America’s Best Pundit: Economix vs. Freakonomics

Economix, a blog run by The New York Times newsroom, is running a fun prediction contest ahead of tomorrow’s election. You have to pick three prediction-market stocks, and the winner is the person with the most profitable portfolio. You can enter here and read more here.

Economix is running this contest partly because David Leonhardt is interested in figuring out whether his readers are smarter than the InTrade prediction market.

Let me offer Leonhardt a friendly Economix vs. Freakonomics bet. I’m willing to bet him the market is smarter than the average Economix contestant. Let’s make the terms a good bottle of wine (plus boasting rights). Do we have a bet?

As usual, my reasoning is all about incentives. Let me explain. If I were trying to choose the three stocks on InTrade that would maximize my expected return, I would pick:

1. Winner: Obama (earning me a 20 percent profit)

4. McCain to earn fewer than 300 electoral votes (earning a 10 percent profit)

11. Obama to win Pennsylvania (earning a 20 percent profit)

But there’s only one prize in this contest, and so even if my picks are exactly correct, my portfolio will yield a “mere” 17 percent return. This is pretty good for a 24-hour period, but it’s unlikely to be enough to win first prize. And so I have a pretty strong incentive to bet on some longshots and hope that I get lucky. In fact, the stocks I picked were:

5. Bob Barr to earn 1 percent or more of the popular vote (for a 300 percent profit)

7. Obama to win Georgia (also a 300 percent profit)

18. Sen. Sununu (R-N.H.) to win re-election (for a 490 percent profit)

Here’s why InTrade will beat the average contestant: the “winner takes all” nature of the Economix contest pushed me to pick longshots, but as I wrote last week, longshots are typically overbet, yielding a negative return. And I think most entrants will follow a similar pattern, leading most of us to pick dud stocks, even if a small minority may get lucky on the longshots.

It’s an interesting irony that I’m betting on InTrade to beat the Economix readers, but my reasoning is not based on the view that prediction markets are fully efficient (they aren’t). Rather, my logic is that the nonlinear incentive in a tournament will act so as to amplify the market’s inefficiencies.

And perhaps this is a parable for our current financial mess, too.


KW

Polls v. Prediction Markets:

Poll Analysis at fivethirtyeight.com shows that McCain has about a 1.9% chance of victory.

Prediction Markets at InTrade show McCain has about a 9.2% chance of victory.

As such, #1 has it completely wrong.

KW

@9: Your point about people hedging for happiness is a great one, especially since there isn't that much money in the intrade markets.

I put down 1,000€ on McCain at a sportsbook, getting 12:1 odds as an emotional freeroll. If McCain loses, I'll be happy. If he wins, at least I get a nice vacation.

The only reason I didn't do this on InTrade was because I found a sportsbook that was offering a better price than what InTrade was paying.

DK1

Since the prediction markets favor Obama even more than the polls, here's a bet: If McCain wins the election, prediction markets can never be mentioned in this column again.

Jerry Tsai

Your reasoning makes sense to me-- it's the same one I applied. The Economix contest's incentive rewards a WIN, not savvy maximization of expected value. The only way to win is to pick unlikely scenarios.

However, the ridiculously crazy longshots you picked may receive too much attention. My strategy was to pick somewhat longshots, maximizing the expected probability of WINNING, by objectively calculating the joint likelihood of the three contests (using current reported odds) and multiplying it by the subjective probability that people will pick outcomes that are as or more unlikely.

Hein

Interesting, in contrast to "gambling for resurrection" (e.g., diversionary war), where the downside is truncated (you can't lose more days in office than you "have") here the upside is truncated. You can't win more than "win" the contest. Neat.

Mark

I'm just wondering if anyone could tell me, before Obama was listed as a candidate, what the prediction markets were paying for a BLACK man to win the 2008 election?

My guess is they would have been some pretty long odds....

jsb

The wild card in this game is that the system is likely to be gamed. The only real question is whether the queer is sufficient to vary more than the margins.

JimJ

#6 David -- The polls can be used as input to a simulation, to see how likely a certain outcome is. http://www.fivethirtyeight.com/ does this, and currently predicts Obama 96.3% of the time.

Of course, there is always room for a bad model, particularly in terms of how it deals with the likelihood of potential new events that would effectively invalidate previous polls. For example, when Senator Stevens was convicted, that changed how likely he was to be re-elected, but it was too late to affect polls that had already been taken.

John the Pundit

According to some of America's top politicians, it's actually "pundint."

Harlan

The thing is, cash (Intrade) and glory (Economix) are not the same. If I were to be in the Intrade markets, I'd probably invest in a McCain victory, to hedge my happiness. If McCain wins, I'd be depressed, but at least I'd be up some bucks. This is very different from an honest guess at the outcome.

DK1

Just to clarify my original point @#1: It's probably not relevant whether the polls or prediction markets assign a higher probability of victory.

My only point was that prediction markets are touted as a "better mousetrap" vs. polls, especially when a contract like this is so heavily traded. So, if McCain wins, the entire foundation of prediction markets will need to be examined.

David

DK1 - how do you know that the prediction markets favor Obama even more than the polls? The prediction markets can be translated into a probability of victory, but the polls cannot be directly translated into anything (maybe an expected popular vote share, but even that is dubious) ... RC is showing Obama with a 7 point lead at 10:30 AM the day before the election, to many people that may translate into a greater than 90.5% chance of victory currently forecast on Intrade ...

Mitch

Actually the prediction markets favor McCain more than they do Obama. As explained above, longshots are more heavily bet than favorites; thus, McCain's actual chances of winning are lower than his InTrade chance of winning.

If a poll shows Obama ahead 52% to 46%, that does not mean he has a 52% chance of winning. He actually has a very significant 6 point lead, putting his chance of winning at (likely) greater than 90%.

Zvi Mowshowitz

The prediction markets do not favor Obama more than the polls. Yes, Obama is trading very high, but he is polling if anything even higher. If the polls are accurate and the election is clean, Obama wins. Period. If anything, Obama is very much underpriced, as Wolfers notes (although he is way up since he calculated the returns). The polls in effect have Obama trading around 96.

Tyler

@#1:

If McCain wins it won't just be the prediction markets that suffer... the polling system, the media, the Democratic party starting with Obama and Howard Dean but ending with... well, everyone, Wall Street, etc... everyone who has relied on pre-election information would end up forced to radically reconsider how predictions, even when money's not involved, work.

However, if Bob Barr wins more than 1% maybe we should still reconsider the prediction market anyway.

Eric

@ #1:

Hear, hear!