Here's who's winning the Republican nomination contest

There's something puzzling happening in the Republican Party. The candidates who have the support of the most voters are not those that are showing signs of winning the nomination. In other words, the standard smoke signals reported in the news about the political horse race are not consistent with the smoke signals that the inner circles of political wonks pay attention to. How can Donald Trump, for example, be ahead in most polls but dead last in fundraising and endorsements?

How can Donald Trump be ahead in the polls but dead last in fundraising and endorsements?

Below, I take a look at five factors often considered to be predictive: polling, elite endorsements, cash on hand, market trading price, and the amount of speaking time candidates had in the most recent Republican debate. There is a startling lack of correlation between these factors, and I offer some explanations for the discrepancies.

Polling

Of course, we do not yet know which candidate will be the 2016 Republican nominee, but the candidate who is ahead in the polls now is not necessarily indicative of this outcome. Looking at polls from the fall of the year before a presidential election, we do not commonly see the parties' eventual nominees leading.

A coin may be able to predict a nominee as well as polls in the fall before the general election

In other words, a coin may be able to predict a nominee as well as polls in the fall before a general election.

I looked at all of the historical polling data on presidential primary races one year before the election. Pollsters haven't always asked questions about voter support so far out from an election, so this data only goes back to 1988. In six of 11 primaries without an incumbent, the candidate leading the fall before the election was the party's nominee.

In 6 of 11 contests, the leader one year out was also the nominee. Data collected by author from Roper Center iPoll.

Early polls are not particularly predictive of a nomination process. As is commonly known, Trump has been leading in national polls for some time:

Trump has been leading the polls for several months. Data collected by author from RealClearPolitics.

Endorsements

As has been argued in this space previously, by my colleagues and others, political scientists have a well-supported theory about how parties settle on their nominees. The endorsements that political elites provide to candidates for office are the most predictive observable indication we have of this process. As more and more political elites coalesce around a candidate and show their support with a formal endorsement, the probability increases that that candidate will be the nominee.

Nate Silver of FiveThirtyEight has been tracking the number of endorsements that 2016 candidates are gathering. The website only tracks endorsements by members of Congress and governors, and it's not clear that these are the most predictive endorsements to track; however, this data provides a decent glance at this relatively behind-the-scenes process.

Bush has been presumed the frontrunner, but hasn't yet swept away the field.

For some time now, Jeb Bush has been a presumptive nominee and has garnered the lion's share of early endorsements (he also has the most Super PAC money). Endorsements have been slow in coming and haven't so far showed signs of magnetism for a particular horse.

Cash on hand

Plenty of research suggests that candidates with the most fundraising prowess tend to win elections, but it's not clear that fundraising totals at this stage of an election are terribly predictive.

Carson, Cruz, and Bush are leading this category. Data from OpenSecrets.org.

The fundraising numbers for the 2016 Republicans are somewhat misleading because Donald Trump's campaign has been largely self-funded. The numbers reported above, from OpenSecrets.org, reflect only the money that candidates' campaigns have raised, not their Super PACs (but see that data here). By the cash-on-hand primary, Carson is doing well, with Cruz and Bush close behind.

Markets

Another way to engage in prognostication about election outcomes is to buy and sell shares of events and then track how prices change in the market. Political markets have come and gone, largely due to questionable legality related to gambling laws, but the website PredictIt has a robust market this year and appears to have considerable volume. As of earlier today, here are the market prices of these candidates:

Rubio has recently become the sweetheart of the political markets. Data collected by author from PredictIt.

The final number I'll throw in here is how much time the candidates eked out of Wednesday night's debate.

As tallied by National Public Radio and reported on Twitter.

Rubio is winning the debate speaking time and market primaries right now, but is significantly lagging in the fundraising, endorsement, and polling primaries. Carson is doing well with polling and fundraising, but terribly in endorsements and fundraising.

So what can we make of these discrepancies?

First, each of these measures captures something a little different.  The polling numbers are likely Republican primary voters, but in truth not many of these are paying a heck of a lot of attention to this contest yet, because no one is asked to cast a vote for at least three more months.

We might expect endorsements and market prices to be somewhat correlated (they're not; r = -0.063) because they might be coming from the same population. If The Party Decides theory of nominations is predictive, then the partisan elites are the knowledgeable, insider players with their fingers on the pulse of the election and electorate. They are helping to solve a gigantic coordination problem in which everyone seeks to back the winner (a.k.a. nominee). The FiveThirtyEight endorsement data attempts to capture this, and the market data presumably does, too. Traders with insightful knowledge may be more likely to be in the market and helping to drive prices, rather than political know-nothings out to take a risk for fun. If insiders are both the endorsers and the market buyers and sellers, then these values would be correlated; however, they're not. I suspect this is for two reasons.

First, many political insiders are still on the sidelines. The elite, knowledgeable players in national and state-level party politics are still waiting to see how things shake out. No one wants to accidentally endorse the wrong (i.e., losing) candidate. It's still very early in the game.

Second, it's not clear that the governors, senators, and House members Silver is tracking are the best sample of the party elite to observe for this purpose. The data used in the above referenced book includes many other party elites such as state and local elected officials, union leaders, large interest groups aligned with parties, and so forth. Maybe the endorsement primary data is both incomplete and not the most predictive sample?

The fundraising data is interesting because it is likely a mix of party elites and the general public. In this sense, it's the noisiest of the indicators.

Finally, I wondered if we could explain candidates' speaking time in the debate as a function of these various indicators. I estimated a linear regression where the dependent variable is the amount of speaking time candidates received and the independent variables are national polling ratings, endorsements, cash on hand, and market price.

The underlying process behind debate speaking time is a bit of black box that has some idiosyncrasies associated with it (moderator question variance, audience reaction, etc.) and has some endogenous elements because the debate rules allow any candidate to follow up when their name has been mentioned. Perhaps debate speaking time is just an indicator of how often one's name is mentioned, then?

However, if debate speaking time is indicative of some strength, movement, or momentum a candidate has, then its relationship with these other standard predictors is meaningful. This is a big if, and I'm not trying to put much weight behind this analysis, but here it is:

Only market prices positively predict speaking times.

The model shows that the market price is the only positive predictor of debate speaking time. A candidate's position in the national polls is not predictive (for this one debate, at least), and candidates with greater endorsements and cash got significantly less speaking time. Doing well doesn't seem to pay off.

What we're really seeing here is sausage being made. The process by which a political party settles on a single candidate to nominate and support in a general presidential election is messy. For hundreds of years in American history, this process has gone on behind closed doors and, quite literally, in smoke-filled rooms. But in the modern era of super-saturated constant communication and reporting, everyday observers can now watch this complicated dance evolve in real time. Like a school of hundreds of fish chaotically swimming about trying to decide which direction to go, eventually a pattern forms; the group coordinates and appears to operate as a whole. The contradictory indicators are our window into this messy process that has not yet found its direction.

I have no skin in the game, but I suspect that markets are picking up on Rubio's strength. Even though he lags in some key indicators at the moment, he is a strong Republican candidate with wide appeal. He is a higher-quality candidate than some (Trump, Carson) because he's held elective office before, and more mainstream in his views than others (Cruz, Paul). His strong recent performances make him one to watch, in my mind. As we see these indicators begin to correlate, we'll know when the party is about to decide.

*An earlier post erroneously had Mondale as the 1988 nominee, where 1988 should have read 1984.

Explanatory journalism is a public good

At Vox, we believe that everyone deserves access to information that helps them understand and shape the world they live in. That's why we keep our work free. Support our mission and help keep Vox free for all by making a financial contribution to Vox today.

One-Time Monthly Annual

$95/year

$120/year

$250/year

$350/year

Other

$

Yes, I'll give $120/year

Yes, I'll give $120/year

We accept credit card, Apple Pay, and Google Pay. You can also contribute via