By Nate Cohn

New York Times

Sen. Bernie Sanders needed a surprise win to show he still has a path to the nomination. He got it.

Sanders managed to defeat Hillary Clinton by 2 percentage points in Michigan on Tuesday night, even though he trailed by at least 11 percentage points in every survey, and even though the demographics pointed toward a significant, if relatively smaller, advantage for Clinton.

Sanders’ win was so surprising that it’s hard to know what to make of it. Are we learning, for the first time, of a big latent advantage in the Rust Belt? Was it a fluke?

Yet one thing is clear: Clinton still has a significant advantage nationwide. Her delegate lead is so solid that she would continue to win, even if Sanders’ ability to defy expectations the way he did in Michigan became a trend.

>> No Great Explanation

It’s not terribly surprising that Sanders outperformed polls showing him down by more than 20 percentage points. It was hard to reconcile that kind of margin with the state’s demographics, which tilt slightly toward Sanders in comparison with the rest of the country.

White voters represent a well-above-average share of the Democratic electorate in Michigan. Black voters, who have essentially been the difference for Clinton in this campaign cycle, represent almost the exact same share of the electorate in Michigan as they do nationally.

Michigan is a mostly white working-class state that has suffered from manufacturing job losses, a receptive audience for Sanders’ views on trade and economics. It’s a relatively liberal state, too, where President Barack Obama has posted an above-average share of the white vote in two presidential races.

But these things don’t explain his victory. All elements considered, Sanders looked like a 12-percentage-point underdog, based on an unpublished demographic model of the results so far.

He beat expectations everywhere.

He did better than expected in Wayne County, home to Detroit, and Genesee County, which includes Flint, because of a relatively strong showing among black voters. He lost black voters by a margin of just 2-to-1, according to the exit polls, about half as much as his margin in the South.

He did better than you would expect in the affluent suburbs of Detroit, like in Oakland County, because he lost voters making more than $100,000 a year by only 5 percentage points. He lost voters making that much by 20 percentage points in Massachusetts and by 36 percentage points in Virginia.

He swept Clinton across most of the white working-class countryside and small towns — communities that look a lot like where he basically tied Clinton in Iowa.

Sanders had occasionally showed this kind of strength among white voters, but mostly in low-turnout caucuses. It’s tempting to look for another explanation, but none add up.

Some have speculated that Clinton voters stayed home or chose to vote in the open Republican primary because they thought the contest with Sanders was wrapped up, but the Democratic race has never looked especially competitive and it hasn’t led to a polling error like this.

It’s been suggested that the Sanders campaign benefited from airing additional advertisements to an extent it hadn’t on Super Tuesday, but nothing like this happened in Massachusetts — where Sanders didn’t air ads just for Super Tuesday but also for the New Hampshire primary, since New Hampshire is in the Boston media market.

You might wonder whether the race broke decidedly against Clinton in recent days, but there wasn’t much of a sign of it in Mississippi: The same model showing Clinton up by 12 points in Michigan showed her up, 84-16, in Mississippi; she ultimately won, 83-17.

>> Clinton Still Has Big Edge

If Sanders’ victory in Michigan could be explained easily by some quirk in the state, perhaps Clinton and her allies could write it off. But without an explanation, one has to wonder whether it will be the start of a trend — a bad sign for Clinton’s chances in other industrial states like Ohio, Illinois, Missouri, Indiana and Pennsylvania. Maybe it’s a reason to take the Sanders campaign’s assertions of strength in California or New York — two states where Clinton ought to be favored — more seriously.

But in the end, none of it seems like quite enough for a victory for Sanders. Clinton has already banked too much of a lead. As an example, the PredictWise chances for a Clinton victory changed only from 95 percent to 93 percent from Tuesday to Wednesday.

Imagine, for instance, a brutal stretch for Clinton, one where she underperforms the demographic projections by as much as she did in Michigan for the rest of the year.

She loses in Ohio and Missouri next Tuesday. States where Clinton was thought to have an advantage — like Arizona, New Mexico, Pennsylvania, Illinois, Indiana, California and Connecticut — become tossups. Clinton wins New York, but by just 8 percentage points.

She gets swept in the West, including big 40-point losses in places like Alaska, Idaho, North Dakota, Utah and Montana, and 30-point losses in Washington and Oregon. She loses by 20 points in Wisconsin and Rhode Island, by 30 in West Virginia and Kentucky.

Clinton still wins — and comfortably. She’s already banked a large delegate lead, and it has nothing to do with the so-called superdelegates.

Forty-three percent of all of the pledged delegates to the Democratic National Convention have already been awarded. She’s won those delegates by roughly a 60-40 margin. To overcome it, Sanders will need to do nearly as well from this point on. Not even the very strong showing for Sanders imagined above would be enough.

In fact, it still wouldn’t be very close. Sanders would basically split the delegates with Clinton the rest of the way — leaving him far short of the big 15-point advantage he needs.

Clinton still is likely to win clear victories in diverse or affluent states like Florida, North Carolina, Maryland and New Jersey — the only four states where she breaks 55 percent of the vote in this projection. The problem for Sanders is that many of these states are far larger than the places where he hopes to excel.

The reality is that the Democratic delegate rules — which award delegates proportionally based on a candidate’s share of the vote — make it extremely hard for Sanders to dig himself out of the hole he is already in. Indeed, he fell short of winning 60 percent of the delegates in Michigan on Tuesday night, despite the upset win. As a result, his burden in future contests grew a little larger.

Sanders needs landslides to counter landslides. Not even the results from Tuesday suggest he’s on pace to get them, at least not in states with enough delegates to counter Clinton’s lead.

© 2016 The New York Times Company