Is Too Much Attention Paid to Small InTrade Contracts?

Harvard economist Greg Mankiw recently posted this chart depicting an InTrade contract on the probability of a U.S. recession in 2008, with the following commentary:

In online betting, the probability of recession is now about two-thirds, compared to about one-half a few weeks ago.

Evidence pointing toward a recession is certainly present, as Larry Summers and others have noted. Still, experts drawing conclusions about the economy’s future based on a single InTrade contract may be risky, particularly given that, as of Jan. 8, less than $40,000 had been traded since the contract’s inception, a tiny amount even by prediction market standards (compare that to the over $3.8 million that had been traded as of the same date for “Hillary Clinton to be the Democratic Presidential Nominee in 2008” — you’ll find more analysis of Clinton contracts here and here).

The numbers beg the question: even beyond politics, are we giving too much weight to smaller prediction market contracts? It’s worth acknowledging that in the prediction-market world, an individual betting a few hundred dollars could raise the “likelihood” of an upcoming recession by 20 points.

(Hat tip: John De Palma)


Actually, the trade volume on the presidential nominations campaign is also small - it's around $10,000 per day on most days. This means two things:

(1) It is east to manipulate the prices. The nominees' budgets are over $100 million.

(2) Serious players don't have an incentive to get involved. The expected profits for a good predictor are in the $1000 per day area. Just getting these good predictions might cost more (think better polls, for example).


I think everyone is missing something: Candidates (plural) can't bias the intrade prediction markets so long as those markets are something people trust. You're forgetting the identity that the sum of the intrade prices for a particular outcome, among potential contracts, must summ to 100%. Hillary's people and Barack's people could spend, literally, their entire war chest moving these contracts, but they either (1) raise both of their odds to 99% (when this is clearly not the true reality), or (2) in the face of the former scenario, any half reasonable, entruprenurial individual with a credit card can 'sell' these very valuable contracts at what he deems to be something like a more reasonable price (there would be a clear, risk free arbitrage opportunity, especially if the bid-ask spread diverged, as it would probably have to). image the contract for Hillary winning and Barack winning are both priced at $99 (for a $100 contract). (presumably this is the buy price). They would have to literally hand out cash to anyone who thought the true odds were less than this: there is a virtually infinite supply of mispriced contracts. So it's not a fair arguemnt to say that only $3 million are in this market, and therefore candidates could easily manipulate it. IF CANDIDATES MANIPULATE IT, and the contract is mispriced, the market will get MUCH bigger VERY quickly.

I can imagine a situation where two candidates are trying to raise their price and lower the others: tens of millions of dollars could change hands, and I believe the market price should still emerge.

These markets do a good job of changing the interpretation of polls to an interpretation of probabilities (how many adults have I met who assume that when candidate A polls 60% and candidate B polls 40% that there is a 40% likelihood candidate B will win... I presume they also expect that had we had 1000 Iowa elections, Bill Richardson would have won 50 times.

As for the impossibility of good information: in a close case, I agree, someone must be relying on information that seems hard to gather...but i wouldn't be surprised if their offering market odds is an externality of already knowing that information (pollsters, campaign-insiders, vote counters, anyone whose business it is to be involved in state-level politics (for the primary)). Much of it is also probably a specific skill that some people have: understanding well the politics of new hampshire. Knowing how to interpret publicly available news, etc.



Robin Hanson

The amount of money in a market may well be weakly correlated with its accuracy, but this only supports a weak criticism. It is much better to just directly evaluate the calibration and accuracy of the market.



Speaking as a professional arbitrager, arbitrage does not require any knowledge at all (other than arithmetic). Read up on arbitrage before you start arguing about it, please.


Bigger prediction markets are the combined knowledge of people who have a lot of money AND are so confident of their knowledge they are willing to put it at risk. Powerful evidence shows prediction markets have better predictive power than polls. So what does it mean to say they "don't work"?


Of course, bigger prediction markets are basically only the combined knowledge of the people who have a lot of money. Maybe they just don't work, period.

Brent Hoag

These markets have had already had attempts at manipulation in the past. The market reacted exactly as Walter explained.

Some more important questions for me would be:

1. Is this prediction market gambling? If not, how would it differ from sports betting? Conversely, I believe that this is one proof that sports betting is not gambling either in the true sense as the sports bettor is betting into a dynamic market, not a static game such as roulette.

2. Is Intrade prohibited by the UIGEA? They may not be able to accept US credit cards, which would decrease their liquidity and their predictive value.

3. Perfect arbitrage situations can exist with these prediction markets. There are no market makers and the contracts can be very illiquid.

4. Intrade (formerly Tradesports) has had settlement issues in the past, such as with the recent North Korean Missile Contracts. With that contract, they required confirmation from the DoD on the NK missile test. However, the DoD never commented in an official capacity so the contract expired as a "no". I do not know the final resolution and whether Tradesports simply unwound and repaid everyone.


Brent Hoag

I went and did some research and it was the Bush election contract in 2004 that had unusual trading activity. Here is the supporting link:


One interesting point is that the Intrade contract only pays out for 2 quarters of negative real gdp growth, which is the standard definintion. The last "recession" of 2000/2001 as declared by the NBER never had 2 consecutive quarters of negative growth. Thus, the Intrade contract would not have paid out since 1990/1991, which was the last time that the U.S. economy met that measure. The data:


I don't know why these prediction markets are given any power for predicting the future. Have we learned nothing from the stock market and the likely hood of predictions? I don't think there is a Warren Buffett of the predictions market, so I'll continue to disregard them. And in that thought, even Buffett got some things wrong.


Considering the amount of stock people have begun to put into InTrade and other similar projection markets, I wouldn't be surprised to find out that candidates are or will soon start to put some of their war chests into maintaining a high price for their victory contracts. It would hurt the integrity of the markets, but what do they care? If it gets mass media and bloggers talking about their rise, and thus suspecting that they must be doing something right to garner such a rise, they will have achieved their end.


Wow, Walter (#4). That was an excellent description on why the markets have to work. I would love to see someone try to do as you suggest and throw a couple million into the market to try and raise the price, only to see it backfire due to normal market operations. I suspect the economic advisor for each candidate would lay out the scenario that you did to try and keep them from doing so, however.


Walter your comment isn't quite answering the right question. The premise is that this market is so opaque, nobody really has a strong prior (at least not a well informed prior) about what is going to happen.

So your entrepreneur, who's going to arbitrage this market to exploit a candidate who is trying to manipulate it, doesn't exist: nobody knows enough to be such an entrepreneur.

Anyway, suppose the "unmanipulated" price is 50% probability of (say) Hillary winning some race. What does it mean for an entrepreneur to know that the right probability is 50%, not 40% or 60%? If this guy knows so much, his probability should be 0% or 100%, i.e. he knows for sure who will win. So he would make his money anyway.

The idea still stands: if the market is small enough, a candidate might find a favorable cost/benefit tradeoff in spending to push up his or her winning probability as implied by this market.


Michael Schmahl

Additionally, I should note that H. Clinton's chances of winning the Presidency (4.5%) are currently higher than her chances of winning the nomination(3.9%). This is apparently another risk-free arbitrage opportunity. But if you sell 1 contract for 2008.PRES.CLINTON(H) and buy 1 contract for 2008DEM.NOM.CLINTON, the transaction fees force you to sell 2008.PRES.CLINTON(H) at an actual value of 4.2% and buy 2008DEM.NOM.CLINTON at an actual value of also 4.2%. The most likely outcome is that Clinton is not nominated, and is not elected -- in which case you break even. This is also your best case.

The transaction and expiry fees on Intrade are high enough that anyone following this market as a spectator must include a 1.5 percentage-point margin of error in any reported prices. Lower, admittedly, that the usual Gallup or Rasmussen poll. But the usual grain-or-three of salt must be taken.


Michael Schmahl

One thing that must be considered is the transaction costs. Looking at the current NEW.DEM.VP.* contracts, the top fifteen candidates are currently trading at around 150% total. With no transaction fees, this would be a dream to an arbitrageur. However, under Intrade's rules, selling one of each contract will cost $0.70 in fees, and the profit is reduced from $1.50 to $0.80.

This is still (currently) a good deal for an arbitrageur, but close to half your profits are going directly to intrade. If the top fifteen vice-presidential candidates had added up to only 129%, there would be no arbitrage opportunity.

Transaction costs create enough friction that the prediction markets should not be taken at face-value. IEM is probably different, since it is not-for-profit.