In the 2022 midterm elections, voting ends when the polls close on Election Day. But that doesn’t mean we will automatically know the winners of all 35 U.S. Senate seats up for grabs, nor the outcome of all 36 governors’ contests, 435 U.S. House seats and a slew of state legislative races also at stake.
The process of calling races — projecting a winner as votes continue to be counted — is tricky.
The Washington Post doesn’t call races on its own. It relies heavily on two nonpartisan organizations with long track records of examining raw vote totals and calling elections: the Associated Press and Edison Research, the latter of which provides vote count data for a group of TV news networks. Although The Post also looks to its independent election model to provide information about how the vote is trending as results are being tallied, it does not rely on the model to make race calls. It’s important to remember the official winner of any electoral contest is not determined until the state government certifies the vote — in many cases weeks after the election itself.
Why is it so difficult for winners to be officially declared on election night? The answer is the United States doesn’t hold a single election for Congress, statehouses and state legislatures — instead, it more accurately conducts thousands of contests in scores of states and counties by local officials in disparate ways. This highly decentralized system is one of the reasons experts say election fraud is so rare — but it also makes it more difficult to reliably predict the results shortly after the polls close.
This year, for instance, because there are so many competitive races for House and Senate, we may not know which party controls either chamber of Congress on election night. That answer could take some time.
The Post is seeking to be as transparent as possible about what goes on behind the scenes in our newsroom on election night. Our top priority is reporting accurate information to our readers, and that might mean we publish race calls more slowly to give our journalists more time to consider whether we have enough data. For a handful of the most competitive races (more on which ones below), we will spend more time analyzing the data from AP and Edison in deciding when we’re ready to publish them.
Below, we break down how we decide when a race call is ready to be reported.
Where does The Post get information about how Americans are voting?
The Post relies on two organizations for vote data: the AP and Edison Research. Our results pages depend on constantly updated tallies from the AP, which has multiple ways of collecting information: reporters who visit and/or call counties or clerks’ offices, maintaining good communication lines with them, and who check county websites or other programmatic feeds. Edison runs its own vote counting system, providing data to the National Election Pool (NEP) consortium of ABC, CBS, CNN and CBS as well as other subscribers.
The AP and Edison also conduct large surveys of early and Election Day voters measuring support for candidates across a wide range of demographic groups as well as opinions on top election issues. AP’s survey is called AP VoteCast, and Edison conducts the exit poll for the NEP. Such polls are usually only one piece of information informing race calls, and competitive contests rely primarily on analysis of vote tallies.
The AP’s race-calling team includes about 60 reporters, editors and analysts in a general election, according to David Scott, the AP’s vice president and head of news strategy and operations.
Its standard for race calls is “certainty,” Scott said, meaning the AP will only call races when it doesn’t see any possibility the leading candidate can lose. The AP seeks to ensure it knows the location of all outstanding votes in a state or county to confirm there isn’t anywhere a trailing candidate might outperform or a trend can be reversed.
Sometimes, the AP calls a race right after the polls have closed. That usually happens in less competitive races, in instances when the trend lines match well with previous elections and mirror the AP VoteCast poll. At other times, a call comes in shortly after polls close, such as when the contest’s expected winner is leading by an overwhelming margin in vote tallies or VoteCast polling. Slower race calls are expected in more competitive races.
Edison Research’s decision desk relies on three teams of 4 to 6 experts each who analyze House contests, turnout levels and consider and then make race calls, said co-founder and executive vice president Joe Lenski. The organization feeds its vote count and exit poll data into multiple computer models. Once a leading candidate has a 99.5 percent chance of winning, Edison’s decision desk will consider making a call.
“Even when the computations indicate we’re at that level,” Lenski said, Edison’s team will be “scrubbing data” to ensure the remaining vote could not change the outcome.
How do absentee or mail-in ballots affect a race call?
Different states have varying rules for when local officials can begin processing absentee or mail-in ballots — opening and verifying the identities of voters — which sharply increased in 2020 because of the coronavirus pandemic. There are even more rules about when such ballots can begin being tabulated.
This means that in states like Florida, Georgia and North Carolina, which allow early processing of mail-in ballots, the first release of vote totals will probably be from mail-in ballots arriving before Election Day and early in-person votes. The next ballots to be counted are often Election Day votes, followed by absentee ballots that arrived on (or in some states a few days after) Election Day.
Other states like Pennsylvania, Wisconsin and Michigan don’t allow absentee ballots to be processed until Election Day. So the first ballots that will be counted and reported in those states probably will be Election Day votes, with the absentee votes being reported later.
The Post thinks carefully about the mix of Election Day to absentee ballots used by the AP and Edison when reporting a race call. If The Post determines that not enough Election Day ballots were included in a projection, we might delay reporting a call.
How will The Post report calls in the most competitive races?
For the large majority of noncompetitive House, Senate, gubernatorial and statewide contests in the general election, The Post will rely on AP race calls. The Post’s results’ pages and its live results modules for these contests will contain AP information and feeds.
We are, however, using a more rigorous process for the most competitive races: roughly 13 Senate contests, 60 House races, six gubernatorial contests, three secretary of state races and two elections in Maryland and the District of Columbia, for comptroller and Council-at-Large, respectively. We are relying on the ratings from the Cook Political Report and pre-election polls to categorize the competitiveness of individual contests.
For the 12 most competitive Senate races, The Post will await a call from either the AP or Edison. Once one organization makes a call, we will review information from both groups about the race, as well as consult our own model data and results maps to see if we’re ready to report the call. We’re dedicating more time to analyzing competitive Senate races because there are fewer of them, and because of each race’s importance in deciding control of the chamber.
For the most competitive House races, The Post has decided to wait for both the AP and Edison to call these contests before reporting a winner. We will apply that same standard — a call from both desks — for the gubernatorial and secretary of state races on which we’re keeping a close eye.
How will The Post report control of the House and Senate?
The Post will report control of the Senate when Democrats have won 50 seats (Vice President Harris would be the tiebreaking 51st vote), or Republicans have captured 51 seats, accounting for the 65 seats that are not up for election this year. The Post will report control of the House when either party has won 218 seats.
Because we are applying a stricter standard to a handful of races, The Post might be slower in reporting which party controls Congress than either the AP or Edison. That’s because we are waiting for both news organizations to project a winner in the most competitive House races, and taking additional time to assess calls in Senate contests.
How will The Post use its election model?
Our election model, which you’ll see across our website on Election Day, estimates the number of outstanding votes in any given race and which candidate or party is most likely to benefit from them. It aims to provide a fuller picture of the results after all the votes are counted instead of a sometimes misleading image of results earlier in the night.
If everything goes well on election night, you’ll also see insights based on our model included in our live coverage and on The Post’s live show. Those insights are based on comparing what the model tells us about how Americans are likely to vote with demographic information from the U.S. Census and records states keep about their voters.
We do consult our model when considering reporting a race call, though solely to confirm that AP or Edison’s projections line up with what our model shows. If our model estimates an entirely different outcome in a given race, we might hold off on reporting the call and dig into the discrepancy.