How Biden Won: Six Hard Truths
Digging into the exit poll data on gender, education, age, and more.
How did Joe Biden win the presidential election? And what lessons does his victory hold for how he should govern?
Early attempts to interpret the results of the election were hampered by the slowness of the counting—the major news outlets didn’t even call the presidential race until the Saturday after the election—and by other 2020 idiosyncrasies. In particular, the exit polls, which in a normal election year can help explain how different political and demographic subgroups voted, were conducted abnormally because of the pandemic. And the pre-election polls, which normally also aid in interpreting election results, were dramatically off this year.
My aim in this piece is to offer an overview—based on an analysis of the data available—of the electoral mathematics behind Biden’s win.
Clarifying the Exit Poll Mess
The exit polling data does not speak with a singular authoritative voice. This year, there were two major exit surveys, the “traditional” one done by Edison Research (based upon a sample of 15,590 in-person and self-described mail-in voters; see CNN and the New York Times) and the AP VoteCast survey (based upon a sample of 110,485 self-described voters; see NPR and the New York Times). The raw data from each of these surveys is “weighted” to reflect the actual election results.
These two exit polls point to different results among key voting blocs, suggesting that one or both of these surveys are suffering from some of the same sampling problems that affected this year’s pre-election polls. To be fair, those conducting these exit polls faced a Herculean task, given the sheer size of the electorate that voted early in person or by mail—respectively 30 percent and 41 percent of the total vote, based on the AP VoteCast estimate. In short, 71 percent of this year’s electorate could not be captured by the exit pollsters upon leaving the polls, which is the easiest and most reliable means of ascertaining whether someone actually voted.
Consequently, the diverging results in these two exit polls will leave us scratching our heads. For example, the Edison data had whites constituting a 67 percent share of the total vote and overwhelmingly favoring Trump, by 58-41, with Biden carrying black voters (13 percent of the total) by 87-12, Hispanics (13 percent of the total) by 65-32, and Asians (4 percent of the total) by 61-34. The AP VoteCast data put the white share of the total vote at 74 percent and going 55-43 for Trump, with Biden winning among black voters (11 percent of the total) by 90-8, Hispanics (9 percent of the total) by 63-35, and Asians (2 percent of the total) by 70-28.
In short, these two exit polls painted different pictures of the electorate. Which exit survey is correct, and in which ways? For example, if Edison were right on the share of the total vote attributed to each racial bloc but AP VoteCast were correct on how those blocs divided their votes between Biden and Trump, that would be significantly different from an electorate in which AP VoteCast got the shares right but Edison got the spreads right.
When you match up the actual voting returns in key states, the much larger AP VoteCast sample appears to be more accurate. That does not mean, of course, that the VoteCast data got everything right. Given the differences between the two surveys, we should resist pretending that either is totally correct—at least until the more detailed census-based studies of who voted and how they voted are completed in the coming months.
While searching for answers to the questions emerging from this election, we must supplement the exit-polling data with findings extrapolated from the actual election returns and relevant demographic data.
With those caveats set forth, we can now turn to the big question at hand: How did Biden win? After melding the exit-polling data with the actual vote counts, six hard truths emerge that should, for reasons I’ll explain, prove unsettling to both Democrats and Republicans.
1. Women Won the Election for Biden
Women outvoted men—by 4 percent in the Edison data and 6 percent in the AP VoteCast survey—and rejected Donald Trump, denying him re-election. According to the AP VoteCast, Biden carried women by 55-44, while Trump carried men by 52-46. Edison had similar numbers: Biden won women by 57-42, while Trump won men by 53-45. The bigger margins and bigger turnout among women were decisive.
The Edison data showed Trump carrying white men (35 percent of the total vote) by a 61-38 margin, while white women (32 percent of the total vote) broke for Trump by a far smaller margin of 55-44. The Edison data also showed black men (4 percent of the total vote) breaking for Biden by 79-19, while black women (8 percent of the total vote) favored Biden 90-9. Among Hispanics, the Edison data showed men (5 percent of the total vote) favoring Biden by 59-36 and women (8 percent of the total vote) favoring Biden by 69-30. Even if those splits in the Edison data are less accurate than the AP VoteCast spreads along race and ethnicity, it is clear that women voters delivered the White House to the Biden-Harris ticket.
2. The ‘Education Gap’ and Its Link to Race and Sex
Second, the emerging education gap worked in Biden’s favor, amplifying the impact of gender and race. According to the AP VoteCast data, voters with high school or less as their education level (27 percent of the total vote), broke for Trump over Biden by 52-46, a figure nearly identical to that group’s Trump vs. Clinton figure from 2016. (The 2016 exit polls are archived here by CNN and here by the New York Times.)
Voters with some college or an associate’s degree but less than a four-year degree (34 percent of the total vote) tipped dramatically away from Trump: In the 2016 exit poll they preferred him by 8 percent, but in the 2020 AP VoteCast data they preferred him by just 2 percent. The narrowing gap by which Trump carried that category of voters opened the door for Biden to win by sweeping the votes of those with higher levels of education. Biden carried college graduates with no postgraduate studies (25 percent of the total vote) by 56-42 and voters with postgraduate studies (15 percent of the total vote) by 58-40. Biden’s landslide margin among the highly educated was much larger than Trump’s margin among the 61 percent of the electorate with less than a college degree.
Gender and race were the crucial building blocks underlying this education divide in the electorate. Most notably, white college-educated men (16 percent of the total vote) voted for Trump by 52-46 while white college-educated women (14 percent of the electorate) went for Biden by 59-39. (It should be noted that this divergence echoes the age and gender divides more generally: Because men formerly far outnumbered women in college, and today women far outnumber men in college, the population of women with college degrees is younger than the population of men with college degrees—and older men skew toward Trump while younger women skew toward Biden.)
White men without college degrees (19 percent of the total vote), gave Trump a landslide margin of 64-34, while white women without college degrees (24 percent of the total vote) gave Trump a large but lesser margin of 60-39. Big though those margins are, they are each smaller than in 2016.
Among non-white voters, by contrast, Biden won regardless of education status (or sex):
among non-white non-college-educated men (7 percent of the vote), he won by 68-30;
among non-white non-college-educated women (10 percent of the total vote), he won by 77-21;
among non-white college-educated men (4 percent of the total vote), he won by 66-31; and
among non-white college-educated women (4 percent of the total vote), he won by a lopsided 80-19.
So Biden swept highly educated white women and he dominated among non-white women regardless of education level (perhaps buttressed by Kamala Harris’s outreach to black and Asian women). And he shaved down Trump’s edge among women with no or only some college education (perhaps aided slightly by Jill Biden’s long career as a community college professor).
Thomas Edsall, in a statistically rich and insightful column, points out that in 1980, Ronald Reagan’s Republicans “won 76 of the 100 counties with the largest share of college degrees.” However, in 2020, Biden’s Democrats carried 84 of those 100 counties.
3. Location, Location, Location
Third, Biden carried the nation’s suburbs by 10 percent in the AP VoteCast survey. From the Eisenhower through the Reagan eras, Republicans held a lock on highly educated and affluent suburban voters. It began eroding in 2004—and cracked apart in 2020. Edsall notes that in 1980 Republicans carried 91 out of the 100 counties with the highest median income level; in 2020, they carried only 43.
Assuming the trend holds, the significance for Democrats is enormous. The suburbs cast 45-47 percent of the total vote in high-turnout elections and 48-50 percent of the electorate in low-turnout elections. If Democrats can forge a coalition that unites the suburbs (45 percent of the total 2020 vote) while continuing to sweep the urban vote (20 percent of the total vote), that is a winning coalition—so long as they can also manage to do just a little better among blue-collar white voters in America’s small towns (17 percent of the total vote) and rural communities (18 percent of the total vote). The small-town and rural vote that Trump’s Republicans have solidified is simply not large enough to hold a majority in either the Electoral College or the popular vote.
Keep in mind that the margins among suburban voters tend to be narrow (Obama carried the suburbs by only 2 percent in 2008 and lost them by 2 percent in 2012), thus Biden carrying the suburbs by 10 percent was a huge factor in his election.
4. The Young and the Faithful
Fourth, the youth vote helped Biden win, even as it did not expand its share of the total vote from 2016. Voters aged 18 to 29, according to the AP VoteCast data, cast the same 13 percent share of the total vote that an in-depth, post-election Pew study showed from 2016. (The 2016 exit polls put the figure higher, at 19 percent of the total vote.) Biden carried these under-30 voters by 61-36.
Nevertheless, even adding in the 23 percent of the vote cast by those aged 30 to 44, only 36 percent of the total vote was from those 44 and younger. Voters 45 and older cast the other 64 percent of the ballots. Consequently, had Biden not cut Trump’s lead among those 45 and older from about 7 percent in 2016 to only 3 percent (51-48), things might have gotten very tight in the Electoral College.
Most of the pre-election public polls showed Biden carrying seniors by close to double digits. That proved inaccurate. Edison put the senior (65 and older) vote at a 22 percent of the overall electorate and showed Trump carrying it by 52-47, whereas the AP VoteCast data had seniors at 28 percent of the electorate, with Trump carrying them 51-48.
That is not to say that the youth vote was unimportant. The 25 percent edge Biden had in the AP VoteCast data among under-30 voters gave him a huge boost, especially since he carried voters aged 30 to 44 by only 54-43. But that large margin from the under-30 voters for Biden mattered only because Biden also succeeded in significantly shaving Trump’s 2016 margin among older voters.
There was a hidden nugget in the exit polls from Georgia, where 20 percent of the total vote was cast by those under 30 in age, higher than the national average, which was obviously a factor in Biden’s surprising albeit narrow victory in Georgia, enabling him to cross 300 votes in the Electoral College.
It is worth charting the key voting blocs enabling Biden’s ability to shave Trump’s edge among older voters. Religion played a key role. Back in 2016, Trump led Clinton among Catholics by 4 percentage points, a margin powered by white Catholic voters. Biden, only the second Catholic to be elected president, cut Trump’s lead among Catholics to 1 percent—a dead heat. Jewish voters (3 percent of the total), voted for Biden over Trump by 68-31. This is down slightly—by 3 or 4 percent— from recent presidential elections, where Obama and Clinton got in the low 70s among Jewish voters. The slight drop may reflect the growing share of more conservative Hasidic and highly Orthodox Jewish voters within the full range of Jewish voters, but it does not represent a wholesale change in partisan loyalties among Jewish voters who still sharply gravitate towards the Democrats. We are probably at mid-passage where the Jewish vote is on the way from a 75-25 split favoring Democrats a couple of decades ago toward a 65-35 split favoring Democrats by the end of this decade. The combined share of the (mostly white) Catholic and Jewish vote has always played an outsized role in the large, swing Electoral College states—not just New York, Florida, and California, but Pennsylvania, Michigan, and Wisconsin, as well as Arizona and Georgia.
5. The Real Enthusiasm Gap
Fifth, after all the talk during the campaign about how Trump’s supporters were more enthusiastic than Biden’s, we can now see in the data that there was an enthusiasm gap at the core of this election—but that it helped Biden, not Trump. In the AP VoteCast survey, 46 percent of voters rated President Trump favorably, while 53 percent rated him unfavorably. Therefore, it should not really surprise anyone that the popular vote now stands at 51.3 percent for Biden and 46.8 percent for Trump. Biden’s achievement was to sponge up the votes of nearly the entire range of those who opposed Trump. Among the 46 percent of voters who reported in the AP VoteCast survey that they “disapprove strongly” of President Trump, Biden won 97-1. On the flip side, meanwhile, the voters who said they “approve strongly” of Trump, who voted for him by 98-2, amounted to a much smaller 31 percent of the electorate.
In other words, the notion that Trump’s base was more passionate about supporting him than was Biden’s is not reflected in the data. Instead the real enthusiasm gap was that 15 percent more of the electorate strongly disapproved of Trump than was passionately in his corner: The 46 percent of the voters who really and truly disliked Trump put Biden on a glide path to victory that MAGA nation, constituting only 31 percent of the electorate, could not overcome. In fact, Biden was able to extend his electoral reach to encompass the nearly the full range of the anti-Trump majority, all the way from the “woke” left to the middle to the Never Trump Republicans. That coalition did not produce a landslide. But it did form a solid majority of the electorate.
This enthusiasm gap was enough to overcome the late move to Trump among the 5 percent of the total vote who told AP VoteCast that they decided who to vote for in last few days. Those late-deciding voters broke for Trump by a margin of 51-38 percent—not enough to change the outcome, but enough to help take Biden’s popular-vote margin down to 4.5 percent, rather than something closer to the 8 to 10 percent level forecast by most of the public polling data in the days before November 3. (Another big reason the public pollsters once again had to scrape egg off their faces: They badly misread how senior voters would land—thinking Biden was leading by 8 to 10 percent among voters 65 and older, when Trump won them by 3 to 5 percent. “This is a deeper kind” of polling error than those of four years ago, writes Nate Cohn. “It suggests a fundamental mismeasurement of the attitudes of a large demographic group.”)
In retrospect, this presidential race was a negative referendum on Trump and his presidency. The Democratic party was not able to erase many voters’ perception that the Democrats would swing too far left (influenced by the Trumpist rants that the Democrats would pursue “socialism” and the defunding of police, despite the fact that neither Biden nor more than a half-dozen or so Democratic candidates for Congress were advocating anything close to that as a platform). The enthusiasm gap—the fact that many of the votes for Biden were really just votes against Trump—was also the reason Biden had no coattails in the congressional races (with the Democrats losing seats in the House and Biden not able to boost the Democrats into a clear majority in the Senate).
That inconvenient political reality for Democrats on coattails can be seen in some interesting data Thomas Edsall obtained from Robert Stein, a political scientist at Rice University. Stein tracked the results of the five House seats in Texas that were at the top of the Democrats’ list of potential pick-up opportunities:
The five districts—two in Collin County north of Dallas, two in Tarrant County (Fort Worth) and one in Dallas County—Stein noted, were “predominately white suburban or exurban districts with above average education and income for Texas.” What happened in these districts on Nov. 3? Biden carried all of them, by an average of 6.5 points, Stein wrote, but all the Democratic challengers for state legislative offices fell short.
The difference between the 2018 midterm elections, when the Republican candidates in swing House districts were defeated, almost savaged, and this year, when the Democrats both lost some swing seats they had won last time and whiffed badly on challenge races to extend their majority, is clear. In 2018, the anti-Trump majority, which included conservative Democrats, moderate independents, and traditional non-Tea Party Republicans, had only one outlet to express opposition to Trump: voting against Republican congressional candidates. In 2020, they could direct their anger straight at Trump—and many of those voters chose to stick with Republican candidates down-ballot. Neither party should take comfort in projecting what will motivate these swing voters in the 2022 races and beyond. Those are difficult-to-read voters whose support will prove a hard-earned prize.
6. Big Turnout But No Big Shift to the Left
Which brings us to the final piece needed to crack the code for how Biden won this election: the defection of Never Trump Republicans to Biden’s banner. The one thing Trump and Alexandria Ocasio-Cortez could agree on was that the impact of Never Trump Republican initiatives—like Republican Voters Against Trump and the Lincoln Project—would be minimal.
That assessment was flat-out wrong.
The dogma on the left—that if only turnout surged, then younger voters would drive the electorate not only in favor of the Democrats but sharply toward the left ideologically—proved not to be true in 2020. Despite the surge from 137 million total votes in 2016 to 158 million in 2020, a modern record of 66.7 percent of the eligible population, there was partisan parity in the turnout: The AP VoteCast survey found that 47 percent of the voters self-identified as Democrats or Democratic-leaning independents, while 48 percent of the voters said they were Republicans or Republican-leaning independents. Edison found that 37 percent of the voters were Democrats, 36 percent Republicans. The turnout surge did not make the electorate more Democratic.
(This, by the way, is another reason the pre-election public polling got it wrong: Most polls wrongly projected that Democrats had a lead of 3-5 percent in who voted, when instead there was partisan parity.)
In terms of ideology, this year’s electorate remained center-right, if the exit-polling data is correct. The Edison survey found that 24 percent of the voters self-described as liberal (favoring Biden by 89-10), 38 percent moderate (going for Biden by 64-34), and 38 percent conservative (favoring Trump 85-14). Despite recent polls suggesting the liberal share of the electorate was rising above 25 percent while the conservative share was drifting down from the 40 percent level (which held steady from 1966-2006) toward 35 percent, in fact the conservative share held at 38 percent. And remember that not all of the self-described “liberals” are pure progressives; many are traditional liberals—so among that 24 percent of liberals, the pure progressives are but a fraction of that in a general election.
Which means that to win not only the presidency but also to gain congressional majorities, Democrats will need moderate and non-Trumpist Republican votes to succeed, outside the large urban centers.
The AP VoteCast data reveals the equation that won the presidency for Biden’s Democrats with partisan parity in place and with conservatives enjoying a solid edge over liberals: Democrats broke for Biden by 95-4 while Republicans went for Trump by 91-8. The narrow band of independents (5 percent of the electorate) broke for Biden by 52-37. In the Edison data, which used a broader, more traditional definition of independents, they broke for Biden by a margin of 13 percent (54-41).
The bottom line is clear: Biden won because the Never Trump Republicans gave him twice as many defections from Trump (8 percent) as Trump was able to pull away from Biden among Democrats (4 percent), enabling Biden’s clear double-digit edge among independents to carry the day.
The robust efforts of those leading this Never Trump Republican assault on Trump may not have moved huge pools of voters, but in terms of the actual significance of the votes that they turned, they moved mountains. Donald Trump liked to denigrate the full gamut of Never Trump Republicans as losers, but their efforts turned Trump into the real and ultimate loser of the 2020 election.