We’re several hours from the Iowa caucus. Let’s take stock of four methods that tell us who might win. All of these methods attempt to lay odds on different outcomes. This is a more interesting way to think about the race than dealing in generalities like “well, Sanders could win” or “Cruz has a chance.”
The tl;dr version: Trump and Clinton are favored, but a lot of uncertainty remains. Both Cruz and Sanders have a significant chance of winning. To see how significant, read on.
Polls
Pollster’s Iowa polling averages show a narrow lead for Clinton (3 points) and a larger lead for Trump (7 points). 538’s polling averages show a slightly larger lead for Clinton (5 points) and a slightly smaller lead for Trump (just under 6 points).
The question is whether the Iowa polls have accurately gauged who is going to show up at the caucus. Iowa polls that are based on calling confirmed registered voters show a larger Clinton lead. Trump’s support is predicated on people who say they will vote but do not have a prior record of participating in primaries. Research shows that people cannot accurately predict whether they will vote.
[interstitial_link url=”https://www.washingtonpost.com/news/monkey-cage/wp/2015/12/07/trump-continues-to-lead-the-polls-heres-why-he-might-not-win-the-election/”]Trump leads the polls. Here’s why he might not win.[/interstitial_link]
For this reason, a poll-based forecast has to come with pretty significant uncertainty. 538’s forecasts convey that nicely. A forecast based only on state polls gives a Clinton a 73 percent chance of winning — substantial but hardly certain.
A similar forecast gives Trump a 58 percent chance of winning. His closest competitor, Cruz, has a 32 percent chance.
Polls plus other stuff
The 538 crew notes that state polls are not the only thing that has helped to predict past Iowa caucus winners. Endorsements by party leaders and national polls are correlated with caucus outcomes too.
When those factors are taken into account, Clinton’s chances increase a little bit, to 78 percent. But the GOP forecast changes quite a bit: Trump now has a 48 percent chance and Cruz a 40 percent chance. This reflects how poorly Trump has fared in earning endorsements.
Prediction markets
Prediction markets create an incentive for people to aggregate information and make accurate predictions. After all, their money is on the line.
In the GOP race, the prediction markets lean more towards Trump than a forecast based only on state polls. Predictwise compiles several different markets and — as of about 10 am EST this morning — gives Trump a 66 percent chance, while Cruz has a 29 percent chance. This suggests some optimism among bettors that Trump can mobilize infrequent voters.
In the Democratic race, the markets basically match the 538 forecasts: Clinton has a 76 percent chance and Sanders as 24 percent chance.
Our forecasting tournament
We’ve partnered with Good Judgment to do crowd-sourced forecasts for 2016. Good Judgment has had success aggregating the predictions of individual forecasts (sometimes beating prediction markets). As of this writing, about 1,300 people have made predictions for the Iowa caucus.
If we simply report the median of every person’s most recent forecast for the Iowa caucus, the Democratic race appears closer than the polls or prediction markets would suggest. The forecast gives Hillary a 51 percent chance of winning, with Sanders at 47 percent and O’Malley at 2 percent.
However, other methods of combining these 1300 or so forecasts help Clinton. Many forecasters have made forecasts about other political events, and if we assign more weight to the most accurate forecasters, Clinton has a higher chance of winning: approximately 63 percent.
On the Republican side, the forecast gives Trump a 49 percent chance of winning, Cruz a 34 percent chance, and Rubio a 12 percent chance. If we assign more weight to the most accurate forecasters, Trump’s chances increase to about 66 percent.
Putting it all together
These various forecasts all point toward Clinton and Trump victories. If I simply average them, Clinton has a 68 percent chance and Trump has a 57 percent chance.
However, 68 percent and 57 percent are far from 100 percent. If I said there was a 32 percent chance or a 43 percent chance you’d be hit by a car the next time you crossed the street, you’d definitely think twice. A lot of uncertainty remains.
And it’s important to note that winning the Iowa caucus is neither necessary nor sufficient for winning the nomination. Iowa helps winnow the candidates, and boosts news coverage and thus poll numbers for candidates who exceed the expectations of political observers.
So a victory in Iowa is nice, but it is hardly enough.