by Ronald Brownstein, The Atlantic
Can Big Data explain the passion and vitriol of American politics? Like almost everything else in modern life, the choices are multiplying for analysts looking to understand how the key groups in American society divide in presidential elections.
Once, researchers and political operatives had only a few options: some postelection academic surveys (particularly the University of Michigan’s American National Election Studies), precinct-level analyses, and, above all, the mainstay of Election Day television broadcasts—exit polls.
Now the choices for understanding the electorate’s behavior have proliferated. The ANES poll has been joined by the Cooperative Election Study (CES), a consortium of academic researchers from some 50 institutions that surveys a huge sample of more than 60,000 voters. Catalist, a Democratic targeting firm, produces its own estimates of voting behavior, based on sophisticated modeling and polling it does with its database tracking virtually all actual voters. The Associated Press and Fox News teamed up with the venerable NORC at the University of Chicago this year to produce a competitor to the traditional exit polls called VoteCast.
Yesterday, the Pew Research Center released its eagerly awaited Validated Voters survey. Pew builds its findings by surveying adults it can identify as definitely having voted in November based on voting records, a methodology many analysts favor. (The CES will soon issue revised results based on a similar process of matching poll respondents to voting records.)
Each of these methods has its fans: Catalist, for instance, has emerged as the data source most trusted by Democratic political professionals, while other politicos and academics swear by Pew or CES. “It is part art and part science,” says the UCLA political scientist Lynn Vavreck, who helped launch the massive Nationscape polling project, which will eventually release its own assessment of 2020 in an upcoming book.
But with yesterday’s release of the Pew results, one thing is now clear: The principal data sources about 2020 have converged to a striking degree in their account of what happened. “As I’ve been looking at our data and comparing it to some of those other sources, I’ve actually been struck by how similar [they] are,” says the Tufts University political scientist Brian Schaffner, a co-director of the CES study. “You get a pretty consistent picture.”
That consistent picture offers both parties reason for optimism and concern in roughly equal measure. The cumulative message from these studies is that we should brace for more years of grueling trench warfare between two coalitions that are becoming more and more inimical in both their demographic composition and vision of America. And to top it off? They appear to be about evenly matched. (While the Democratic coalition is clearly numerically larger—having won the popular vote in an unprecedented seven of the past eight presidential elections—Republicans have some offsetting advantages, some structural, others manufactured, that could allow them to control Washington nonetheless.)
Here are some other big conclusions from the studies:
GOP constituencies are shrinking, but the party’s hold over them is tightening.
A consistent message in these data sources is that the GOP’s core groups—particularly white people without a college degree—are declining as a share of the electorate as the nation grows more diverse, better educated, and more secular.
The major election studies differ on the share of the vote they believe was cast by white people without a college degree, from a high of 44 percent in the Catalist data, to 42 percent in the new Pew results, to just under 40 percent in the recent registration and turnout study from the Census Bureau (the first time the group has fallen below that threshold in census data).
But whatever absolute level of the vote the studies attribute to those noncollege white people, Catalist, Pew, and the Census Bureau each found the same relative movement, with the share of the vote cast by them in 2020 dropping two percentage points from 2016. That continues a long-term pattern: Working-class white people have declined as a share of the vote between two to three percentage points in each election during this century. That may not sound like much, but it adds up: In census data, they were still a 51.5 percent majority of voters as recently as 2004, before falling just below half in 2008 (almost certainly for the first time in American history) and continuing down to their current level.
Other groups important to the GOP are also shrinking. According to Pew, white Christians fell to 49 percent of total voters in 2020, down from exactly 50 percent in 2016; that’s also likely the first time in American history those voters didn’t constitute at least half of the electorate. Rural communities are also contracting as a share of the total vote (and population) in most states.
The countertrend is that the GOP last year continued to amass commanding margins with all of these voters. Even Joe Biden, a 78-year-old white Catholic who touts his working-class background in blue-collar Scranton, Pennsylvania, achieved only grudging gains among white voters without a college degree: Pew found that he won 33 percent of them, just slightly better than the meager 28 percent Hillary Clinton captured in Pew’s 2016 survey. (The exit polls and Catalist, which also put Biden’s share with noncollege white voters at about one-third, recorded similarly small gains.) Likewise, while Pew found that Biden narrowed Clinton’s deficits among both white Catholics and white mainline Protestants, Donald Trump still carried both groups by roughly 15-percentage-point margins. All of the major data sources found that Trump also carried about four-fifths of white evangelical Christians. Similarly, Pew and Catalist both found that Biden remained stuck at the modest one-third of the vote Clinton won in rural areas.
These findings underline the trade that Trump has imposed on the GOP: He’s bequeathed Republicans a political strategy based on squeezing bigger margins out of shrinking groups. Many GOP strategists believe that’s an utterly untenable long-term proposition. “That’s not a formula for winning majorities and winning most of the time,” says the longtime GOP pollster Glen Bolger, who notes that Trump lost the popular vote twice and “got beyond lucky” to win the Electoral College in 2016. But that doesn’t preclude the GOP from continuing to win power in the near term with that approach—given that the Electoral College and Senate magnify the influence of states where those shrinking groups remain more plentiful (more on that below), and the determination of red-state Republicans, through their wave of restrictive voting laws, to suppress the influence of the rising groups that generally favor Democrats.
Class inversion is here to stay.
The new Pew data, like the earlier 2020 assessments, underscore the durability of what I’ve called “the class inversion” in each party’s base. In the ANES studies, the longest-running of these sources, every Democratic presidential nominee from Adlai Stevenson through Jimmy Carter ran better among white voters without a college degree than among white voters with one. But as cultural issues supplant economic concerns as the principal dividing line between the parties, every Democratic nominee since Al Gore in 2000 has run better among white voters with a degree than among those without one.
The class inversion hit a new peak in 2016, with Hillary Clinton running at least 15 points better among college than noncollege white voters in most of the major data sources (including a breathtaking 27 points better in Pew’s assessment). In 2020, Catalist and the exit polls showed the gap widening, while Pew found it slightly narrowing, but the class inversion remained enormous in all three; each study also found Biden winning a majority of college-educated white voters. (Those gains were central to his strong showing in white-collar suburbs around major cities.) He was especially strong among college-educated white women: “We have the ability to make [them] a base group,” says Celinda Lake, who served as one of Biden’s lead campaign pollsters. But ominously for the GOP, all three sources also showed Biden gaining significantly over Clinton in 2016 among college-educated white men, who historically have been a much more reliable Republican constituency. And while white people without a college degree have been steadily shrinking as a share of the vote, these college-educated white people have slightly grown since 2004 (from about 28 percent to 31 percent of the electorate, per the census). Especially valuable for Democrats: They are highly reliable midterm voters.
Voters of color may be diverging.
Pew’s study found that Biden won 92 percent of Black voters last year, and the other major data sources gave him only slightly smaller shares. Democrats may need to keep an eye on Black men, among whom Trump performed slightly better in 2020 than in 2016, but their support among Black women—which reaches as high as 95 percent in some of these analyses—provides an immovable obstacle to broad GOP gains.
Asian Americans, the fastest-growing nonwhite community, also look solid for Democrats. Although Republicans have strong beachheads in some Asian communities sensitive to arguments against Democratic “socialism” (such as Vietnamese Americans and some Chinese groups), the major data sources agree that Biden still won about two-thirds or more of Asian American votes last year, even as their turnout soared.
Hispanics, though, could be emerging as a wild card. Pew put Biden’s vote among Hispanics at only 59 percent; that’s lower than any of the other major sources, but they all agree that Biden fell off measurably from Clinton (and Barack Obama before her). The decline was most visible among Central and South Americans in South Florida and rural Mexican Americans in South Texas, but it extended far beyond that, Catalist and others found. Trump may have raised the party floor with Hispanics by attracting more of the culturally conservative among them; the yellow light on that prediction, as I’ve written, is that almost every incumbent president ran better, as Trump did, with Hispanics in their reelection campaign than in their first race. The clearest conclusion is that both parties view Hispanics as more of a contested community after 2020 than they did before—and will spend their campaign dollars accordingly.
The generational cavalry is arriving for Democrats.
Both Pew and Catalist found that the racially diverse, well-educated, and highly secular Millennials (born from 1981 through 1996) and Generation Z (born from 1997 through 2014) cast almost 30 percent of the votes last year, up substantially from 23 percent in 2016. Both sources also found Democrats winning about three-fifths of the votes from those two generations combined. If Democrats can defend their lead with that group, it will pay compounding dividends: The nonpartisan States of Change project forecasts that the two generations combined will cast 37 percent of the vote in 2024 and 43 percent in 2028. “You add those two [generations] together and you are talking about permanent structural change,” Lake says. Because these generations are the most racially diverse in American history, this current of new, young voters has been key in increasing people of color from about one-fifth of the electorate in 2004 to nearly three-tenths last year, according to census data. They are also swelling the numbers of Americans unaffiliated with any religious tradition, and Pew found Biden winning more than 70 percent of such “seculars” (even as they cast one-fourth of all votes.)
Conversely, the preponderantly white Baby Boomer generation, which has aged from its 1960s roots into a Republican-leaning cohort, is receding: While Catalist and Pew agree that Boomers outvoted Millennials and Gen Z in 2020, States of Change projects that the younger groups to outvote them for the first time in 2024. (Generation X is projected to remain constant through the 2020s, at about one-fourth of the electorate.)
Two factors might dilute this potential Democratic advantage. One, Schaffner notes, is if the turnout of these two younger generations, which spiked to historic levels in 2018 and 2020, slackens with Trump off the ballot in 2022 and potentially 2024 as well. The other, cited by Vavreck, is that these generations might become more receptive to GOP arguments on issues such as taxes and crime as they move further into middle age, with families and mortgages.
But Lake, like many Democrats, is optimistic that the GOP focus on stoking their base through endless cultural conflict (on everything from undocumented immigration to critical race theory) will leave Republicans very limited opportunity for gains among the younger generations. “Young people are very turned off by the racism, by the climate deniers,” she says. “So everything they are doing to solidify their base, and everything they are doing to try to win 2022, is digging them into a deeper hole for 2024 with young voters.”
A big challenge for Democrats is that the broad demographic changes favoring them—growing racial diversity, rising education levels, increasing numbers of secular adults not affiliated with organized religion—are unevenly distributed throughout the country. Adding to that challenge: The two-senators-per-state rule and Electoral College magnify the political influence of smaller interior states least affected by these trends (particularly the increase in racial diversity). Red-state Republicans are moving to systemically reinforce those advantages with the most aggressive wave of laws restricting access to the ballot since before the Voting Rights Act in 1965, and they are gearing up for equally aggressive gerrymanders of state legislative and congressional districts in states they control.
As I’ve written, the unequal distribution of racial and cultural change leaves Democrats facing something of a conundrum. The minority population is growing fastest across the Sun Belt, but the party generally doesn’t win as large a share of the vote among white people in those states as they do in the Rust Belt states, where minority growth has been much slower. Until Democrats can consistently win Senate seats and Electoral College votes in the diversifying Sun Belt states, that means they still need to win some of the Rust Belt states (particularly Wisconsin, Michigan, and Pennsylvania) where noncollege white people compose a much larger share of the vote than they do nationally. Democrats lately have made progress in the Sun Belt: Biden won both Georgia and Arizona, and the party now holds all four of their Senate seats. But Democrats’ Sun Belt gains aren’t yet expansive or secure enough to eliminate their need to hold the key Rust Belt battlegrounds—and for that they need to win a competitive share of working-class white voters.
The grooves are deeply cut.
The major data sources do show some noteworthy shifts in voter preferences from 2016, such as Trump’s gains with Hispanics and Biden’s with college-educated white voters. But given all that happened during Trump’s tumultuous presidency, including a deadly pandemic, most analysts are struck by the extraordinary similarity in how voters behaved across the two elections. “Continuity is the big story, consistency,” says Alan Abramowitz, an Emory University political scientist. Not only did the 2020 result “highly correlate” with the 2016 outcome both demographically and geographically, he notes, but presidential preferences also predicted how people voted in House and Senate races more closely than ever before.
Biden and his advisers clearly have a vision of how to break this stalemate: They hope that by delivering kitchen-table benefits, such as stimulus checks, infrastructure jobs, and expanded child-tax-credit payments, while muting his personal engagement with hot-button cultural issues, they can improve his standing among working-class voters of all races, including white voters. But that strategy faces unstinting GOP efforts to highlight the cultural issues that alienate those voters (especially white voters but also some Hispanics and Black men) from the Democrats. Ruy Teixeira, a veteran Democratic analyst, argues that even if Biden delivers material benefits for blue-collar families, downplaying cultural issues such as crime and immigration won’t be enough. “You are going to have to draw the line a little bit more sharply against parts of the party and policies that are anathema to these voters,” Teixeira says.
Still, almost all of the analysts I spoke with believe that however the parties position themselves through 2024, change in these durable voter alignments is likely to come only around the margins.
Big outside events could shatter that assumption, of course, but the striking message from all the data sources studying 2020 is that America remains deeply but closely divided. Wide partisan fissures by race, generation, education, and religion are combining to produce two coalitions that are matched almost equally, with a Democratic edge in overall numbers offset by a geographic advantage (potentially reinforced by restrictive voting laws) for Republicans. “It is going to be super, super close again in 2024, I can tell you this right now,” Vavreck said firmly. “And I don’t even need to know who the candidates are going to be.”
Ronald Brownstein is a senior editor at The Atlantic.