The role of trade in the 2016 U.S. election is pretty confusing. Both the Democrats and the Republicans have one candidate who talks about trade all the time. Donald Trump claimed in the March 10 debate that he doesn’t differ much from the Republican Party tradition except for “one primary respect, and that’s trade.”

Many attribute Bernie Sanders’s upset victory in Michigan to his anti-trade message. Some argue that Trump and Sanders are “pulling different ends of the same troubled American heartstring,” suggesting that Trump may ride a bi-partisan populist coalition to victory in November.

The other candidates rarely mention trade unless they are asked about it. Trade barely registers on lists of most important problems in national surveys of Americans. And Americans on average are actually pretty upbeat about trade (although overall support is a bit ambivalent). So what’s going on?

1. Democrats who oppose trade don’t like Trump any better. But Independents do.

Donald Trump argues that he can pick off Bernie Sanders’s voters on trade if the Democrats choose Hillary Clinton as their candidate. That’s an important part of Trump’s argument that he has appeal beyond traditional Republican voters.

But what does the evidence say?

The pilot 2016 American National Elections Survey asked 1,200 survey respondents: “Do you favor, oppose, or neither favor nor oppose the U.S. making free trade agreements with other countries?” The survey also asked people to rank each candidate on a 100-point thermometer scale, where higher values indicate warmer feelings. The chart above plots the average candidate’s thermometer rankings by partisan self-identification and positions on trade.

Within parties you pretty much see what you’d expect. Republicans who oppose trade agreements (about 25 percent) have the warmest feelings towards Trump. Sanders scores better than Hillary Clinton among Democrats that oppose trade agreements (also about 25 percent), but only just. Sanders lags quite far behind Clinton among Democrats in favor of free trade agreements.

But Trump is wrong. This difference in opinion doesn’t carry over across the party divide. Democrats who oppose trade agreements and those who favor them are about equally unsympathetic towards Trump. That works in the other direction, too. Sanders doesn’t get a boost among Republicans who oppose trade.

One explanation is that Trump and Sanders supporters oppose trade for different reasons. Sanders supporters are skeptical of capitalism. They believe that workers must be protected from the effects of free markets, including free trade. Trump supporters are nationalists. Sanders supporters are not. Unlike Trump supporters, Sanders’ supporters look favorably upon immigrants even when they are skeptical about trade.

The story looks different among independents. Trump gets the highest average sympathy rating of all candidates among independents who oppose new trade agreements. Moreover, independents who do not support trade also give Clinton very low average thermometer scores. This can be important come November: Thirty percent of self-identified independents in this sample say they oppose new trade agreements.

2. Free trade agreements really have hurt some American workers. But it’s not clear new protectionism will help them.

Paul Krugman put up this graph to show that manufacturing jobs have long been on the decline in the U.S. Recent trade agreements that opened up trade with Mexico (1994) or China (in 2000) can hardly be held responsible for something that’s been going on since the 1950s.

Krugman looks at manufacturing jobs as a proportion of all employment. That tells us something important about the relevance of manufacturing for the U.S. economy as a whole. But it doesn’t tell us is whether manufacturing jobs have disappeared. The graph below plots the absolute number of manufacturing jobs. Here you see a dramatic drop since 2000.

A new article (forthcoming in the American Economic Review) by Justin Pierce and Peter Schott attributes much of this decline to the U.S. granting permanent normal trade relations to China in 2000. This may not matter much for the U.S. economy as a whole, but it affects some local communities a great deal.

For example, these same economists also find that counties that were more exposed to competition from China suffered increases in mortality rates.

David Autor, David Dorn and Gordon Hanson similarly show that increased trade with China has severely harmed many U.S. workers.

It is not at all clear that new protectionism or “better trade deals” will do anything for the people who have lost manufacturing jobs. But some voters are legitimately upset about what trade deals have meant for them.

Public opinion scholars find that education correlates most consistently with opinions about trade. Highly educated individuals are better able to compete in a global labor market. Especially the service industry benefits heavily from free trade. A college education also exposes individuals to ideas about the benefits of free trade.

By contrast, less-educated Americans are most likely to see their jobs being outsourced to China, India or elsewhere. Trump draws a lot of support from non-college educated voters. That’s not the case for Sanders.

3. Politicians rarely campaign on foreign trade. But when they do they say they want less of it.

2008 elections for House of Representatives. Graph by Peter Sima.

Bernie Sanders and Donald Trump highlight their opposition to past and future free trade agreements at seemingly every opportunity they get. The other candidates mostly avoid the issue. When pushed in debates, they are typically critical of specific aspects of trade agreements, including ones they have voted for in Congress or have otherwise supported. No one explicitly campaigns on a pro-free trade platform.

This turns out to be a pretty common pattern in recent U.S. politics. One of my very talented PhD students, Peter Sima-Eichler, examined TV advertisements by candidates for the 2002, 2004, and 2008 congressional elections. Only 12 percent of these ads mentioned trade. Almost all of them advocated for more protectionism. The only exceptions were a few ads by  former South Carolina senator Jim DeMint and New Hampshire Sen. Jeanne Shaheen. Many congressmen who ran anti-trade ads actually had pro-trade voting records.

One explanation for this is that the costs from trade are usually much easier to see than the benefits. The map above shows that trade ads were heavily concentrated in parts of the country that lost many heavy manufacturing jobs. It’s not that hard to directly link these hardships with specific trade agreements. This is why, as I blogged last year, there is evidence that job losses that result from trade have bigger consequences in elections than other job losses.

By contrast the benefits of trade are more spread out and more difficult to associate with agreements. Americans would pay more for their iPhones, cars, and many other products if there were no free trade agreements. But few people make that link. And it’s a difficult story to tell in a TV ad or campaign rally.

Elites in both parties are much more favorable towards foreign trade than the general public. Both Trump and Sanders voters are anti-elitists. That’s the main (perhaps the only) thing they have in common.