New Jersey’s Democratic Lt. Gov.-elect Sheila Oliver and Gov.-elect Phil Murphy. (Lucas Jackson/Reuters)

Given the year they’ve had, Democrats are understandably euphoric about their wins in Virginia, New Jersey and elsewhere Tuesday. A man dragging himself across a desert will be deliriously happy to find a bottle of water. But he shouldn’t assume that one bottle means there’s a river over the next dune — nor should the Democrats be overly optimistic about what lies ahead.

I say this not just because a year is an eternity in today's politics, although it is. Donald Trump went from novelty candidate to president-elect in the year between November 2015 and November 2016, and in the 12 months after that, the Republican Party went from a historic high-water mark to a shambolic mess. Only a fool or a political pundit (I know: Writers should avoid redundancy) would predict what will happen over the next 365 days. Still, I feel safe in forecasting that it will be a lot.

I could also warn Democrats to heed what might be the only ironclad rule of contemporary American politics: Conventional wisdom is always wrong. I was pretty sure Democrat Hillary Clinton would eke out a win over Trump despite her dreary campaign — until I noticed that everyone else in politics and the media seemed to think the same thing. With the Acela crowd so certain, Clinton was doomed.

More substantial reasons for Democrats to remain cautious are found in a deeply researched paper published Nov. 1 by the liberal Center for American Progress. Political scientists Rob Griffin, Ruy Texeira and John Halpin set out after the 2016 election to determine who voted — by race, age and education — and in what proportions. Their months-long project drew strands from a wide range of data sources and wove them into a picture quite different from the one painted by the imperfect art of Election Day exit polling.

"Voter Trends in 2016: A Final Examination" suggests that the coalition of college-educated progressives and people of color on which Democrats have staked their identity may be weaker than most party strategists believed. And as they continue their crawl through the political wilderness, they may find that efforts to strengthen the coalition prove counterproductive, as they did against Trump.

I was struck by two sets of data from this rich trove of findings that may add up to a cautionary tale. First, the white electorate is larger and less educated than exit polls would have us believe. The pollsters calculated that 71 percent of voters in 2016 were white and that more than half of them had four or more years of college. But the CAP team came to a very different conclusion: The turnout was nearly 74 percent white (a significant difference in a razor-thin election), and only about two out of five of these voters had a college degree.

Overall, 45 percent of voters in 2016 — by far the largest segment — were whites who either did not attend or did not complete college. This was not entirely a Trump-driven phenomenon. The authors found that exit polls greatly underestimated the voting power of non-college-educated whites in 2012, too.

Second, whatever strength Democrats have gained from identity politics appears to have reached a natural ceiling. Candidate Trump built his campaign on his willingness to offend people. He bashed immigrants, linked Mexicans to violent crime, dog-whistled to white supremacists. Yet when the votes were counted, Trump outperformed 2012 GOP nominee Mitt Romney among African American voters and matched Romney among Latinos.

There’s no denying that a significant source of the energy in the Democratic Party comes from people for whom identity politics are highly salient. But these findings suggest that further sharpening these issues will not gain Democrats much of anything. To the extent that some white voters are alienated by these issues, identity politics may backfire, driving votes away.

A lot of pixels have been devoted to the theory that Clinton would have won the election had she matched Barack Obama in African American turnout. The CAP study confirms that this is true. But the study also shows that she would have won had she matched Obama among whites without a degree.

Once the party of the working class, Democrats have lost their connection to the largest bloc of voters in America. Democrats had an edge in 1992 of more than five points over Republicans in the registration of white voters with only a high school diploma. By 2016, Republicans had flipped that advantage and widened it to more than 25 points.

No party should feel sanguine heading into an election so glaringly weak with the plurality of the electorate. Democrats will celebrate in 2018 and beyond only if they begin reconnecting with the white working class. How? By assuring them that their concerns matter — not more than, but as much as, anyone else’s.

Read more from David Von Drehle's archive.