A 2016 Review: There’s Reason to Be Skeptical of a Comey Effect
|
Version 0 of 1. On Friday, Oct. 28, James B. Comey, the F.B.I. director, sent a letter to Congress about new evidence in the investigation into Hillary Clinton’s emails. Politicians, analysts and journalists are still debating whether the letter cost Mrs. Clinton the presidency. It’s certainly possible. But I am not at all sure, in part because of the final Upshot/Siena College poll in Florida. I had learned the results of our survey that morning. It showed Donald J. Trump ahead of Hillary Clinton in the state by four percentage points, 46 percent to 42 percent. At the time, the poll looked like a bust. There wasn’t much reason to think the result was even in the ballpark. Mrs. Clinton was ahead by six points in national polls and ahead by a similar margin in states worth 270 electoral votes, suggesting Mrs. Clinton was probably up by a few points in Florida. I can’t say I was thrilled about having to write an article about a poll that looked flat-out wrong. But it’s now clear that Mrs. Clinton was weaker heading into Oct. 28 than was understood at the time. Several other polls were conducted over the same period that showed Mr. Trump gaining quickly on Mrs. Clinton in the days ahead of the Comey letter. And the timing of these polls — particularly the gap between when they were taken and when they were released — has probably helped to exaggerate the effect of Mr. Comey’s letter on the presidential race. The case for a big Comey effect hinges on the large decline in Mrs. Clinton’s lead in the polls that followed the letter. In the FiveThirtyEight model, for instance, Mrs. Clinton had nearly a six-point lead heading into the Comey letter, but just a three-point lead one week later: an apparent three-percentage-point shift against Mrs. Clinton. It seems reasonable, as many have argued, to attribute much of that decline to Mr. Comey’s decision. But the Upshot/Siena poll of Florida is one of several surveys that challenge this interpretation. That poll was completed the night before the Comey letter, but it was not released until Sunday, two days later — a longer lag than usual, since Sunday is seen as a better day for news media coverage than Saturday. Some analysts have used poll aggregators or forecasting models to measure the effect of the Comey letter, and they have implicitly treated this Upshot poll, and others conducted before the news but released after, as evidence of a Comey effect. But it can’t be; for example, none of the people we polled for our survey knew about the letter. Unfortunately, there is not much polling from this narrow period before the Comey letter and well after the third presidential debate. But it was accepted at the time that Mrs. Clinton’s lead was slipping heading into the morning of Oct. 28. The ABC/Washington Post tracking poll conducted over the same period as the Upshot/Siena poll of Florida, for instance, showed Mrs. Clinton’s lead at just two points, down from a double-digit lead after the third debate. That poll was also released after Mr. Comey’s letter. Most important, the polls taken before the letter were as bad for Mrs. Clinton as those conducted after it. Again, there aren’t many of these polls, but taken at face value there’s a case that Mrs. Clinton had nearly or even completely bottomed out by the time the Comey letter was released. Even if she had not, the trend line heading into the Comey letter was bad enough that there’s no need to assume that the Comey letter was necessary for any additional erosion in her lead. These polls are consistent with an alternative election narrative in which the Comey letter had no discernible effect on the outcome. In this telling, Mrs. Clinton had a big lead after the third presidential debate, when the ABC/Washington Post poll opened with her ahead by 12 points and an Upshot/Siena poll of North Carolina gave her a seven-point lead. But her advantage dwindled over the following week, as post-debate coverage faded and Republican-leaning voters belatedly and finally decided to back their traditional party’s nontraditional candidate. Even if the Comey letter did affect the race at that point, the effect might have faded in the final days of the campaign. Mrs. Clinton’s national lead in the polls grew over the weekend ahead of the election. In retrospect, there is virtually no evidence to support the view that Mrs. Clinton really had a six-point lead by Oct. 28, even if it was a very reasonable interpretation of the polls that had been released to that point. She didn’t have a six-point lead in any of the 16 (sometimes low-quality) national surveys that went into the field on or after Oct. 23 and were completed before the Comey letter, including her steadily shrinking lead in the ABC/Washington Post tracker. A new report from the American Association of Public Opinion Research on 2016 polling reached a similar conclusion. This doesn’t mean that Mr. Comey didn’t or couldn’t have played a pivotal role. The fairly sparse polling makes it hard to be sure of just how much Mrs. Clinton’s standing fell before the Comey letter. Maybe our Florida poll was a dud after all. Mr. Trump won the state by only a point, although many of the trends evident in our poll — like lower black turnout, a less-than-record-setting showing by Mrs. Clinton among Hispanic voters, and Mr. Trump’s surge among Republican and white working-class voters — held true on election night. It’s hard to rule out the possibility that Mr. Comey was decisive in such a close election. Mr. Trump won Michigan, Wisconsin and Pennsylvania by less than a percentage point. Even if there were no evidence to support a shift after Mr. Comey’s letter, there would still be reason to wonder whether his actions were decisive. The story dominated the news for much of the week before the election. One could imagine how Mr. Comey’s letter might have swayed voters who remained undecided heading into Election Day. But in such a close election, anything and everything could have plausibly been decisive. |