Skip to content
 

A flaw in prediction markets? Or simply lack of markets?

This blog has previously included discussions of prediction markets.

A new article in Slate claims that Intrade and perhaps some other markets have an “anti-Obama” bias. According to the article, a wide swath of people (including Intrade participants, economists, Fox and CNBC financial pundits, and the Wall Street Journal) systematically underpredict the Administration’s likelihood of advancing its agenda, and are systematically gloomy about the outcome of the Administration’s successes. The article points out that in the year since a Stanford economist wrote an op-ed entitled “Obama’s Radicalism is Killing the Dow,” the Dow is up 72%.

A few things come to mind. First, if the article’s author (Daniel Gross) really believes what he’s saying, then he should have first made a few thousand dollars (or a few hundred thousand dollars, if that’s possible) doing Intrade wagers to take advantage of this inefficiency. Perhaps he has already done so.

Second, most of what that Slate article about doesn’t have anything to do with political futures markets. What an economist says in a Wall Street Journal op-ed, or what a bunch of economists say when asked to predict next year’s GDP, is very much NOT the same as having these people participate in a market. A market is supposed to evince an accurate reflection of people’s actual predictions, and to provide an incentive to improve those predictions; having some pundits mouth off does nothing of the kind. Only the Intrade case would be an example of bias in the markets, and unfortunately Gross gives only a single example of what he claims is bias: he says that even quite late, the market assigned a rather low probability to the passage of the health reform bill.

Finally, I think Gross may be cherry-picking. He cites examples of people who have been pessimistic about the Administration and the economy, but there are also people out there who have been optimistic. The individuals might be biased, but this doesn’t necessarily indicate that economists, or pundits, are biased as a group.

That said, something does look funny about the InTrade wagers. Right now, they’re assigning about a 25% chance that Sarah Palin will be the next Republican nominee for President!

9 Comments

  1. John S. says:

    So what is your personal estimate of the odds, and why? And more importantly, are you putting your money where your mouth is?

  2. John S. says:

    The prices on Intrade are supposed to be probabilities, aren't they? If so then why do the prices of the potential Republican nominees for the 2012 presidential election add up to more than 100?

    Maybe that's what explains this chart.

  3. zbicyclist says:

    "assigning about a 25% chance that Sarah Palin will be the next Republican nominee for President"

    Too low? ;)

  4. Grant Cavanaugh says:

    Presumably, the relatively rich people betting on Intrade do skew right politically, as do rich Americans in general. That could easily explain why there might be a slight initial bias against positions that rely on Democratic policy wins.

    The fact that these biases persist might be little more than than a function of thin markets.

    You couldn't take big money bets against a Sarah Palin nomination even if you knew with certainty that the market was biased because the total contract volume is so low that any substantial bet would move the market against you.

    In the "real" financial world most new futures contracts fail because they never reach the critical level of liquidity needed for the contract of offer reliable price discovery or the opportunity for large position hedges.

    Why would Intrade be any different?

  5. Peter Twieg says:

    How could you not make big bets against Palin if you knew she would not be nominated? You know that the value of the asset in 2012 will be zero. You could sell as many Palin shares as you want knowing that you'll never have to pay out on them.

    Even if the market moves in her direction, it doesn't affect that result in the end. Prediction market bubbles can't exist in perpetuity because eventually the markets will close.

  6. Phil says:

    Regarding Intrade specifically:
    1. I signed up for an account yesterday. All I have to do is deposit some money, and I can place bets. But…that might not be legal, and is definitely a pain. Intrade is based in Ireland, and whether it is an illegal gambling site is unclear to me and to Intrade: according to Wikipedia, "In July 2008… Intrade sent a letter to the Commodity Futures Trading Commissions asking for clarification of the legal status of intrade.com …, saying "While Intrade serves a global community and has registered members from 162 countries, our 82,000 plus membership are predominantly resident in the United States … it is perversely unclear as to whether Intrade, and indeed myself, are considered persona gratis by the United States." The Intrade website says most U.S. credit card companies will not approve a payment to Intrade (I didn't try mine). They suggest other ways to get money to them. But both the hassle and the legal worries surely greatly reduce the number of U.S. participants, which may cause a major problem with the market, especially for U.S. political events.

    2. My intent was to bet something like $500 against Palin being the Republican nominee in the 2012 election (currently about a 25% chance on Intrade). That's about the most I can wager without feeling that I have to consult my wife, who I fear would put the kibosh on the whole thing, at least at that cost level. I'm sure Palin has less than a 10% chance of being the next nominee, and in fact I would probably take the No side of a 5% chance (but the Yes side of a 1% chance). But given the ambiguous legal status of Intrade, I haven't yet decided whether to put my money where my mouth is.

    3. If John S is right about the Republican nominees adding up to substantially more than 100%, then there is obviously a big money-making opportunity: bet against all of them, and you can't lose. (Actually I'm not sure about Intrade's fee structure, perhaps transaction costs make this a losing strategy). Intrade's market must be _really_ inefficient if this opportunity exists.

  7. Does betting require that you put money away now and forgo interest income for a couple years?

  8. As for the >100% sum, I've noted that in every commercial sport bets, once you interpret the bet ratios as Prob and add up, you get towards 120% or 130%. I ignore where this surplus is originated and discharged (inefficiencies, transaction costs), but it is on this order of magnitude – at least in cycling where you have a large number of runners.
    I personally adopt this simple routine: I reproportionate uniformly the exp. prob. from sum = 125% e.g., to sum = 100%.

  9. Nick Firoozye says:

    This is something that has come up as well on UK elections as well–the (unusual) fact that prediction political markets can actually have risk premia due to the correlation between outcomes and states of the world/future consumption.

    Let's review risk premia in general: In a C-CAPM or in CAPM (now you may have problems with both, granted but there's valuable intuition in each), if a security performs well in a low-consumption state, it is considered to be an insurance policy. Insurance policies are in high demand since everyone wants to hedge those low-consumption states and insurance policies are consequently expensive, trade at a premium or have low expected return.

    The opposite is true of a high-beta asset, or one which pays off well in a high consumption state, pays off poorly in a low-consumption state (e.g., an equity). Because there is a plentitude and this does not help to smooth or hedge consumption, there is less demand for this and it must trade at a discount (and the expected return must be large).

    In the UK election, everyone seems to believe that a hung parliament is the worst outcome, that equities will dump, that Gilts will underperform and that the currency will dive. (I think it's a bit overstated, but there it is–it's in the news all the time, just that view). Consequently, the BET that pays £1 when there is a hung parliament is a good insurance policy against that possibility, and thus should trade at a premium to the ACTUAL EXPECTATION. And, thus, the market-implied probability probably OVERSTATES the probability of a hung parliament.

    Now in general, I would not expect there to be a risk premium in political markets. The outcome of the next election is probably more casued by the economy than can the next election actually move the economy one way or another. If the correlation between a security's payoff and consumption is 0, there is no premium.

    In the case of Obama's health care reform, it could easily be that like it or not, market participants foresaw this as good for the economy (seriously…again, this is not my opinion, just one that makes the argument work). If it was good for the economy, the wager has a higher beta, and must trade at a discount which means the implied (i.e., risk-neutral) probabilities should UNDERSTATE the actual (physical world/subjective) probabilities.

    Again not my opinion and I personally am not in favour of big govt under any circumstances. I personally believe that even changing the policy of litigiousness, (e.g., make it harder to win malpractice lawsuits) necessarily lowers the cost of healthcare and all the seemingly unncessary tests that doctors perform. If we want to keep up quality of service, at the same time, possible criminal malpractice cases (e.g., Michael Jackson's physician) could replace these, and still prevent the worst abuses. It's possible to change the system in seemingly small ways with much bigger impact than to have a sledge hammer to hit it with.

    It's pretty clear that as soon as prediction markets move into clearly economic or financial arenas, the rules will be different and you can't just say the risk-neutral measure is the same as the physical measure (i.e., the implied probabilities need not be the same as the participants subjective probabilities). Having a risk premium of course means that if we repeat the experiment again and again, we'll see that the market will 'get it wrong'.