What We’ve Learned From 2016

A Look Back At Election Forecasting

A German paper reporting on the expected Clinton win that didn’t happen

It’s been about a month now since the shocking night of November 8th, 2016. It started out with a mix of anticipation and expectations. Polls told us it would be a status quo year that favored a modest Hillary Clinton victory. But things started to suddenly shift at about nine o’clock. By midnight, it was clear the race would be more of a 1960/2000 style result than the 1928/1988 type result most folks were expecting (me included). By three in the morning, shock gave way to numbness as the political world found out Donald Trump had pulled off the biggest upset in Presidential politics since Harry Truman’s infamous 1948 stunner. And he did it by winning states like Michigan and Wisconsin that no one had on their radar as possible flips.

The initial aftershocks seemed to be more focused on the big polling miss. As someone who wrote electoral outlooks (Articles No Longer Exists) on here and had an electoral data site based on what polling was telling us, I was one of the most surprised watching the vote tallies come in. Data failed this year and in turn I failed as well in seeing this impending result. For that I apologize. Almost everyone involved in the data world, even well taught amateurs like me, missed it.

While there were some models that told us Trump would win, they all predicted he’d win the popular vote — which he lost. Most forecasts were right that Hillary would win the popular vote, but were off by nearly 100 electoral votes either way on the electoral college. Even generally accurate measures like forecast aggregation, citizen based forecasts, and betting markets missed it. Heck, even the Scholastic mock-election involving kids was wrong for the first time since 1960.

And no, I don’t think it’s fair to claim Trump supporters or journalists that were writing about those supporters were the ones who saw this. Let’s be frank, a lot of these supporters were the same ones who expected a hidden Romney vote in 2012 that never came, and some of these journalists were trying to push upset narratives long before Trump came along.

The closest thing we have to what I consider a legit warning about what we saw happen on November 8th, 2016 was Sean Trende’s article back in July 2015 on the possibility we were overdue for another 1948 style upset. When it comes down to it, this was a legitimate upset of the likes that we may not see again in our lifetimes, depending on the future of polling.

What makes the miss so frustrating for data geeks like me isn’t just the fact we couldn't cheat our way to knowing who’d be President before the votes came in for the seventh decade in a row but that we have little clue as to exactly why the result was what it was. We can speculate of course, but we won’t have as much data as hard evidence like we have in past races. Maybe Trump was always the one to beat? Maybe Hillary really let it slip away late? Maybe the race was always close and down to the wire from start to finish? We’ll never know, we’ll just speculate and attach our pet causes to the result.

What we can deduce is this, national polls were actually not too bad and exits confirmed that findings like President Obama’s rising popularity in the last year did happen (53% approval according to national exits with actual voters). State polls were good in some critical swing states like Colorado, Nevada, New Hampshire, or Virginia; but the misses in places like Florida, North Carolina, Michigan, Pennsylvania, and Wisconsin threw off the electoral college forecasts.

This election has set certain precedents: Given the right perfect storm of things, a deeply unpopular candidate who alienates minorities and women can achieve victory. Pure celebrity can get you elected to the Presidency now, a George Clooney 2020 run suddenly sounds much more serious than it would have four years ago, ditto a Kanye West run. Maximized rural and suburbs performance can help offset the heavily populated urban areas, even as the white share of the electorate gets smaller. Every state that is within single digits in polling should be taken into consideration as a tossup even if most polls tell us a state is leaning one way.

This election also showed us some interesting trends: Minnesota still continues its streak of refusing to vote for a GOP nominee even when that nominee can appeal to the rust belt. The GOP continues to have a tough path to the White House; since the 80’s they’ve only won narrowly. Though Trump was able to knock down the Rust Belt’s “blue wall”, he did worse than Romney in Arizona and Texas and lost Colorado, Nevada, and Virginia. The map is changing on both sides and you should expect the next Democrat President to be elected with a new state or two to their coalition.

So where does election coverage go from here? For my part I shut down my election data site. I will probably dip my feet back into the elections analysis business sometime in the future, but not as a hard predictor. Those who decide to stick around in the prediction game are going to have to learn to be more humble and cautious with their predictions until polling begins a new streak of correct calls.

I am cautiously optimistic that pollsters will learn from this like like they did in 1948. Some online pollsters did catch a much closer then expected race in the Rust Belt, so there’s a start. But even if 2016 is an outlier like 1948 was, polling will take decades to ever gain back the trust and confidence it used to have. Though there remains the possibility that we are in for more polling misses that are common, like in the UK.

Ultimately the big lesson about this election is that we are a polarized and divided country. We have close Presidential races where winning the popular vote is not enough. We are divided by class and race to the point our politics have become dangerously tribal. Until we learn to unite and cast less blame on one another, this division and polarization that has gotten stronger the last two decades will only grow. We now live in an era where revenge can be just as much a motivator as inspiration to the casual voter and that is is the most depressing lesson from 2016 of all.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store