To Clinton supporters, I offer my congratulations on what can only be described as a smashing victory! You can be narrowly happy about the course the Democratic primary is taking. But unfortunately, the way that Clinton won is not good news for her General Election prospects. And if she does indeed become the nominee (which was more likely than not and is now even more so), it is not good for the General Election prospects of Democrats generally. Regardless of what else happens, we must come together to address that. It is necessary to find a way to change that, or the consequences could be grim. It would have been much better for all concerned — including Hillary Clinton — if she had won by more like 20% or 30%, with a larger electorate with a demographic composition more along the lines of what polls had generally projected in their likely voter models.
For Sanders supporters, I know, it would have been nice if things had gone differently. The results can in no sense be described as good for Sanders. If we understand the reasons for what happened, there is mixed news. The bad news is that the same thing is about to happen across the rest of the south, and especially in the deep south. Another piece of bad news is that while I have provided some reasons to be skeptical of how important "momentum” is, if there is one thing that could be enough to throw all that out the window, it would be an unexpectedly large 50 point loss. But there is at least one sliver of good news to be found in all this. And that is that there is at least some reason to think that this will be confined, at least mostly, to the South and especially the Deep South. That is, if the drop in White turnout in South Carolina was due mostly to the Old Dixiecrat effect, and not very much due to the Trump effect. Sanders will need that to be the case in order to still have any sort of realistic chance of winning the nomination.
What Happened in South Carolina
There are two primary reasons why Hillary Clinton won South Carolina by a much larger margin than expected.
- First, African American vote share was up significantly as compared to the 2008 Democratic primary.
- Second, Bernie Sanders received a substantially lower vote share among African Americans than polls indicated (both state polls and national polls).
If only one of those two things had happened, Sanders might have some (very small) chance to forward with an (extremely uphill) path to victory still at least theoretically conceivable. But the one-two punch of both of these factors together was devastating.
Both of these two things are in large part explainable by low Democratic voter turnout in South Carolina.
- In 2008, 516,853 people voted in the South Carolina Democratic primary. According to voter file vote history data (which is accurate and not an estimate) 225,536 of those people were White and 291,317 were Non-White (most of whom were African American.
- In 2016, only 368,577 people (a decrease of 29%) voted in the South Carolina Democratic primary. We don't have exact numbers on race yet, but can estimate using exit polls that around 129,000 of those were White and around 240,000 were Non-White.
- This means that both White and African American turnout fell. But White turnout fell much further (-43%) than Non-White turnout (-18%).
There are fundamental reasons why this happened, and it is all but certain to be repeated across the deep south.
South Carolina Polls
South Carolina state polls badly missed the actual composition of the electorate and significantly overestimated the percent support Bernie Sanders would receive from both African Americans and Whites. As we will see, the way that they missed is very much consistent with overestimating turnout. The only (partial) exception was Clemson — which had the race at 64-14 and got the right sort of margin, but with a caveat that we will return to later.
The chart below shows race crosstabs of support by race for all South Carolina polls conducted in February. Numbers highlighted in yellow mean that those numbers were not directly reported by the pollster, but were derivable/could be estimated (within a percentage point or two because of rounding error) from other data that they did give. Clemson and Gravis did not have crosstabs:
2008 Exit polls showed an electorate that had been 55% African American and 43% white. On average, these polls had African American vote share at 56.5%, and had Clinton winning African Americans 66.7%-22.3%. In reality, or at least insofar as the exit polls are roughly correct, African American vote share was about 5 points higher, and African Americans voted for Clinton 86-14. White vote share was similarly overestimated (by about 6 points), and on average estimated that the race would be close among White voters (whereas exit polls say Clinton won by 9 points with Whites).
The only pollsters that had African American vote share about right were Selzer/Bloomberg, CNN, and CBS/Yougov. But each of those (and all other polls) substantially overestimated Sanders' support levels. Since on average they had found Clinton up, one would have expected Sanders to be likely to get at least a bit more than 22.3% (getting at least some of the undecideds). Instead, he got substantially less — only 14%.
Moreover, South Carolina polls certainly seemed to show that Sanders was gaining ground with African American voters. On Feb 18, DKE commenter nimh had an excellent open thread comment (worthy of a diary) analyzing the trends. The whole thing is worth reading even now, but the most important point for our purposes is this chart, showing the crosstabs of support among African Americans in South Carolina over time from PPP and Yougov:
And this is the same chart with CNN’s poll added in (nimh notes that the trend for CNN is perhaps a bit messed up because their first poll included Biden):
nimh also noted the same sorts of trends in national polls (again, the whole comment is worth a look), though they were less clear:
Was it momentum what done it?
So why were the polls wrong? Was it momentum what done it? I don't think it really was.
I have seen some people argue that Clinton did so well in South Carolina because of positive momentum following Nevada. Or alternatively, that the reason Clinton did so well was because Sanders “abandoned South Carolina” (despite having 200 staff, spending a lot of money, and spending some though certainly not all of his time there). It would be too strong a claim to say that momentum had no positive effect for Clinton, it seems unlikely that this made more than a few points worth of a difference.
That is actually a bad thing for Sanders. Why? Because it means that this is not something that he could practicably reverse simply by “regaining the momentum.” The causes lie deeper than that.
Exit polls estimate that:
- 60% of voters in South Carolina made up their minds prior to this last month, and voted 76-24 for Clinton.
- 15% made up their minds in the last few days, and voted 71-29 for Clinton.
- 23% made up their minds in the last week, and voted 76-24 for Clinton.
South Carolina Exit polls on when people decided who to support
It does not appear from this that Clinton did better among late deciders than she did among (indeed, it's even a bit the opposite).
Furthermore, 53,986 of the total votes were absentee ballots, mostly cast well before Nevada. 76% of those , and Clinton did especially well with absentee ballots. Absentee ballot results reported first, which is why Clinton’s lead started out even larger than it ended up at the end of election night.
What assuredly happened there is that the Clinton campaign pushed early/absentee voting, to great effect — and hats off to them. This is not surprising, because much of Clinton’s campaign staff is made up of former Obama staffers who turned banking early votes into an art form. But the point is that these votes were not in any way affected by “momentum” from Nevada.
In sum, it just doesn't seem plausible to think that the result would have been much different at all in the alternative Universe where Bernie Sanders won Nevada and Clinton didn't get any “momentum” from Nevada. It is possible the late deciders might have broken a bit more evenly (which at most would have been worth a few points). It is also quite possible, though, that they would have broken in pretty much the same way.
Historical South Carolina Primary Turnout
In my previous South Carolina Primary Preview, I posted this chart with the actual demographic data for actual Democratic primary voters in South Carolina between 1984 and 2014. See that for more detail.
Data is only available for one Presidential primary — the 2008 Presidential primary. Note that they changed the age definitions in 2008, which is why I split this into two pieces. The rest of the data is all for regular (non-presidential) primaries:
Now, look at that chart and consider the Clemson poll (which had Clinton up 64-14). The Clemson poll had an extremely tight likely voter screen. They only included people who had voted in 2 of the last 3 primaries (2014, 2012, and 2010). If you look at the numbers, the turnout in those primaries was 132,321, 111,493, and 210,504. So at the very most, the electorate Clemson was polling included 210,504 people (if everyone who voted in 2010 also voted in either 2012 or 2014). In other words, at the very maximum, only 57% of the people who actually voted even had a chance to be included in the Clemson poll.
Now look at the demographics of the people who voted in those elections, and compare them with the demographics of the people who voted in the 2008 Presidential primary. You can see that the people who voted in 2010, 2012, and 2014 were much more heavily Non-White (59%, 63%, and 68%, as compared to 56% in 2008) and were substantially older than the larger electorate that voted in the 2008 Democratic primary (29% 34%, and 37% Senior, as compared to 22% in 2008).
So a lower turnout electorate would be more similar to the 2010/2012/2014 electorates, and would especially consist of a disproportionately large share of older African American women. A higher turnout electorate, on the other hand, would be relatively more similar to the electorate that turned out in 2008, and would include greater shares of relatively younger people and of Whites. And since support for Clinton or Sanders correlates very strongly with both race and with age, that difference between a low turnout electorate and a high turnout electorate is highly significant.
So those polls that found Sanders getting 22% or so of the African American vote (not including undecideds) were probably not incorrectly registering support among African Americans. The people that they interviewed probably did indeed support Sanders. And they probably did not change their minds, nor were they affected by "momentum." Instead, Sanders’ 14% support rate among African Americans was lower than the 22% estimated by polls because the African Americans who supported Sanders were the least likely to vote.
The problem with the polls was that in a low turnout election, those African Americans who were more likely to support Sanders were also the same African Americans who were least likely to actually vote. That's also probably the reason why South Carolina polls seem to have overestimated Sanders’ support among White voters — in particular because the White voters who they polled were somewhat younger than the White voters who actually voted.
So back to the Clemson poll.
I would bet that of the people Clemson polled, they did vote about 64-14 for Clinton. Now, you might say that this was spot on — they voted 64-14 for Clinton, and the 22% undecideds broke pretty much evenly, leading to a 75-25 estimate for the poll, which closely matches the actual result.
However, I don't think that’s how the undecideds there broke. I think those 22% undecideds in the 64-14 Clemson poll broke about 3 to 1 for Clinton — consistent with the exit polls, which show late deciders voting about the same as everyone else (3 to 1 for Clinton).
In that case, the 64% — 14% Clemson poll turns into about an 80.5% — .5% result. But the actual result was significantly better than that for Sanders. It was 73.5% — 26%. So where did Sanders' extra 6.5% come from that got him up to 26%? It came from the (at bare minimum) 158,073 people who did in fact vote but who were ineligible to be included in Clemson's likely voter screen because they had not voted in 2 of the last 3 primaries. And those people were both more likely to be younger and more likely to be White than were the people included in the Clemson poll. That’s where Sanders’ extra 6.5% came from.
It is true that not nearly as many such people voted But still, some of them did vote. Returning back to the Clemson poll, the 64-14 Clemson poll’s likely voter screen was too strict. In order to get a correct result, a likely voter model should have been less strict than Clemson but more strict than all the other pollsters.
This means that the Clemson poll was not actually that much more accurate than all of the other polls. Instead, the Clemson poll was inaccurate in a different way. It had a likely voter model that was much too strict. All the other polls, by contrast, had likely voter models that were much too lenient. The Clemson poll only appears to be more accurate (it got a 50 point margin, compared to the actual 48 point margin) because the undecideds broke in such a way to make its incorrect result seem to match the actual result (if you just looked at the margin, or if you broke undecideds evenly).
In order to really be correct, a poll would have had to have a likely voter screen somewhere in the middle of Clemson's and the broader screen that was being used by every other pollster.
Did state polls conducted after Nevada show momentum to Clinton? Not really.
Now, you may be thinking to yourself, "but the Clemson poll was picking up Clinton's momentum from Nevada! It was showing a shift in the race to Clinton!” Like Marco Rubio, I am going to have to try to dispel that notion.
On its face, this does sound plausible and should not be dismissed. However, what is the evidence actually standing behind it?
There were only two polls conducted after Nevada (Clemson and Emerson). Look at South Carolina polls from RCP:
The previous Clemson poll was conducted all the way back in October. Nonetheless, we can learn something by looking at it. It had Clinton up 43-6. At the time, that poll just looked like it had an absurdly high share of undecideds. But its overall margin (37 point Clinton lead) was not that different from all of the other polls conducted at about the same time.
But what was different was that Sanders had a much lower absolute level of support (6%) than other polls conducted around the same time (18%, 25%, 15%, 21%, etc). Why were those other polls picking up more Sanders support? Because they were polling a much broader electorate, while Clemson was polling its ridiculously tight 2-of-the-last-3 likely voter screen.
After this, Sanders climbed in all of those other polls up to the 30s or low 40s. If Clemson had conducted another poll at that time, they would have found Sanders in the low teens.
So very little changed over this whole time — Sanders picked up a reasonable amount of support among the broader potential electorate in South Carolina. But he picked up much less support among the more narrow electorate of the most consistent Democratic primary voters who were most likely to vote in a low turnout election, who were also disproportionately older and African American.
Why were those other polls picking up more Sanders support? Because they were polling a much broader electorate, while Clemson was polling its ridiculously tight 2-of-the-last-3 likely voter screen. There was no sudden shift here — it was just polls using different likely voter screens and people gradually making up their minds, lowering the numbers of undecideds.
I looked casually at some other states and don’t see much signs of large sudden shifts pre and post Nevada, in cases where the same pollster had polls both before and after Nevada.
The incorrectness of 2008 SC exit polls on age
In the South Carolina preview post, I also noted that 2008 exit polls were flat out incorrect in their estimation of the age of the electorate, when you compare it to the actual vote history data (which is correct, and not an estimate):
- The exit polls said that voters aged 18-24 made up 8% of the electorate. In reality, they made up 3% of the electorate.
- The exit polls said that voters aged 18-44 made up 40% of the electorate. In reality, they made up
These are all facts, not speculation.
Obama did not do well in South Carolina thanks to young voters. Instead, he did well thanks to support from older African American voters, and especially from older African American women.
Effects influencing African American Vote Share
And after posting that historical turnout data, I also noted that this data raised the very important question of what share of the vote would be cast in 2016 by African Americans. And I noted the following effects that could influence that:
The Aging Af-Am population effect — Aging of the African American population and the general trend towards increased African American turnout that began even before 2008 should continue. Advantage Clinton.
The Obama effect — With Obama on the ballot, African American turnout should be somewhat lower than otherwise. Advantage Sanders.
The old Dixiecrat effect — The decline of Dixiecrats and southern whites switching to the GOP should decrease the white vote share and increase the African American vote share. Advantage Clinton.
The non-Dixiecrat white liberal effect — This doesn't alter the African American vote share, but the decline of Dixiecrats should mean the remaining white voters are a bit more liberal. Advantage Sanders.
The Trump effect — Since the South Carolina GOP primary was first, some voters (especially white independents) may have voted in the GOP primary who would otherwise have voted in the democratic primary are more likely to have already voted in the GOP primary. Advantage Clinton.
I don't know how those competing effects will shake out. If I had to guess, I would say that African American vote share is more likely to go up a bit than to go down. But I doubt anyone else really knows either. If you think you do, let’s hear it in comments!
If the more pessimistic polls for Sanders turn out to be correct, I suspect it might actually be more because of an increase in African American vote share than because of any sudden shift in the dynamics of the race.
Now that we have seen the results from South Carolina, we can evaluate how those competing effects did in fact shake out. First, there was a substantial negative effect on African American turnout of not having Obama on the ballot — it dropped 18%.
However, that effect was dwarfed by the combined power of the “old Dixiecrat effect” and of the “Trump effect” which combined to reduce White turnout by much more — a drop of 43%.
This disproportionate drop in White turnout as compared to Non-White turnout not only caused the share of the electorate to become more heavily African American than in 2008. But because the voters who did not show up in 2016 but whom would have voted in 2008 were disproportionately younger, and because there was such extreme polarization in the vote by age, that meant that not only did the lower turnout mean higher African American vote share, but it meant that the African Americans left making up that larger vote share (with lower turnout) were precisely those African Americans who
And so that’s why we got the double whammy of higher African American vote share combined with lower than estimated support from African Americans for Sanders. And because of that double whammy, Clinton didn't just win by the 20% or even 30% that pretty much everyone was expecting, but she won by an unexpectedly large margin that the polls did not anticipate.
To see how and why this was, we will now turn to Georgia, which has some more detailed demographic vote history data.
Georgia 2008 Democratic Primary Turnout
This chart of data from the Georgia Secretary of State shows the distribution of voters in the 2008 Democratic primary by age and race. 1,056,251 people voted. 418,760 of them (39.6% of the total) were white, 585,023 (55.4% of the total) were Black, and very few were Hispanic, Asian, or Other race:
And this shows the same data visually:
2008 GA Dem Primary Turnout by Age and Race
Here’s the same data converted to percentages:
25% of all White 2008 Democratic primary voters were age 65+. By comparison, only 12% of African American 2008 Democratic primary voters were age 65+. In the same sort of chart, we can see the same thing. The age distribution was skewed much more heavily towards older voters for Whites than for African Americans, Hispanics, Asians, or Other race:
Share of GA 2008 Dem primary voters in each age group by race
Or viewing things the other way, we can compare what percentages of each age group were White, Black, etc. Younger voters were disproportionately African American, while older voters were disproportionately white:
Look through the pie charts, and you can see the White share of the pie increasing exponentially as you get to the oldest age groups:
Share of each age group that was each race:
So from all this, you can see a striking difference between the age distribution of White 2008 Democratic primary voters and Black 2008 Democratic primary voters. The age distribution of White Democratic primary voters were extremely heavily skewed towards Seniors, while the ages of Black primary voters were distributed around middle age. There were many more Senior white voters than Senior black voters, and among all other age groups, African Americans substantially outnumbered Whites.
Why do we see this sharp difference? Because a lot of the White Age 65+ voters who voted in the 2008 Democratic primary were Dixiecrats who grew up in the age of Jim Crow segregation. Younger whites in the South were more likely to be Republicans, while older whites were more likely to be Dixiecrat remnants.
It is an unfortunate fact of life that people die, and are more likely to do so as they get older. So 8 years later, a lot of these people are now deceased. Of those who are still around, a lot of them abandoned the Democratic party when Obama got the nomination. More still abandoned the Democratic party in 2010, and throughout the rest of his Presidency.
And now, even more are abandoning the Democratic primary in order to vote in the GOP primary for Trump.
It should be at least as extreme in the rest of the Deep South as in Georgia
Now, this is Georgia we are talking about. Compare that to South Carolina. In Georgia, metro Atlanta has a reasonable number of white progressives. In South Carolina, that is more limited. The African American population is also younger than the national average (55.8% under age 45) in Georgia, but older than the national average (51.1% under age 45) in South Carolina.
For those reasons, the sharp difference in age distributions we can see in the more detailed Georgia data would almost certainly be even more extreme in South Carolina — and likewise stronger for Mississippi, Alabama, and Louisiana.
And for these same reasons, we should expect the same sort of drop in White Democratic primary vote share. And with lower African American turnout, the African American electorate will be substantially older than the African American electorate that showed up in 2008.
Implications for Super Tuesday
So, it should be manifestly clear to everyone that very much the same thing is about to repeat all across the South, and especially in the Deep South.
There is no reason to think that turnout across the rest of the South will be any higher than in South Carolina. If anything, it is more likely to be lower, because the campaign has been much less active in those states. The lack of an “Obama effect” increasing African American turnout means lower African American turnout. That means that the African Americans who do vote will tend to be older and will give a lower share of their vote than would African Americans generally. And the “Old Dixiecrat effect” and the “Trump effect” both will mean lower turnout among whites. And there is every reason to believe that just as those twin effects overwhelmed the lack of an “Obama effect” on Black turnout in South Carolina, they will do so elsewhere. That means that the double whammy that led to Clinton’s enormous South Carolina margin is about to be repeated — and may well be exceeded. At this point, this is entirely unavoidable and cannot be changed. If that's not clear to anyone, I think it will become so in the early evening of March 1.
In fact, it is all but certain that in the rest of the Deep South African American vote share will be higher than 2008 (but unfortunately for all of us with lower overall turnout). It is at least as likely as not that African Americans across the rest of the Deep South will vote even more heavily for Clinton than was the case in South Carolina.
African American Support for Sanders in Southern Super Tuesday States
PPP’s polls of Southern Super Tuesday states (and Louisiana, Michigan, and Mississippi, which vote later) found that African support was as follows:
African American vote in PPP polls
Remember, that 23% in South Carolina was a substantial overestimate. And it was an overestimate because it is included a greater share of younger African Americans than the African Americans who actually voted.
But even so, that 23% from South Carolina is the highest of any of the states polled (matched by Texas). What does this mean? It means that those already low African American vote percentages that Sanders is receives in these PPP polls are also more likely than not to be overestimates of the support that he will get from African Americans if turnout is low (restricted mostly to the disproportionately older women African Americans who vote regularly in Democratic primaries). If turnout were hypothetically higher (like in 2008), then Sanders might indeed get (for example) 14% in Georgia.
But we have no reason to expect that turnout should be higher, given that it was lower in South Carolina, and also was lower in previous states with Democratic primaries. First, Obama is not on the ballot. That means that the African Americans who do vote will tend to be older than the ones in the broader electorate that pollsters are polling. So as a Bernie Sanders supporter, right now I would happily take those levels of African American support that PPP has. I am afraid it is actually going to be lower than that.
So I would not be at all confident that Sanders will even break 10% among African Americans in Alabama and (to a lesser extent) Georgia. And at the same time, African American vote share is likely to increase as compared to 2008 for the same reasons it did in South Carolina (Old Dixiecrat Effect and Trump Effect) as it did That would be even more devastating for Sanders than what occurred in South Carolina, as it is quite possible (if not likely) that Sanders will be held under 15% in particular African American majority Congressional districts across the South. If this happens, Sanders will not meet the 15% viability threshold and will get zero delegates from those districts. The reason for this is not just momentum, but fundamentals. That is to say, even if Clinton receives no “momentum” benefit whatsoever from South Carolina, we should still expect this to happen. Any momentum effect will only help her further.
The rest of the country is not the Deep South
The good news for Sanders, at least insofar as we can call “good” that which is not outright dire, is that there is at least some reason to think this will be less extreme in the rest of the country. Even in the outer South (i.e. Virginia, North Carolina, Florida) there are still more White Democrats. So the effect of the absence of the old Dixiecrats should be less extreme there, and in the North, Midwest, and West, it should be relatively more absent.
But not entirely. It is no secret that a lot of working class whites defected to the GOP starting in 2010. And now, we have seen a great many of them flocking to Trump's banner in the primaries that have already occurred.
When the race gets to states like Ohio, Michigan, Missouri, Illinois, and Pennsylvania, this will merit watching very closely. And Virginia merits watching very closely as well, as a Southern state that is not in the Deep South, but which is also a swing state. Will there also be higher African American vote share combined with lower turnout? If so, then the same sort of one-two punch will apply to some degree.
Why this is not just bad for Sanders, but also for Clinton and Dems generally in November
But if so, it would not only be very bad for Sanders, it would be very bad for Democrats more generally. It would be very bad for the General Election, regardless of who the nominee is.
It would be very bad news because it would mean a few things:
- The effect of having Obama on the ballot in 2008 and 2012 on African American turnout was large. That means that a large chunk of the increase in African American turnout that we saw in 2008 and 2012 was not something that would have happened otherwise (since African American turnout had been increasing in 2000 and 2004). That means that without having Obama on the ballot in 2016, we should expect substantially lower African American turnout in the General Election (unless we can figure out how to turn that around).
- At the same time, it would mean that the decline in White Democratic primary turnout we saw in South Carolina was not just attributable to the Old Dixecrat effect, and doesn't just apply. Moreover, if we can also see this effect in the Midwest (Ohio, Michigan, Pennsylvania) and elsewhere in the country, then it suggests that the Trump effect is strong. And that would be a very bad thing in the General Election.
Those two factors add up to a one-two punch that could be every bit as devastating to the Democratic nominee in the General Election as was the one-two punch in South Carolina to Sanders.
So whether you are a Clinton supporter or a Sanders supporter, you should really be hoping hard that when the Super Tuesday exit polls come out for Virginia, they somehow show a steady or reduced African American vote share, combined with high turnout in the Democratic primary. If that happens, it means that these effects will just be limited to the Deep South. It means that the one-two punch we saw in South Carolina will mostly have resulted from the Old Dixiecrat effect and the Obama effect, and not from the Trump effect. If you are a Clinton supporter, yes, that means Clinton won't win Virginia in the primary by quite so much. But it means that if she (or Sanders or anyone else) is the nominee, then she has a much better chance of actually winning Virginia in the General Election.
But if the exit polls in Virginia show an increased African American vote share combined with significantly lower Dem turnout, with a greater decrease in white turnout than in African American turnout, it will mean that the Trump effect is strong. And if the Trump effect is strong now, it will probably also be strong in the General Election. Meanwhile, without the Obama effect, we have every reason to expect lower African American turnout. And if the Trump effect is strong in Virginia
This is something that we absolutely need to turn around
So it's absolutely imperative that we find some way to do three things:
- Bring back some form of the Obama effect on African American turnout, so it does not fall (as it did in South Carolina).
- Give younger voters the motivation to actually turn out and vote.
- Effectively fight the Trump effect, and stem the anti-establishment exodus of working class whites to the GOP, and to Trump.
If we can’t find a way to do that, then in the outer South:
- North Carolina will not be remotely competitive.
- Virginia won’t lean Dem, and could be a true tossup or even lean R.
- Florida won’t really be a tossup, but will probably lean reasonably R as in 2004 (unless gains among Hispanics are fully strong enough to offset the Trump effect in North Florida and the drift of older retirees to the GOP).
That's enough by itself to return the electoral college map to something more similar to what we had in 2000 and 2004. And if the Trump effect is strong in places like Ohio, Pennsylvania, and Michigan, then we could have a real fight on our hands, without any clear reason to think we have the upper hand.
In other words, with Trump on the GOP ballot and Obama off the Dem ballot, the Obama coalition could come tumbling down and crash into pieces. That “blue wall” we liked to think made America safe from another George W. Bush? Gone. History.
That's not to say that the GOP (with Trump or whoever else) would necessarily win. It would probably be competitive — similar to 2000 or 2004. But the clear electoral college advantage that we enjoyed in 2008 and 2012 could be gone.
Super Tuesday Preview and Projections
So given the results in South Carolina, I decided to drop all state polls from the South from my model, because I think they are almost all missing the true composition of the likely electorate, and as a result, probably overstating how well Sanders will do. Instead, I am relying on the pure demographic/exit poll side of the model for the South. For my South Carolina prediction, the demographic side of the model had Sanders lower than did state polls, and would have been more accurate if it had African American vote share and support rates right (it overestimated the former based on polls and overestimated the latter based on 2008 exit polls).
Similarly to in South Carolina, I expect the combination of the Old Dixiecrat effect, the Obama effect, and the Trump effect to combine to form a lower turnout electorate with significantly higher African American vote share than in 2008, and with the African Americans who do vote being disproportionately older and less likely to support Sanders. Specifically, I expect African American vote share in the South to be something like this:
That comes from assuming that the Old Dixiecrat effect, the Trump effect, and the Obama effect will play out similarly in the rest of the Deep South as in South Carolina, and from assuming that their combined force will be at least somewhat less strong in the Outer South.
I am still including state polls in my model outside of the South, because state polls did not seem to so badly miss the composition of the electorate in Iowa, New Hampshire, and Nevada. Outside of the South, if turnout is lower it does not hurt Sanders quite as much as it does in the Deep South — though it still very probably hurts him.
Scenario: Current RCP Average (50-40 Clinton)
This is roughly what we might expect the outcome to look like if the current Real Clear Politics average (showing Clinton ahead 50-40) is about right. Under this scenario, Clinton wins on Super Tuesday by about 506 delegates to 352.
If anything, I am afraid that these projections for the Deep South may be overoptimistic for Sanders. That is because they are based off of the 14% level of support Sanders received from African Americans in South Carolina. But as we have explained above, there is reason to think that Sanders could get even lower rates of support among African Americans than that. At least in South Carolina he had a very active campaign with 200 staffers, and he did spend a considerable amount of time campaigning there.
Scenario: The CNN poll is right, Clinton has a large national lead
CNN’s latest national poll has Clinton up by 55-38. The crosstabs are a bit screwy (they have Sanders at 39% among non-whites, which given the results in South Carolina suggests that they are still including too many younger non-whites in their likely voter screens, at least for the Deep South). But in any case, the race could look something like this if Clinton has gotten a big momentum bump from South Carolina that has changed up the race. While I have argued that there is reason to be skeptical of the effects of Nevada momentum on South Carolina, the effects of an unexpected 50 point win are another animal entirely. Alternatively, this could be how things turn out if Sanders had never actually gotten much closer, and if the polls that showed him closer were wrong. In this case, Clinton would do very well on Super Tuesday — winning by to 518 delegates to 340. If the Super Tuesday results are something like this, then it is all but impossible to see how Sanders could have a realistic chance to win the nomination absent some black swan event that totally shakes up the state of the race.
Scenario: Sanders gets blown out in the South, but does much better in the rest of the country
Given this, is it possible to conceive of any sort of scenario in which Clinton wins by as much as the South Carolina results suggest she will across the South, but in which Sanders is still competitive nationally?
This is a static scenario in which that happens. A scenario like this will be more likely if voter turnout is relatively high (especially in CO, MA, MN, and OK). It seems clear that Republican turnout will be high, but it is not clear if Democratic turnout will also be high.
If the results are close to this, I for one would count it as as good of a night for Sanders as could be realistically hoped for, considering the scale of Clinton's South Carolina win. This would require Sanders to overperform expectations in Massachusetts, Minnesota, Colorado, and Oklahoma. Under this scenario, Clinton comes away with just a bit more than 100 delegates more than Sanders on Super Tuesday. But a result similar to this would show that Clinton’s strength is really just limited to the South, and particularly the Deep South. With results not too dissimilar from this, Sanders would clearly show that he could win everywhere else:
If something like this happens, I would expect that Sanders would probably do better in Minnesota (where the only available poll, which is out of date, significantly drives down the model's projection) and worse in Massachusetts than this projection — because of the strength of the Party machine in MA and because it would dramatically overperform recent polling.
Alternatively, if Sanders were to do a bit worse but then were to pick up ground in national polls (and do better than this static scenario shows in later states), then the same sort of eventual outcome might be possible. But the amounts by which he would have to win most states that are not in the South by in order to get a majority of the pledged delegates is very large indeed, and are generally well above what available polling and expectations indicate is most likely. So while that might not be mathematically impossible, strictly speaking, it is a very steep hill to climb. Also, this does not show Sanders actually winning — it just keeps the delegates competitive. In order to win the nomination while getting blown out by this much in the South, Sanders would have to do a bit better than this scenario envisions somewhere.
Where do we go from here?
As I have tried to spell out, we have a problem. The one-two punch that doomed Sanders in South Carolina also exposed a looming problem for whoever becomes the nominee in the General Election (and that looks more likely than not to be Hillary at this point). Low turnout of the sort seen in South Carolina, with African American turnout plummeting and White Democratic turnout falling through the floor even more (with streams of defectors to Trump), is a potential recipe for disaster. We had all better hope that it is limited to the Deep South, and that an awful lot of it is caused not by the Trump effect, but by the Old Dixiecrat effect.
And as I said, it's absolutely imperative that we find some way to do three things:
- Bring back some form of the Obama effect on African American turnout, so it does not fall (as it did in South Carolina).
- Give younger voters the motivation to actually turn out and vote.
Effectively fight the Trump effect, and stem the anti-establishment exodus of working class whites to the GOP, and to Trump.
So far, neither Clinton nor Sanders has accomplished item #1. Clinton obviously managed to earn very high levels of support from African Americans, and is set to do so across the rest of the South. But her higher-than-expected 84-16 margin was a direct result of lower than expected African American turnout. If African American turnout drops in the General Election, as the South Carolina primary results suggest that it has a very good chance of doing, that will make things much more difficult. I don’t know what it would take to reverse this — maybe an African American for VP. But in Clinton's case, if that were someone like Cory Booker or Deval Patrick, that would just further expose her to attacks from Trump about her coziness with Wall Street. And someone like Kamala Harris would probably be considered too inexperienced before having a term or two in the Senate under her belt. It is probably too much to hope for that Obama merely urging African Americans to vote will be enough by itself. He did this in 2010 and 2014, without the sort of effect that would be needed.
On item #2, Sanders has excited younger voters, but it is unclear just how much he has been able to actually turn that into turnout. But we shouldn't just look at the exit polls, and conclude from that that Sanders is either succeeding or failing at turning out young voters. As we have seen in the case of Georgia, the exit polls on age in 2008 were flat out, demonstrably incorrect. That is not arguable, it’s a fact. At least in Georgia, young voters did not turn out at high rates, even in 2008. I don’t know how far that extends to other states, but we need to find out.
On item #3, Sanders offers at a counter-appeal to Trump's demagogic xenophobia, which is at least more effective than anything else anyone else has been able to accomplish. He re-frames this angry impulse in an alternative, more positive direction — into economic populism and into hope for a better future. Clinton has recently begun to mimic this, which has made her a better candidate. She would not have done so without Sanders in the race. Whether that is enough to offset the Trump effect is an open question — and one we assume is resolved at our own peril.
Unless and until we figure out how to resolve these problems, it is not just Bernie Sanders and Sanders supporters that have a problem — it’s all of us who don't want to see Donald Trump or possibly Ted Cruz in the White House.
So even under the worst case scenario of results for Sanders, I would argue that it is in all of our interests — and even in Hillary Clinton's interest — for Sanders to stay in the race and to continue doing what he can to gin up enthusiasm amongst the base.
If Sanders doesn't do so, then media attention will be sucked away and transferred to Trump, and we’ll be left going into the General Election without having resolved these 3 problems, and without any clear plan to resolve them. In particular, any contribution Sanders might have made to resolving items #2 and items #3 will be lost.
If we do that, and if we do that with Clinton as the nominee, she could still win — and may well even be favored. After all, Trump is Trump, with negatives that I don't even need to begin to describe.
But that's a very big risk to take without having a clear conception of how things will turn out.
Finally, here are demographics on all the Super Tuesday states: