On Election Night, viewers tuning in to their favorite news networks are often anxiously waiting for one thing: to find out who won. But as the results roll in, the magic moment often comes not from official election boards, but from the news organizations themselves projecting the winners. For many, it raises a question—how do news organizations call races, and are their projections accurate?
Here’s a behind-the-scenes look at how projections work, the methods newsrooms use, and whether these predictions stand the test of time.
The Projection Process: Data-Driven Predictions
News organizations typically project winners based on a combination of vote totals, historical data, and statistical models. Instead of waiting for every ballot to be counted, major networks like CNN, the Associated Press (AP), and others rely on teams of data analysts, statisticians, and political experts who track voting trends in real time. These teams closely monitor how the votes are breaking in each district, comparing them to past elections to spot trends and make educated predictions about who is likely to win.
For each race, the process starts with exit polls—surveys of voters as they leave polling places. These early surveys provide insight into voter demographics and the issues that matter most, along with initial signs of candidate support. However, exit polls alone are not enough. Instead, they serve as a supplement to the actual vote count, with analysts using these initial numbers to get a feel for how the race might unfold.
As real vote totals begin to come in, statisticians compare current data to historical voting patterns in each district or state. A Democratic-leaning precinct that’s breaking for the Republican candidate, for instance, may signal a larger trend. Specialists then combine these observations with sophisticated statistical models that predict the outcome based on current returns.
Role of Decision Desks and Statistical Models
Central to the process is the “decision desk,” a team of experts focused exclusively on monitoring data and calling races. Decision desks for major networks and the AP work independently from each other, often using slightly different models or criteria to make their calls. This independence is one reason why some networks may project a winner before others or occasionally disagree in their projections.
These desks use advanced statistical models that weigh multiple factors in real time. For example, in a governor’s race in a battleground state, the decision desk would look at:
• The percentage of votes counted,
• Where those votes are coming from (rural or urban areas),
• How each county has historically voted,
• Trends in mail-in versus in-person voting.
Only when the model is highly confident, sometimes above 99%, will the decision desk “call” the race.
Are Projections Accurate?
Historically, the answer is a resounding yes—most of the time. Major news organizations are generally cautious, knowing that even a single inaccurate call could damage their credibility. According to the AP, which has been calling races since 1848, its projections have been accurate over 99% of the time.
However, the 2000 presidential election between George W. Bush and Al Gore spotlighted the challenges and limitations of projections. That year, some networks prematurely called Florida for Gore, then retracted it, and finally projected it for Bush before ultimately conceding that the race was too close to call. The incident led to significant changes in how projections are made, with news organizations adopting even stricter standards to avoid premature calls.
Today, networks often refrain from calling races until they have overwhelming statistical certainty, especially in states with close margins. Improved statistical methods, more extensive polling, and additional data sources have helped news organizations reduce errors in recent years. Moreover, projections for most down-ballot races—such as House and Senate seats—tend to be highly reliable due to the sheer volume of data available from past elections and voting trends within specific districts.
Why News Outlets Can Call Races Sooner Than Officials
Even as news outlets call races on Election Night, official counts from election boards often continue for days or weeks afterward, especially in areas with high numbers of mail-in ballots or complex recount laws. This delay can lead some voters to wonder why they hear about winners from the news before the official results are in.
The reason lies in how different the goals of news organizations and election officials are. News organizations aim to inform the public as quickly as possible, while election officials must ensure absolute accuracy and compliance with legal requirements. It’s not uncommon for election results to be certified weeks after the vote; however, news organizations use predictive models that allow them to call races with confidence much sooner.
Can You Trust Projections?
For voters watching at home, the good news is that projections are remarkably reliable, especially when coming from experienced news organizations. However, in extremely close races, news outlets may hold off, advising viewers that it’s “too close to call.” When this happens, it’s not an indication of uncertainty in their process—it’s a signal that, statistically, the margin is just too narrow to responsibly project a winner.
By using a balance of statistical rigor, historical data, and real-time voting information, news organizations aim to give viewers an accurate and timely look at election outcomes. And while no system is foolproof, major networks have consistently proven that their projection methods are among the most accurate predictors of electoral outcomes in modern democracy.
