Editor’s note, 11/3/20: This article was last updated on November 6, 2018, and some details may be out of date. For our most recent coverage of the US election, visit our 2020 election hub.
The New York Times clearly thinks of its famous election needle, the speedometer-like readout that predicts the results of the midterm elections as vote totals trickle in, as a scientific instrument.
But millions of political junkies — who either have a healthy concern for the fate of the American republic or an acute inability to chill and patiently accept some uncertainty, depending on whom you ask — turn to the needle in the same way people addicted to intravenous drugs would: as a fix.
The NYT election needle debuted in 2016, an election cycle in which sophisticated modeling of opinion polls (which went mainstream in 2012) had become so accepted that it seemed totally reasonable to use numbers to predict the future. Nate Silver of FiveThirtyEight and Nate Cohn (who, at the Times, is one of the needle’s masterminds) sniped at each other about the assumptions built into their models; everyone got really huffy about correlated errors for a few days. Of course, this all seemed like a good idea because it seemed like Hillary Clinton was definitely going to win ... and when she didn’t, the shock tarnished a lot of liberals’ (and reporters’) faith in polls at all.
The needle was born of that Trump-era agita. It’s appeared five times, and each has inspired waves of excitement and nausea among a certain type of extremely online politics nerd. Instead of a static certainty born out of polls done before the election, or the slow plodding of a live results page, it jitters constantly as new information comes in — inspiring the confidence in math that one expects from 21st-century political hobbyists, and the violent mood swings of the most hardcore partisan.
A lot of the emotional response to the needle has to do with the circumstances of its debut: an election that Clinton was heavily favored to win but that Donald Trump managed to pull out, with the needle inexorably tracking the way the candidates’ fortunes changed. Many people’s memories of the moment they realized Trump was actually going to be America’s 45th president look something like this:
I mean this literally. That image is a screenshot of the needle Vox staff captured on the night of the 2016 presidential election.
The Times has continued to tweak its model and display for the needle over the past two years, and to explain (repeatedly) how exactly the needle works and why you shouldn’t develop too much of an attachment to its predictions.
But it is far, far too late. Making predictions is the entire point of watching election night returns live — and getting attached to those predictions is an inevitability for anyone who is watching the returns because they care about the outcome. Election night in America has become an emotionally fraught communal experience. The needle doesn’t provide a check on that emotion; it can merely channel it into partisan joy or equally partisan despair.
The New York Times election needle is a way to fill in the dead air of election night with science instead of punditry
The needle is a new(ish) solution to a problem that’s existed since the dawn of the election night broadcast: The biggest news story of any Election Day is the outcome of the elections, but we don’t actually know the outcome of the elections until the wee hours of the morning on Wednesday. (At best. If control of the House comes down to a handful of races in California, we might not know for weeks.)
That is fine for print newspapers, but TV and internet outlets have to fill several hours between the closing of the polls and the decision on the final outcome. And there are tons of people — way more than traditionally pay attention to news on a Tuesday night, or to politics at all — paying attention during those hours.
The election night audience really only wants to know one thing: the results. That is the one thing the media cannot provide. Instead, media outlets have found a lot of ways to talk about what appears to be happening as vote counts keep rolling in slowly across a district or state.
The New York Times’s needle is one of those strategies. The Times’s best needle explainer/apologia, from March 2018, compares the needle to the work that political analysts do as the precinct data rolls in.
As results are tallied, these analysts identify surprises where votes that have already been counted — a larger or smaller than expected turnout in a given area, an unexpectedly strong or weak showing by a particular candidate. They then identify the areas to watch where votes haven’t been counted yet where the trailing candidate will need to do particularly well. (If you have ever heard a joke on politics Twitter about “crucial Waukesha County,” congratulations, you now understand it, kind of.)
But instead of having a human perform this analysis for the audience, the needle performs it. As the Times wrote:
Our live forecast is just a formal means to do for online viewers what analysts like [MSNBC’s Steve] Kornacki or [CNN’s John] King have been doing for television viewers for years. It looks at where votes remain to be counted, and makes an educated guess about how those votes will break based on past election results and trends evident in initial returns.
The Times, characteristically, understates the needle’s appeal. It is free and accessible (the Times takes down its paywall on Election Day) and doesn’t require a cable subscription; it is constantly on display and near-constantly updating; it allows you to understand who’s winning at a glance (without having to turn on the sound or gain an encyclopedic knowledge of Georgia political geography).
Most importantly, in an age of media distrust, it is a lot harder to accuse the needle of bias than it is to accuse a human being.
Americans often say they want the straight facts, or that they want to discover things on their own, because they don’t trust any intermediary not to have “an agenda.” But when they get the raw information, they rarely have the expertise they’d need to figure out what it means.
The needle represents fully digested election information — a bewildering array of numbers collapsed to a single point. But because that work happens behind the scenes, it allows people who prize their independence of thought to believe they’re receiving “just the facts.”
The needle’s motion is generated by using actual votes to make predictions about expected ones
If you’re wondering how the needle works, you should probably let the Times explain that to you. But if you insist:
The needle predicts the likelihood of a given election outcome by using information from the votes that have been counted to try to predict the votes that haven’t.
The model that drives the needle is programmed with a set of assumptions about what will happen in each district or county it’s tracking: both how many people will vote in total (turnout) and how many of them will vote for each candidate (vote share). At the beginning of the night, when not many votes have been counted, that’s what the needle is mostly working off: those assumptions.
Obviously, as votes start coming in for a given area, the model can pivot from assuming how that area will vote to knowing how it will vote. But the first areas to come in aren’t necessarily representative — in general, for example, large urban areas will count votes more slowly, which could present an illusory Republican lead.
The needle model addresses this by comparing its assumptions about turnout and vote share to the reality in places where votes have already been counted. It then updates its assumptions for similar areas that haven’t counted votes yet — assuming that if it underestimated turnout for the Democrat in one rural blue-collar precinct, for example, it’s probably underestimated turnout for the Democrat in other rural blue-collar areas of the state.
Maybe that’s wrong; the second-biggest concern the Times’s needle masters have about it, they write, is that the first precincts to report will just be unrepresentative, period. That means the needle will try to fix something that isn’t broken.
The biggest concern the Times has, in case you’re wondering, is that some states report early voting when they release their first day-of votes, but early voting tends to skew Democratic. So it might be hard to tell a blue wave on Election Day from a phenomenon in which Democrats simply voted early and Republicans turned out in greater force by the end.
The needle can’t “get it wrong” — but it can mislead
The needle isn’t like the polling averages of Nate Silver or other analysts, who made final predictions about the vote outcome based on polls taken before the election. It is based on the final vote tallies — once the last vote is counted, the needle will settle at maximum certainty on the side of the victor (if it’s still on display by then).
But the point of the needle isn’t to watch where it settles. It’s that in the midst of election returns, while everything else is just noise, the needle points to where we will most likely end up. The point of the needle is to make viewers feel they know the outcome even when the outcome is unknowable.
Over the course of the night, the needle might vary widely in its assessments. That’s less of a problem for the main needle, which simply measures likelihood, than for the needles that project what the margin of the total vote will be and how many seats a party is likely to pick up. In both of those cases, most of the predictions the needle makes will be wrong, and it’s just a question of how wrong they’ll be.
The Times measures “wrongness” by magnitude — their biggest error, according to Cohn and Josh Katz, was the 2017 special election for US Senate in Alabama, when for five minutes they estimated that Democrat Doug Jones would win the popular vote by 8 points and he ultimately won by 1.5.
But the public, as the Times noted, treated the Jones call as a victory — because the Times had correctly predicted that Jones would win, even as Republican Roy Moore was way ahead in the actual vote tallies.
Pollsters and political scientists care about margins of victory and margins of error. But to most of the Americans watching, what matters is the binary result. When people expect their candidate to eke out a win, and they lose barely instead, that feels like a betrayal of expectations much more than an unexpectedly close win does.
This is why it’s impossible to talk about the needle without talking about 2016. The needle didn’t get the 2016 election wrong — to the contrary, it pulled to Trump’s side even as the vote totals were still showing Clinton ahead. It’s that a lot of people watching real-time vote results saw the needle tilted toward Clinton for the first few hours of the night, and then pull, seemingly inexorably, toward Trump, And it felt like an admission of error.
The needle is just the name we give to the things we choose to do together
There is no particular reason anyone would need to watch live election results at all.
By noon Wednesday — long before anyone elected on Tuesday has actually been sworn in — someone who stays up until 2 am Eastern and crashes while the votes are still being counted in El Paso will not actually know more than someone who went to bed when the polls closed in Kentucky and woke up the next morning to read about the results.
As a matter of fact, the second person might know more — not just because they’re better rested, but because they haven’t been misled by illusory trends. In 2016, the first person might have fallen asleep with the sense that Donald Trump truly had captured the hearts of the country — while the second person would be able to look at the popular vote total and note dispassionately, without the hours of agony, that Hillary Clinton remained ahead in the popular vote.
But election night is a ritual, and that means it’s not really about information. It’s about the communal experience of everyone finding out information together — the same way a live sporting event or a Twitter-favorite television show is.
And more importantly, as many people have started tying their own identities more closely not just to politics per se but to one of the two major political parties, election night really can feel like a glimpse into the national soul. Rejoicing, supporters of the winners might feel that America is living up to its highest ideals; processing a defeat, their opponents might wonder if they ever really understood their country to begin with, or start identifying who’s to blame and spend the next several years villainizing them.
Nothing about this is emotionally healthy. But policy matters, and policy is enacted through politics, and so it’s impossible to tell people to care less. On election night, the only thing that someone who cares a lot can do is sit and wait for results to come in — and seek out the comfort of a needle that will tell them how this ends.