Bernie Sanders’s base isn’t the working class. It’s young people.

After Bernie Sanders crushed Hillary Clinton in the West Virginia primary last week, the national media was ready with an explanation: the white working class.

The New York Times and The Atlantic, for instance, both attributed Sanders's win to his strength among low-income white workers. "White Working-Class Voters in West Virginia Pick Sanders Over Clinton," read NPR's headline.

This trope has become the conventional wisdom in the media, with the Wall Street Journal, the Nation, The Huffington Post, and a host of other outlets (including me at Vox) stating as fact that downscale whites have formed a crucial piece of Sanders's base.

This interpretation makes for an interesting narrative, but it's missing the real story. Sanders's victories aren't being powered by a groundswell of white working-class support, but instead stem from his most reliable base since the start of the primary: young voters.

Because young voters also tend to have lower incomes, the massive age gap between Sanders and Clinton has sometimes looked to observers like a gap in economic class, according to political scientists Matt Grossmann and Alan Abramowitz.

But the most salient divide in the primary is not between rich and poor. It's between young and old — and between white and black.

"What we found in Michigan is that the differences between the candidates were all based on age, and it just happens that younger people are poorer on average," said Grossmann, a professor at Michigan State University. "Age differences in Michigan were the key to the Democratic primary, and income differences are an artifact of that split."

Is Sanders winning the "white working class?" Well, it depends.

There are two main ways of trying to define the "white working class," and it's useful to clearly distinguish between them.

One is to say that the white working class includes every white American voter who is currently below a certain income threshold — say, $30,000 — or currently lacks a college degree.

But under this criteria, the white working class also includes a large group of young people and enrolled students who will soon join the middle or upper class and aren't necessarily facing any real material deprivation. (A junior at Harvard with a job lined up on Wall Street may have an income of $0 right now, but she's hardly destitute.)

If this is your definition of white working class — essentially, "all white people not currently making a lot of money" — Sanders really does appear to be winning it. And that's how we've gotten a ton of media coverage about his "white working-class" base, which tends to be based on exit polls showing white people with low income levels or partial college educations supporting his candidacy.

The problem with this definition is that it's not the one journalists have tended to evoke when writing about Bernie Sanders's "white working-class" base.

Instead, as this Jacobin feature illustrates, the national media (again, myself included!) have often said or implied that Sanders is winning with a very different definition of the white working class: middle-aged and older Americans who are at the bottom of the economic ladder, struggling financially, and generally denied opportunities in the job market.

"Even in losing, Sanders has shown that a candidacy based on economic populism can win back some voters who long ago deserted the Democratic Party," wrote the New Yorker's George Packer.

The data supports the first definition of Sanders's "white working class" base — all white people earning little money. But according to the political scientists I spoke to closely studying the question, the second conception of Sanders's "white working-class" base doesn't have the evidence to support it.

Accounting for age, is Bernie Sanders winning among poorer voters?

There's a big limitation to the exit polling data most election watchers use to draw conclusions about demographic trends in the primary: It doesn't let us dig deep enough.

The polling firms release data saying which candidate has won with which age group, and which candidate has won with which income or education group. But they don't give out the results for the income groups within each age group, according to Abramowitz, a professor at Emory University.

Political scientists have a couple of ways to try to get around this problem. One, taken by Grossmann in Michigan, is to use separate survey data to bore into what the exit polls don't reveal: Who does better by class within each age group?

If Sanders's "white working-class" voters aren't just college students, you'd also expect him to be doing better among downscale middle-aged white voters than rich ones.

But this turned out not to be true: Low-income white people in their 40s, 50s, 60s, and 70s did not break for Sanders. There was little difference in support by income among older voters, with higher-income older white voters actually more likely to support Sanders, according to Grossmann's Michigan data.

"My main concern is that the image of Bernie-supporting older poor people who've lost their factory jobs to trade is not supported," Grossmann says. "I'm least supportive of the idea that there's a population of white, older workers who lost their jobs and are now supporting Sanders. There's very little evidence of that."

Similarly, Abramowitz ran a multivariate analysis to help figure out this question. Abramowitz looked at a large survey data set and asked: What forms of identity actually predict support for Bernie Sanders or Hillary Clinton?

"It was age, and beyond that nothing mattered. Maybe ideology mattered a little bit," he said. Income was not a factor.

(Of course, as Slate's Jamelle Bouie has pointed out repeatedly, the white working class does not reflect the working class overall. And according to a Washington Post analysis in late April, Clinton has won among all households earning less than $50,000 by an 11-point margin.)

Polling data: Hillary Clinton is winning among older white working-class voters

I took another approach. As evidence of Sanders's working-class base, several of his supporters have pointed to a Reuters online polling data set that does indeed show him doing better than Clinton among white voters with low incomes and among voters who don't have college degrees — by a 37-30 margin.

But if you look a little more closely at the data, it turns out Sanders only does better among downscale whites who are also young.

For instance, he commands a massive 68-10 lead over Clinton among voters younger than 29 without a college degree. He holds similarly massive leads among young voters earning less than $25,000 per year.

However, once you get to the middle-aged and the elderly, Sanders ceases to be the candidate preferred by the majority of low-income white voters.

Among white voters over 40 without a college degree, it's Clinton who leads by a 43-35 margin, which is about the same size as her national polling lead. Among white voters in their 50s without a college degree, Clinton is up by a 46-35 margin.

This suggests that Grossmann's reading of the Michigan data is correct: that "the image of Bernie-supporting older poor people who've lost their factory jobs to trade is not supported."

The wide age split in the primary can cut both ways. Much news coverage has said Clinton is beating Sanders among black voters, which is true in the aggregate. But as I've written , several polls have put Sanders ahead of Clinton among young African-Americans; in the Reuters polling data, for instance, Sanders beats Clinton by 25 points among black voters aged 18 to 29.

The case for believing Sanders really is winning the white working class

Some Sanders supporters are ready with a response to the idea that he's actually losing older downscale whites: that it doesn't make sense to exclude young people from an analysis of the white working class.

For example, blogger Carl Beijer has argued for months that Sanders is ultimately the candidate favored by more Americans with lower incomes.

"Clinton is the candidate of the rich, winning clear majorities with Americans who make $75,000 or more," Beijer writes. "And Sanders is the candidate of the poor, gaining slim majorities with voters who make $50k or less."

This is true, and a coherent way of viewing Sanders as garnering more white working-class supporters. When I asked the political scientists about it, they didn't dismiss the idea out of hand.

"Young people are more economically disconnected and less secure in their jobs," Grossmann said. "I don't want to discount that perspective entirely."

We don't have a great way of figuring out what percentage of young people in the polls are also likely to soon get good jobs. There does appear to be some evidence, according to Grossmann, that rich young people tend to favor Sanders slightly more than poor young people.

But based on the data we have, it’s clear that when we talk about Bernie and the white working class, a little more precision could go a long way.