Nate Silver had a post up talking about possible impact. Relevant snippets:

https://www.natesilver.net/p/could-haley-stevens-cost-democrats

Michigan Democrats really need to knock off the infighting

You can watch the Stevens video for yourself, but to me at least, she comes across as fairly petty and vindictive. To be fair, though, so has El-Sayed, who disinvited Stevens from campaign events. There’s not much maturity on display here. If Democrats had to do it over again, maybe they’d have nominated Mallory McMorrow, or begged Gary Peters to run for another term.

AES does have solid support at this point, which helps him, and most voters have already said that they’ve decided

Good news for AES part #1a: he’s actually been expanding his lead over Rogers in recent surveys. Indeed, Michigan has been one of the better positive surprises for Democrats lately in polls that have mostly been fairly flat during the stretch run. Part #1b: he already has almost 50 percent of the vote. In fact, Michigan has among the fewest undecided voters4 of any Senate race at 4.8 percent; the lowest is Maine at 3.5 percent.

Earlier on, when AES and Stevens were fighting each other, AES wasn’t polling so well. This is bad for AES.

Bad news for El-Sayed: look at the trajectory of the polls in that first chart. He was actually trailing Rogers back when he and Stevens were fighting in the summer before the primary was complete. So there’s at least some evidence that voters in Michigan are sensitive to Democratic unity, or the lack thereof.

To summarize my understanding of the impact of endorsements: the political science literature generally says endorsements aren’t that important unless they’re surprising, and this one was surprising. Or at least surprising-ish: tensions between the candidates had already been escalating and, as I wrote above, it hadn’t seemed to hurt El-Sayed in the polls; in fact, he was gaining ground.

The impact is probably limited by the endorsement being so late; this helps AES

There’s also the fact that roughly 1.9 million Michiganders have already voted, which might reflect something like 40 percent of the eventual turnout there, not counting voters whose ballots are in the mail but haven’t been counted yet. Now, generally speaking, undecided voters don’t vote early: rather, committed partisans do. But that blunts the effect, to some degree. It would probably have been worse for El-Sayed if this happened last weekend, in which case there would also have been more time for the story to metastasize in the national media. But now, newsroom resources are stretched thin as everyone prepares for Tuesday.

Silver says that the fact it could give AES an “anti-establishment” image makes the impact of the endorsement harder to predict, as Michigan favors that

El-Sayed has been running against the Democratic establishment in Michigan, a state that has long had a penchant for anti-establishment, outsider candidates. Bernie Sanders won the state in 2016, Jesse Jackson did in 1988, and it was one of John McCain’s few wins against George W. Bush in 2000. Stevens famously became an avatar of the said establishment in a race where a major question was about AIPAC’s heavy spending on behalf of Stevens.

Then again, El-Sayed barely won that primary in what nearly was a massively wrong call by the polls. He’s frankly one of the weirder candidates we’ve seen in recent years: quite charismatic and very bro-y. But also he’s someone who can really get his back up instead of turning the other cheek. As someone who can be a bit confrontational myself, I sort of admire it in a weird way, but there’s a reason I’d never run for United States Senate. I don’t think there’s clear evidence on how this plays with voters; it would have been clearer if he’d won the primary by more.

Silver’s guess, based on some estimation and his model, is that this has maybe a 5% chance of causing the Michigan Senate seat to go Republican when it otherwise would have gone Democratic, and maybe a 1% chance of turning control of the Senate Republican when it would have otherwise gone Democratic.

So let’s say the Stevens endorsement of Rogers reduces El-Sayed’s margin by 1 percentage point. If he’d been on track to win by 4 percent, now he’ll win by 3 instead, for instance. Or, if he was going to squeak by with an 0.5-percent victory, now he’ll lose by 0.5 points.

That 1-point swing is just an educated guess, and might be toward the higher side of my prior. Again: not many undecideds in Michigan, many people have already voted, the story might read positively for some voters, and many purported “game-changers” prove to be nothing of the sort. Weigh that against the fact that El-Sayed’s numbers did seem to suffer during the heat of the primary campaign in the summer. Let’s just stick with 1 percent.

How often, in FLIPR, did El-Sayed win Michigan by 1 point or less? In Saturday’s simulations, about 5 percent of the time. Again, AES does have some cushion in the polls. But if the polls are off, they might be so off that he was bound to lose anyway. It’s not quite that Rogers has to hit the bullseye, but it’s a pretty narrow target.

Next question: how often, in FLIPR’s simulations, did El-Sayed beat Rogers by less than 1 point and Democrats finish with exactly 51 Senate seats? In other words, a narrow win in Michigan that decided control of the Senate? That happened in 316 of our 40,000 simulations. So about 1 percent of the time, if you round up slightly.

(All of this is fairly linear, by the way. If you think the endorsement will hurt El-Sayed by half-a-point, the chances that it flips Michigan against him go down to about 2.5 percent, and the odds that it flips the Senate to 0.5 percent.)

So there’s a chance, indeed, that Haley Stevens just cost Democrats the Senate. It’s just not particularly high, unless your estimate of the endorsement’s importance is higher than mine.