The lessons for political journalists from the presidential election of 2016

By J.P.P. | NEW YORK

SINCE Donald Trump beat Hillary Clinton, there has been a series of “why-oh-why didn’t the mainstream media see this coming” articles, and almost an equal number of attempts by journalists to claim that they had, against much available evidence. What follows is not an attempt to add to those, but rather to look at what we political journalists might do differently next time.

The role of the polls
In the final issue before the election, The Economist ran two stories about different demographic groups. The first pointed out that there was a widespread feeling among blue-collar whites that the odds of success in America in the 21st century had become stacked against them, and that this was accompanied by a belief that Mr Trump could tilt those odds back in their favour. The second described low levels of enthusiasm for Mrs Clinton among African-American voters. Had you read these two stories and nothing else, you would have concluded that Mr Trump was going to win. Yet that was not the impression we gave in our coverage overall.

Why was that? It was not because our reporters are out-of-touch, or because they misunderstood America’s mood. The Economist believes in shoe-leather reporting: our correspondents spend much of their time in those parts of the country that mainstream media is supposed to ignore. Instead, it was because they did what good political journalists are supposed to do. When talking to potential voters, reporters are always aware that the sample of views in their notebook is small and unrepresentative in all sorts of ways. So they check what they are hearing against the much larger samples of people that pollsters talk to. This is good journalistic practice. It also turned out to be unhelpful in 2016, because the polls in Pennsylvania and the Midwest were so wrong. Next time we will probably give more weight to what is in our notebooks and less to what the polls say. That approach in turn might misfire. What’s right for one election can be wrong for the next.

Guessing at probabilities
The election-forecast models that assigned overwhelming probabilities to a victory for Mrs Clinton have received some criticism, and deservedly so (at The Economist we resisted the temptation to build our own forecasting model this election cycle). There is a misunderstanding about what such models were saying which their makers failed to make clear enough. When you roll a dice, the probability of getting a six really is one in six, so long as that dice is true. When an election model says that the odds of a Trump victory are one in six, this is not an actual probability. It is merely a guess at what the probability is. These are fundamentally different animals and should not be confused. Guesses at probabilities are of limited usefulness when dealing with one-off events. The data-journalism moment in political writing grew out of sports statistics. In basketball you can have a good idea of the chance that a particular player will score from a shot taken behind the three-point line, because there are hundreds of past data points to use and because he will probably try it again in a few minutes. Presidential elections are not like that. They happen only once every four years and, to extend the basketball metaphor, the court changes size and shape from one to the next.

When writing about politics, knowing what happened in the past usually gives you an advantage. Very occasionally, though, it is a hindrance. The Democrats have a blue-firewall in the Midwest until all of a sudden they don’t. If you make a habit of predicting that there is about to be a break with the past you will be wrong most of the time. If you pay attention to what happened previously you will be right more often—and then spectacularly wrong if you find yourself alive in in 1778, 1847 or 1913. At its most self-confident, predictive political data-journalism reminds me of those Enlightenment thinkers who believed that the universe functioned according to a set of rules that could be discovered with enough study. It doesn’t.

The ones who didn't vote
Voter anger, particularly among whites without a college degree, was among the most-covered stories of 2016. In retrospect, I wish we had paid more attention to apathy. Anyone who wrote about this election believed, probably rightly, that it was the most consequential election of their lifetime. Yet about half of eligible voters believed it didn’t matter enough for them to find their way to a polling station. Some of them are probably protesting about the election result right now.

All political reporters know how complicated what goes on inside the brains of individual voters is. One afternoon during Florida’s Republican primary I stood outside a polling station in West Miami talking to people who had just cast ballots. An elderly Cuban émigré told me he wanted Mr Trump to win but had voted for Ted Cruz. Asked how he got there, he explained that America needed a dictator, which is why he supported Mr Trump. But he didn’t want to blame himself if it got one, hence the vote for Mr Cruz. Another voter, who worked for the federal government, explained that she was voting for Marco Rubio, a noted small-government conservative, because she was fed-up with cuts to the federal budget. Non-voting is probably just as complicated. Next time I want to read more about non-voters. They, not the 60m who voted for Mr Trump, are the silent majority.

Our first duty
As an analyst of the horse race, my own biggest failure was to assume, based on both reporting and polling, that more college-educated white women would abandon Mr Trump’s Republican Party. In the end he won 45% of them. I still do not really understand what happened there. Saying that Mrs Clinton was an uninspiring candidate without a compelling message doesn’t quite cover it. The main responsibility of political journalists, though, is to write about the candidates, their proposals and their fitness for office. Guessing the outcome of an election is a secondary activity.