“Secretary Clinton cleaned our clock in the Deep South, no question about it,” Bernie Sanders said during Thursday night’s Democratic debate in Brooklyn. “That is the most conservative part of this great country,” he continued. “But you know what, we’re out of the Deep South now. And we’re moving up.”
I have a few problems with this line of argument, which seems to imply that Democratic voters in the Deep South don’t reflect the larger Democratic electorate. (The remarks Thursday night echo previous comments made bySanders and his campaign.) Consider Sanders’s reference to the term “Deep South,” which traditionally describes Alabama, Georgia, Louisiana, Mississippi and South Carolina: These are five of the only six states, along with Maryland, where at least a quarter of the population is black. Given the United States’ history of disenfranchising black voters — not to mention the importance of black voters to Democrats in November — it’s dicey for Sanders to diminish Clinton’s wins there.
But the Deep South isn’t Sanders’s only issue. His problems in the rest of the South are what really dooms him. Clinton’s largest net delegate gains over Sanders came from Texas (+72) and Florida (+68), two states that are within the South as the Census Bureau (and most other people) define it. Clinton also cleaned Sanders’s clock in Virginia and North Carolina. Overall, Clinton gained a net of 155 delegates on Sanders in the five Deep South states, but she also added 211 delegates to her margin in the rest of the region.
|Deep South*||225||70||Clinton +155|
|Other Southern states||493||282||Clinton +211|
|Rest of country||589||745||Sanders +156|
In addition to being important to the Democratic Party’s electoral present and future, Florida, Virginia, North Carolina and Texas are quite diverse. They’re diverse ideologically — Miami and Austin aren’t exactly “the most conservative part” of the country — and they’re diverse racially. They contain not only a substantial number of African-Americans but also Hispanics and, increasingly, Asian-American voters.
In fact, these states are among the most demographically representative of the diverse Obama coalition that Clinton or Sanders will have to rely on in November.
Although it will be a couple of decades before the electorate as a whole is majority-minority, the Democratic vote is already getting there. In 2012, only 55 percent of President Obama’s voters were white, according to the national exit poll. Our demographic projections of this November’s electorate, which account for population growth since 2012, calculate that the white share of the Democratic vote will tick down another percentage point, to 54 percent. The rest of the Democratic vote will be black (24 percent), Hispanic (15 percent), or belong to Asian or other races (7 percent), according to our projections.
So let’s take those projections as being maximally representative of the broader Democratic electorate as it stands today. In which primary or caucus states has turnout come closest to those ratios?
In 21 states to have voted so far, we have data on this from exit polls. Seehere for Virginia, for example, where Democratic turnout was 63 percent white, 26 percent black, 7 percent Hispanic and 5 percent Asian or other when it voted on Super Tuesday. That’s pretty close to the Democratic electorate overall, although with too few Hispanic voters. In the other 29 states — those that haven’t voted yet or where no exit poll was conducted — I’ll estimate the Democratic electorate based on our demographic projections, with an adjustment for the fact that the Democrats who vote in primaries are somewhat whiter than those who vote in November.1
Then I’ll calculate the root-mean-square error (RMSE) for each state — a measure of the difference between the demographics of its primary or caucus turnout and the projected Democratic electorate in November. A lower RMSE is better for our purposes, because it means the state’s demographics are more representative of the national Democratic coalition.
|DEMOCRATIC PRIMARY/CAUCUS ELECTORATE|
|STATE||WHITE||BLACK||HISP./ LATINO||ASIAN/ OTHER||RMSE*||OUTCOME|
|North Carolina||62||31||3||4||16||Clinton +14|
|District of Columbia||45||49||4||2||29|
|South Carolina||35||61||2||2||44||Clinton +47|
|New Hampshire||93||2||1||4||47||Sanders +22|
|Projected Democratic electorate in November||54||24||15||7||—|
The most representative state by this measure is New Jersey. We expect its primary electorate to be about 57 percent white, 26 percent black, 11 percent Hispanic and 6 percent Asian or other, quite close to the national Democratic electorate. New Jersey won’t vote until June 7, although Clinton was well ahead when the last poll was released there in February.
After New Jersey comes Illinois, which Clinton won narrowly — and then Florida, where Clinton won going away. Then there’s New York, which votes Tuesday, and where Clinton is 15 percentage points ahead in our polling average. Virginia, another Southern state, ranks as the next most representative; Clinton won it easily. Then there’s Nevada, another Clinton state, before we go back to the South to North Carolina, also won by Clinton. The next group of four states (Maryland, Tennessee, Arkansas and Michigan) are roughly tied and include some further representation for the South, along with, finally, one state (Michigan) that Sanders won.
In other words, Clinton has won or is favored to win almost every state where the turnout demographics strongly resemble those of Democrats as a whole. This shouldn’t be surprising — Clinton is winning nationally by about 14 percentage points in the popular vote. So if you’re in a state that’s well-representative of Democrats’ national demographics, you might expect her to win it by a solid margin too.
It’s true that a couple of states in the Deep South, such as Mississippi and Louisiana, rate as being not well-representative by our definition. But overall, that’s more likely to be true of places where Sanders has won. New Hampshire ranks as the 46th most representative out of 50 states (and Iowa ranks 44th — maybe those states shouldn’t hold the first two contests?). Wisconsin, which Sanders won last week, is below average.
And the sort of wishful thinking Sanders is engaged in can cut both ways. Yes, Clinton’s lead would be considerably narrower (although she’d still be winning) without delegates from the Deep South. But what if you excluded delegates from caucuses, where Sanders has gained a net of 150 delegates on Clinton? Without those delegates, Sanders couldn’t even maintain the pretense of a competitive race. Not only are most of those caucus states extremely white and therefore poorly representative of Democrats’ national demographics — many of them (such as Idaho and Nebraska) are also quite red. Furthermore, caucuses tend to disenfranchise voters by making it harder to vote. Our demographic modeling suggests that this has hurt Clinton and that Sanders wouldn’t have won by the same enormous margins if those caucus states had held primaries instead.
But overall, the math is pretty simple. Sanders is winning states that are much whiter than the Democratic electorate as a whole, Clinton is winning states that are much blacker than the Democratic electorate as a whole, and Clinton is winning most of those states that are somewhere in the middle, whether they’re in the South (like Virginia) or elsewhere (like Ohio or Nevada). That’s why she’ll probably be the Democratic nominee.
- For example, in Illinois, our demographic projections expect 54 percent of Democratic voters in November to be white, but 58 percent of voters were white in the March Democratic primary there, according to the exit poll. This is a typical example; the share of white voters in a Democratic primary is usually a few percentage points higher than the white share of Democratic voters in November. Because our demographic projections were originally intended to reflect the November electorate, I’ve adjusted the number of white voters slightly upward so that they reflect a primary electorate instead.^
Nate Silver is the founder and editor in chief of FiveThirtyEight.