Skip to main content
ABC News
Bernie Sanders Doesn’t Need Momentum — He Needs To Win These States

The media narrative of the Democratic presidential race is that Bernie Sanders has lost momentum to Hillary Clinton. After nearly beating Clinton in Iowa and then crushing her in New Hampshire, Sanders had a setback on Saturday, the story goes, losing Nevada to Clinton by 5 percentage points. And this weekend, Sanders is about to lose South Carolina and lose it badly.

All of this is true insofar as it goes. But it doesn’t do nearly enough to account for the demographic differences between the states. Considering the state’s demographics, Sanders’s 5-point loss in Nevada was probably more impressive than his photo-finish in Iowa. It was possibly even a more impressive result than his 22-point romp in New Hampshire. Meanwhile, a big loss in South Carolina would be relatively easy to forgive.

That doesn’t mean Sanders is in great shape, however. Based on the polling so far, Sanders is coming up short of where he needs to be in most Super Tuesday (March 1) states, along with major industrial states like Ohio and Pennsylvania where he’ll need to run neck and neck with Clinton later on.

These conclusions come from a set of state-by-state targets we’ve calculated for Sanders and Clinton, which are based on some simple demographic factors in each state. As has been clear for a long while, Sanders performs better in whiter and more liberal states. But the abundance of new polling from Super Tuesday states, along with the Nevada result, gives us the data to establish more accurate benchmarks than the ones we set before. (See last week’s article “Bernie Sanders’s Path To The Nomination” for our previous estimates.) In particular, although Sanders might not have won the Hispanic vote in Nevada, he’s clearly made up ground among Hispanic voters. African-Americans, in contrast, remain overwhelmingly in Clinton’s camp. There may also be an urban/rural divide in the Democratic vote, with Sanders performing better in more rural areas.

Here are the latest numbers. I must emphasize that these are not predictions of what will happen in these states. Instead, they’re estimates of what would happen if the national vote were evenly divided between Sanders and Clinton (which it isn’t yet). In other words, they tell us whether each state is Sanders-leaning or Clinton-leaning relative to the national average. Sanders will need to beat these targets to have a shot at the nomination, especially since a tie would probably go to Clinton because of superdelegates. As you can see, however, Sanders is currently running behind these benchmarks in states with recent polling.


Take the Super Tuesday states, for instance. Our benchmarks suggest that Sanders ought to win Vermont, Minnesota, Colorado, Massachusetts, Oklahoma and Tennessee to be on track for the nomination. Sanders is going to rout Clinton in Vermont, of course; he’s also slightly ahead in Massachusetts polls, although not by as much as our targets say he “should” be. There hasn’t been enough recent polling in Colorado or Minnesota for us to make forecasts of the caucuses there, but we’d probably consider Sanders the favorite in those states also.

Sanders trails in polls of Oklahoma (narrowly) and Tennessee (badly), however, when he probably needs to win those states too. Meanwhile, he’s losing states such as Georgia by a wider margin than our benchmarks suggest he can afford. The Democrats’ delegate allocation is quite proportional, so these margins matter; underperforming his targets on Super Tuesday would mean that Sanders would have to make up more ground later on with less time left on the clock.

But the March 15 states don’t look great for Sanders either: He trails Clinton in Ohio when that’s a state where he should be able to fight her to a draw. There’s less polling for the contests beyond March 15, but the states where we have recent numbers, such as Pennsylvania, are fairly discouraging for Sanders also.

So Sanders is doomed? If he doesn’t beat these polls, then probably yes — Sanders is not going to win the Democratic nomination if he’s losing Ohio by 13 percentage points. And if Clinton has a really good night on Super Tuesday — by winning Massachusetts, for instance — that would take almost all the suspense out of the race.

But Sanders still has time to make up ground. And as we said at the outset, he’s made progress so far.

Let’s step back a bit. The benchmarks in this article are similar to the ones we published last week, but with some important exceptions. For instance, our earlier benchmarks reflected a combination of polling, demographics, Facebook “likes” and fundraising, but this time, they reflect demographics only. For a variety of reasons, we think this is a methodologically superior approach.1 For example, we’re now treating African-Americans differently from Hispanics and other non-white voting groups. Whereas Clinton leads emphatically among African-Americans, her advantage among Hispanics is more ambiguous. See the footnotes for more detail on our methodology.2

These new benchmarks make Sanders’s performance in white, liberal and rural Iowa look worse than before; they suggest that he should have won Iowa by 19 percentage points instead of essentially tying Clinton. In New Hampshire, Sanders won by 22 percentage points, but the benchmark says he should have won by 32 percentage points, a 10-point gap.

In Nevada, however, Sanders came within 5 percentage points of his benchmark. And what if Sanders loses South Carolina by 25 percentage points, as polls now have it? That’s only slightly behind his benchmark of a 20-point loss. It might not seem like it on the surface, but by these numbers, he’s been gaining ground.

But, again, I don’t want to make this out to be incredibly encouraging for Sanders. His polling in the Super Tuesday states looks pretty bad, even after allowing for the fact that they aren’t a great set of states for him. Still, follow the numbers in these states and not the talk about who has “momentum.”

Listen to the latest episode of the FiveThirtyEight politics podcast.



  1. The problem with setting benchmarks partially based on polls is that it risks confusing the effects of the campaign with the inherent demographic advantages or disadvantages of each state. Polls correctly predicted a (very) narrow Clinton win in Iowa, for instance. But that win came despite Iowa’s looking very favorable to Sanders demographically.

  2. As before, the estimates are based on a regression analysis, where the dependent variable is the average of polls in each state since Jan. 1. These estimates are then re-calibrated such that Sanders and Clinton each get 50 percent of the vote in the average state. However, there are a few methodological differences from last week’s version:

    • I’m now using all polls from our polling database since Jan. 1 and not just the state-by-state data generously provided to us by Morning Consult. Polls are adjusted for house effects and the timing of the poll (accounting for the fact that Sanders has gradually gained ground on Clinton since Jan. 1).
    • The inputs to the regression are 1) what share of Democrats are African-American in each state; 2) what share are liberal; 3) what share are Hispanic, Asian-American or of another non-white race and 4) what share live in rural areas. Note that variables No. 3 and No. 4 are strongly correlated: Rural states tend to have far fewer Hispanic, Asian and “other” voters than do urban ones. Because of this, I ran two versions of the regression — one with variables No. 1, 2 and 3 and one with variables No. 1, 2 and 4 — and then averaged them together.
    • In Vermont, Sanders’s home state, and Arkansas, Clinton’s former home state, the demographic model doesn’t do enough to account for favorite son/daughter effects, so they are excluded from the regression. Instead, the benchmarks in those states are based on polling.
    • In all other states, however, the benchmarks reflect demographics only.

Nate Silver is the founder and editor in chief of FiveThirtyEight.