Despite the robust record of these models, they didn’t get much attention leading up to 2016. Mr. Trump was such a nontraditional candidate — in his fund-raising; support for atypical G.O.P. positions on trade and on protection of Social Security and Medicare; minimal advertising; and seemingly arbitrary ground game — that politicians and analysts alike thought he would knock the election out of the equilibrium where structural factors like party and the economy mattered. “If it were a normal year” or “with any other candidate” were common refrains that came as fundamentals models were dismissed.
In actuality, the state of the nation’s economy — growing slowly — was predicting a very close election in 2016: a contest that either party could win.
For all of the exceptionalism of 2016, the outcome was remarkably typical. The final vote share was very close, as the low growth rate and lack of an incumbent suggested it would be.
But just because the outcome was as close as the fundamental conditions predicted it would be doesn’t mean that many of the articles about how Mr. Trump won are wrong. Could it be true that misogyny played a role, as Hillary Clinton suggested a few weeks ago? Yes. What about the ascent of white voters without college degrees in the Rust Belt? Or the increasing rates of drug-related deaths and the increasing polarization around race, religion and ethnocentrism? With an election that turns on roughly 75,000 votes in three states, a lot of things can be pivotal.
Good work is being done trying to figure out who was in the 9 percent of the electorate that moved from Barack Obama in 2012 to Mr. Trump in 2016 (and the smaller share that moved from Mitt Romney to Mrs. Clinton).
Yet we also need to look at the 90 percent of the electorate that was necessary — if not sufficient — to produce the outcome: those who stuck with their party. A lot of identities may have been triggered for a lot of voters in 2016, by both candidates, but the one identity that mattered most was the identification almost all voters have with one of the two political parties.
Self-described Democrats, including those who lean just a little bit toward the party, almost always vote for the Democrat, and the same pattern applies for Republicans and Republican leaners. Partisans did not shy away from their party nominee in 2016 any more than they have in recent years. But in the past, some moves away from party have been pronounced.
In 1952, more than 90 percent of Republicans cast ballots for Dwight Eisenhower, according to the American National Election Study (A.N.E.S.), the longest-running repeated collection of attitudes on American politics. The same was true for Republicans in 1956, 1960, 1968, 1972 and most elections to date. Republican voters couldn’t muster the same loyalty to Barry Goldwater in 1964 (just over 70 percent said they voted for him). But even in 1976, 1992 and 1996 — the other years exhibiting a dip below 90 percent for Republican in-party voting — the rates were over 80 percent.
For Democrats, the pattern has been more dynamic. The party’s presidential candidates between 1952 and 1980 didn’t engender the same level of loyalty from Democrats, largely because of regional differences in rates of support. Southerners were less supportive of the nominee than Northerners in the years before and after the party’s passage of the Civil Rights and Voting Rights Acts. But by the 1990s, Democratic candidates were securing the support of roughly 90 percent of Democrats. This loyalty has persisted in both parties in every presidential election since.
In 2016, according to the A.N.E.S., roughly 92 percent of Democrats voted for Mrs. Clinton, and about 92 percent of Republicans voted for Mr. Trump. This is down just 1 percent in each party from 2012. Even when party identification is measured a year before the election — in 2015 — as it was in the RAND Presidential Election Panel Survey, the rates of in-party voting remained in the 90 percent range.
This pattern of support raises many compelling questions for those who study politics. Has the party label become an efficient shortcut for voters, helping them decide which candidates best meet their priorities and goals? Or is support for a party more like support for a favorite sports team, devoid of any content other than entertainment, drama and identity? The truth is probably a mix of the two, but figuring out how much of party loyalty is purely cheering for your team and how much is rooted in policy views is a tough task.
What’s more clear is that the election was always predicted to be close based on things we know historically shape outcomes. Mr. Trump was atypical; voters were not.