Second Choices and Delegate Allocations: Why Primary “Electability” Is a Wild Guess

Let me declare my biases up front. I voted for Elizabeth Warren yesterday in Minnesota. She was my first choice among the huge field of Democratic contenders. Like hers, my main priority is seeing as many of her plans as possible enacted. My next four choices dropped out before I got to vote. If I have to choose between Sanders and Biden (I don’t), my choice is Sanders, but the margin is very narrow, based mostly on candidate negatives on both sides, and easily shifted by things like VP choice.

I also spent nearly a decade doing actuarial (pension) math for a living. I’m not a trained actuary, but I have a lot of experience in determining probable outcomes in situations with a lot of variables. More importantly for this post, I’ve spent an awful lot of time identifying and making explicit the assumptions that go into these calculations.

Yes, I’m writing this (or compiling it from various places I’ve talked about the question over the last few days) because of the pressure for Warren to drop out of the race. At best, it’s based on assumptions about what would happen to her votes that don’t match the best information we have. At worst, it’s counter-productive to seeing progressive issues move forward.

Let’s start with the assumptions being made about what will happen if Warren drops out. The two big assumptions are:

  • People who are prepared to vote for Warren will switch predominantly to Sanders.
  • Those votes will roll up to enough of a gain in Sanders delegates to prevent a brokered convention.

There’s also a third assumption I’m seeing in my social media that Warren really wants delegates so she can throw them to Biden specifically to prevent Sanders from becoming the nominee. I’ll talk about what can happen at the convention later, but I’m not going to address this assumption directly because:

  • It’s coming from a small minority even of Sanders supporters.
  • Those are the same people who think snake emojis are solid politics.
  • Their reaction to this post will be the same no matter what I say.

On to the main assumptions.

Second Choices

Would Warren supporters move primarily to Sanders if she dropped out? Not according to them.

The late-February poll from Morning Consult, not terribly different from the other limited data we have on fallback preferences, shows Sanders would gain a plurality of votes from Warren supporters. That is, he’d get more of those votes than any other candidate polled. However, only one of those other candidates is still in the race.

If you add fallback support for all the candidates who endorsed Biden to his support, you’re looking at 44% for Biden. That’s more than the 40% for Sanders, but within the margin of error.

But wouldn’t Warren endorsing Sanders shift that? We don’t know.

Wouldn’t some of the people who had Klobuchar or Buttigieg as their second choice prefer Sanders as their third choice? We don’t know.

Truth be told, we don’t even know how many of these people really have a strong preference after Warren or who gave an answer just because they were asked. We have no idea how stable or predictive of actual results polling on second choice candidates is, and we’d be hard-pressed to find out. That’s even more true this year, when we’ve seen large swings in polling from week to week and between polls and actual votes.

We can say that second choices in this primary (at least as expressed in polls) aren’t primarily about policy positions. You can see that among Warren supporters and among Sanders supporters. In the Morning Consult poll, 35% of Sanders supporters have Warren as a second choice. The fallback support for Biden and Bloomberg among those same supporters, by contrast, totals 37%.

What would happen to Warren votes if she dropped out? We don’t know. We can guess, but if we’re going to be honest with ourselves and each other, we have to admit we can’t know.

All we really know is that weekly fluctuations in Warren support haven’t produced mirrored fluctuations in Sanders support.

Delegate Allocations

Delegate math is so much fun.

No, it’s not. One of the delights of my life is building complicated modeling spreadsheets. I find delegate math to be kind of a necessary annoyance.

Yes, I said necessary. I could be persuaded that choosing a presidential nominee should happen differently, particularly in a world with the Electoral College, but that’s not the only thing we choose delegates to the national conventions to do. They also work on the party platform. You’d have a tough time persuading me we shouldn’t have strong geographic diversity among the people building the platform.

As long as we also have those people vote for the nominee, this math for assigning pledged delegates is going to be complicated. This is why we don’t have full delegate counts yet in states with essentially all their returns in. I’m not even going to talk about caucus math, though you can find Iowa explainers if you want them. I’ll talk more about unpledged delegates—superdelegates—when I get to the convention.

Firstly, how this is done varies from state to state. State parties choose what geographic levels to assign pledged delegates at and how to split their delegates among various levels. I’m going to simplify things and talk about proportional delegates assigned at the state level and delegates assigned a level below that (by congressional district in Minnesota). If you’re sharing this, and you know enough to tell people that things are different in your state, please do.

When we’re talking about delegates assigned at a geographic level below the state, the important things to remember are:

  • We’re assigning whole, not fractional, delegates to a candidate with a small number of delegates in each region.
  • Delegates get distributed among all the candidates that get at least 15% of the primary vote in each region.
  • We often have no polling data at this level. The data we do have may have large margins of error due to small sample sizes.

In practice, this means that regional delegate splits are only roughly similar to the vote splits in that region. For example, a candidate who gets 15% of the vote in a region with four delegates will get 25% of the delegates. If that happens consistently across several regions, it can skew delegate counts significantly at the state level.

Also, you’re not going to be able to tell how close any given candidate is to the 15% threshold in a region or to any other important tipping point in the delegate math until people vote. You can guess. You can assume your region will go the way of a larger poll or the endorsements in your area or your social group or your pessimism, but it’s a guess. It isn’t based on the data, because there isn’t any good systematic data at this level.

This process repeats at the state level. The important differences are that there’s more polling data and typically a larger group of delegates to split.

The polls are still subject to the dramatic shifts we’ve seen this year. The larger group of delegates at the state level means we can work in something closer to percentages. I did this last night for a friend in Texas to demonstrate the difference Warren hitting 15% in a state can make if those votes aren’t scared off by “electability” arguments. I’ll do it again here with updated (but not final) results.

Assumptions (extreme rather than probable):

  • All changes to Warren’s votes come from or go to Sanders only.
  • Warren got no early votes

As things currently stand, Biden has 34.3% of the vote, Sanders 30.0%, and Warren 11.4%. Warren gets none of Texas’s 79 state-level delegates, so they’re split between Biden and Sanders:

  • Biden gets 42.
  • Sanders gets 37.

If Warren had dropped out under these extreme assumptions, the percentage for Sanders would be (roughly) 41.4%. Meaning:

  • Biden gets 36.
  • Sanders gets 48.

If Warren had hit the 15% threshold under these extreme assumptions, Sanders would be down to 26.4%. Delegates:

  • Biden gets 36.
  • Sanders gets 28.
  • Warren gets 15.

Either way, the progressive delegate total is the same, even under our most extreme assumptions favoring Sanders. Again, I’ll get to what happens at the convention shortly, but first, a more realistic scenario that still favors Sanders.


  • Warren’s votes come from or go to Biden and Sanders at a ratio 33%/67%, giving Sanders a substantial share of Warren partisans who listed Klobuchar or Buttigieg as their second choice in the Morning Consult poll.
  • Warren would have about 11% of the 13% of Texas votes cast early but no votes on Election Day. This means there still would have been about 1.4% of the primary vote for her that couldn’t go to another candidate.

Current allocations would still be:

  • Biden gets 42.
  • Sanders gets 37.

Allocations if Warren dropped out:

  • Biden gets 40.
  • Sanders gets 39.

Allocations if Warren hit the 15% threshold:

  • Biden gets 35.
  • Sanders gets 29.
  • Warren gets 15.

The progressive delegate total here is 44 to Biden’s 35. This is the only scenario under more reasonable assumptions where this happens.

At the Convention

All right, we can finally talk about the convention. If you’re tired of math, you can relax here. We’re down to scenarios. This is another area where we’re largely assuming the possibilities are more limited than they are.

In scenario 1, one candidate gets an outright majority of delegate votes based on pledges in round 1. We’re done. It’s over.

Except this scenario isn’t the most likely. Having Warren in the race and reaching thresholds for delegates in various states keeps scenario 1 less likely. But if it does happen, FiveThirtyEight estimates the winner is about four times more likely to be Biden than Sanders. I can’t begin to count the assumptions that go into that model or decide how reasonable I think they are, but that’s where the information we have is pointing.

In scenario 2, Sanders’ delegates and Warren’s delegates together total more than half the pledged delegates. In this case, Warren has the option to look at the totals going into the first vote and direct her pledged delegates to change their votes. It’s not really precedented, but we’re talking about the candidate who wants to radically restructure our political procedures for greater fairness. If you think she’d blink at doing this because it isn’t traditional, I don’t really know what to tell you.

That doesn’t mean a guaranteed Sanders win. Her delegates might balk, though the longer she maintains active campaigns in the states, the more say she has in choosing delegates loyal to her. She might also decide Sanders would be a disastrous candidate in the general election, though given the fairly mild criticism she’s lobbed at him, I’d think something drastic would have to occur.

In that case, we’re in basically the same situation as scenario 3. Here, Warren doesn’t have enough votes to make the difference between the candidates. Buttigieg, Klobuchar, Blomberg, and Gabbard delegates will decide. Or perhaps superdelegates will in the second vote. It’s very much not my favorite scenario for several reasons, but it’s not the disaster many people paint it as. Did you know The Squad are superdelegates?

You still want her to have a significant number of delegates in this scenario, because those delegates are still useful to Biden and his supporters. Their support lends legitimacy to him as a candidate. They can be used to negotiate VP, cabinet positions, changes to the platform, and more. They are still power in progressive hands. Would you rather have them in Warren’s hands than Sanders’? That’s up to you, but as someone who supports Warren as a damned good politician, I know I do.

That’s it for the scenarios based on standing candidates, but there is a scenario 4. What happens if either Biden or Sanders is incapacitated before the convention? I don’t rate this as close to the most likely scenario, but the possibility is not insubstantial. If you want Sanders (and don’t think snake emojis are solid politics), you want Warren there to take up the progressive mantle if something happens to him.

You also want her there as a legitimate candidate who isn’t Sanders if something happens to Biden. “Never Sanders as nominee” has a significant constituency within the party. Some of the reasons for that are worse than others, but it exists. “Never Warren as nominee” does not, and her presence and delegates make it all but impossible to build a case for bringing in another candidate because enough people don’t like Sanders or think he can win.

And really, I don’t even want anyone tempted to that scenario. It would be disastrous. I’d much rather she be right there if someone has to start looking around.


If I had to boil all this down to a single point, it’s this: We know a lot less about all this than we think we do. Someone making a decision that’s different from yours doesn’t necessarily want something different. We’re all starting from uncertain assumptions and making more about every stage of the primary process and the general election.

Be easy on yourself. Your vote in something this large and complicated is a butterfly’s wing. It’s important, maybe even critical, but you’re going to have a hell of a time calculating the outcome when you do it.

Be easy on others too. People who are pushing in the right direction are doing the same work you are, and they’re about as likely to be right about what they should do. If you get anxious over individual votes at this level, find somewhere to push where the results are more direct and predictable. More-local elections are a great choice, and some of them give you the opportunity to elect those superdelegates.

Whatever you do, ease up on that certainty. You haven’t earned it, not because you’re unworthy, but because none of us can get there.

Second Choices and Delegate Allocations: Why Primary “Electability” Is a Wild Guess

4 thoughts on “Second Choices and Delegate Allocations: Why Primary “Electability” Is a Wild Guess

  1. 1

    But the second choice of Butyigieg supporters, according to the same Morning Consult poll, is Biden. Below that, it’s a tie between Biden and Warren. Once it bounces back to Warren, it’s back to a 40% second-choice preference of Sanders. Will Buttigieg’s endorsement sway voters? Sure, but it won’t necessarily overrule most the preferences of people who have Buttigieg as their first choice, let alone those who have him as their second.

  2. 2

    By looking at how Biden vs. Sanders might perform using Hillary Clinton’s 2016 primary performance as a starting point, we can see how the math currently does not favor Sanders after Super Tuesday.

    The Electoral Vote web site folks looked at five possibilities:

    ** Biden doing 10% better than 2016 Clinton vs Sanders
    ** Biden doing 5% better
    ** Biden doing just the same
    ** Biden doing 5% worse
    ** Biden doing 10% worse

    The only time with these five possibilities where Sanders wins at least 1991 delegates and the nomination is when Biden does 10% worse than Clinton.

    Here is their closing summary on this:

    We know, it’s a big table with a lot of numbers. But this does not look good for the Vermont Senator. It appears that Sanders is set to lose the nomination if Biden performs about the same as Hillary Clinton did in 2016. If Biden overperforms her, he will leave Sanders in the dust. Even if Biden underperforms her by 5%, he still claims the nomination. All other things being equal, Sanders needs to improve his 2016 numbers by at least 8-10% to outpace Biden.

    Source —

  3. 3

    Biden can’t remember what day it is, whether he was arrested or not, if he wants to cut social programs, or even what his wife looks like. If he’s chosen he will lose to the orange tiny handed monster full stop.

    I’d wish you folks south of the border good luck, but this system you’ve got going on has only been able to come to pass because vast numbers of Americans have worked(?) for it. So enjoy the crops you’ve sowed.

Comments are closed.