In September, Cook Political Reports elections guru David Wasserman argued that Democrats would be foolish not to court non-college-educated white voters in 2020. That group may make up 45 percent of the electorate nationwide, he wrote, but it represents a majority of the electorate in key battleground states—Wisconsin, New Hampshire, Michigan, Minnesota, and Pennsylvania—and casts almost half of the ballots in North Carolina.

That set off another round of debate about whether and how the Democrats should “win back” the non-college-educated whites who played a crucial role in delivering the Rust Belt, and ultimately the Electoral College, to Donald Trump in 2016.

Some analysts argue that progressives who urge Democrats to focus on turning out their core base—people of color, unmarried women, and younger voters—are too cavalier about the consequences of continuing to lose less-educated whites. Those progressives in turn worry that the pundits and moderate Dems who obsess over working-class whites rarely define what appealing to those voters would look like in practice. Does it mean shifting the party closer to the center and putting less emphasis on issues that matter to the base, like discriminatory policing, reproductive health care, and LGBTQ rights?

But these debates miss quite a bit of evidence, direct and indirect, that Democrats have already “won back” enough white working-class voters to compete next year. Since the 1980s, Democratic candidates have proven that they can win elections while losing whites without a college degree by a significant margin. Obama won 36 percent of their votes in 2012. Bill Clinton averaged 41 percent in his two victories. And in 2020, the candidate will likely need to win a smaller share of white people without a degree, because that group has long been declining as a share of both the electorate and the broader population. According to Gallup, their share of the population has declined by three percentage points since 2014. And a study released by the Center for American Progress in October projects that next year their share of the electorate will be 2.3 points lower than it was in 2016.

The reality is that the Democratic candidate is unlikely to do as poorly with this group as Hillary Clinton did. In 2016, despite winning the national popular vote by a significant margin, she won just 28 percent of these voters, according to Pew, and that wasn’t enough to deliver Michigan, Wisconsin, and Pennsylvania. According to a data set that combines survey and voter registration data with election results, Clinton lost non-college-educated whites by a 28-point margin in 2016, significantly worse than Obama’s 10-point deficit in 2008 or his 21-point gap in 2012.

A similar analysis looking back further found that Al Gore lost working-class whites by 17 points in 2000, and they went for George W. Bush over John Kerry by 23 points in 2004. Clinton also fared significantly worse among this group in 2016 than Democrats did overall when Republicans crushed them in midterm waves in 2010 (by 23 points) and 2014 (by seventeen points).

There are three reasons to believe that Clinton’s performance with non-college-educated whites in 2016 was an outlier, and that Democrats’ support among this group has already reverted to its longer-term trend line—one of gradual decline that has been offset by demographic shifts in the electorate.

First, Trump’s election represented a perfect storm that’s unlikely to be repeated. The Clinton campaign’s e-mails were hacked and dribbled out strategically over the final six weeks of the campaign; then–FBI Director James Comey broke agency protocol by announcing that he was reopening an investigation into Clinton’s own e-mails 11 days before the vote. According to an analysis by FiveThirtyEight, that story dominated the news down the final stretch and correlated with a significant decline in Clinton’s poll numbers—decreasing her support by between 1 and 4 percent nationally.

Second, the Democratic Party lost non-college-educated whites by 21 points in the 2018 midterms, the same margin as Obama did when he won in 2012. Midterm electorates aren’t the same as in a general election, when many more voters go to the polls, but 2018 featured historically high turnout for a non-presidential cycle and was more similar to a general presidential election than typical midterms.

And polling data supports the idea that 2018 can tell us something about the mood of the public. Polls have found Trump losing significant ground among whites without a college degree relative to his performance in 2016. The decline has been especially apparent among less-educated white women—a group he carried by 27 points in 2016 but which broke for a generic Democrat by six points in the August NBC/Wall Street Journal survey. Last month’s Fox News poll found Trump beating Biden by just one point among these women, topping Sanders by three points and losing to Elizabeth Warren by three points.

Remember that Trump won Michigan, Pennsylvania, and Wisconsin, which gave him enough Electoral College votes to win the White House, by a margin of less than one percentage point.

Meanwhile, Trump’s approval rates have cratered among all voters in all of the battleground states. Between Trump’s swearing in and the middle of last month, his net approval numbers (approval rate minus disapproval rate) fell by 23 points in Wisconsin, 21 points in New Hampshire, 21 points in Michigan, 18 points in Minnesota, and 19 points in Pennsylvania, according to Morning Consult’s polling. While those numbers aren’t broken down by race and educational attainment, those are the states where non-college-educated whites make up a majority of the electorate, according to David Wasserman’s data. (Trump’s net approval has also declined by 19 points in Ohio, 21 points in North Carolina, and 24 points in Florida.)

One final point on this debate: The only demographic that saw a significant decline in turnout between 2012 and 2016 was African Americans. The conventional wisdom is that this was because Barack Obama wasn’t on the ballot, but the reality is that this defied a longer-term trend of increased black turnout—black turnout was slightly lower in 2016 than it was when John Kerry was on the ballot in 2004. Voter suppression certainly played a role. If all else were equal, it’s possible that even with 28 percent of non-college-educated whites, Clinton might have eked out a win if African Americans had been more motivated to vote or hadn’t faced relentless voter suppression in places like Milwaukee and Detroit.

Given the paper-thin margins by which Trump won, Democrats probably don’t need to win the 36 percent of whites without a college degree that Obama got in 2012 to put together a winning coalition. With Trump firing up the Dems’ natural base, a shift of just a couple of points among this group could be decisive. And there’s every reason to believe that kind of movement is already baked in as we head into 2020.