Last month, four studies were published by the Voter Study Group that, taken together, provide a serious challenge to some of the most common—and perhaps cherished—beliefs of what went down during the 2016 presidential election. (One might be tempted to write that they “destroyed,” or “blew up” the conventional wisdom, but that would be overstating it.)

The studies used a unique database that offers some advantages over previous efforts to determine what motivated voters last year. Rather than surveying random samples of, say, 1,000 voters at various points in the race, the Voter Study Group researchers, in conjunction with the polling firm YouGov, repeatedly surveyed the same panel of 8,000 voters before and after the 2012 election, and then again before and after the 2016 election. This allowed them to see how individual voter’s preferences changed over time.

Ruy Teixeira and Robert Griffin, co-authors of one of the papers which looked at Trump’s “appeal” to his supporters, explained why this approach offers a real advantage over individual polls or analyses of registration data. “As counterintuitive as it may seem,” they write, “there is a substantial amount of evidence that voters shift their opinions to more closely match those of their preferred candidate.”

So, suppose that we find a relationship between support for Trump and a given attitude in a typical survey. We have to ask ourselves: Does this person support Trump because they had this attitude or does this person have this attitude because they support Trump? As a specific example, did Trump tap into latent racial and cultural resentments or did voters adopt some of his positions once they had decided to support him?

The most vivid example of this phenomenon may have been the dramatic increase in support for Russia, Vladimir Putin, and WikiLeaks among Republicans over the course of Trump’s campaign. That was a rather clear effect of Trump’s rhetoric on the stump.

Similarly, polls conducted during the campaign found that there were significant differences on the issue of trade between Trump’s primary voters and those who backed other Republican candidates, and then between Trump and Hillary Clinton supporters in the general election. By the end of the campaign, a Pew survey found that Trump voters were twice as likely as Clinton voters to say that “free trade agreements have been a bad thing for the US,” and “free trade agreements have definitely/ probably hurt your family.” This led to dozens of stories from areas hit hard by offshoring, and a lot of analysis about how the politics of trade were reshaping the race, and perhaps the two major parties’ coalitions.

But those surveys offered a snapshot of public opinion after a campaign in which Trump made trade a central issue in both the primaries and the general election. When Teixeira and Griffin looked back at how the same voters viewed trade years before the 2016 campaign got underway, they found a different story.

Teixeira says that “it isn’t that Trump’s voters weren’t unenthusiastic about trade, or skeptical of trade, or that they didn’t generally think trade was causing problems with jobs—all of that was true. But the problem is that you couldn’t really differentiate them from the rest of the voting population on that basis.” In the Republican primaries, Teixeira says that only John Kasich’s supporters seemed to differ from the rest of the Republican coalition in their views of trade. And then, “in the general election, there were some differences between Trump supporters and Clinton supporters on the issue but they were relatively small and didn’t appear to have much influence on voters’ choices.” He cautions that this doesn’t mean that trade didn’t play a role in the campaign, but “its salience as an issue may have been exaggerated.”

One of the most significant findings from the Voter Study Group came from John Sides, a political scientist at George Washington University. His paper, “Race, Religion, and Immigration in 2016,” found that those white working-class Obama-to-Trump voters who have been the subject of so much talk over the past year were already moving toward the Republican Party long before Donald Trump descended down that escalator and suggested that a large share of Mexican immigrants are “rapists.”

“The movement of non-college educated whites to the Republican Party prior to Trump is the piece that people are missing,” Sides says. “Some of those voters were not going to be Democratic supporters in 2016 regardless of who the candidates were. And it suggests that some of their shift away from the Democrats had to do with the election of Barack Obama and the dynamics of his presidency rather than the dynamics involving Trump and Clinton in 2016.”

That last part isn’t the whole story. It’s likely that Trump’s xenophobic rhetoric and racial demagoguery did help clarify the differences between the major parties’ respective coalitions. But Sides says that the Obama presidency was itself illuminating in that regard. You might look at the movement of a chunk of non-college-educated whites to the Republican Party in 2016 as the tail end of a long and non-linear realignment that began during the civil-rights era. At that point, whites in the South bailed from the Democratic Party en masse. As for non-college-educated whites, Democratic presidential candidates have only broken 40 percent among this group once since 1980—Bill Clinton did it in 1996—but those numbers obscured the fact that there were significant regional differences. Even as late as 2012, Democrats enjoyed an advantage among whites without a college degree on the Pacific coast and in the central Northern states, and were competitive in other regions.

But in 2008 and 2012, there was a lot of talk about the rise of the “Obama coalition,” which was led by educated urban whites, women, and people of color. This helped cast the differences between the parties in sharp relief. Sides found that “the percentage of white non-college voters who perceived that the Democratic Party was to the left of the Republican Party on the issue of how much the government should help improve the status of African Americans grew dramatically over the Obama years. The Obama administration was clarifying on this issue, and that may have hastened their departure from the Democratic Party.”

But that wasn’t the only shift. It appears that Obama’s presidency caused some movement within both parties’ coalitions. According to Sides’s study, around 9 percent of Obama’s 2012 voters, many of them less-educated whites, went for Trump in 2016. But at the same time, 5.4 percent of Romney voters pulled the lever for Clinton in 2016, and many of these voters were white folks who had at least some college education.

Overall, Sides found that the share of the electorate that voted consistently for candidates of one party or the other in both elections was typical, and even a few points below the share that switched parties between 2008 and 2012. Consider how much ink has been spilled, and how many hands have been wrung on the question of how Democrats might win back those white working-class voters. Was it the messaging? The messenger? Some magical formula that Trump possessed? We didn’t hear this kind of concern about the college-educated voters who went in the other direction.

More importantly, both Sides and Lee Drutman, a political scientist at Johns Hopkins university and a New America Fellow who authored a third report, “Political Divisions in 2016 and Beyond,” told me that most of those Obama-to-Trump voters are now Republican partisans and are unlikely to move back. (Ruy Teixeira was somewhat more hopeful of Dems’ recapturing the voters who abandoned them this year if Trump and the Republicans cause a lot of damage to the economy and/or the health-care system.)

Even if there was already movement within the coalitions, Trump certainly tapped into something with his campaign. On the question of whether Trump voters were motivated by economic or cultural anxiety, Sides says, “If you look at this long list of attitudes that were measured in late 2011, and you ask how strongly were these attitudes associated with a vote for Obama versus Romney, and then how strongly were they associated with a vote for Trump versus Clinton, only three became more strongly associated with a vote in 2016: your views of immigration, your views of African Americans, and your views of Muslims.”

Those who thought the economy was doing poorly tended to vote for the Republican candidates in both elections, but the economy didn’t become more salient an issue between the two cycles. So views of the “other,” says Sides, was the “distinctive aspect of this election.”

And yet, the economic versus cultural anxiety debate was also a false dichotomy. The data show that that they were intertwined. Teixeira and Griffin found that those who were pessimistic about the economy in 2012 became more negative toward immigration over the course of the study. And Sides says that in his unpublished data, “a really good predictor of whether you supported Trump in the Republican primary wasn’t so much ‘Are you worried about losing your job?’—pure economic anxiety—as it was ‘Do you think that whites lose out on jobs because they have to be given to minorities?’” He concludes that “if economic anxiety mattered, it was because it was bound up with concerns about my group’s status versus this other group’s.”

Another finding that seems to run counter to some well-worn narratives of the 2016 election comes from Lee Drutman’s paper. While we hear a lot about how there’s an existential battle between progressives and moderates for control over the Democratic Party, Drutman found that “there’s not a lot of divide on ideology. Across a wide range of issues, Clinton’s primary voters looked a lot like Sanders’s voters. The party is more united on first principles than the Republican Party is.”

That’s somewhat consistent with the aggregated exit polls from the primaries. They found that most Democratic primary voters wanted continuity with Obama’s policies, and Clinton won those voters handily. Sanders won by a large margin among those who wanted more liberal policies, but one in three in that group supported Clinton, and among those who wanted the next president to pursue more conservative policies, Clinton won a narrow 49-45 victory over Sanders. The two groups’ demographic skews dwarfed their ideological differences, with Sanders crushing Clinton 71-28 among voters under 30, and Clinton beating Sanders by the same margin among nonwhites.

According to Drutman, the one area “where Sanders’s supporters look quite different from Clinton’s supporters is in their general support for the political system overall.” He says that “the stereotype that Clinton was the establishment candidate and Sanders was the outsider candidate is absolutely true,” and believes “this insider/outsider divide is quite significant in the Democratic Party, because it is a question of personnel. And there’s a feeling among Sanders supporters that the political system has failed us, and that may become more pronounced if that issue isn’t addressed. And it’s unclear who will address it.”

There were also some differences on trade—Sanders supporters held somewhat more negative views—and on Islam, which Clinton’s supporters tended to see in a more positive light. But an important caveat here is that, unlike the studies by John Sides and Teixeira and Griffin, Drutman’s questions were only asked of the panel in 2016. So there may be a campaign effect on the issue of trade, which Sanders emphasized throughout his run. As for their voters’ differing views on Islam, that’s probably due to the nature of the candidates’ respective coalitions. Drutman notes that “Sanders did better among disaffected rural whites, who tend to have some negative views toward Islam.” At the same time, around a quarter of American Muslims are black, and Islam has always been part of the fabric of African-American communities.

In the immediate aftermath of an election, the demand for analysis of what just happened outstrips the supply of solid information. That’s especially true after a race that produces surprising results. Despite having limited data, political reporters and pundits rush to explore what shaped the race. They interview voters, offer up their gut feelings and extrapolate from what data are available. And the conventional wisdom of a race often becomes set, and hardens, before more granular data become available.

At a minimum, the Voter Study Group’s research, which includes a fourth paper, by the CATO Institute’s Emily Ekins detailing a sort of taxonomy of Trump voters, call a number of widely held conclusions about last year’s race into question. And they’re planning to go back into the field and ask those same panelists more questions in the future, so we should get an ever-better idea of what informed Americans’ political preferences.