Deep-Pocketed Manipulators are a Prediction Market’s Friend.
Michael Giberson July 8th, 2007
The re-publication of the Tim Hartford column on possible manipulation of the Clinton prediction market at Intrade provides an occasion for additional thought. (As noted by Chris Masse, on July 7 Slate published the Hartford column that initially appeared June 29 at the Financial Times.)
On this topic, Slate is late and even the FT was a little behind the story. When the FT published last week, the price of the Clinton contract had already fallen most of the way back to the levels that prevailed before the alleged manipulation, and by July 7 the price was all the way back to the mid $20s.
It is difficult to prove manipulation in almost any market, and manipulation of prediction markets may be much harder to demonstrate that most . To try to get a handle on the somewhat slippery notion of manipulation, consider as an example the standards for proving market manipulation under the U.S. Commodities Exchange Act. As summarized in a recent Senate subcommittee report, four elements are required:
(1) the accused had the ability to influence market prices;
(2) the accused specifically intended to influence market prices;
(3) artificial prices existed; and
(4) the accused caused the artificial prices.
In a continuous double auction, such as Intrade, it isn’t too hard to “influence market prices” in the short run – all you need is enough backing to buy or sell your way through the queue of standing orders until you reach your desired price. To maintain your price, you need pockets deep enough to satisfy all comers. An examination of trading patterns may be sufficient to show ability to influence price and may provide indirect evidence of intention to influence price.
But what about this notion of “artificial prices”? In a general sense, an “artificial price” seems to be a price deviating from fundamentals in ways not readily justified. Perhaps this sort of issue is easy enough to examine for a commodity futures market, but it seems to be much harder to investigate the issue in the case of prediction markets. In particular, how would a market observer – even one with full access to trading records and other data – differentiate between a deep-pocketed trader who seeks to manipulate the price and a deep-pocketed trader who simply has a significantly different estimate of the underlying probabilities?
I think the potential for manipulation troubles some prediction market folks because it puts an implied asterisk next to any prediction market price that says “caution: the price you currently observe in this market may not truly represent the collective opinion of the population of traders, but instead may reflect an attempt to manipulate the market outcome.” The possibility of manipulation taints the purity of the price as a predictor.
Nonetheless, in the longer run deep-pocketed manipulators are the prediction market’s friend. As the theory paper by Robin Hanson and Ryan Oprea (”Manipulators Increase Information Market Accuracy“) notes, the net effect of a manipulator is to motivate informed traders to collect more information and subsidize participation of informed traders in the market. Take a closer look at the volume bars in the Intrade chart above. Notice how volume perks up dramatically after the alleged manipulation began and stays higher than before?
More information being brought to market, more trading, and more informed participation: all good things for the market.
[Edited version of post originally placed at Knowledge Problem. I also blogged at Knowledge Problem a few days ago about the initial column by Hartford, though that post doesn't cover any ground that readers here won't already be familiar with.]
Excellent blog post.
If that’s “manipulation”, then the fact that it lasted two months is very troubling to me. I will wait till Koleman Strumpf and all the others make up their mind on this.
CFM: If someone gets it in his head to start manipulating Clinton prices, it takes a while before traders can load up the requisite ammunition (money) to fire back, because of myriad liquidity constraints imposed upon Intrade by — quelle surprise! — the US Congress.
Alex Forshaw, with all due respect to you, your answer doesn’t satisfy me. In the past, we were told that manipulation attempts didn’t last long. This one (if proven) did. This would be very disturbing (if manipulation were to be the case here).
Why doesn’t it? An “efficient market” has zero capital controls, meaning that if something aberrational occurs, outside money can easily flow into the market to correct the perceived aberration. With Intrade and markets like it, that is clearly not the case, thanks to American legislation.
So if someone swamps one of these markets with a lot of money to create some kind of imbalance, it takes a lot longer to correct it.
Hmm… I don’t buy your argument. Iowa Electronic Markets is a very thinly traded betting exchange, and it took shorter periods of time to overcome manipulation attempts, if I’m correct. Again, I say that two months sounds horribly too long to me.
Mike Giberson,
Have you compared the two Hillary Clinton event derivatives, InTrade vs. BetFair?
CFM: No I haven’t done the comparison between Intrade and Betfair. (In a few minutes of clicking around Betfair I didn’t find historical data, but it is early here on the east coast and I’m only half though my coffee.)
IEM is thinly traded, but I suspect that the limits on initial account size and other restrictions there mean that no single account has the ability to enter a market and move it in the same way as happened at Intrade with the Clinton market.
The size of the manipulation experiments conducted on IEM and reported by Rhode and Stumpf was likely comparable to average account size (does IEM report statistics on such things?). It is likely that in the experimental economics research on prediction market manipulation, subjects were provided equal or near equal trading funds. My guess is that the new money brought into Intrade for the Clinton contract beginning around May 12 was substantially larger than the average account.
I agree that two months sounds like a long time for the market to recover (from either a manipulation attempt or the arrival of a new, well-funded, naive speculator), but I’m sympathetic to Alex’s position and don’t discount the possibility of it taking that long for counter-speculators to assemble the information and funds to overcome the initial price swing.
Another factor: very informed market observers were driven to blog about the possibility of manipulation and to making offers to bet off exchange, rather that simply putting their money down and taking the alleged manipulator’s money. Had these observers taken the direct approach, I’m sure it would have cut at least a week or two off the time to recovery.
Betfair historical data
http://data.betfair.com/
“the new money brought into Intrade for the Clinton contract beginning around May 12 was substantially larger than the average account.”
Indeed.
Your last paragraph seems to be in accordance with Robin Hanson’s views, if I’m correct.
I’m trying to build up my Hanson fanboy street cred.