Sure, Donald Trump has led the charge against agreements like NAFTA, but Hillary Clinton has also adopted an anti-free trade stance after contending with the popularity of that position in her primary battle with Bernie Sanders.

But the narrative that the electorate is pushing elected officials towards trade-skeptical policies doesn’t actually jive with poll numbers, which show that a majority of Americans think that increased trade is good for the economy.

So why have the candidates moved left on trade, even if the electorate has not? Economists Bradford Jensen, Dennis Quinn, and Stephen Weymouth of Georgetown’s McDonough School of Business may have an answer. In a new working paper shared with Fortune, the economists show that international trade “directly influences U.S. presidential elections.” The incumbent party tends to win counties with more highly skilled service sector jobs, not the type that are typically lost to trade, and lose counties in which there is a high concentration of trade sensitive low-skilled manufacturing.

“In national-level models, we show for the first time that increasing imports are associated with decreasing presidential incumbent vote shares,” the authors write. “These effects are large and politically consequential. We find an Electoral College incentive to protect the manufacturing sector and to oppose free trade agreements.”

That’s because the states that happen to be swing states, meaning they are crucial to an electoral-college victory, also happen to be places where there is a high concentration of manufacturing employment, like Ohio, Michigan, Iowa, and New Hampshire. The basic idea is that workers in these sectors have been hurt economically by deindustrialization and have been expressing their anger at this trend by voting against politicians in office. Politicians running for office, whether they are an incumbent or a challenger, try to leverage or blunt this effect by criticizing status-quo trade rules.

But there are also structural forces that work in the favor of greater free trade too. That’s because rising exports can increase jobs, as much as rising imports are killing them in traditional manufacturing sectors. The authors find that workers in high-skilled exportable services and high-skilled exportable manufacturing increasingly vote for the incumbent candidate. This reality, combined with the fact that just 10% of the labor force is now employed in manufacturing, means that there is plenty of electoral support for policies aimed at increasing trade. This support just happens to be more concentrated in non-swing states.

What’s more, as the authors point out, the sectors in the U.S. economy that benefit from more trade, rather than are hurt by it, are rising. For example, services now account for roughly 30% of all exports, and that this number is on the rise. These data explain why poll numbers show most Americans are in support of free trade at the same time that presidential candidates—who are fighting over the support of undecided voters in a few swing states—are increasingly not.

They also explain why, despite protectionism punching above its weight as an issue in presidential elections, there has been little movement on the issue in Congress over the years. Indeed, the last big vote on the free trade issue went free-traders way, when Congress voted to give President Obama fast track authority to negotiate the Trans Pacific Partnership Agreement.

What we are seeing in the 2016 presidential election, therefore, may be the last, dying gasp of protectionism. Sure, those who feel they have been hurt by free trade agreements are probably more motivated by their losses than the benefactors of free trade are motivated by their gains. Like many battles in modern politics, there are benefits to having a very angry minority in your corner as opposed to a majority that is happy, but not ecstatic, with the status quo. But this research out of Georgetown shows that the possible of voters who are angered by free trade is on the decline, and 2016 could be their last stand.