Election-year polls understandably focus on likely voters. Then, after the election, the attention turns to actual voters, mainly using exit polls. But getting good data on Americans who didn’t vote is more difficult. That’s why the SurveyMonkey poll, which interviewed about 100,000 registered voters just after Election Day, including more than 3,600 registered voters who didn’t vote, is so useful. It’s still just one poll, and so its findings aren’t gospel, but with such a big sample we can drill down to subgroups and measure the demographic makeup of nonvoters to an extent we couldn’t with a smaller dataset.
Registered voters who identified as Democrats and independents were more likely than Republicans to stay home.
It seems reasonable that many of these voters stayed home because they didn't know who to vote for or didn't want to vote for the person their party selected. And in fact, favorability ratings of either candidate were lower among people who didn't vote. However, Enten argues that because of the high correlation between party affiliation and voting, and because other polls show people who didn't identify with either candidate tended to vote for Clinton over Trump, these registered voters who stayed home would probably have voted for Clinton if they had actually voted on Election Day.