This is the fourth article in a series that reviews news coverage of the 2016 general election, explores how Donald Trump won and why his chances were underrated by the most of the American media.
I’ve contended in this series that national news outlets haven’t candidly explored the reasons they were so sure Hillary Clinton would become president. Blaming polls or “data” is something of a red herring given that the data told a complicated story. Perhaps the closest thing they’ve offered to an actual explanation is that the failure reflected a lack of shoe-leather reporting. One day after the election, for instance, New York Times Executive Editor Dean Baquet said that the biggest flaw in his paper’s 2016 coverage was in not having enough reporters “on the road, out in the country, talking to different kinds of people than the people” they usually talked to.
There are versions of Baquet’s idea that I agree with. The Times and other outlets would have been better off if they’d put less emphasis on the insider’s view of the race, as articulated by campaign consultants in New York and Washington. Better yet would be if local and regional news outlets, which face increasingly difficult economics, had more of a place in the national conversation, serving as a potential check against groupthink in the Northeast Corridor. The voters who proved decisive in the Electoral College were mostly in “flyover country” instead of on the coasts.
But if the idea is that reporters should have spent more time talking to people in the field as opposed to looking at the polling, then I’m wary — and I don’t think it goes very far in explaining why reporters failed to foresee a Trump presidency. Partly this is because it can be easy to misread the vibrations on the ground, as the Times did in 2012 when it unironically cited “yard signs on the expansive lawns of homes in the well-heeled suburbs, and … the excited voices of Republican mothers” as reasons to think Mitt Romney would win Pennsylvania. But more importantly, polling and ground reporting ought to be compatible. Reporting can provide context and depth and nuance, and we’ve done more and more field reporting ourselves at FiveThirtyEight, having sent reporters to such places as Ashtabula County, Ohio, and Presque Isle, Maine. But polls are also a way of “talking to human beings,” only in a relatively unbiased fashion that potentially reaches a more representative sample of voters than a reporter probably could.
Instead the problem was in a failure to draw a connection between Clinton’s problems in the heartland — which were evident both from polling and from reporting — and her vulnerable position in the Electoral College overall. Take, for example, this Sept. 29 Times article, which correctly noted that Trump was running strongly in Ohio. Shouldn’t it have been a bad sign for Clinton that she was faring so poorly in such a traditionally important state? Not really, the article asserted, instead reassuring readers that Clinton’s more diverse coalition give her a greater number of paths to 270 electoral votes:
And the two parties have made strikingly different wagers about how to win the White House in this election: Mr. Trump, the Republican nominee, is relying on a demographic coalition that, while well tailored for Ohio even in the state’s Democratic strongholds, leaves him vulnerable in the more diverse parts of the country, where Mrs. Clinton is spending most of her time.
It is a jarring change for political veterans here, who relish being at the center of the country’s presidential races: Because of newer battleground states, Mrs. Clinton can amass the 270 electoral votes required to win even if she loses Ohio.
It was fine to point out that Ohio, which is generally somewhat Republican-leaning relative to the country as a whole,1 wasn’t a must-win state for Clinton. But the article didn’t contemplate the possibility that Clinton’s poor position in Ohio could also portend problems in Michigan, Wisconsin and Pennsylvania, which probably were must-wins for her and which, like Ohio, had plenty of white voters without college degrees. In that sense, Ohio could still be a bellwether — a leading indicator of trends elsewhere in the country and the region — even if it wasn’t likely to be the decisive state.
Instead, the Times went out of its way to assert that Ohio had lost its bellwether status, describing it as a bygone curiosity “decreasingly representative of contemporary America” whose days at the forefront of presidential politics were behind it. The Times’s deeply held conviction that the electoral map favored Clinton was impervious to whatever evidence its reporters were picking from the Buckeye State.
However, the Times’s error paralleled a major mistake made by some of the statistical models of the campaign. That’s in the tendency to treat states as being independent from one another when in fact their outcomes are highly correlated. You can make a lot of inferences about how people might vote in Michigan and Pennsylvania based on the polls in Ohio, for instance. Moreover, when the polls are in error in one state, they tend to miss in the same direction in similar states. Clinton would have to be awfully unlucky to lose Ohio and Pennsylvania and Michigan and Wisconsin, the models were saying, not recognizing that problems in any one of these states probably implied she was having problems with them across the board. Get that one assumption wrong, and you could mistakenly calculate that Clinton’s win probability was in the high 90s instead of the 70 percent range, where FiveThirtyEight’s forecast had it.
Furthermore, like their more traditional counterparts, data-friendly journalists were sometimes too quick to dismiss the implications of Clinton’s awful polling numbers in states such as and Ohio and Iowa. If Clinton was really trailing Trump by 7 percentage points in Iowa, as the final (usually highly accurate) Des Moines Register poll had said, how safe could her position be in nearby states like Wisconsin?
It would have helped to have more high-quality polls in other swing states such as Wisconsin and Michigan. (Traditional telephone polls commissioned by local newspapers have dwindled because of budget cuts.2) But based on the data the public did have available, there were a lot of Electoral College warning signs for Clinton, especially after FBI Director James B. Comey’s letter to Congress on Oct. 28 — and people mostly ignored them. Clinton might have been able to win the Electoral College without Ohio, but not without the Midwest.