A Lot of People Owe Nate Silver an Apology

Matthew Black


America may be a highly polarized country, but there does seem to be one thing that everyone can get on board with: bashing on Nate Silver. Although he is not well-known outside the political world, those who closely follow US politics tend to have strong feelings about him, most of which are undeserved.

Hillary’s defeat

Back in 2016, Nate Silver got a lot of flak from liberals because his website, FiveThirtyEight, consistently gave Hillary a lower chance of winning than most of the other election forecasts. Whereas other forecast models gave Hillary a greater than 90% chance of winning, by November 8th, FiveThirtyEight had Hillary at only 71%.

This led to a ton of backlash among liberal circles against Nate Silver, accusing him of being a needless contrarian, of being a fearmongerer, of turning his site into clickbait.

Nate Silver’s reasoning for his bearish take on Hillary’s odds was pretty simple: Hillary was in the lead in the polls, but all it would take was a normal-sized polling error to cost her most of those swing states. There was no real way to tell if a polling error in Trump’s favor was going to happen, but if it did happen it would likely be correlated in other states with similar demographics. If the polls were off in Pennsylvania, they’d most likely be off in states like Michigan and Wisconsin as well.

As Nate Silver himself has stated, over and over again, a 3-4 point polling error is extremely common. A poll is much more likely to be off by a couple of points than to be spot on, and he took that into account when creating his model. That is why Nate insisted that although Hillary was still a favorite, Trump very well could pull off a victory.

And sure enough, the scenario that Nate Silver spent months warning us could happen ended up happening: Hillary won the national popular vote by about as much as the polls predicted, but there was just enough of a correlated polling error in certain swing states to give Trump the lead.

And yet, all those people who bashed on Nate for fearmongering did not apologize. They did not say, “Oh, geez, I shouldn’t have been so overconfident. Maybe I should’ve listened to Nate Silver when he warned us every single day for months about the likelihood of a systemic polling error.”

Instead they spent the next four years bashing Nate Silver for “getting it wrong.”

If you look at any of Nate’s tweets from November 9th, 2016 to November 3rd, 2020, you will inevitably see a barrage of responses saying something like: “When are you going to apologize for 2016?” “Why can’t you just admit that you were wrong in 2016?” “How can we trust you after 2016?”

Somehow, the narrative surrounding Nate Silver and FiveThirtyEight flipped overnight. Whereas before, people spoke of Nate Silver as someone who was giving Trump way too high a chance, now he was lumped in with all the other over-confident pollsters who said there was no way Trump could possibly win.


Biden’s victory

Then the 2020 election came, and this time FiveThirtyEight gave Biden a 90% chance of winning, which was a little lower than most other forecasts gave him. Nate Silver’s reasoning this time was similar to his reasoning in 2016: the polls look good for Biden, but there was a significant chance of another polling error in Trump’s favor. However, Biden was doing well enough that even a 3–4 point polling error wouldn’t be enough to cost him the election.

And sure enough, that’s exactly what happened. There was a polling error in Trump’s favor, but not a big enough polling error to let him win. Did Nate Silver get any credit for predicting this outcome? No. He was, yet again, bashed for “getting it wrong.”

Today’s narrative among a lot of circles is that Nate Silver “predicted” a “Biden landslide,” which is pretty objectively not true. Nate Silver claimed that a landslide was possible for Biden, but he also spent every damn day on his podcast and in his articles stressing the fact that there could easily be another polling error in Trump’s favor.

It’s clear, at this point, that there are a ton of misconceptions around Nate Silver, and no matter how much Nate himself tries to clarify, his detractors are not interested in listening. There is a reason why so many people try to discredit Nate Silver using bad faith tactics, but first, let’s make it clear what it is Nate actually does.

Nate Silver is not a pollster

Perhaps the biggest misconception is the idea that Nate himself is the one conducting all these polls. That would explain why people would think back to all those times they saw a poll with Biden leading the national vote by double digits and blame Nate Silver when Biden only wins by 4 to 5 percent.

Nate did not make the polls telling us that Biden would win in a landslide.

Rather, Nate (and his team at FiveThirtyEight, of course) look at all the professional polls being run throughout the country, average them together — giving extra weight to the polls they deem more reliable, and less weight to the polls they deem less reliable — and then try to guess from that what the most likely outcome will be.

This is a simplified version of what actually happens at FiveThirtyEight — there’s a whole complex algorithm they’ve got that would take hours to explain — but it’s very important to understand that Nate’s job is not to tell you who will win. His job is to look at all the data available and tell you what the odds are of any given event happening. He doesn’t tell you what will happen; he just tells you the odds.

From percentage points to odds

One of the big changes the website made after 2016, utilized mainly in the 2018 midterms, was to express their data in odds rather than percentage points.

Rather than saying that a candidate had a 83.3% chance of winning, they would say that the candidate had a 5 in 6 chance of winning. They were still a heavy favorite, but the odds of the underdog pulling off a victory was about the same as rolling a die and landing on a one.

People have no trouble understanding this concept in a nonpolitical context, even when the odds are much slimmer. The odds of winning the Mega Millions jackpot is somewhere around 1 in 300,000,000, yet whenever news comes out of somebody winning, nobody goes, “Wow, what the hell, the stats lied!” People understand that unlikely events happen all the time.

Although FiveThirtyEight went back to percentage points in 2020, they still made sure to express Biden’s chances in those kinds of terms. They compared Trump’s odds of winning to be slightly higher than the odds of it raining in LA on any given day. Describing probability in these terms tends to help people understand FiveThirtyEight’s purpose a little better.

Polls themselves do not claim to be 100% accurate

When you see a poll that says something like “X candidate is at 52% of the vote, Y candidate at 47%,” remember that the poll isn’t claiming that X will definitely win 52%. They’re claiming that as of the time in which they conducted the poll, they’ve estimated that 52% of the voting population (plus or minus several points) would vote for X, and 47% of the voting population (plus or minus several points) would vote for Y.

They are not claiming that this is how the election day results will turn out. They’re saying that this is what the electorate looks like like at the current moment, with the understanding that things can change fairly quickly. They are not trying to trick you into thinking they have all the answers, nor are they trying to “predict” what will happen on election day.

Nate Silver doesn’t base his estimates around his political preferences

As someone who reads his articles and listens to his podcast, my guess would be that Nate’s political beliefs are left of center. However, it seems pretty clear, after listening to him for four plus years, that his political beliefs don’t factor in at all to the way he runs his model.

This does not stop people from all over the political spectrum from claiming bias against them. During the 2020 primaries he was accused of being biased against Bernie for asserting that Bernie having a heart attack wouldn’t be good for his poll numbers, or that Bernie would have trouble in the primary reaching voters outside his current base. Both of these claims turned out to be 100% accurate.

He was accused of being a Biden shill during the primary, because he repeatedly asserted that Biden had maintained a large, consistent lead throughout the entire primary while the other candidates constantly rose and fell.

He was accused of being anti-Hillary in 2016 because he gave her a lower chance in the polls than others did, and he was accused of being anti-Trump throughout the 2020 election because he pointed out how Trump’s notoriously low approval ratings would hurt his re-election chances.

People of all political stances seem to get defensive when it comes to the topic of their preferred candidate’s electability, which gives Nate the unfortunate position of being attacked by that candidate’s supporters every time the stats don’t look good for that particular candidate.

This is an understandable impulse, and there are some legitimate criticisms to be made of the way discussions around electability can itself affect a candidate’s electability, but we really need to learn to resist the impulse to kill the messenger.

The fact is, Nate Silver has no incentive to tell you anything other than the facts as he sees them, and he has every incentive in the world to be as accurate as possible. His career depends on it. If he says your candidate isn’t doing well, he is not doing this to mess with you.

Why Nate’s work matters

When you point out that low probability outcomes happen all the time, that inevitably leads to the question of “what’s the point?”

If FiveThirtyEight gives a candidate a 10% chance of winning and that candidate ends up winning, that doesn’t prove FiveThirtyEight was wrong to give them a 10% chance. But if it doesn’t, then what does? What’s the point of an election model that doesn’t take a clear stance in telling you what exactly is going to happen?

For political junkies, the point is to give us an idea of what to expect going into election day. For campaign advisors, the point is to give them data in which to base their campaign strategies. And for people who care about the health of American democracy, the point of FiveThirtyEight is that it helps reporters and election officials sniff out voter fraud.

Keeping track of polling data helps us to find voting irregularities, so if voting fraud does happen we’ll have an easier time finding it.

This is why Trump supporters and pundits in particular seem to be working so hard to dismiss people like Nate Silver. It’s not just to protect their egos; it’s to discredit the results of the election, and to discredit the results of future elections that don’t work in their favor.

If people don’t trust sources like FiveThirtyEight at all, it becomes easier for bad faith actors to make baseless claims of massive voter fraud. A significant number of Trump supporters may believe the election was stolen from them, but those claims don’t hold up among the majority of Americans because there are well-documented databases like FiveThirtyEight that show that the 2020 results were in line with expectations.

Meanwhile, if Trump had ended up winning the popular vote by several points, the sheer unlikelihood of that event would’ve tipped reporters and investigators off to possible voter fraud. People would have an easier time uncovering the fraud if it existed, and the media would have an easier time convincing people of it, because there was a well-known, trustworthy source that showed how unlikely this event was to happen without some kind of interference.

This is why what Nate Silver does is so important. It’s why flinging baseless attacks at him every time the polls are slightly off, or whenever he interprets the data in a way you disagree with, is counterproductive at best and dangerous to democracy at worst. Nate isn’t here to tell you exactly what’s going to happen. He’s here to tell you the range of possibilities within a fair, free election. Don’t expect anything more than that.

Comments / 0

Published by

I write thoughtful articles about politics, TV/film, and internet culture.

Binghamton, NY

More from Matthew Black

Comments / 0