Huge turnout is expected in 2020, so which party would benefit?

It is commonly assumed that Democrats benefit from higher turnout because young and nonwhite and low-income voters are overrepresented among nonvoters. And for decades, polls have shown that Democrats do better among all adults than among all registered voters, and better among all registered voters than among all actual voters.

But this long-standing pattern has become more complicated in the Trump years. The president is strong among less educated white voters, who are also overrepresented among nonvoters. And Democrats already banked many of the rewards of higher turnout in the midterm elections, when the party out of power typically enjoys a turnout advantage and did so yet again, according to 2018 Times/Siena data.

Nationwide, the long-standing Republican edge in the gap between registered and actual voters all but vanished in 2018, even though young and nonwhite voters continued to vote at lower rates than older and white voters.

At the same time, the president’s white working-class supporters from 2016 were relatively likely to stay home. Voters like these are likeliest to return to the electorate in 2020, and it could set back Democrats in crucial battleground states.

Democrats have an opportunity to gain by tapping into another group: the voters on the sidelines of American politics who haven’t voted in recent elections or aren’t registered to vote at all.

This group, by definition, does not usually factor into electoral analysis, but a high enough turnout would draw many of them to vote. Analysts have speculated about a 70% turnout among eligible voters next year, based on the very large 2018 turnout — the highest in a midterm since 1914 — and on polls showing unusually strong interest in the 2020 election.

These adults on the periphery of American politics are probably more favorable to Democrats than registered voters are, but the story here is complicated as well.

They are not quite as favorable to Democrats as often assumed, in part because polls of adults include noncitizens, who are ineligible to vote. A large increase in voter registration would do much more to hurt the president in the national vote than in the Northern battleground states, where registration is generally high and where people who aren’t registered are disproportionately whites without a college degree.

The registered voters who stayed home in 2018

The voters who stayed home in 2018 were not much more or less likely to approve of the president than those who actually turned out, based on data from nearly 100 Times/Siena surveys, linked to records indicating who did or did not vote.

Overall, the president had a 47% approval rating among Times/Siena respondents who voted, excluding those who did not offer an opinion about the president. But he had a higher approval rating (48%) among all registered voters in the nearly 60 battleground districts and a handful of Senate contests surveyed before the midterms.

The Republicans lost their typical midterm turnout advantage, even though they didn’t give up some of their traditional demographic advantages. Young and nonwhite turnout was markedly higher than it had been in 2014, but still lower than that of older and white voters. Registered Republicans were likelier to turn out than registered Democrats, according to data from L2, a nonpartisan political data firm.

These traditional Republican demographic advantages were canceled out, and in some cases reversed, by two new Democratic advantages. The low turnout among whites without a college degree bolstered Democrats in much of the country, allowing college-educated whites to make up a larger share of the electorate.

As a result, the voters who turned out in 2016 but stayed home in 2018 were more likely to approve of the president: He had around a 50% approval rating among those nonvoters in Times/Siena data.

The increase in turnout among the young in 2018 came overwhelmingly from anti-Trump voters, giving the Democrats a wide advantage among voters under age 45. The advantage was largest among those 18 to 24: The president’s approval rating was 28% for voters in that group and 45% among those who stayed home.

It’s important to emphasize that the Times/Siena data is not representative of the country. The 2018 battleground districts were disproportionately white, well educated and Republican-leaning. Urban areas were almost entirely unrepresented and black voters were underrepresented as well.

After accounting for the differences between the battlegrounds and the country, the Republicans held a narrow turnout advantage on a national scale. The fundamental turnout shifts were similar, but the lower turnout among nonwhite voters hurt the Democrats more nationwide than it did in the relatively white battleground districts.

Overall, the president’s approval rating was 45.3% among registered voters and 45.7% among likely voters, according to our estimates, based on national voter file data, the Times/Siena polling and a district-by-district estimate of the president’s approval rating based on national election surveys.

The opportunity for Democrats, however small, is fairly clear here: It’s reasonable to assume higher turnout would draw from a pool of voters who are relatively likely to disapprove of the president.

The opportunity for Republicans is somewhat more subtle, but clear as well. The voters who turned out in 2016, but stayed home in 2018, were relatively favorable to Trump, and they’re presumably more likely to join the electorate than those who turned out in neither election. In a high-turnout election, these Trump supporters could turn out at a higher rate than the more Democratic group of voters who didn’t vote in either election, potentially shifting the electorate toward the president.

Those who aren’t registered but still might vote

A high-turnout election would draw from another group of voters: those who aren’t yet registered.

These voters are hard to measure. They are underrepresented in public opinion surveys, and there’s reason to wonder whether those who do take surveys are representative of those who don’t. They are also less likely to hold opinions on current events, including on the president. (For ease of comparison, those without an opinion of the president have been excluded from measures of the president’s approval rating.)

With those caveats in mind, the president’s approval rating among nonregistered voters stood at just 37% in an Upshot compilation of 12 surveys, conducted between December 2017 and September 2018, by the Pew Research Center and the Kaiser Family Foundation. Trump’s approval rating was at 43% among registered voters in the same collection of surveys.

The data includes over 14,000 registered voters and nearly 3,200 voters who aren’t registered, allowing for a fairly detailed analysis and comparison of the groups.

The president’s weakness among nonregistered voters is consistent with a long record of polling showing Democrats fare better among all adults than among registered voters, including in today’s FiveThirtyEight averages.

The potential for Democrats is obvious. But in general, these figures — and other polls comparing the adult and registered voter populations — exaggerate the opportunity available to Democrats because they include noncitizens, who aren’t eligible to vote.

People who aren’t citizens represent 22% of the nonregistered adult population, according to the Current Population Survey, and they’re very different demographically from citizens who aren’t registered to vote.

Just 11% of noncitizens are white and non-Hispanic, compared with 59% of eligible but nonregistered voters. This means that the pool of potential but not-yet-registered voters is more white and non-Hispanic than it might appear.

And because the Pew/Kaiser data indicates that almost all of President Donald Trump’s weakness among nonvoters is attributed to demographics — that is, nonwhite people tend to like him less — the political difference between registered and nonregistered voters shrinks considerably without noncitizens.

The uncertain consequences of higher turnout

Of course, not all eligible voters, or even all registered ones, will vote in 2020. It’s impossible to guess just who will. Either side could draw a relatively favorable group of voters to the polls.

Even if every single citizen were to turn out, the effect on the presidential race would not be clear. The president’s approval rating would probably sink by around a point, compared with the 2018 electorate. But the effect on individual states could vary widely.

The major Democratic advantage among nonvoters, their ethnic diversity, would do little for Democrats in the Midwest, where the population is more white and where nonvoters are likelier to be working-class whites who appear to view the president relatively favorably. Democrats would gain more in the diverse but often less competitive states.

In the Times/Siena-based estimates, Democrats appeared to be at a turnout advantage in the Rust Belt in the midterms but at a disadvantage in the Sun Belt. The difference between the groups of states might seem small, but it is not. A hypothetical full-turnout election among registered voters would cut this difference in half, and a full-turnout election among all eligible voters might eliminate it entirely.

This is consistent with state-by-state surveys of adults, like a 2019 compilation of Gallup polling data that showed the president’s approval ratings in Florida, Texas, Wisconsin, Michigan and Pennsylvania all crowded together between 41% and 43%, a few points higher than the 40% he held nationwide in the poll.

The danger for Democrats is that higher turnout would do little to help them in the Electoral College if it did not improve their position in the crucial Midwestern battlegrounds. Higher turnout could even help the president there, where an outsize number of white working-class voters who back the president stayed home in 2018, potentially creating a larger split between the national vote and the Electoral College in 2020 than in 2016.

There’s nothing about the composition of nonvoters that means a higher-turnout election would invariably make it easier for Democrats to win the presidency, or for Republicans to keep it.

This article originally appeared in The New York Times.