Financial markets are supposed to be great aggregators of dispersed information. No one individual may know everything about a company’s future performance, but because the stock market richly rewards people for being right and for knowing something other market participants do not, stock prices generally do a pretty good job of reflecting the information that’s out there about a company’s well-being.
So what about election betting markets? These tend to show Donald Trump with much higher odds of reelection than you would infer from polling or that are reflected in poll-based models like Nate Silver’s FiveThirtyEight model. As of midday on Friday — the day after a debate that was widely seen as Trump’s last best hope in a race where polls show him well behind — the PredictIt market gives Donald Trump 40 percent chance of winning, far higher than the 12 percent odds in Silver’s model. It seems like the prediction markets are informed more by hobbyist enthusiasm than political reality. Do the markets see something the polls and the modelers don’t?
Don’t bet on it.
The problem with election-prediction markets has to do with what they don’t have in common with the stock market and other traditional financial markets. In order for a financial market to work efficiently, you need depth and liquidity. That is, investors need to be able to get their money in and out, buy in large quantities with low transaction costs, and turn a small piece of knowledge into a large profit. Those features make it worthwhile for savvy market participants to respond when prices are out of whack.
Say, for example, that you know something about oil prices that other market participants don’t. Even if the price is just a little bit off from where you think it should be, you can make big bets that will produce a substantial payout if things move your way. And if the price of oil happens to move against you on a given day, you can make the bet again tomorrow, and the next day, and the day after that — and if you truly know something special, you’ll make money in the long run.
But if you know better about an election than other participants in an election-prediction market, what good does that do you? You can sell a contract for 60 cents that will earn you $1 if Trump loses the election, which sounds like a fat profit. But because there are regulatory limits on how much money U.S. participants can put into betting markets, you can only bet so much. And if you bet through PredictIt, the most prominent U.S.-based site for placing political bets, you’ll have to pay a fee of 10 percent of your profits, plus 5 percent of all the money you withdraw from the site after you win.
And elections are infrequent. It’s plausible that Trump is so mispriced that you could make a profit betting against him despite all those fees. But then what do you do after you win? There isn’t another presidential election for four years, and the election markets may not be so out of whack next time. A hobbyist might find it fun to exploit an error in the election markets, but you can’t build a professional investing operation around it in the way you could with stocks or commodities or bonds. The low frequency of betting opportunities creates inefficiency even in European markets — run by the same oddsmakers that take sports bets — that don’t face the same restrictions on getting funds in and out.
And there are hobbyists who have fun even if they don’t have a good strategy to make money by betting. They like Trump, they think he’s going to win, and they bet on him. Because of the caps on trading in these markets, they don’t stand to lose that much money for being wrong, so their behavior is no crazier than that of someone who goes to Las Vegas and plays craps even though he knows that, on average, he will lose. In the stock market, such hobbyists usually get steamrolled by more dispassionate investors (though Tesla kind of looks like a hobbyist-driven stock), but in the election markets, without professionals against them, they are likely to just drive the price up until Election Day.
None of that is to say that the polls are necessarily correct, or that Trump cannot win. I just would not assume there is any particular wisdom in the prediction markets. After all, look at 2016 — on the day before that election, the PredictWise aggregator of election markets gave Donald Trump just an 11 percent chance of winning. Silver’s model, much misremembered by the public, gave him three times those odds.