Skip to main content
ABC News
What Defines The Sanders Coalition?

Sen. Bernie Sanders’s coalition isn’t that big, at least not right now — he won about a quarter of Democratic primary voters in Iowa and New Hampshire, plus about a third of the first-alignment vote in Nevada,1 and about a quarter to a third of Democratic voters say they support him in most state and national polls. But Sanders’s backers are worth understanding because of what they’ve accomplished: making a democratic socialist who is not officially a Democrat the front-runner to lead the Democratic Party, over the objections of virtually all of the party’s establishment and many of its voters.

So who exactly are Sanders’s voters?

Yes, as you have probably read, many of his supporters are young and identify as very liberal. But you can be old and liberal or young and moderate. So is Sanders exclusively the candidate of people who are both young and liberal, very liberal people of all ages, young people of all ideologies, or some other demographic? The answer, from our analysis of exit/entrance polls and other surveys, is … it’s complicated.Iowa, New Hampshire and Nevada). And Iowa and New Hampshire are far whiter than the Democratic primary electorate at large, which limits the conclusions we can draw about voters of color from polls of those two states. Also, the surveys of the country as a whole and individual upcoming states that we are relying on are of hypothetical electorates, not actual voters.


Here’s what we found.

His base is younger voters and the very liberal

Younger Democrats (those under 45) are more likely to be very liberal than older Democrats,3 and Sanders is very popular where these two groups overlap. But very liberal Democrats under 45 make up a small bloc of the electorate (10 percent of the exit poll sample in New Hampshire and 16 percent in both Iowa and Nevada).

But he’s also popular, though not quite so overwhelmingly, with the nonoverlapping parts of these two groups. The entrance and exit polls in Iowa, New Hampshire and Nevada showed the Vermont senator either leading or tied at the top among voters under 45 who identify as somewhat liberal or moderate, and in New Hampshire and Nevada, he led with very liberal voters who are over age 45. (In Iowa, he was second only to Warren with older, very liberal voters.4 So Sanders’s coalition is not solely age-based or solely ideological; being either young or very liberal makes you likelier to support Sanders, even if you’re not both.

The effects are mutually reinforcing. In fact, when we used ABC News/The Washington Post polls to test what types of people tended to be Sanders supporters, both being young and being very liberal strongly correlated with voting for Sanders — and what’s more, those effects did not diminish when we included them both in the model together, nor when we controlled for additional factors such as race, income and education.

Still, if we had to say what most defines Sanders’s supporters, we would say age. The left-leaning polling firm Data for Progress, in a survey released shortly before the Nevada caucuses that was fairly close to the final results, found that Sanders was winning 66 percent of somewhat liberal respondents under 45, compared to 38 percent of those who are very liberal and over 45. That suggests that younger voters are Sanders’s strongest demographic, even more so than those who are very liberal.

He won Latinos, but his coalition is multiracial

The focus on the African American vote in the 2020 primary stems from 2008 and 2016, when one candidateBarack Obama in 2008, Hillary Clinton in 2016 — won an overwhelming majority of black voters (around 80 percent) in many states. Black voters are around 20 percent of the Democratic primary electorate, so that kind of support provided those candidates a decisive advantage in the nomination race. At least right now, no candidate has seems to have an advantage of that size among black, Latino or white voters.

But Sanders seems to be the leading candidate among Latino voters, according to both national polls and those in states with a large number of Latinos. And the entrance polls in Nevada suggest he won 50 percent of the Latino there, well ahead of his competitors. (Biden was the second most popular candidate among Nevada’s Latino voters, at 17 percent.)

But while many Latino voters are backing Sanders, we found that a voter’s ethnicity was still less predictive of support for Sanders than their age or ideology. In our analysis of recent ABC News/The Washington Post polls, once we accounted for age, Latino voters were about as likely to support Sanders as voters in other racial or ethnic groups, perhaps because Latino respondents in the poll, just like Latino voters in the U.S., skew younger than the overall population, and younger voters tend to break for Sanders. So it could be that Latino voters’ age explains Sanders’s strength with that group more than their ethnicity.

It’s also worth noting that Latinos represent a fairly small bloc of the national Democratic electorate (around 12 percent), compared with white (59 percent) and black (19) voters. In Nevada, 17 percent of Democratic caucus voters were Latino and 11 percent were black, but part of the reason for Sanders’s large margin of victory in the state is that he also easily won the white vote, which was 65 percent of the electorate. So although it seems Sanders’s base is hardly all white, it’s likely that no matter who wins the nomination — whether it’s Sanders or anyone else in the race — a clear majority of their votes will probably come from white Democrats.

Many of his backers don’t have a college degree

In Iowa, New Hampshire and Nevada, Sanders was more popular with voters without college degrees than voters with degrees. He was the clear leader among voters without degrees in all three states, but Nevada was the only state where he was well ahead among college graduates. And those two dynamics — Sanders being more popular with non-college-educated voters than college graduates and leading the field among those without degrees — generally also show up in national surveys and polls of upcoming states. In terms of raw numbers, there are probably more Sanders supporters without degrees than Sanders supporters who identify as very liberal, since the former is simply a much larger group. (About half of the voters in the three states that have voted so far didn’t have college degrees, for example, but only 20 to 30 percent of voters in those states identified as very liberal.)

Sanders does better among voters without a college degree

Share of voters with and without a college degree who voted for each candidate, from entrance or exit polls in Nevada, Iowa and New Hampshire

Nevada New Hampshire Iowa
Candidate College degree (51%) No degree (49%) College degree (54%) No degree (46%) College degree (53%) No degree (47%)
Sanders 28% 40% 23% 31% 16% 30%
Biden 17 18 7 10 15 16
Buttigieg 17 13 24 24 23 21
Warren 16 9 12 6 19 14
Klobuchar 13 7 25 14 17 8
Steyer 7 11 2 5 2 2
Gabbard 0 0 3 5 0 0
Yang 2 3 5 7

Sample size was 2,768 in Nevada, 2,911 in New Hampshire and 1,628 in Iowa.

Source: Edison Research

But — and you may be starting to sense a theme here — we tend to think that age and ideology are driving the story here more than diplomas. In a poll FiveThirtyEight conducted jointly with Ipsos on the eve of the New Hampshire debate, college degrees only seemed to make a difference in voters’ likelihood of supporting Sanders if they were between the ages of 45 and 65 — those without degrees in that age range were significantly more likely to be considering voting for Sanders than those with degrees. But for voters under 45 and over 65, education didn’t matter much — people under 30 were open to backing Sanders whether they graduated from college or not; people over 65, degree or no, largely were not considering backing him.5 And in ABC News/The Washington Post polls conducted in January and February, the effect of education on Sanders support disappeared entirely once we controlled for ideology, age, income and race.

It’s hard to say how much income matters

In New Hampshire, Sanders won by 24 percentage points over his nearest rival among those with family incomes of less than $50,000. That was about a quarter of voters in the state. A recent Quinnipiac University poll showed Sanders had a bigger lead among households that make less than $50,000 than among any other income group. And in surveys from both Ipsos and ABC News/The Washington Post, people with a household income over $100,000 were less likely to say they were supporting or considering voting for Sanders even after we controlled for age, education and race.

But we are hesitant to conclude too much from these findings. Most other surveys, including the entrance polls in Iowa and Nevada, have not broken down Democratic voters by income. And the income data may be at least partly explained by other factors. People without degrees tend to have lower incomes than those with degrees. Income generally goes up with age (before falling again for people over 65 — retirement age).

How might all of this matter? In at least three ways. First, it suggests that Sanders has the potential to grow his support with older voters who are very liberal and younger voters who are more moderate. We know those traits — liberalness and youth — make someone more likely to support Sanders, and while Sanders is already doing well where the two groups overlap, he has plenty of room to improve among voters who have one trait or the other but not both.

Second, looking at Sanders’s support helps explain his tensions with the Democratic Party establishment. Very liberal Democrats, those under 45, those without degrees and those with incomes below $50,000 represent a substantial portion of the Democratic electorate — enough to basically lift a candidate like Sanders to the top of the polls (especially in a field as crowded as this one). But most establishment Democrats don’t fit into any of those categories. Outside of the four House members known as The Squad, the number of Democratic members of Congress who would describe themselves as very liberal is probably fairly small. The Washington Post in early 2019 calculated that the average Democratic member of the House is nearly 60 years old, and at the time, fewer than 50 of the 234 Democrats then in the House were under 45 (and a few more will have aged past that line in the past year). More than 90 percent of members of Congress have college degrees. And members of Congress make at least $174,000 per year. So one explanation for Sanders’s lack of endorsements from members of Congress is that the kind of people who really like Sanders are generally not in Congress.

[Our Latest Forecast: Who Will Win The 2020 Democratic Primary?]

Finally, the deeper you look at Sanders’s support, the more you realize how complex and overlapping the dividing lines in the Democratic primary are. There is a bit of a diploma divide, a generational divide and an ideological divide, all happening at once and reinforcing one another. When we get more data from upcoming primaries, we could learn that there is also a racial divide (if black voters mobilize behind Biden or another candidate while Latino voters stick with Sanders) and an income divide (if Sanders is clearly more popular among voters with lower incomes). We’re not suggesting these divides are so large or so deep that Democrats will find it impossible to unite behind a candidate in the general election. But part of the reason that the Democratic primary is so tense, particularly between those who like Sanders and those who don’t, is that these differences aren’t just about ideology — age and social class also factor in to some extent. Those demographic differences may mean that voters start to see supporters of other candidates not just as people who disagree with them, but as people who are, in certain ways, not like them. The fight between Democrats and Republicans is increasingly about identity — and the intraparty disputes between Democrats may now be happening along those lines too.

FiveThirtyEight Politics Podcast: Biden fights for a win in South Carolina debate


  1. After voters backing nonviable candidates were asked to realign, he ultimately wound up with over 40 percent of the popular vote in Nevada.

  2. It’s important to note a few limitations of this analysis. First, exit polls are still polls, and we only have three (Iowa, New Hampshire and Nevada). And Iowa and New Hampshire are far whiter than the Democratic primary electorate at large, which limits the conclusions we can draw about voters of color from polls of those two states. Also, the surveys of the country as a whole and individual upcoming states that we are relying on are of hypothetical electorates, not actual voters.

  3. In New Hampshire, for instance, 29 percent of voters under 45 said they were very liberal, compared to 16 percent of voters over 45; in Iowa, the numbers were 36 and 17 percent, respectively.

  4. Voters who identified as conservative were too small a group to break out candidate support by age.

  5. This data has two downsides. First, it doesn’t include ideology, so we can’t see how ideology affects each of the other variables. And second, it asks a slightly different question — which of the candidates respondents are considering supporting (allowing for people to name multiple candidates), rather than which they would vote for — which could also be affecting the results.

Laura Bronner is a senior applied scientist at ETH Zürich and FiveThirtyEight’s former quantitative editor.

Perry Bacon Jr. was a senior writer for FiveThirtyEight.