Bernie Sanders won Michigan in a shocking upset largely because he capitalized on a grassroots backlash against American free trade deals.
That, at least, is how the Vermont senator's surprise victory was interpreted by the New York Times, CNN, US News & World Report, the Nation, and scores of other news outlets (including Vox).
Similar interpretations followed Sanders's big victory in Wisconsin on Tuesday, with Sanders himself arguing that Hillary Clinton's "support for corporate trade deals" was a key ingredient in her defeat there.
The line has become repeated often enough to become something of a truism about the Democratic primary. There's just one problem with the narrative: It's not clear if it's true.
The counties hit hardest by manufacturing loss did not break big for Sanders
As pointed out by Slate's Jordan Weissmann, there's one obvious problem with this idea: The parts of Michigan particularly hit by manufacturing job losses weren't actually the most disposed to vote for Sanders.
Evan Soltas, a Princeton economics student who has contributed to Bloomberg View and Vox, created the below scatter plot to track how changes in Michigan counties' manufacturing levels tracked with Sanders's vote share. (The below graphic is shared with Soltas's permission.)
Soltas wrote he was surprised to find no connection between manufacturing job losses and Sanders support. "No matter how intense the local decline in manufacturing, Sanders won about half the Democratic vote," Soltas said.
A similar analysis was done by Brendan Greeley at Bloomberg, who compared gains in manufacturing over 14 years by county. Sanders won big in places like Ottawa County, which gained 7,000 manufacturing jobs since 1990, and lost big in places like Detroit's Wayne County, where manufacturing jobs fell by more than 71,000, according to Greenley.
"Counties that put proportionally more new people on shop floors were slightly more likely to vote for Sanders," Greenley said.
Mother Jones's Kevin Drum also notes that Sanders only won union households by a 2-point margin (about the same rate at which he won the state), and actually lost among voters who said they were most concerned about the economy (though most unions have also endorsed Clinton). If voters were drawn to Sanders because of his message on trade, you'd expect him to win that latter demographic.
Warning: Exit polls are very unreliable
The best argument that Sanders drew on a groundswell of anti-trade sentiment comes from exit polls in some states where he triumphed.
In Michigan, for instance, Sanders did about 10 points better than Clinton with voters who say "trade with other countries ... takes US jobs away," and 10 points worse among voters who said that such trade created jobs. There were similar results in the exit polls from Wisconsin.
But we have a massive chicken-and-egg problem here. Were these voters first drawn to Sanders because of his message on trade, or did they reflect his message on trade only after they decided to support him?
"There's an association on nearly every exit poll question. It doesn't mean it's causal," says Matt Grossmann, director of the Institute for Public Policy and Social Research at Michigan State University (and a Vox contributor). "Often voters choose the candidate first, and then they answer the survey question second."
If we thought Sanders was pulling in voters because of his message on trade, we'd expect him to be winning trade skeptics even in states that he's losing.
But that's not the case. In fact, Sanders has lost among voters suspicious of trade in several states, like Ohio and North Carolina, where Hillary Clinton also won the vote overall.
Another way of thinking about this: According to the exit polls, Sanders has consistently won among voters who say he'd be better at handling racial relations than Clinton by at least 10 points. But few would use that data point to conclude that Sanders's support is being driven by his stances on racial issues.
A far more plausible interpretation is that voters bought Sanders's message and then came to believe he was also the most trustworthy on this question.
Sanders's most loyal constituency tends to be overwhelmingly pro-trade
This doesn't mean, of course, that there's no evidence for the idea that Sanders is winning on his trade message.
But if it really were driving his support, we'd want signs of a large, untapped anti-trade constituency in the Democratic primary electorate that exists independent of Sanders's exit polling. And the overwhelming evidence here suggests that this just isn't the case.
Grossmann, the political scientist, pointed to national tracking data that shows a vanishingly small number of Americans consider free trade to be one of their top political priorities.
"Trade has never been mentioned by more than 2 percent of the population as the most important issue," Grossmann said. "Before this campaign, there was no evidence that trade was becoming a more important issue to the American public or that public opinion was becoming decidedly against international trade."
Additionally, what is by far Sanders's biggest constituency — young people — has repeatedly expressed support for free trade deals. (One Pew study from 2014 found that 69 percent of people under 30 think positively of free trade agreements, Grossmann noted.)
"If you want an explanation for Bernie Sanders's support, it needs to explain why young people are overwhelmingly flocking to him," he said.
Then there's a new Pew Research Center report recently highlighted by Vox's Matt Yglesias. It showed that Sanders's supporters view free trade agreements as a "good thing" for the US by a 55-38 margin.
We've seen overinterpretation of exit polls before. It's a dangerous thing to do.
In 2004, voters leaving the ballot box were given a simple exit poll. It asked them to identify their most important issues and gave them just a handful of options to choose from — including "moral values."
About 20 percent of people said they chose on the basis of "moral values," and these voters broke overwhelmingly for George Bush. A huge number of articles proceeded to say that Bush beat John Kerry because of his lead among "moral values" voters.
But that interpretation has since been picked apart by political scientists and later denounced by the pollster who conducted it. Professors D. Sunshine Hillygus and Todd G. Shields later analyzed the results and found that the notion of "moral values" was only "a very minor part of citizens' voting calculus in the 2004 presidential election."
This doesn't mean all exit polls are worthless. But Grossmann makes the case that to use them effectively, they should be more carefully checked with preexisting evidence about the state of the electorate.
On that front, luckily, we do have a clear explanation for the Democratic primary.
Rising income inequality is not just listed as one of Sanders's supporters top concerns. It's also showed up repeatedly as one of the biggest areas of concern for the country for years now.
"I think it's much more plausible that the concern with inequality is driving support for Sanders," Grossmann says. "That's been Bernie Sanders's signature issue from the beginning."