Doing political analysis can feel like listing off truisms. Democrats win big with African American, Hispanic and Asian American voters! College-educated white people are moving left! White evangelicals vote GOP! Nonreligious people love Bernie! Republicans now own the white working class!

Those assumptions didn’t emerge out of nowhere — Republicans really do win with non-college-educated people and evangelicals, and Democrats do get the majority of black, Hispanic, Asian and (more recently) college-educated white voters.

But these stereotypes of the party’s voters are, well, stereotypes. And they can lead people to see exaggerated, cartoonish version of the parties instead of the real things. Oversimplifications aren’t merely errors: They can lead to party strategists and campaign officials making bad calculations about how to win over demographic groups, with serious consequences for election outcomes. And Republicans and Democrats alike shouldn’t forget that there are plenty of non-college-educated white Democrats — and that keeping them in the party might require different things than winning back their counterparts who went for President Trump.

Make no mistake — Trump did win bigly with non-college-educated white voters. But these voters make up a huge percentage of the overall electorate, and they remain a big chunk of the modern Democratic Party as well. Just as Democrats shouldn’t make the mistake of assuming that immigration means Hispanic Republicans will soon be rarer than unicorns, both parties should be clear that working-class white Democrats are a politically and demographically distinct group, rather than confused Republicans. And ignoring them would be a mistake for both parties.

After the 2016 election, demographers Ruy Teixeira, John Halpin and Rob Griffin estimated that 62.5 percent of white voters without a four-year degree voted for Trump, and 31.6 percent voted for Hillary Clinton. That’s a huge margin, not that far below Clinton’s 36.5 percent win margin with Latinos. And other data sources agree with that general picture: the Cooperative Congressional Election Survey said Trump won non-college white voters 60 to 34 percent; the exit polls put the results at 67 percent to 28 percent; and Pew Research put it at 64 percent to 28 percent.

It would be easy to mentally round that 65-ish percent GOP support up to 100, and the 30-ish percent Democratic support down to zero. That would be a big mistake. White voters without a college degree made up roughly 45 percent of the electorate in 2016, meaning that when Democrats get a third of the group, they get millions of votes.

In fact, Pew recently found that 33 percent of Democratic and Democratic-leaning registered voters were non-college educated white voters, a figure that eclipses the percentage of Democrats who are college-educated white voters (26 percent), black (19 percent) or Hispanic (12 percent).

Put simply, Democrats aren’t starting from zero with the white working class. They start out with a real base that they should try to maintain (or expand on) if they want to win in 2020. And to do that, they need to understand these voters, how they differ from Trump’s contingent of white non-college-educated supporters and why keeping them in the Democratic camp might require a distinct strategy from one intended to win people who flipped Republican.

There are some easy demographic tells for why certain blue-collar white voters stuck with Clinton in 2016 instead of switching to Trump.

According to the Cooperative Congressional Election Study, an academic survey of more than 50,000 Americans, about 1 in 5 Clinton-voting blue-collar whites were born after 1985, compared to only 13 percent for Trump, who tended to do better with older voters. They also differ on religion — almost half of white working-class Republicans are evangelical Christians, and only 1 in 6 white working-class Democrats are. Forty-eight percent of voters who are blue collar but also on the left said they were Protestant, Roman Catholic or Mormon, but these Democrats aren’t as overwhelmingly Christian or evangelical as their equivalents on the right. Unionization wasn’t a big dividing line among these voters, however. Clinton’s blue-collar white supporters were more likely to be women, but the gap wasn’t enormous.

And there are important political and policy divides that determine whether working-class white voters break left or right. On immigration, 34 percent of blue-collar Clinton voters wanted to increase the number of border patrols on the U.S.-Mexico border, while 80 percent of Trump voters from the same demographic did. One in 5 blue-collar white Clinton supporters wanted to “identify and deport illegal immigrants” while roughly 3 in 4 Trump voters from this bloc did. Though there is some agreement on law-and-order issues — a strong majority of both groups felt safe with the police and supported background checks on all gun sales.

The two groups are also divided on abortion. More than 80 percent of Clinton-voting non-college whites said they’d support a policy that would “always allow a woman to obtain an abortion as a matter of choice” — a policy that 68 percent of Republicans in that same demographic group opposed.

And the groups are polarized on racial issues. About 3 in 4 Clinton-voting blue-collar whites agreed that “white people in the U.S. have certain advantages because of the color of their skin.” Six in 10 Trump supporters from this same demographic disagreed.

There is a real chicken-and-egg question here. Voters sometimes pick a party based on some strong identities, their policy instincts or a few important issues and adopt that party’s views on other issues. If Trump gained a voter’s trust by channeling his or her anti-establishment feelings, that voter might adopt Trump’s position on issues they know or care less about. And if Clinton grabbed a voter by successfully arguing that she was more experienced than Trump, that voter might also shift his or her positions to line up with hers.

But these numbers underline why we shouldn’t think of non-college educated whites as just one group. The label would theoretically include both a 70-year-old conservative evangelical retired welder in the rural South and a 26-year-old Bay Area socialist agnostic woman who dropped out of college to work at a start-up.

Neither party’s base is in perfect lockstep on every issue. It’s possible to imagine Trump losing some culturally right, economically left voters if his opponent successfully runs as a populist and hits Trump hard for bills such as tax reform. It’s also possible that if a Democrat neglects the working-class white voters who stuck with the party or intentionally tries to trade them for some other voters, a Republican will take that trade and again surprise the political world by winning on blue-collar white strength.

Some level of stereotyping is inevitable in politics. There’s nothing wrong with statements such as “Democrats win Hispanics by a solid margin” or “Republicans rely heavily on the white working class” — and exceedingly general language such as that can be necessary (or even helpful) for describing a country of more than 300 million people. But parties who turn shorthand into mental shortcuts are in danger of misunderstanding the electorate and losing winnable elections. And the obsessive post-2016 attention to working-class white voters risks mistaking the denizens of a couple of Midwestern diners for a huge and complicated voting bloc.

Note: Thank you to Shiro Kuriwaki and Brian Schaffner for walking me through the basics of using the 2016 CCES data.

Read more: