For the past several months, some Republican politicians have been questioning the validity of the November election, and millions of citizens think it was fraudulently decided. This week, those beliefs had tragic consequences.

Does this signal a course change in American politics? Some pundits suggest that President Trump’s behavior is just one illustration of the Republican Party’s slow transformation into an authoritarian party. Others worry that his efforts to overturn the election illustrate the decay of U.S. democracy. The Jan. 6 invasion of the U.S. Capitol, Congressional Republicans’ challenge to the electoral count of selected states and last weekend’s revelation of the president’s efforts to strong-arm the Georgia secretary of state have only amplified these concerns.

Our research leads us to believe that the past four years aren’t just a predictable reflection of long-standing changes in American politics. Rather, they illustrate something that pundits and political scientists have a harder time understanding than do historians and complexity scientists: Chance events — things that could have gone differently under slightly different circumstances — can have big consequences in politics. And that makes us worry about an underappreciated long-term trend in American democracy: The U.S. is shifting in ways that magnify the consequences of apparently random events.

Politics depends on random events but in non-random ways

Chance plays a big role in all areas of life. However, some kinds of organizational structures can make chance even more important. In situations with many independent actors, individual differences wash out, making outcomes predictable. Life insurance companies can predict how many people in a given category will die early, even if they can’t predict which individual will die when. But where everyone is bound together in tightly integrated organizations, certain chance events can have big consequences.

During the Cuban missile crisis, the lucky assignment of one particular Soviet submarine officer to one particular boat helped avert a nuclear weapons exchange. Similarly, the U.S. Civil War could have unfolded differently in 1862 had Gen. Robert E. Lee’s Special Order 191 not been lost and later found by Union soldiers, ruining Confederate campaign plans.

This principle helps explain politics as well as war. Donald Trump’s election was not inevitable. If other Republican candidates had realized in time that he might win, they could have coordinated against him. But Trump is an outlier on several of the dimensions identified by the now-standard five-factor framework of personality psychology. As president, his strikingly low degree of conscientiousness (meaning diligence and the internalization of rules and norms) and unprecedented encouragement of civil unrest have had massive consequences.

Random shocks are becoming more important

Long-term trends in our system are making it more vulnerable to disruption by chance events such as Trump’s election.

First, the party elites have much less power in selecting candidates than they once did — and that makes it more likely that outlier candidates will make it onto the ballot. Elites long helped keep would-be demagogues off general election tickets. As they stepped back, ideologues gained influence.

Second, the polarized media environment and the electoral strategy of restricting access to voting make it easier for outliers to win if they make it to the general. As others have pointed out, voters who consume partisan media are more likely to be misinformed — and may be less aware of the failings of their party’s candidates. Making voting harder could increase the influence of dedicated partisans, again making it easier for politicians who violate long-established norms to win.

Third, outlier leaders have more influence once they have won. The “unitary executive” theory adopted by many courts increases deference to the president. That limits federal agencies’ independence and makes Congressional oversight harder. Presidents become more likely to select senior officials through acting appointments and appointments not requiring Senate confirmation. This reduces Congressional influence, decreases bureaucratic effectiveness and renders agency leaders more subject to presidential whim.

Finally, the electoral college has become an institutional amplifier of chance events. Just a few voters can swing presidential elections. Trump’s 2016 victory by 80,000 votes in three states surprised most pollsters and election forecasters. His loss was also a chancy affair. Biden’s victories in Arizona, Georgia and Wisconsin rested on 45,000 votes. Without those states, each candidate would have gotten 269 electoral votes, sending the contest to the House, with unpredictable consequences.

Lest we forget, a swing of 60,000 votes in Ohio in 2004 would have given John F. Kerry a victory — even though he’d have lost the popular vote by 3 million votes. And in 2000, mere hundreds of votes in Florida handed George W. Bush the presidency, a result probably driven by the peculiar ballot design used in one Florida county.

Four of the last six elections could easily have gone the other way.

We can expect more of the unexpected

Democratic norms won’t necessarily protect the U.S. system from extreme leaders. Trump still won in 2016 after violating widely shared norms. And while pessimistic predictions regarding the 2020 elections were inaccurate, the post-election period has turned as dark as many feared.

Researchers have a hard time predicting how relatively simple contagions such as the coronavirus will spread. It is much, much harder again to predict how quickly new ideas and approaches to politics might spread through a large network such as the executive branch — new approaches that might include, for instance, a leader unconstrained by facts. Such systems can be quite vulnerable in some ways, yet surprisingly resilient in others.

Forecasting large-scale political events is so hard exactly because they are the product of institutions and organizational networks that magnify the impact of chance and contingency.

But we do know that the unexpected is becoming more likely. For the past decade, trends in U.S. government and politics have made the system more susceptible to random shocks. Those who value political stability may want to think about how such uncertainty can be reduced again.

Jonathan Bendor is professor of political economy at the graduate school of business and professor (by courtesy) in the department of political science, Stanford University.

Jacob N. Shapiro is a professor of politics and international affairs and directs the Empirical Studies of Conflict Project at Princeton University