. Jim Comey didn’t say, ‘Don’t put any resources into Michigan until the final week of the campaign,'” Axelrod said. “One of the things that hindered her in the campaign was a sense that she never fully was willing to take responsibility for her mistakes, particularly that server.”

Nate Silver, one of the few polling analysts to have given Trump a chance of winning on Election Day, wrote on Wednesday (the day before the study’s release) that the Comey letter did cost Hillary the election. Silver has not yet responded to the study from AAPOR on his blog, but his argument in the post is that he sees a “Little Comey” effect that was nonetheless enough to “probably” cost her the election. Hillary’s defenders are trying to make a “Big Comey” argument that doesn’t really stick, Silver writes:

First, they said the letter’s impact was larger in Midwestern swing states such as Wisconsin because there were large numbers of undecided voters there, especially among white voters without college degrees. And the Clinton campaign claimed that the second Comey letter — which he issued late in the afternoon on Nov. 6 and which announced that the emails on Weiner’s laptop hadn’t turned up anything new — hurt Clinton because it put “FBI,” “Clinton” and “email” back in the headlines. This is hard to test because the second Comey letter came so late in the campaign that there wasn’t time for polls to pick up its effects.

But it’s plausible that Clinton’s underperformance versus the polls on Election Day had something to do with Comey — either lingering effects from his original letter or new effects from his second letter. The “Big Comey” case might attribute a 4-point impact to him nationally — accounting for the swing between Clinton’s 6-point lead on the morning of Oct. 28 and her 2-point popular vote margin on Election Day — and slightly more than that in the swing states.

My personal views are more toward the “Little Comey” side of the spectrum, since I think there would have been a fair amount of mean-reversion even without Comey. That’s because Clinton and Trump had alternated better and worse months in the polls in a way that tracked with the news cycle. Clinton had been in a strong position in the polls in June, August and — until the Comey letter — in October, while Trump had drawn close to her in May, July and September (and therefore might have been “due” for an uptick in November). This pattern may have reflected some sort of complicated feedback loop in media coverage. After some initial stimulus — say, a strong debate — there was a frenzy of favorable coverage for a candidate and negative coverage for her opponent, with news events framed against a backdrop of rising or falling polls. Then after a few weeks, the reporting on the story exhausted itself, the polls stabilized and the press was eager to look for a reversal of momentum. Comey’s letter came at a time when the campaign press may have been itching for a change in the narrative after several tough weeks for Trump. If not for the Comey letter, perhaps some other story would have blown up in Clinton’s face. Still, this theory is speculative, and those other stories might not have had the kryptonite-like effect that email-related stories had on Clinton’s numbers.

One potential problem with the “Little Comey” theory is that it postulates that the state-based polling was accurate. AAPOR’s study found that national polling was largely accurate, and that the results of the election fell pretty much along the projected lines, with Hillary narrowly winning the popular vote. However, polling in critical states turned out to be unreliable, thanks to poor modeling and insufficient data. As NPR reports, the AAPOR found that state polling was “historically bad”:

First off, only some polls were off, and it wasn’t the national polls. Hillary Clinton won the popular vote by 2.1 percentage points, and polls had her winning the popular vote by an average of 3 points. That’s not much of a gap at all, compared to past presidential polling.

But state polls were off by an average of 5 points, the largest average since 2000. This is where the researchers drilled down into the whys of what went wrong.

And where did it go wrong the most? Back to Politico:

In Ohio — the perennial bellwether Trump won easily — surveys underestimated the Republican by 5.2 points. In Wisconsin — which hadn’t awarded its electoral votes to a Republican since 1984 — the polls underestimated Trump by an even greater margin of 6.5 points.

In Pennsylvania and unexpectedly close Minnesota, the polls underestimated Trump by between 4 and 5 points, while the polls in Michigan were off in the same direction by about 3.5 points.

Small wonder, then, that the media got caught by surprise on Election Night. What happened? State-based polling didn’t accurately capture late deciders, AAPOR concluded:

Altogether, around 13 percent of voters nationally made up their minds in the final week, according to Pew data the researchers reviewed. That’s in line with past elections. However, in the swing states of Wisconsin, Florida, Michigan and Pennsylvania, those late-deciding voters were far more likely to vote for Trump than Clinton.

Nationwide, 45 percent of the late-deciding voters ended up voting for Trump, compared to 42 percent for Clinton. But in Michigan, for example, it was 50 percent for Trump and 39 percent for Clinton — and that was the smallest margin of these four states. In Wisconsin, meanwhile, it was 59 percent Trump, 30 percent Clinton.

The turnout models didn’t work in these states either, the study argues, and pollsters failed to adjust effectively for education demographics. Whatever happened, the result was that polling was off in the states that turned out to be critical battlegrounds, and as a result most people — with Silver as a notable and derided-at-the-time exception — predicted Hillary would win easily.

Had those state polls been more accurate, the election result would have been far less of a surprise, and the notion that Comey significantly impacted the result would have little traction. The real story of this election is that Hillary Clinton ignored middle-America voters, doubled down on the progressive tilt that had cost Democrats the House, Senate, and massive numbers of state legislature seats, and was in general a terrible candidate. As Axelrod said this week, “It takes a lot of work to lose to Donald Trump, let me tell you.” Let’s give credit for that work where it’s truly due.