Be careful what you wish for.

New York magazine's Jonathan Chait hopes his fellow liberals will cheer on the possibility of Republicans nominating Donald Trump for president. Chait's preference will make no difference at all to the result of the GOP race. But still, Chait's essay is important for what it tells us about how at least one smart liberal is thinking about 2016 and the stakes involved in who becomes the Republican standard-bearer.

And what it tells us isn't good.

The GOP is an unstable (but electorally very successful) amalgam of an ethno-nationalist base with a wealthy anti-government and pro-immigration donor class. Republican presidential candidates normally work very hard to smooth over the tensions between these very different constituencies. Trump refuses to do this. Chait argues that by explicitly rejecting the outlook of the donors and siding unambiguously with the base, Trump's campaign has already begun to make mischief within the Republican electoral coalition.

If he won the nomination, the chaos would increase enormously. And that is an appealing prospect for a liberal. As Chait puts it, "A Trump nomination might not actually cleave the GOP in two, but it could wreak havoc. If, like me, you think the Republican Party in its current incarnation needs to be burned to the ground and rebuilt anew, Trump is the only one holding a match."

Let's leave aside the possibility that burning down the current incarnation of the GOP would also destabilize the Democratic Party's own incoherent electoral coalition. If we could be close to certain that Republican nominee Trump would lose the general election, I could see accepting the risks and even cheering him on as a catalyst for fundamental change in the Republican Party.

But can we be so certain? Chait seems to think so. His first reason why liberals should support a Trump nomination is that the billionaire "would almost certainly lose." I'm not so sure. Yes, it's true that Trump is "massively — indeed, historically — unpopular, with unfavorable ratings now hovering around 60 percent." But Trump's most likely general election opponent — Hillary Clinton — doesn't do much better, with an average unfavorable rating in the low 50s and two recent polls showing her as high as 55 and 56 percent. That's not a big difference.

Chait argues that the only thing that could enable the wildly unpopular Trump to overcome this obstacle and eke out a victory would be a "landscape-altering event." Like what? Chait names a recession. But recessions aren't once-in-a-century catastrophes. They happen on average at least once in a decade — and the last one (the Great Recession that hit in the run-up to the 2008 election) ended nearly six years ago.

But maybe even a Trump win in November isn't something to be overly concerned about. That is Chait's surprising third reason why liberals should cheer him on in the GOP nomination contest: Not only would a President Trump "probably end up doing less harm to the country than a Marco Rubio or a [Ted] Cruz presidency," but a Trump presidency "might even, possibly, do some good."

Here I think the normally sharp and sensible Chait careens off the rails, basing his entire argument on a presumed (and fanciful) parallel with Arnold Schwarzenegger's two terms as governor of California: The grossly unqualified non-politician with few ties to the Republican Party at first acted like an imbecile but then became a flexible and highly effective governor. Might not Trump do the same?

Never mind that Schwarzenegger left office with a 23 percent approval rating and a massive hole in the state budget. The ominous fact is that a president is exponentially (and when it comes to nuclear weapons, infinitely) more powerful than any state's executive officeholder. Which means that the stakes in a race for the presidency are exponentially higher as well.

Though he doesn't make the case explicitly, Chait presumably thinks that Trump would do less harm than a President Rubio or Cruz because he has distanced himself from the ideology that dominates the Republican Party — and because his wealth places him beyond the reach of manipulation by the party's big-money donors. But that independence — the same independence that led him to blow off the final Republican debate before the Iowa caucuses — makes Trump more dangerous than standard-issue Republicans, not less.

A President Rubio or Cruz governing with congressional majorities would do lots of things that Chait and I think are bad for the country. But they would be quite predictable things: tax cuts for high-income earners, big increases in defense spending, massive deficits, the repeal of ObamaCare, and so on.

What would a President Trump do? Aside from rounding up and deporting 11 million undocumented immigrants, building a massive wall along the southern border, (somehow) making Mexico pay for it, and forbidding Muslims from entering the country — each one of which would be quite bad — it's impossible to say. Untethered from the constraints that traditions, parties, donors, and other establishment institutions normally impose on politicians, Trump really would be his own boss, relying solely on his own temperament and judgment to determine which policies to pursue.

Even if Trump hadn't already demonstrated in a thousand ways that he possesses the temperament and judgment of a childish, vindictive bully, this would be an alarming prospect.

As it is, we simply have no way to know how Trump would govern. And that should be more than enough reason to stand against him with everything we've got.