Then-Republican presidential nominee Donald Trump shakes hands as he arrives at a campaign rally on Aug. 30, 2016, in Everett, Wash. (Evan Vucci/Associated Press)

Most of our assessments of the electorate in 2016 are dependent on estimates. Polling before the election that suggested where people were leaning; exit polling after the fact that gives us some sense of who actually turned out. When more than 137 million people vote, understanding exactly who they were and why they voted the way they did necessarily involves some guesswork.

On Thursday, though, Pew Research Center released an unusually robust survey of the 2016 electorate. In addition to having asked people how they voted, Pew’s team verified that they did, giving us a picture not only of the electorate but also of those who didn’t vote. There are a number of interesting details that emerge from that research, including a breakdown of President Trump’s support that confirms much of his base has backed him enthusiastically since the Republican primaries.

The data also makes another point very clear: Those who didn’t vote are as responsible for the outcome of the election as those who did.

As we noted shortly after the election, about 30 percent of Americans were eligible to vote but decided not to, a higher percentage than the portion of the country who voted for either Trump or his Democratic opponent, Hillary Clinton. Pew’s data shows that almost half of the nonvoters were nonwhite and two-thirds were under age 50. More than half of those who didn’t vote earned less than $30,000 a year; more than half of those who did vote were over age 50.


Pew’s data allows us to see very specifically how voter turnout and candidate support compare. By looking at the preferred candidate in a demographic group and then comparing the density of that group in the population that voted with the density in the nonvoting population, we get a sense for how nonvoters determined the 2016 results.

We’ll start simply. Women tended to prefer Clinton to Trump and made up a higher percentage of the voting population than the nonvoting population. That split alone helps explain Clinton’s popular-vote victory.


But “women” contains multitudes. Black, working-class Democratic women; white, wealthy Republican women. The split by party shows how that makes a difference: Republicans made up more of the voter pool than the nonvoter pool and, unsurprisingly, broadly supported Trump.


(Remember: We’re not comparing actual turnout to the pool of registered voters — we’re comparing percent of voters to percent of nonvoters. In a world where voting doesn’t vary by demographic, the percent of voters from any group would be the same as that group’s percentage of nonvoters. If a demographic group votes less than another, though, it will be below the centerline while the other group is above it.)

Looking at race and ethnicity, we see how the heavier turnout of white voters affected the contest. Black and Hispanic voters voted much more heavily Democratic than white votes backed Trump, but they turned out less.


While half of nonvoters were white, 74 percent of voters were.

An even more dramatic example of that comes when we look at age groups.


People under 30 preferred Clinton by 30 points but made up much more of the nonvoter population than the population that actually voted. A third of nonvoters were under 30; only 1 in 8 voters was in that age group.

Here’s what the income divide in Pew’s chart looks like.


Each of these demographic groups contains fragments of the others, of course. Some of those earning under $30,000 in income are Republican or black or female. If we overlap race and income, the dots above separate a bit.


Whites making more than $30,000 a year skewed Republican and made up more of the voter pool than the nonvoter pool. Poorer whites and nonwhites generally made up more of the nonvoter pool than the voter pool.

Income and education generally correlate, but the chart looking at education and race is remarkably different from the chart above. College graduates leaned toward Clinton — but whites without college degrees voted heavily for Trump. Nonwhites without a college education were 40 percent of the nonvoter pool and only 1 in 5 actual voters.


Evangelicals were the most strongly pro-Trump of the religious groups of voters, and they represented more of the voting pool than the nonvoting pool. Black Protestants and Hispanic Catholics made up less of the voting population than the nonvoting population — and strongly preferred Clinton.


If we step back and look at the bigger picture — all the demographic data points from Pew’s analysis, including those above — the expected trend emerges.


Demographic groups that preferred Trump were three times as likely to be a bigger part of the voter pool than nonvoters. Among groups that preferred Clinton, they were about 50 percent more likely to be a bigger part of the nonvoting community.

Clinton nonetheless won the popular vote. But an increased turnout of under-30 voters in, say, Wisconsin, Pennsylvania and Michigan could easily have changed the results of the history.