One of the most unnerving things about analyzing American politics in recent years has been the need to start thinking in terms what financial analysts call "tail risk" — the likelihood that events normally considered rare or unlikely will actually happen.

Ten years ago, most pundits would have described as exceedingly small the likelihood of someone like Donald Trump — a real-estate mogul, reality-show star, and promulgator of racist conspiracy theories who had no political experience at all — being elected president. Yet it happened.

Two years ago, epidemiologists would have given a range of probabilities on the question of whether a pandemic would soon kill millions around the world in the space of 10 months. Because that's exactly what's happened, those on the more alarmist side would appear to have been vindicated, with those who were more sanguine shown to be insufficiently attuned to the dangers we faced.

What additional tail risks does the United States confront today? And which of them is most likely to become a reality?

A lot of commentary over the past four years has focused on the question of whether or not Trump is a fascist on the verge of becoming a ruthless authoritarian. I don't think there's much doubt that Trump personally would love to be a dictator — but there's also abundant evidence that he's never been anywhere close to making himself one, mainly because he's far too ignorant, lazy, and inept to outsmart and depose America's longstanding democratic-republican institutions.

But that's not the only reason why Trump was never going to be able to impose fascism on the United States: American culture is too deeply hostile to tyranny, even to a fault. We're a country born of a tax revolt, after all, and many of us regularly mistake the ordinary exercise of government power as an existential threat to individual liberty. That means any effort to impose dictatorial rule would be maximally likely to inspire an equal and opposite reaction that results in greater disorder, breakdown, and chaos.

That's why the tail-end scenario that worries me most by far is the outbreak of a civil war.

I've written about this before, always feeling a little skittish when I do. I worry about sounding like a hysteric, which is the opposite of what I strive to do with my writing. Critics of my talk of civil unrest usually raise one of two objections.

Some say that 21st-century Americans are just too slothful and self-absorbed to fight an actual civil war. We're a nation of couch potatoes, unhealthy, overweight, hooked on painkillers, and pathologically fixated on our computers and phones. That doesn't sound like the kind of people liable to brandish weapons and risk their lives for a cause.

Then there are those who say that a civil war couldn't possibly break out today because, unlike in 1861, when one entire region of the country went to war against another, our battle lines are too scrambled in the real world to make combat viable. Where would the front lines be? What territory would be fought over and conquered?

Until quite recently, I was largely persuaded by the first point. Ross Douthat's recent book on decadence, which I favorably reviewed last February, helped to persuade me that nearly everyone waging digital warfare online is indulging in a fantasy, immersed in virtual-reality combat, and maybe even blowing off steam that might make a real-world conflagration less likely than it would otherwise be. But last week shattered this assumption, showing that it's far more likely that this analysis is itself a fantasy. Thousands of people traveled to the nation's capital to participate in a physical assault on Congress. This shows that there absolutely are Americans in 2021 with the means and motivation to fight for a cause.

As for the second objection, I think it's an error to assume that any civil war that might arise would need to resemble the one that tore the country apart from 1861 to 1865. Or that it would look like the West's most prominent recent civil war, the one that turned the former Yugoslavia into a charnel house during the 1990s. Both of those civil wars had a strong territorial component. The first was of course a conflict between the northern and southern regions of the country over the institution of slavery. The second was sparked by the reassertion of ethnonationalist and religious attachments in a country that had suppressed or blended them for decades.

When pondering the worst-case American future, it makes more sense to think in terms of several different forms of violent intranational conflict. One is the English Civil War that raged from 1642-1651, pitting the crown against parliament, cities and towns dominated by a rising commercial middle class against the aristocratic countryside, and the staid religious convictions of the ruling class against the theologically driven radicalism of more demotic religious sects.

Another model of civil violence is The Troubles that rocked Northern Ireland for 30 years beginning in the late 1960s, with factions aligned with the (Catholic) Irish Republican Army, which sought unification with Ireland, squaring off against those allied with the (Protestant) Unionists (backed by English troops), who wanted the territory to remain part of the United Kingdom. There were some conflagrations in this conflict that resembled traditional military battles. But most of the time the republican side waged its side of the war through acts of terrorism at home and abroad, while their opponents used brute force to crack down on the roiling insurrection.

A final possibility is the Spanish Civil War that shattered the Iberian Peninsula into a multitude of factions — from anarchists, Stalinists, and anti-clerical absolutists on the left to fascists and Catholic authoritarians on the right — between 1936 and 1939. The viciousness of the fighting was provoked both by the depth and sheer number of cleavages in Spanish society, but also by the fact that the war became a proxy for the intense ideological storms raging across Europe at the time, with outside governments and partisans from other countries joining the battle on every side.

What makes all of these models germane to thinking about the current American situation is that, although they all dealt in part with territorial disputes, they were not primarily conflicts in which one army fought another over the political meaning of lines on a map. They also involved other disputes — over deep cultural and religious differences — and they were waged in many cases by launching (rhetorical and literal, conventional and unconventional) attacks on institutions that each side thought was aligned with the other.

One way in which all of these alternative forms of civil war differ from our own situation is that the parties on the different sides of those earlier conflicts fought for concrete political goals. When one side or the other achieved its aims, measured by standards both sides could agree on, the fighting could come to an end. Our own battles, by contrast, increasingly revolve around disputes over the character of reality itself — who actually won the 2020 election? — and are fueled by conspiracies and painstakingly cultivated hatred of the other side. (How can either side "win" a dispute like this? By capturing all of its opponents and subjecting them to psychological deprogramming?)

That might make any conflagration that arises in our country over the coming weeks, months, or years the first fully postmodern civil war in history.

Last week was the first time I began to think something like this really might unfold in this country — and even to ponder whether, if it does, historians will look back at the events of Jan. 6, 2021 on Capitol Hill as the time and place it got underway.

Will that be our fate? Is a civil war bound to happen? There are good reasons both to doubt that it will and to suspect that it might.

The case for optimism would point to Congress, tech companies, and the national guard moving to tamp down on the potential for violence in advance of Joe Biden's inauguration. It would also highlight Trump's latest video message — released on Wednesday night — in which the president denounced violence more forthrightly than he had up until then. Maybe that will lead the transition to the Biden administration to proceed more smoothly than many of us thought likely as recently as a few days ago. That could help begin to settle the country down, which is something the new president will be doing his best to continue.

On the other hand, Trump has said nothing to counteract the poisonous, conspiratorial lies about election fraud that whipped up the insurrection in the first place. So it's hard to know what will unfold over the coming days and weeks, with three-quarters of Republicans still convinced the election that is putting Biden in the White House was fraudulent. (If you believed this, would you be willing to sit back and let the inauguration proceed without doing some damage?)

If there are additional attacks by the right, they will likely also be inspired in part by reaction to the very moves that have justifiably been made since Jan. 6 to counter the insurrection — the tech crackdown and second impeachment of the president, in particular, which Trumpist Republicans are already calling evidence of burgeoning left-wing oppression. New acts of insurrection would then prompt more justified moves to quash the violence — including harsher restrictions on the right and perhaps their application to a broader swath of conservatives. Which would inspire more acting out by members of the MAGA militia.

That's the iterative process that's unfolding all around us. A version of it gets enacted whenever a civic conflagration starts to spread. Each side has its list of grievances that the other dismisses. Each side's escalation provokes the other's acting out in the name of self-preservation, until the limits of normal politics are cast off and violence becomes the new normal.

How great is the risk of America going completely off the rails in the short-to-medium term? I wouldn't want to bet big money on it. But the odds are improving all the time.