In brief
On Tuesday night, U.S. newsrooms began calling winners in several closely fought midterm races, using vote counts from the Associated Press and the firm Decision Desk HQ, while some networks added exit polls and statistical models. The strategy has raised questions about public confidence, as only a third of Americans report high trust in news outlets to report accurate results. The process draws on a long history of AP vote‑counting and a growing network of reporters and data firms to supply timely results.
On Tuesday night, U.S. newsrooms began announcing winners in some of the country’s most tightly contested 2026 midterm races, long before the final tally of every ballot. The calls, made the instant a voter reached the polls at 6 p.m. ET in Indiana and Kentucky and at 7 p.m. in several southeastern states, rely on a combination of real‑time vote data and statistical judgment that many viewers see as both critical and controversial.
At the core of these race‑calling engines is the Associated Press, the nonprofit newsroom that has counted votes in every presidential, congressional and statewide contest since 1848. For this election, the AP said it would deploy more than 4,000 reporters to county election offices and polling sites across the nation. These reporters feed data into the decision desks of ABC, CBS, CNN, and NBC, turning raw precinct counts into the feeds that the public watches on live television.
Personalized data vendors also play a role. Decision Desk HQ, the firm that supplies the political analyses of several cable networks, provides its own tabulation models that are fed into the same real‑time feeds. In the final months of coverage, the firm’s analysts combine early‑vote totals, demographic factors and historical patterns to estimate how many votes remain uncounted and whether a trailing candidate can close a theoretical margin.
While most networks use the AP counts as their baseline, they differ in how they decide when to call a win. The Wall Street Journal and MS NOW rely solely on the AP’s data stream, waiting for the AP’s internal thresholds to be met before issuing a call. Fox News and CBS, on the other hand, mix the AP vote totals with exit‑poll interviews, early‑vote totals and proprietary statistical modeling. They say they make their projections only once the data indicates a statistical certainty, but the models are not standardized across the networks.
Exit polls themselves are a key piece of evidence. Several major networks consult SSRS, a research firm that acquired the exit‑polling business Edison Research in 2025. SSRS runs interviews with a sample of voters at the last moments of voting, providing an estimate of how people lean on the day of the election. The networks then contract these exit polls with the AP data to refine projections in real time.
Readers have grown wary of this high‑stakes forecasting, however. An October poll conducted by the Medill Local News Initiative and the Alliance for Trust in Media found that more than 60 percent of Americans felt news organizations could report results accurately, but only 32 percent said they had high confidence. The study also showed that Democrats were more trusting than Republicans, a split that suggests partisan lines smear the consensus on media reliability. The contrast between high overall trust and low high‑confidence trust can influence viewers’ reactions when a forecast turns out to be wrong.
Recording history of errors adds to the tension. When networks are wrong—when the call is too early, or the wrong candidate is announced—viewers become skeptical. That tension is part and parcel of the process. Networks typically underline that the call is a forecast, not a final result, and that the official certification will still take place weeks later.
These procedural details run in the background of a larger civic conversation about accountability and transparency. Voting, court certification of results, and subsequent legal challenges all depend on the initial picture that the media offers. The accuracy of early calls can shape public sentiment, mobilize or demobilize voters, and influence Congress’s internal negotiation. The process also shapes journalistic standards, prompting media organizations to refine their data pipelines and to publish more explicit descriptions of their analytical methodology.
On the night of the calls, the sound of ticking clocks represented more than just the passing of time; it represented the calculation of thousands of votes, the coordination of distant reporters, and the austere limits of data science in a field where a single miscall can echo through the political landscape of a state and across the nation.
