On the night of the 2016 U.S. presidential election, the New York Times introduced the “needle,” a data visualization reminiscent of a car’s speedometer.
This tool aimed to inform readers about the potential implications of early vote counts on the final outcome. As election night unfolded, millions of Americans watched the needle with a mix of curiosity, anxiety, and dread. The Times’ needle indicated a likely Donald Trump victory hours before the Associated Press officially called the race for him.
According to Wilson Andrews, the Times’ deputy graphics editor who has overseen election results since 2016, the concept behind the needle is straightforward.
“We begin with pre-election expectations for how votes might play out across the country, down to the county or precinct level,” he explained. “We then create estimates based on polling data, historical results, and demographics. Our aim is to establish a benchmark for where we believe the election may lead in a particular area.”
The needle starts with these pre-election estimates. If activated before sufficient votes are counted, its forecast reflects these initial assumptions. However, as votes come in, the model adapts, adjusting its predictions based on real-time results.
For instance, if a candidate exceeds expectations in a given county, it may indicate that they will perform similarly in comparable areas. Thus, the model continually processes insights, prompting the needle to move throughout election night.
“That’s the essence of the model,” Andrews said. “It transitions from a pre-election estimate to a real-time model that considers the votes counted and those remaining.”
**Will There Be Multiple Needles?**
Since its initial introduction in 2016, the election needle has evolved. In 2020, amid a record number of absentee ballots, the Times only published needles for three states—Georgia, Florida, and North Carolina—due to the variability in data reporting.
This year, Andrews and his team are still determining how many needles to deploy on election night. Ideally, they hope to forecast the presidential race alongside seven additional needles for key swing states, but this will depend on the availability of data during the election.
“Each state and county has unique processes, and we must navigate these effectively,” Andrews noted. “We will only publish forecasts when we are confident they will provide reliable information for our readers. We don’t release them just to meet expectations; we ensure we have sufficient data to offer a helpful guide on election night.”
Polls indicate that this year’s presidential race is incredibly close, a primary concern for Andrews and his team as they approach the final stretch.
The Times’ polling page emphasizes that the election could sway in either direction.
“We want to convey that polls are not infallible and can often miss the mark,” Andrews explained. “We’ve incorporated reporter insights alongside these graphics to enhance understanding. Our goal is to educate our readers on interpreting polls accurately.”
In 2020, Joe Biden was not declared the winner until four days post-election. Given the tightness of this year’s race, the Times is preparing for coverage that may extend beyond election night.
“2020 taught us valuable lessons about covering elections that take longer than one night,” Andrews remarked. “We are gearing up for 2024 as an election week rather than just an election night, investing in tools to explain the voting process more directly.”
Election results pages typically focus on live data displays, often lacking context and analysis. The Times aims to change this by integrating real-time insights from experts.
Context is essential. “We’ll inform readers about the anticipated data from specific states and the order in which it will be reported,” Andrews stated. “Our model will update in real time, guiding readers through the vote count process. Cable news has traditionally excelled in this area, and we strive to improve our support for readers.”
Andrews’ team is also working to ensure the needle remains functional beyond election night, providing a cautionary note to readers: “This is our way of indicating that early vote counts in a swing state should be viewed critically, as reported votes might skew in various directions. That’s the core message we aim to communicate.”
**What Underpins the Needle?**
Over 60 contributors are involved in crafting the forecast that will power the needle on election night. This diverse group includes traditional reporters, statisticians, software engineers, and designers who create the visual displays. They’ve dedicated the past two years to refining the data behind the needle, significantly more reporting than many might assume.
“It requires building strong relationships with election officials to understand what types of votes they expect to count and when,” Andrews said. “We gather data through both programmatic and manual reporting, supplementing the information we receive from the Associated Press. Our aim is to enrich this data with additional sources, including our Siena poll, detailing who is voting and how.”
The needle’s model relies not only on datasets but also on the diligent work of numerous reporters on the ground in key states.
“There’s a substantial amount of on-the-ground reporting, both before and after election day, that contributes to our success,” Andrews shared. “We want to provide insights into critical locations like Maricopa County or Philadelphia, especially if the election remains unresolved post-election night—areas that were pivotal in 2020.”
**Will the Times Call Any Races?**
Through the needle, the Times plans to offer a real-time model reflecting the state of the race throughout election night. However, it will not make race calls, a responsibility historically held by the Associated Press since 1848.
“It’s a crucial distinction,” Andrews emphasized. “The needle’s purpose is to help readers understand how the vote could unfold. We do not use it as a race-calling mechanism.” The Times primarily relies on the AP for race calls, monitoring competitive states and districts closely to decide whether to publish those calls on a case-by-case basis.
A notable instance occurred in Arizona during the 2020 election; the Times refrained from making a call because it lacked sufficient information to determine whether Trump or Biden would win. Consequently, they only announced the winner days later than other outlets.
“We prioritize accuracy,” Andrews stated. “Historically, we have not been the first to call results, and there is no internal pressure to be first. Our primary focus is getting it right.”
**How is the Times Preparing for Potential Challenges?**
With many state races too close to call, there is a real possibility that one presidential candidate may claim victory before all votes are counted, potentially framing the final result as fraudulent and attempting to seize power. This scenario makes the initial hours of vote counting particularly susceptible to misinformation, underscoring the importance of providing a nuanced account of real-time developments.
How does the Times plan to handle this? “We are producing pre-election guides detailing counting patterns in each state and incorporating that information into our results pages,” Andrews explained. “We want to help readers understand that there are still votes to be counted and that this aligns with our expectations based on insights from officials. We’ve already conducted reporting in every state to better understand how this election may differ from 2020.”
Vote counting processes may vary from four years ago, with some states potentially slower or faster, which will be reflected in the forecasts and results displays.
Andrews identified two factors that could affect how quickly races are called in each state. The first is whether local rules permit election officials to begin counting ballots before election day. States that allow this often complete their counts on election day, while others that do not may take longer. States like Pennsylvania, Arizona, and Nevada experienced delays in 2020.
The second factor is the closeness of the race. For example, Georgia counted votes relatively quickly in 2020, but the narrow margin necessitated thorough verification of all votes before calling the race. “Even in states like North Carolina or Georgia, which typically count faster, we may have to wait until Thursday or Friday to call a race,” Andrews noted.
**How Will the Times Analyze the Results?**
The line between displaying election results and providing analysis is increasingly blurred, with the Times planning to offer real-time insights from reporters and data journalists throughout election night.
“If you turn to us on election night, you won’t just see live results; you’ll also discover how different counties are voting and what shifts are occurring in the presidential race compared to 2020,” Andrews mentioned. This analysis will inform more extensive pieces over the week, as journalists strategize on priority questions and establish data pipelines to deliver quick responses.
**Advice for Smaller Newsrooms**
The Times will approach the U.S. election with more resources than any other newsroom. When asked what advice he would offer to colleagues in smaller teams around the world, Andrews replied:
“The crucial takeaway is not to interpret early returns as indicative of the final outcome. Patience is essential, particularly in swing states. Comparing against previous elections can provide insights, especially when large counties haven’t reported significant data yet. However, caution is necessary; 2020 may not serve as the best reference this time, as states might report differently or at varying speeds.”
As for framing the election for their readers, he advised, “Journalists should emphasize that counting votes may take time and that election night might extend into an election week. This was certainly the case during the 2022 midterm elections, where the control of the House of Representatives wasn’t determined for eight days, and a similar situation could very well arise again.”
