The major news networks and newspapers have called the 2020 presidential election for Joe Biden. For the past two days, they had declined to make a formal call, even though their analysts made it clear that they thought Biden’s election was inevitable. So why did they delay for so long, and then all suddenly make the same decision at once?

While we don’t yet have full information (very likely we will find out more over the next couple of days), given what we do know, there is a highly plausible explanation based in game theory, an approach to understanding the world used by economists and political scientists. Here is what it is:

Game theory is about coordination

Game theory is all about understanding when people coordinate, and when coordination fails. How is it that people can sometimes figure out to cooperate with each other, even if they can’t communicate easily? Why do people sometimes fail to cooperate, even when cooperation would be in everyone’s collective interest? These are the kinds of puzzles that game theorists study. While serious game theory can be intensely mathematical, people can grasp many of the basic insights pretty quickly.

One of these core insights was developed by Thomas Schelling, a Nobel Prize-winning game theorist, whose ideas are still at the core of U.S. nuclear strategy. Schelling wanted to figure out how people could coordinate on one particular solution out of many. He laid out a thought experiment: Imagine you have to meet someone in New York City on a given day, but you aren’t able to communicate with that person, and you don’t know exactly where you are supposed to meet and at what time.

There are tens of thousands of possible places you might meet, and many different times of the day. But Schelling suggested that people would be able to narrow down the list of possible choices pretty easily. He believed (based on playing out this scenario with his students) that people would end up meeting at midday at Grand Central Terminal. This is because midday and Grand Central Terminal are what Schelling calls “focal points.” They would leap out as obvious solutions, where everyone knows that everyone else knows that they are the obvious time and place to meet.

It likely explains the dilemma that networks faced

There are many coordination problems in social and political life, and one of them is calling an election. Because the U.S. electoral system is complicated and decentralized, and takes a long time to reach official decisions, major U.S. news organizations have come to play an important political role in calling the winners of presidential elections. But this role can be politically risky. What if one news organization calls a victor in a presidential election and others disagree? Even worse, what if a news organization calls an election and later finds out that it is wrong when all the votes are counted. This would massively damage the news organization’s reputation.

This was a particularly serious dilemma for news organizations in 2020. Fox News and the Associated Press called Arizona early, believing that it was sure to go to Biden. However, other news organizations declined to make the same call, and it became clear over time that the call was premature, as Biden’s lead narrowed (he will likely still carry Arizona, but it will be close). No one wanted to make the same mistake — especially because they didn’t want to be the only news organization to go out on a limb.

This resulted in what game theorists call an “assurance game.” That is a game in which people only want to coordinate on a particular action if they are sure that everyone else is coordinating on the same action. If they don’t feel confident that others will do the same thing, then they are likely to decline to act. News networks didn’t want to call the general election for Biden unless they were sure that everyone else was going to do the same thing.

That is probably why they continued to decline to call the election, even as their analysts became increasingly certain that Trump had no electoral path to victory. Each was privately convinced that Biden had won, but none of them wanted to be the first to take the risk of calling it for him, in case others failed to come in. Furthermore, it was obviously difficult for these organizations to explicitly coordinate with each other, which would surely result in accusations from the Trump campaign that the networks were trying to fix the election.

There were many possible “solutions” to the problem — points at which the election might be called. As FiveThirtyEight′s Nate Silver notes, there was no fixed moment at which a Trump victory slid from “highly unlikely” to “nearly impossible,” making it difficult for the networks to figure out a point at which they could all coordinate without risk of going out on a limb.

It also explains the plausible solution

So how did this dilemma get solved? Here’s one theory that fits the facts as we know them right now. Over the past 36 hours, gossip began on Twitter and elsewhere about one possible point at which news organizations might call the election: when Biden’s lead over Trump in Pennsylvania exceeded 0.5 percent. This was not entirely arbitrary — it is the point at which you can’t call for a recount. But in theory at least, a possible recount was just one source of uncertainty.

However, if this explanation is right, such quibbles were mostly irrelevant. The 0.5 percent cutoff provided a focal point at which networks could call it for Trump — one that every network knew that every other network was paying attention to. This in turn solved the coordination problem — a news organization that called it for Biden at the 0.5 percent cutoff knew that others would follow. And that, indeed, is what seems to have happened. To be clear: This is just a theory, but it appears to fit the facts.