Donald Trump’s startling refusal to promise he’d accept the election results was just the latest in a series of seemingly casual dismissals of long-held American political norms. Trump has suggested widening libel laws in a way that would blatantly infringe the freedom of the press; he’s also flouted the First Amendment in his vow to ban Muslim immigrants and to close down mosques. Meanwhile, his threat to jail his opponent if elected has raised the specter that he might use his office to quell legitimate dissent.

As much as his critics wring their hands, however, a consensus has emerged that even if Trump were elected, the American system would survive it. We have the rule of law. We have checks and balances. If Trump overshot the bounds of his authority, the system would constrain his actions in much the way it has on occasion done to past Presidents from John Adams to Barack Obama.

But would it? The uncomfortable truth is: We can’t be so sure. For the past three years, I’ve been studying the way in which populists use democratic elections to undermine liberal protections like the rule of law—and what I’ve found is that modern democracies, including America’s, are far more vulnerable to hostile takeover than you might think.

There are three interlocking reasons why our confidence in the system is naïve. For one, we’re in genuinely uncharted territory with Trump: we’ve simply never seen a candidate with this much disregard for typical Constitutional values get this close to the White House. There’s no precedent for what might happen if he got there. For another, if you look at how our system of checks and balances is really built, it has relatively few resources to stop an authoritarian president from violating the Constitution and getting away with it. And the third reason may be the most unsettling of all: In a democracy, the final brake on the tyrannical exercise of power is public opinion. And polls suggest the American public has never been as skeptical of democracy or as open to authoritarian alternatives like military rule as it is right now. If a President Trump really blew down the walls of our system, a worryingly wide swath of the public would likely stand behind him.

This might not matter this year; Trump is behind in the polls, and it’s unlikely that he’ll get to test his promises from anywhere more powerful than a cable news studio. But there’s now no question that a future populist of his ilk, with more discipline and less personal baggage, could take an even more serious run at the White House. And what Trump is exposing is just how fragile our system might be if that happens.

***

Through U.S. history many presidents have pushed the limits of their constitutional authority, some overstepped it, and a few even engaged in straightforwardly illegal activity. Even so, every President in the history of the United States has espoused a real commitment to the Constitution, and has at times proved willing to subordinate his immediate self-interest to his political values. Even Richard Nixon, as close to an out-and-out crook as the White House has known, finally resigned when Congress moved to impeach him. It’s simply not clear that Donald Trump would do the same.

While there have at times been real conflicts between different branches of government—especially the Presidency and the Supreme Court—the country’s checks and balances have never been put to the test in quite the way that Trump now threatens to do. To understand how the American government might buckle under this kind of stress, it’s worth asking exactly what such a collision might really look like. What if a President Trump really did decide to impose big penalties on critical newspapers, or prosecute Hilary Clinton, or refuse to comply with a Supreme Court ruling?

If you game it out, there are at least three likely scenarios that lead to a crisis. The first case would be Trump ordering the federal bureaucracy to do something blatantly unconstitutional—like, say, closing down mosques or prosecuting political opponents. It is likely that many senior bureaucrats would refuse to comply with such an order, either resigning in protest or simply disobeying the relevant instructions. In a political world in which leading politicians and most voters have a deep commitment to democratic norms, this is an effective form of resistance: The press would be likely to cover the story prominently. Public opinion would rally against an administration that insists on issuing illegal instructions. The President would back down. According to Constitutional scholars like Jack Goldsmith, a Harvard Law School professor who has written extensively about limits of executive power, this is exactly the kind of mechanism by which, despite its vastly expanded powers, the executive has effectively been constrained since 9/11.

But all of this assumes that we can still count on a shared commitment to democratic norms, and have a President who is sensitive to widespread outrage. So what would happen if a President Trump decided that he would rather persist with an unpopular course of action than risk looking weak? Or if public opinion didn’t swing against him in the first place? (Since many recent polls show broad support for discriminatory policies against Muslims, this is hardly an unimaginable scenario.)

In essence, there would then be little to stop the President from a simple power move: firing all bureaucrats who disobey his orders, and replacing them with loyalists. There is, in fact, a clear precedent for this in American history. In the so-called Saturday Night Massacre, Nixon instructed Attorney General Elliot Richardson to fire special prosecutor Archibald Cox, who had been appointed by Congress to look into his illegal activities during the Watergate scandal. When Richardson resigned his post in protest at the President’s attempt to interfere in judicial proceedings in such a blatant manner, Nixon instructed Deputy Attorney General William Ruckelshaus to do his bidding. When he, too, resigned in protest, Nixon turned to Solicitor General Robert Bork—who duly complied. By the end of the night, Cox was gone.

If a President Trump took a leaf out of Nixon’s playbook and repeated the Saturday Night Massacre at an even bigger scale—including rank-and-file bureaucrats as well as political employees in his purge—it would undoubtedly sow chaos and deplete federal agencies of much-needed expertise. But finding a team of hacks to follow his orders, however haphazardly, would not prove difficult. If he wanted to close down mosques, or have his cronies prosecute political opponents, he probably could.

The dynamic looks surprisingly similar in the second scenario: if a President Trump ordered soldiers rather than bureaucrats to engage in illegal activities. What would happen, for example, if Trump ordered the Pentagon to kill the families of terrorists. or to engage in interrogation techniques that have unequivocally been defined as torture? In a first step, senior officers would publicly state that they refuse to carry out his illegal orders—a serious and essentially unprecedented constitutional crisis in itself. But if this was not enough to deter Trump, they would have only two options: stall and wait to be fired, or resign.

Either way, his position as commander-in-chief of the armed forces would once again give Trump very strong cards. If the officers resigned, he would have the constitutional authority to promote his own cronies, or to go down the chain of command until he found a willing henchman. If they quietly refused to follow his orders, he could fire them. Either way, the military might be weakened and the public outraged—but his orders would be carried out.

Another avenue of resistance to all of this would of course be legal. The victims of unconstitutional discrimination could sue against the state. So could rank-and-file bureaucrats—and military personnel—who were fired because they refused to carry out illegal orders. This is the third scenario, and there can be little doubt that the Supreme Court would be a bulwark against unconstitutional overreach, even if the seat that is now vacant were to be filled with a died-in-the-wool conservative. Is this, then, the backstop that would save the system?

Not necessarily.

Legal proceedings tend to take a long time to work themselves out: Fired employees might, for example, eventually win compensation for wrongful dismissal or even be reinstated in their jobs. But this is likely to happen years after the fact, when the damage had already been done. And even if the Supreme Court directly struck down a key part of his agenda, Trump could hold a big press conference in the Rose Garden and dramatically announce that he simply didn’t recognize its right to curtail his agenda. Supreme Court justices would howl in protest. They might even denounce Trump. But their power ultimately rests with the willingness of the executive branch to defer to their moral authority—or the willingness of an overwhelming majority of the public to turn on a President who refuses to do so. The Supreme Court's role as arbiter of what’s constitutional is ultimately just a matter of tradition, and Trump has already proved his willingness to flout tradition when it happens to suit his interests.

Each scenario thus culminates in the same question: to save the republic, Trump would have to be impeached. But impeachment requires both a simple majority in the House and a two-thirds majority in the Senate. Considering how scared most GOP officials have been of taking on a Presidential candidate with authoritarian tendencies, it is far from certain that many of them would prove more courageous in the face of an authoritarian President. And by the time the will to impeach him has built, and the highly complex proceedings completed, it would in any case be far from clear how much of the republic there would be left to save.

***

There is a long tradition of believing that a well-calibrated system of checks and balances can make democracy stable and guarantee the rights of its citizens. “The problem of organizing a state, however hard it may seem,”Immanuel Kant wrote in Perpetual Peace, “can be solved even for a race of devils.”

This may be true. Perhaps America’s institutions are so intricately designed that they would successfully guard even against a President hell-bent on advancing his own interests and amassing personal power. But this optimistic hypothesis, just like its more pessimistic counterpart, has never been tested in the United States. And the international comparison suggests great caution: countries like Hungary, Russia or Iraq did have carefully designed institutions with intricate checks and balances; but because the politicians who inhabited these systems had little commitment to liberal-democratic norms, they failed to protect democracy.

This suggests that it isn't primarily the legal genius of the Founding Fathers that has made American democracy uniquely stable, but, at least as importantly, the deep commitment of ordinary Americans to liberal norms like the separation of powers. It is the people's love for democracy, not the protections their democracy affords them on paper, that is the final check on the tyranny of the majority which the founders so feared.

In light of this, it is all the more worrying that Trump’s willingness to challenge basic democratic norms comes at a time when Americans are less invested in their political system than ever before.

It’s long been obvious that American citizens are unhappy with the performance of political elites: They have grown less and less satisfied with particular governments, less and less trusting of major institutions like Congress, and less and less likely to say that politicians mostly try to do the right thing.

But that’s only the beginning. As I showed in a recent paper with Roberto Foa, this general dissatisfaction with political performance has started to color Americans’ views of the democratic system itself. While over two in three older Americans say that it is essential for them to live in a democracy, for example, fewer than one in three young Americans share their deep allegiance to the democratic system. Even more worryingly, the number of Americans who are open to straightforwardly authoritarian alternatives to democracy, while still low, has been growing at a rapid clip: twenty years ago, one in fifteen Americans believed it to be a “good” or “very good” thing to have the army rule; today, it is one in six.

All of these surveys were taken long before this year’s ugly, fractious, fratricidal election campaign. And as might be expected, the first signs seem to indicate that the past months have deepened the loss of public trust in the country’s political institutions. Asked their view of U.S. democracy in a Washington Post poll earlier this week, for example, an astounding 40% responded that they had “lost faith”; another 6% stated that they had “never had faith” in the system in the first place.

In the end, it therefore seems likely that America’s remarkable stability has depended on both an ingenious institutional set-up and a deep commitment to democratic norms. “Ultimately,” Eric Posner, a leading constitutional scholar who has written extensively about the soft limits on Presidential powers, told me last week, “how dangerous the Trump Presidency would be depends on your view about public opinion: Would people act collectively to defend liberal norms of decency?”

Americans tend to think of checks and balances, or the rule of law, as entities that are somehow outside the political process—and can step in to save the day when normal politics fails in some significant way. In times when conflicts are minor, and the attachment to liberal norms runs deep, there is something to this. But that is no longer the world we live in. At this point, the Constitution cannot save us from our politics. On the contrary, we need to fix the deep political crisis of our time—a crisis that is likely to persist well beyond the Trump candidacy—to save the Constitution.