Prediction Markets: Be Careful What You Wish For

NYT's David Leonhardt has a nice column on the shortcomings of prediction markets, 364 days after singing their praises.

That's not a knock. It's fair to say those of us who were taken by the Wisdom of the Crowds-appeal of prediction markets have been disappointed by how messy they are in reality.

While there is plenty of evidence that the markets are better predictors of future outcomes than polls and experts, they seemingly failed miserably when traders gave Barack Obama a 93 percent chance of winning the New Hampshire primary. That opened the door for critics to let loose.

In response, all the pro-markets crowd could do was point out that a 1-in-14 chance for Hillary meant that it was highly improbable, but not impossible, that she'd win. After all, it's not like the markets gave Obama a 100 percent chance of victory.

If you believe in the many-world theory, then we lived in one of those 14 worlds where Hillary was the victor. But even if this is mathematically accurate, it's not exactly inspiring.

Getting back to Leonhardt, he believes that the solution for getting better results from the markets is to have more liquidity.

...it's certainly fair to say that Intrade isn't as advanced as some of us had thought...As more traders try to exploit Intrade's inefficiencies, those inefficiencies will become rarer and rarer.

But I'm not convinced that's the case, largely because of an interesting study done by Paul Tetlock of the University of Texas which I saw presented at the AEA's in January. (Incidentally, Steve Levitt, who was also presenting a paper at the session, argued that research into prediction markets was a dead-end for academics: "Once you've proven they work, what's next?")

Investigating the market efficiency at Intrade's sister site TradeSports, Tetlock found that highly liquid markets showed "significant pricing anomalies," and that surprisingly, illiquid markets were "remarkably efficient." Tetlock thinks that liquidity may actually be a proxy for noise trading:

...illiquid markets have fewer noise traders, and periods of illiquidity prevent arbitrageurs from profiting on short-term trades that would destabilize prices.

So, having more traders on Intrade might not actually solve the problem, because Tetlock's results suggest that exploiting inefficiencies is actually quite difficult. To be fair, Leonhardt doesn't call for more trading, but for more arbitrageurs. The only problem with that is that growth in arb trading is likely to be correlated with growth in noise trading, so more traders could actually make pricing anomalies worse.

A key factor in whether arbitrageurs will bet on fundamental information is whether they expect to be able to liquidate their positions before a security expires. In markets that are persistently liquid, arbitrageurs may actually destabilize prices because they can liquidate their positions before expiration at prices that may differ from fundamentals.

So, as prediction markets become more popular, it's important to also keep our expectations of their abilities in check: they're not going to get it right every single time, but they're the best available tool out there for divining the future.