The false claim that Bernie Sanders was sunk in 2016 by black voters
I’ve heard the claim repeated dozens of times that the reason Bernie Sanders failed to win the 2016 Democratic Primary1 was because he wasn’t able to get enough support from black voters. This has become such a truism among some pundits that attempting to refute it smacks of a conspiracy theory, but I hope to show that it is actually false convincingly in this article. It turns out that the claim hangs on math that is actually fairly unintuitive, so much so that even after doing the calculations for many states, I still found myself unable to guess what level of support Bernie Sanders got among black voters versus white ones in any particular state when looking at exit polling data.
This may sound absurd. After all, the exit polls can look straightforward at first glance. For example, in South Carolina, the exit poll data contains something like the following table:
|Candidate||White (35%)||Black (61%)|
Nothing could be simpler, right? 35% of the voters were white, 61% were black, and of the white voters, 54% went for Clinton, 46% went for Sanders. Of the black voters, 86% went for Clinton, 14% went for Sanders. Sanders has a gap among white voters of 8%, and a gap of 72% among black voters. Repeat this process on all 50 states, some of which are much closer than South Carolina, and you can trivially generate your hot take for MSNBC from there.
What’s wrong with this analysis? Well, what I’m interested in when I ask the question “what’s Sanders’ relative support among black and white voters?” is whether, if you asked every single voter leaving a primary election in 2016, a greater proportion of black voters would support Sanders than white voters, or vice versa. Or to put it in simpler terms, if you know 14 random white people, and 14 random black people, which group is going to have the greater number of Sanders supporters? I hope that seems like the obvious thing to be interested in to you too.
It will probably surprise you then to learn that the answer in South Carolina is that two in fourteen black voters support Sanders, and only about one in fourteen white voters support Sanders.
How can this be? It’s because of a very simple fact that the exit poll is unintentially obfuscating: if you know fourteen white people in South Carolina, about one of them will support Sanders, one will support Clinton, and twelve of them are Republicans! This is the kind of demographic fact that exit polls don’t capture, because they’re not designed to. Polling results are divided and reported separately for Democrats and Republicans, even though the elections and exit polls are (usually) held simultaneously.
Fortunately, official election results and exit polls do provide enough data to pretty reliably piece together what the actual political distribution looks like. The actual distribution of South Carolina voters looks like this:
As this table suggests, South Carolina is extraordinarily bifurcated along racial lines. White people in this state are extremely far-right, to such an incredible extent that in a primary election where 75% of voters were white, 61% of Democratic voters were black. While rarely to such an extreme extent, this is true of just about every state, and has a similar distorting effect on the results of exit polls, and therefore a similar distortion on political commentary that is based on those polls.
I went through the exit poll data, and put together a complete summary based on every state I could get data for. Let me briefly explain how the math here is done, using South Carolina as an example. Feel free to skip over this paragraph entirely if you’re not interested in this. The number of votes for each candidate is a matter of public record. I used The Green Papers as my primary source here. This site records that 740,881 votes were cast in the Republican primary, 370,904 votes in the Democratic primary. Now we look at the exit poll data. In South Carolina3, in the Republican primary 96% of voters were white, 1% were black. So we estimate that there were 7409 black Republican voters, 711,246 white Republican voters. The same procedure for the Democrats reveals that 226,251 of their voters were black, 129,816 were white. The exit poll data shows that 46% of white Democrats voted for Sanders, while 14% of black Democrats did. So this means there were about 31,675 black voters for Sanders, and 59,716 white voters. The total number of white voters in the election was 841,0624 and the total number of black voters was 233,660. So 13.6% of black voters went for Sanders, and only 7.1% of white voters did.
Obviously there will be some degree of error in the exit polls, and therefore in these results. But it’s not that severe: for example, if the proportion of Republican voters who were black was changed to 0% or 2% (from 1%), this would make a difference of about half a percent in Sanders’ support among black voters. Taking all states together should have the effect of evening out the errors, although some systamatic errors may remain. I’m not too bothered by this, since the point of this article is to counter a false view that the pundits take themselves to have learned from these very exit polls. If the polls themselves are untrustworthy, then their conclusion is unsound too.
Anyway, on to the results. They’re based on the total vote of 21 states in which all of the following were true: they held primaries in 2016, an exit poll was taken in them by the major media organizations, and the exit poll had a sufficient number of black respondents to draw conclusions about who they supported. (The primary qualifier is important because in caucus states like Iowa, the total popular vote count was not the official result reported by the election.) These states are South Carolina, Alabama, Arkansas, Georgia, Oklahoma, Tennessee, Texas, Virginia, Michigan, Mississippi, Florida, Illinois, North Carolina, Ohio, Wisconsin, New York, Connecticut, Maryland, Pennsylvania, and Indiana. I would have liked to include California, but they voted so late in 2016 that the media didn’t take an exit poll. Here is a spreadsheet with the math.
Here are the results:
In other words, my claim holds for all states in which there is data. On the whole, black voters are at least as likely to support Sanders as white voters. (The difference between the two is +0.8% for black voters, but I suspect that’s within the margin of error of this kind of research.)
Now, a certain kind of pundit might be inclined to respond as follows: “If you look just at the relative support for Clinton vs. Sanders among white voters, you’ll see that Sanders edges out, and so it remains true that Sanders lost the race because of poor support among minorities.”
I find this sort of analysis rather unhelpful. To put it simply, what we are imagining is disenfranchising all minorities … in which case, yes, Sanders would have won the 2016 Democratic primary, and then would have gotten utterly crushed in the general election because the Democrats depend on minority support for their basic viability as a party. It’s wrong in another way too: the pundit (at least rhetorically) takes the point of view of Sanders, and decides that “blame” needs to be parceled out to various Democratic primary demographic groups according to the degree to which they failed to support him. (Alternatively, a pundit might take a rhetorical position opposing Sanders and blame him for failing to reach out to these groups.) This isn’t really what’s happening in a primary. The reason that moderate and conservative black voters play such an enormous role in the Democratic primary is that almost two thirds of white voters are so far right that they don’t vote in the Democratic party primary at all!
Now, you might imagine a less racialized (and simpler) country in which the major political parties were basically in alignment with the range of political views along a left-right spectrum. There would be a lot more black Republican voters. The question of why we don’t live in something closer to that world is an interesting one; FiveThirtyEight took this question on directly in a recent article.5 Their conclusion was that “social pressure is what cements that relationship between the black electorate and the Democratic party”. The word “cements” is doing a lot of work here. Social pressure certainly can’t explain the majority of the effect; the same article says that 85% of black respondents identified as Democrats in an online poll where social pressure was not a factor.
It seems plausible to me that another significant factor is a response to the racialized politics of the Republican party, as the extreme proportion of white supporters in its ranks attests. If this is true, though, why wouldn’t the party take the pragmatic approach by toning down its rhetoric to pull in the many conservative minorities who are aligned with them on policy questions? Certainly, part of the answer is that they haven’t needed to so far, and that the rhetoric may serve to energize part of their white base, but what this research may suggest is that it can actually be helpful to a political party to have a large number of people consistently voting to nominate moderates in the opposing party’s primary process.
The promise of Sanders all along, of course, was that the supposed left-right spectrum is a lie. If people (and their candidates) do fall on a simple spectrum like that, then you can trivially show that the Condorcet winner will be a centrist. Even in a complicated two-party system like that of the United States, a centrist is expected to be the strongest candidate the majority of the time. (Obviously, the Electoral College throws a wrench into this.) But Sanders, and Trump to some extent, represent a claim that the true views of most voters are not well represented by the current two-party system, and that in fact someone very far to the left (or right) on the current spectrum might be more acceptable to the median voter than a centrist.
How else to understand Sanders’ candidacy at all? So far, he has not shown signs of being able to win a Democratic primary, suggesting (but not proving) that he’s too far left for many Democrats. If this is true, then he’d be sure to lose a general election that introduces an almost equal number of Republican voters. However, he has surprisingly performed at or near the top of recent head to head polls against Donald Trump, compared with other Democrats. What does that mean?
I suggest that the one explanation that suffices includes multiple factors. One very important reason why Sanders would stand a chance in a general election is polarization. Most regular voters in this country are loyal to one party or the other, and loath to switch parties based merely on the ideology of their candidates. (Moreover, there are slightly more Democratic voters than Republicans.) So if Sanders wins a Democratic primary, most of his support will come from loyal Democrats who don’t necessarily approve of all his policies. That said, it’s notable that Sanders has consistently performed at or near the top of these polls. I suggest this means that there must be some truth to his claim to represent those who do not find themselves cleanly on the left-right American political spectrum.
It’s important to notice that these two explanatory factors pull in opposite directions. On a strict party-loyalty hypothesis, it wouldn’t matter at all who gets nominated. This seems to be mostly true (for the small number of candidates who actually stand some chance of being nominated), but it’s not the whole story. Sanders represents the possibility of pulling support beyond mere party loyalty, and he’s succeeded to some extent at that, but perhaps not enough to win a primary election.
In the final analysis, this shows exactly why the exit poll based criticism of Sanders is misguided. Among Democrats, black voters are much less likely to support Sanders than white voters. But this is largely because of partisan demographics that Sanders can’t help: the Democratic party pulls in a number of surprisingly conservative black voters, while the Republican party presumably has a corresponding effect on many white voters who might be open to Sanders’ policy aims, but are more at home in their party’s racial antagonism.
On the whole, Sanders’ problem is not with black voters; they support him at equal or greater rates than do white voters. His problem is that his promise of pulling voters from both parties and those currently unaligned has not yet come to fruition. He has not been able to shift the majority of white voters away from their Republican or independent allegiances. The hope for left wing Sanders supporters must be that time and voter education will cause a realignment, and that people like Sanders will begin to see support across the political spectrum and from current non-voters. His high level of support from young voters does suggest some promise. But if the American political spectrum did accurately reflect the distribution of its voters, there would be little hope for candidates like Sanders in the near future. America simply has too many white people on the far right for that.
And now the 2020 primary election as well. This article focuses on the 2016 election specifically, because in this election Sanders faced only one challenger, giving a clearer picture of where voters stood than in 2020, when he faced a very broad field. ↩
This is neglecting third party voters. There are very few of them, and they appear to be disproportionately white, so if they were included, they would further reduce Sanders’ support among white voters. ↩