Photo: Well This Is News
NPR and AP Partner to Call Races with Data-Driven Precision
We haven't found a right-leaning take on this story yet.
Breaking stories sometimes publish before every side has covered them. We'd rather show you that than invent a view.
How Major News Organizations Call Election Winners on Race Night
Key Takeaways
- The AP and networks calling races have never disclosed the statistical confidence threshold they use to declare a winner, so voters cannot distinguish between 99 percent certainty and 95 percent certainty when they hear a race is called.
- Major news organizations have not published historical accuracy data, error rates, or instances where early race calls were revised in previous elections, leaving the track record of this methodology publicly unknown.
- Calling a race before counting finishes shapes how voters and subsequent elections perceive the outcome, but this electoral consequence receives no discussion in news coverage explaining the process.
The Analysis
Major news organizations have developed a standardized process for calling election winners that involves releasing projections before all ballots are counted, yet neither the available reporting nor public discourse fully separates the statistical methodology from the editorial judgment embedded in when and how networks choose to make those calls.
The AP, which provides race calls to NPR, PBS NewsHour, and other outlets, bases its projections on analyzing reported votes against historical turnout and demographic patterns specific to each precinct. The AP counts actual reported ballots and models the remaining uncounted votes to calculate the probability that a trailing candidate can still win. This is a mathematical exercise, not a matter of opinion. The AP does not call a race until it determines statistically that the trailing candidate cannot overcome the gap based on remaining ballots and their likely composition. NPR's reporting describes this as a partnership in which NPR relies on AP analysis rather than conducting independent race calls.
What the available reporting does not fully establish is the threshold question: at what level of statistical confidence does the AP actually make the call? The documentation available does not disclose whether the AP uses 99.5 percent confidence, 99 percent, or a different standard. This matters because a race with a 99 percent statistical likelihood of outcome differs materially from one with a 95 percent likelihood, yet both could be described with identical language: race called. The reporting from both NPR and PBS emphasizes the rigor of data analysis without addressing the editorial choice embedded in confidence thresholds.
The left-aligned framing, represented by NPR, emphasizes partnership, methodology, and precision. The language centers on "how we do it," positioning the news organization as transparent about process. This framing serves a confidence-building function: by explaining the mechanics of analysis, NPR suggests the outcome is not opinion but mathematics. What this framing leaves out is any discussion of error rate, historical instances where early calls proved wrong, or the distinction between statistical projection and the actual completion of counting. If such errors exist in AP or network history, they are not mentioned.
Neither outlet provides data on the accuracy of AP race calls across multiple election cycles, the frequency with which early calls have had to be revised, or comparative analysis of how different networks or news organizations weight statistical confidence differently. One plausible interpretation is that the absence of this historical track record from both NPR and PBS reporting reflects either genuine confidence in AP accuracy or a preference to leave that question unexamined in public-facing explanations of the process.
The underlying structural question the reporting does not address is whether the public distinction between "calling a race" and "declaring a race won" remains meaningful when millions of voters watch networks project outcomes before counting is finished. The AP's methodology may be sound, but the timing and presentation of those calls shapes electoral narrative and potentially voter behavior in subsequent elections, a consequence the available coverage does not foreground.
Networks deploy statistical models to project winners before all ballots are counted, yet neither NPR nor PBS discloses the confidence threshold that triggers an actual call, creating a gap between methodological transparency and editorial accountability. AP may operate at 99.5 percent confidence or 95 percent confidence; identical language masks this material difference. The reporting emphasizes mathematical rigor while omitting historical error rates, revised calls, or comparative analysis of how different organizations weight statistical certainty differently. This matters because early projections shape electoral narratives and potentially influence voter turnout in subsequent contests, particularly in states with staggered voting windows or overseas balloting still arriving. The public cannot evaluate whether news organizations are calling races at thresholds appropriate to electoral stakes if the thresholds remain undisclosed. This structural opacity persists despite detailed explanation of data methodology, suggesting either institutional confidence in accuracy or institutional preference to leave precision standards outside public scrutiny.