What did the prediction markets expect from Tuesday?
These contract prices are aggregated on my website, PredictWise.com, and turned into probabilities of any candidate winning any election.
On March 14, the market-based predictions for the Republicans were:
- Donald Trump 86% for Florida
- Trump 73% for Illinois
- Trump 54% for Missouri
- Trump 99% for North Carolina
- John Kasich 71% for Ohio
At the same time, the market-based predictions for the Democrats were:
- Hillary Clinton 93% for Florida
- Clinton 54% for Illinois
- Bernie Sanders 68% for Missouri
- Clinton 90% for North Carolina
- Clinton 66% for Ohio.
Depending on the final outcome in Missouri, nine or 10 of these predictions pointed to the eventual winner. But there was uncertainty in the some of the elections; they were not all 100 percent; and the elections eventually breaking in a certain way did make a difference.
How did the predictions change after the election?
Not surprisingly, the likelihood of Republican nomination moved meaningfully towards Trump. Trump started the day at 74 percent and ended the day at 80 percent. Kasich, with his big win in Ohio, started the day at 11 percent and ended the day at 8percent. Ted Cruz was down from 14 percent to 11 percent. Marco Rubio, highly likely to lose Florida, was essentially already discounted going into the day.
The bigger surprise is that the likelihood of Kasich being the nominee dropped, despite his big win in Ohio.
Why? One clue here is that the probability of a brokered convention, defined as the convention going to the second ballot, also dropped to around 36 percent, down from 42 percent the day before. Kasich and Cruz would need a second ballot to win the nomination. Thus their successes are hollow if they cannot block Trump from getting a majority of delegates before the convention.
And while Kasich’s win in Ohio blocked some delegates from Trump, it also ensured a three-way race for the rest of the primary contests. Trump benefits from Cruz and Kasich splitting any anti-Trump votes.
Trump is the only candidate with a non-negligible probability of winning the Republican nomination on the first ballot. Thus the markets currently have Trump at 64 percent likely to win the nomination on the first ballot, with a 36 percent likelihood of a second ballot.
When it comes to predicting the outcomes of a brokered convention, the markets have Trump at 16 percent, Cruz 11 percent, and Kasich 9 percent to win. The remaining 1 percent goes to a unity candidate who did not run this year: Paul Ryan. Should Trump fail to win the nomination outright, the markets think he is about 40-45 percent to win the nomination at the convention.
The Democratic nomination moved much more on the strength of Clinton’s strong victories over Sanders. She started the day valued at 90 percent likely to win the nomination and finished the evening at 96 percent to win the nomination. This puts her back to where she was before Sanders’ surprise win in Michigan. Clinton has been in control of this nomination from the beginning; the lowest point for her, since Joe Biden announced he would not run in October, was 81 percent, just after the New Hampshire primary.
What about the general election?
The markets are also assessing whether it’s likely that the eventual Democratic nominee or the eventual Republican nominee will win the general election. After Trump’s and Clinton’s strong showing on March 15, the markets predicted that the eventual Democratic nominee will win, at a probability of 72 percent.
By comparison my fundamental model, based on joint work with Patrick Hummel (using presidential approval, economic indicators, incumbency, and past election results to predict the outcome) has the generic Democratic candidate at 48 percent likely to win against the generic Republican candidate.
Thus comparing the two models, the actual likely Democratic nominee is outperforming the generic Democratic nominee by 24 percentage points.
Here’s another point of reference. At this point in the 2012 presidential race, Barack Obama running for reelection was valued at a 60 percent likelihood to win.
For even the strongest candidates, the two-party system, with just 13 or so truly “swing states,” makes any nominee instantly competitive before the multibillion-dollar general election campaign. For March 16 of an election year, 72 percent is a very high probability of winning.
Notes on methodology: There is a three step process for aggregating the prediction market data on PredictWise.
- Step 1: Construct prices from the back/sell, lay/bid, and last transaction odd/price in the order book. I always take the average of the highest price traders are willing to buy a marginal share and the lowest price people are willing to sell a marginal share, unless the differential is too large or does not exist.
- Step 2: Correct for historical bias and increased uncertainty in constructed prices near $0 or $1. I raise all of the constructed prices to a pre-set value depending on the domain.
- Step 3: I normalize the probabilities to equal 100% for any mutually exclusive set of outcomes.
David Rothschild is an economist at Microsoft Research. Find him on Twitter @DavMicRot and online at PredictWise.com.
Note: Because of an editing oversigh, an earlier version of this story included the wrong graphs. They have now been updated. We regret the error.