The biggest question of the political season is whether Donald Trump will get enough delegates to win the GOP presidential nomination before the convention. Prediction markets, which allow people to bet on future events using real money, estimate an average 61 precent chance of a contested Republican convention with two or more votes required. The chance Trump will fail to get to the required 1,237 delegates before the convention, they estimate, is 69 percent.
This is exactly the type of situation where election “prediction markets” can be most valuable — rare events where we can’t rely on recent history and when polls aren’t terribly useful. Most people don’t understand the nuances of delegates and conventions, which means polling on what will happen is tricky and potentially unreliable.
On the other hand, people betting in prediction markets usually know quite a bit about the topics they’re betting on. Those markets are “not just absorbing polling,” said David Rothschild, an economist and researcher at Microsoft Research who combines data from multiple betting markets to produce predictions on his site PredictWise. “They’re absorbing a lot more dispersed information, a lot more idiosyncratic information. Primaries are a great example of where prediction markets thrive.”
The fundamental question behind these betting markets isn’t what bettors would like to see happen — it’s what they think will actually happen. This is a stark contrast to polls, which typically ask what the respondent will do or would like to see happen. The way to make money in a such a market is to accurately predict what ultimately happens.
Another critical difference is that, where polls attempt to accurately represent voters, prediction markets are not determined by anything approaching a representative sample. The betting markets don’t restrict participation to Americans, and all are headquartered overseas — making it likely many non-Americans participate.
That said, the markets do sometimes resemble the polls, and bettors undoubtedly consider polls in making their wagers. “Once we hit the general election, you’ll see the prediction markets really start to converge with polling because there won’t be as much additional information out there for markets to aggregate,” Rothschild said.
Betting on politics works much the same way as betting on anything else — for example, on two of the major sites, PredictIt and Hypermind, you can bet for or against any candidate. Prices are set based on the demand for the candidate, so if most people are betting that Hillary Clinton will win the Democratic nomination, the price to bet on her winning goes up. Conversely, it becomes very cheap to bet that she won’t win — but there’s the opportunity to win more money. Another site, Betfair, uses an odds-based betting system.
The odds assigned on Betfair and the prices on PredictIt and Hypermind can be interpreted as the probability of Hillary Clinton winning the Democratic nomination — and then these probabilities can be used to forecast the elections.
A brief history of using prediction markets to forecast elections indicates that they do a pretty good job and often indicate winners long before polls do. However, there are a lot of unknowns that go into the bettors’ decisions — what information beyond polls they’re considering, where they’re from, their experience and expertise with American politics — which make these prediction markets a potentially shaky source.
But where polls take several days to conduct and register changes in the political landscape, the markets are highly susceptible to changes in conditions. On the night of the Iowa caucuses, when it became clear Trump would lose, his probability of winning the nomination dropped dramatically within minutes. It rebounded in the next few weeks as Trump won contests, and he’s favored at 73 percent to win the nomination as of this writing — even though his odds of avoiding a contested convention are low.
Rothschild thinks the ability to account for events and changing conditions is a key advantage of prediction markets. “When you look at some candidates and you see a much higher percent chance that they are going to have a major gaffe, or a much higher percent chance that they will have a health risk, or a higher percent chance that something’s going to happen between now and election day that’s going to positively or negatively affect their outcome — this is the type of thing markets take into account.”