Election Update: What Are The Chances Of A Brokered Convention?

A brokered convention is not a likely occurrence, exactly. But it’s not unlikely, either. It’s roughly as likely as the Tennessee Titans beating the Baltimore Ravens earlier this month, or Donald Trump winning the Electoral College despite losing the popular vote. According to the FiveThirtyEight primary model, there’s a 15 percent chance that no Democrat wins a majority of pledged delegates:

So does that mean we’d be headed to a BrOkErEd CoNvEnTiOn — the first brokered convention of the modern primary era, which began in 1972?

Well, not necessarily. It’s time for a little bit of a detour into a brokered convention explainer. (As a bonus, there is a Venn diagram coming.)

What is a brokered convention, exactly? And is it the same as a contested convention?

First, consider the rather precise phrasing that our model uses. It forecasts the chance that a candidate wins “more than half of pledged delegates.” To be more exact, it forecasts the chance a candidate wins a majority of pledged delegates as of the end of the day on June 6, when the last state or territory (the Virgin Islands) votes.

Winning a majority of pledged delegates is not quite the same thing as winning the nomination. Nor does the failure of any candidate to win the majority of delegates by June 6 necessarily imply a contested or brokered convention.

This is because of a dirty little secret: Democratic delegates are pledged to certain candidates based on the results of primaries and caucuses. But they are not actually legally bound to vote for them, even on the first ballot.

In theory, this means Democratic delegates could subvert the will of the voters and deny a candidate who had won the majority of pledged delegates the nomination. But this is highly unlikely. (Barring, perhaps, a candidate dropping out between June 6 and the convention because of a scandal or health problem.) Among other things, candidates pick their own pledged delegates which increases the chance that delegates will be loyal.

It is reasonably likely, however, that a candidate who didn’t have a majority of pledged delegates on June 6 could nonetheless wind up with one before the convention in Milwaukee. Consider a case like this … after the Virgin Islands votes on June 6, the pledged delegate count is as follows:

• Elizabeth Warren has 45 percent of pledged delegates;
• Joe Biden has 25 percent;
• Bernie Sanders has 20 percent;
• the remaining 10 percent are scattered between other candidates.

Our model would treat this as a “no majority” outcome. Basically, the model doesn’t consider anything that happens beyond June 6, including the possibility that delegates vote for a candidate other than the one they are pledged to. So what might happen in a case like this is that — in recognition of Warren’s fairly emphatic plurality — Sanders would ask his delegates to vote for Warren on the first ballot. Or, heck, maybe Biden would too. The delegates, of course, wouldn’t be under any obligation to follow Sanders’s or Biden’s instructions. But if enough of them did, Warren could win a majority of delegates on the first ballot. Moreover, all of this might be decided well in advance of Milwaukee, and the convention itself would be anticlimactic.

I should pause here to note that the definitions of “contested convention” and “brokered convention” — and whether those terms mean slightly different things or are entirely synonymous! — are open to debate.

“Brokered convention” is sometimes regarded as an outmoded term because it implies the use of brokers — which isn’t a terribly good description of the convention process as it exists today. Rather, a contested convention could be a relatively open process with individual delegates — not party bosses wheeling and dealing behind closed doors — having most of the agency to determine the outcome.¹ With that said, unless one is trying to be very precise, I don’t have any problem with using “brokered convention” as a near-synonym for “contested convention.”

Is our primary system democratic?

As to the best definition of a contested convention, I have stronger feelings about that. I’d say a contested convention is a case where the nomination outcome is genuinely in doubt at the time the convention begins.

I realize that this introduces a little bit of subjectivity (what does “genuinely in doubt” mean?).² The problem is that the most common alternative definition — that a contested convention is any convention that requires multiple ballots — doesn’t capture the spirit of a contested convention very well.

That’s because there could be situations where a convention nominally required multiple ballots to resolve, but the outcome wasn’t contentious. Say in the scenario above — where Warren had won a clear plurality of pledged delegates — everyone agreed that the best way to ensure party unity behind Warren would be to have Biden and Sanders delegates respect the original primary and caucus vote on the first ballot and vote for Biden and Sanders. Then everybody would get behind Warren as a show of force on the second ballot. Moreover, say all of this was scripted and widely disclosed to the press weeks ahead of the convention. It’s hard to think of this being a contested convention in any meaningful respect.

Conversely, it’s possible to imagine the outcome being genuinely in doubt — but the relevant negotiations to resolve the deadlock take place before the first ballot is cast. Say, for instance, that after the Virgin Islands, the delegate count was Sanders 32 percent, Biden 28 percent, Amy Klobuchar 25 percent, and Warren 15 percent. There’s a lot of doubt about the identity of the nominee when everyone gathers in Milwaukee. But Biden offers Klobuchar the vice presidency in exchange for Klobuchar instructing her delegates to vote for Biden on the first ballot. Klobuchar agrees and almost all of her delegates go along, so Biden is nominated with 53 percent of the vote on the first ballot. To me, that ought to count as a contested convention, even though it technically required only one ballot.

Got all that? Here’s a Venn Diagram to help:

Again, the key thing to remember here is that the model’s no-pledged-delegate-majority scenario does not necessarily imply a contested or brokered convention. It’s probably fairly close to an upper bound on the likelihood of a contested convention, but a 15 percent chance of no majority of pledged delegates on June 6 might imply, I don’t know, only a 10 or 12 percent chance of an actual contested convention, as I’d define it.

The best time for Democrats to avoid a contested convention is before Super Tuesday

Whether it’s 10 percent, 12 percent or 15 percent, I suspect the probability our model spits out for a contested convention will strike some of you as high and others of you as low.

On the one hand, a contested convention has historically been a sucker’s bet. Pundits and reporters love to speculate about the possibility. But out of 18 competitive nomination processes since 1972, none has resulted in what’s uniformly regarded as a contested convention, although some arguably were. (I think the 1976 Republican race probably meets the definition of a nomination whose outcome was uncertain when the convention began. Even that was 44 years ago, however.)

On the other hand, a number of other nominations — including the 2008 Democratic race — have come fairly close to resulting in contested conventions. The 2016 Republican convention could also plausibly have been contested if Republicans had used Democrats’ rules. (Trump got a big boost from winner-take-all and winner-take-most states, which Democratic rules do not allow for).

[Our Latest Forecast: Who Will Win The 2020 Democratic Primary?]

Moreover, the 2020 Democratic primary is among the more open-ended nomination processes we’ve seen, with at least four — and possibly five (if you count Klobuchar) — plausible winners of Iowa. But it isn’t necessarily the most open-ended nomination; the 2008, 2012 and 2016 Republican nominations were all fairly similar, as were the 1988, 1992 and 2004 Democratic nominations, and none of those produced contested conventions. The once-very-large number of candidates doesn’t seem terribly relevant any more either; many candidates have already dropped out, and several of the ones that remain have little to no chance of accumulating delegates.

Still, it’s not hard to imagine cases where the primary is at a stalemate between three or more candidates. And that’s key: almost all no-majority scenarios require three or more Democrats to continue accumulating delegates on a sustained basis.

There are three main factors that work to prevent such a scenario from happening, however:

• First, Democratic rules only award delegates to candidates who get at least 15 percent of the vote in particular states or districts. It’s not a perfectly bright line — if you’re at, say, 14 percent, you’ll still likely get delegates in some districts but not others. But it’s a fairly bright one. Maybe Andrew Yang or Tulsi Gabbard can command 5 to 6 percent of the vote with their own relatively distinct constituencies — but it won’t net them any delegates.
• Second, the early states serve to winnow the field. Only Biden and to a lesser extent Sanders are comfortably above 15 percent in national polls, and even they’re not so far above as to necessarily be in a position to stay there following poor showings in the early states. Meanwhile, Warren and Buttigieg (who isn’t above 15 percent nationally) would almost certainly be in trouble in the absence of early state wins.
• And third, our model assumes that candidates are reasonably likely to drop out if they don’t have a shot at the nomination, even if they could complicate life for other candidates. This is the most debatable assumption, so we’ll revisit it in a bit.

More precisely, the key to Democrats ensuring there’s not a contested convention is for a lot of this winnowing to take place before Super Tuesday — that is, either candidates fall below 15 percent because their supporters shift to other candidates or they drop out.

Remember: What happens in the early states doesn’t matter very much to the overall delegate math. Iowa, New Hampshire, Nevada and South Carolina collectively account for only about 4 percent of Democratic delegates. But Super Tuesday accounts for roughly one-third of all pledged delegates. So if three or more candidates are still winning delegates as of Super Tuesday, that doesn’t necessarily make a contested convention a likelihood, but it does make it a more distinct possibility.

Our model hopes to sort its way through all of these scenarios, so let’s look at the output from 10,000 simulations in our model run as of late Thursday morning. For each simulation, I tallied up which candidates the model forecasted as having at least 15 percent of the vote (after allocating undecided voters proportionally) as of Mar. 2 — the day before Super Tuesday. This is a good proxy for how many candidates might be expected to pick up a material number of delegates on Super Tuesday.

Here’s a table listing those scenarios, grouping the simulations by how many candidates had at least 15 percent of the vote in each one; as a shorthand, I’ll refer to these as viable candidates. (I acknowledge this table might be slightly on the confusing side, but let me show you the numbers and we’ll go from there.)

First of all, it’s unlikely that the first four states will completely winnow the field. As you can see in the first row in the table, only 4 percent of simulations resulted in their being only one viable candidate as of Mar. 2. That means even if someone — say, Biden — totally sweeps the early states, someone else will probably hang around. But if it were the case that we had just one viable candidate as of Mar. 2, that results in a delegate majority more than 99 percent of the time.

On the other hand, the early states narrowing the field to two candidates is a fairly likely occurrence. In 52 percent of simulations, that’s what we got: exactly two viable candidates heading into Super Tuesday. By far the most likely combinations are Sanders + Biden (a 24 percent chance) or Warren + Biden (13 percent).

Some of what you’re seeing here is related to the model accounting for “lanes.” In other words, it’s likely that either Sanders or Warren survive the early states — and it’s possible that they both do — but fairly unlikely that it’s only the two of them and not also Biden or one more of the moderates (Buttiigeg, Klobuchar or Michael Bloomberg).

But in these two-candidates scenarios, it should be fairly easy for Democrats to avoid a contested convention. In these simulations, in fact, someone got a majority of pledged delegates 94 percent of the time. It’s not completely out of the question that two candidates could divide up the map almost evenly, in which case a third or fourth candidate picking up a small handful of delegates could act as a spoiler. — but it’s unlikely.

However, three-viable-candidate scenarios still occurred 41 percent of the time. And the most likely one — Sanders + Warren + Biden — occurred in 19 percent of all simulations. Combinations involving Buttigieg and two of the other three leading candidates are also plausible if Buttigieg wins Iowa or New Hampshire. There are also see a few scenarios in the table involving Bloomberg if he gains ground before Super Tuesday.

And in these cases, a contested convention becomes quite thinkable — but it’s still less likely than you might assume. About 74 percent of Sanders + Warren + Biden scenarios result in someone winning a majority of pledged delegates, for example. That could be true for a variety of reasons. It could be because one candidate is way ahead of the other two. Or it could be because someone has a poor Super Tuesday and drops out after that. Or it could be because Democratic voters gravitate toward one candidate late in the process to avoid a contested convention — as Republican voters sort of did late in 2016 when they coalesced around Trump.

There’s also a slight fudge factor in the Democratic delegate math that could be at play here. If a candidate drops out between when a state’s primary occurred and when statewide delegates are actually assigned — which can occur weeks, or even months, after the primary pending on the state — those delegates are reassigned to other candidates still in the running. This doesn’t make a huge difference³ but it does help a bit at the margins for a candidate who is very close to the majority threshold. The probabilities listed as calculated by our model (and as described in this article) account for this reallocation process.

Finally, there are those unlikely scenarios when there are still four or more viable candidates before Super Tuesday; these occur only about 4 percent of the time. Why so unlikely?

First of all, it’s because it would entail the field getting more crowded than it is right now; currently, only 3 candidates (Biden, Sanders, Warren) have at least 15 percent of the vote nationally or are anywhere close to it. So this scenario where there are four viable candidates is pretty unlikely, because if someone else joins this group, it probably means that someone else struggled. Say Buttigieg were to win Iowa, for instance. That necessitates Warren not winning it — in which case she might need a lot of other things to go right to stay above 15 percent nationally.

Even still, these four-candidate scenarios produce someone winning a majority of pledged delegates just shy of the time. Keep in mind that part of this is because four candidates still in the running doesn’t necessarily mean they’re all on equal footing. If national polls before Super Tuesday are something like Sanders 42 percent, Biden 20 percent, Warren 15 percent, Buttigeg 15 percent, then — well, a lot of things could happen — but Sanders is going to wind up with a majority a lot of the time.

Why our model might be underestimating contested convention odds

Maybe you’ll find all of this impressive … or maybe you don’t. Maybe you’re worried about the chance of garbage in, garbage out.

The model is trying to take a lot of history and encode it into a statistical form, which is what most statistical models do. But the risk, of course, is if candidates don’t behave the way that they have in the past, our model could misestimate the likelihood of a variety of different scenarios. In particular, the model is somewhat sensitive to the logic of when candidates drop out. And small changes in our assumptions about this could make a reasonably large difference.

For instance, although as of Thursday morning, the model had a candidate failing to win a majority of pledged delegates in “only” 15 percent of all simulations, there were a lot of close calls. In another 11 percent of simulations, the winning candidate got more than 50 percent of pledged delegates but under 55 percent. And in a further 11 percent of simulations, the winner had more than 55 percent of pledged delegates but still under 60 percent.

One assumption that could be sensitive is the chance that candidates drop out while they’re still a threat to accumulate delegates. So I looked at how often our model predicted candidates to drop out when: (i) they were polling at least 15 percent nationally at the time they quit;⁴ and (ii) whether it was before Apr. 28, when New York, Pennsylvania, Maryland, Connecticut, Delaware and Rhode Island vote — the last truly major opportunity to accumulate delegates on the calendar:

• Biden quit under these circumstances in 12 percent of simulations;
• Sanders in 11 percent of simulations;
• Warren in 7 percent of simulations;
• Buttigieg in 2 percent of simulations.
• And for other candidates, this rarely happened.⁵

In other words, the model has a candidate quitting when they still have a chance to accumulate more delegates — even though they might have little or no chance to actually win the nomination — about 30 percent of the time.

Should we be worried about that? Well, the model’s estimates of candidates’ likelihood of dropping out are derived from historic data — and historically, candidates don’t usually drop out when they have a real shot at winning. But they don’t necessarily hang on until the bitter end either. So far in 2020, that pattern has held, too. Candidates mostly have quit (think Cory Booker or Kamala Harris or Julián Castro) when their situations became dire but not necessarily totally, embarrassingly hopeless.

On the other hand, there are at least some reasons to think some candidates might hang around longer this time. Biden and the other moderates could stay in the race in an effort to stop Sanders, with the party establishment starting to fret (to put it mildly) about the possibility of Sanders winning. Meanwhile, Sanders hung around until the last states voted in 2016 and could do so again, perhaps out of the hopes of forging some alliance with Warren.

Then there’s Bloomberg, whose strategy is hard to figure out, but which mostly seems to hinge on accumulating a sizeable number of delegates on Super Tuesday and beyond. So far, he remains some ways away from the 15 percent he’d need in national polls and in Super Tuesday states. But at the rate he’s spending, it isn’t that crazy to imagine him continuing to grow in the polls and accumulating delegates in some states, even if he’s little threat to actually win the nomination. This sort of delegate accumulation or spoiler strategy doesn’t have a ton of recent precedent⁶ and isn’t necessarily something the model is equipped to handle terribly well. But in any event where Bloomberg increases in strength, a contested convention would be that much more likely.

Of course, all of this is getting a little bit ahead of ourselves. Okay, let’s be honest — way ahead of ourselves. But the delegate math, including the possibility of a contested convention, is something that affects every candidate’s strategy from the very start of the race. Our advice is not to assume there will be a contested convention just because things look a little chaotic at the moment — many primaries have looked chaotic at this stage of the race. But at the same time, you shouldn’t dismiss the possibility out of hand just because there hasn’t been one recently.