Strategic Manipulation of the Information Markets

Just over a week ago, Nate Silver, the founder of the excellent political blog FiveThirtyEight.com, wrote about potential manipulation of Intrade’s information market on the 2008 presidential race. It’s difficult to gauge whether this manipulation is in fact occurring, in part because it’s hard to imagine that very many traders would be willing to invest thousands of dollars in moving a relatively meaningless information market more than a month before the election. Late last week, however, Intrade’s political market almost certainly fell subject to strategic manipulation, this time with a much more prosaic goal: to exploit overlapping markets in order to turn an easy profit.

A bit of background is in order. Last week, Intrade opened up a new information market designed to gauge which candidate—Joe Biden or Sarah Palin—was most likely to “win” the vice presidential debate, in the sense of boosting the fortunes of Barack Obama or John McCain. This new “debate” market was pegged to the longstanding market that measures Senator Obama’s chances of winning the election (which I’ll call the “reference” market). The debate market compared trading on the reference market before the debate with trading on the same market after the debate; to be precise, it compared the volume-weighted average price of trades on the reference market from 6:00pm – 9:00pm Eastern time on Thursday night (the debate began at 9:00pm Eastern) against the volume-weighted average price from 12:00am – 12:00pm Friday morning (after the debate had finished). If the post-debate reference price finished lower than the pre-debate price, then people who had bought stock in “McCain to win” on the debate market were paid at 100 cents on the dollar, and people who had bought shares in “Obama to win” received nothing.

This structure presents an opportunity for a strategic trader to manipulate the two markets for substantial profit, especially if the debate is viewed as close to a tie. (Or, more accurately, if neither candidate substantially exceeds expectations, since expectations should already be priced into the reference market.) This trader need only invest heavily in one side or other of the debate market, and then make just enough trades on the reference market to determine the outcome. The trader may lose small amounts of money on these reference market trades, but if she has bought enough stock in the debate market her gains there will overwhelm her losses.

This appears to be precisely what occurred. Attached  here is a spreadsheet showing all trades made on the Obama-for-President market (the reference market) in the time surrounding the debate, which began at 9:00pm on October 2^nd. Recall that the debate market measures the weighted average of Obama stock from 6pm-9pm EDT Thursday against the weighted average from 12am-12pm EDT Friday. Here is how the reference market behaved:

1. Between 6pm and 9pm on Thursday, the market cruised along at about 67.4 (meaning that Senator Obama was given a 67.4% chance of winning the election), with a high volume of trading in that vicinity.
2. When the debate ends at around 10:30pm, the reference market price was still in that neighborhood, at around 67.4.
3. There was immediately some selling that drove the price down into the 66s and even high 65s over the next hour or so, and then at 11:58pm there is a price increase of nearly a point, to 66.6. It is during this period that all of the “news” regarding the debate—by which I mean polls and experts purporting to decide who won—reaches the public.
4. Within the first half hour after midnight, the first half hour after the start of the relevant time period, there is immediately a sell-off that drives the reference market down 2 points to 64.6.  (There is essentially no news during this period.)
5. The reference market then hovers in the range of approximately 65 to 66.4 for all of Friday morning, with small oscillations up and down.
6. At around 11:25am a run-up begins, and by 12:00pm the reference market is actually at 68, higher than before the debate.  (It appears that the strategic trader, knowing that victory was in hand, loosened her grip on the market.) But it is too late to make much of a dent in the weighted average: there were 2865 shares traded from midnight until 11:25am, and only 1126 shares traded after 11:25am, of which only 176 shares were traded at above 67.5. The traders who bet on McCain to get a post-debate bounce have won.
7. By 1:30pm, the reference market was trading at approximately 68.

Imagine for a moment that our strategic trader was responsible for trading all of the 2865 shares purchased between 12:00am and 11:25am, trades which drove down the reference market price. Despite those trades being made at below-market prices, they are nonetheless within 2.5 or 3.5 points of the “market” price (the price at which the market equilibrated by 1:30pm). In other words, the market manipulation could not have cost the trader more than 2865 x 3.5 x $0.10 = $1003, and this estimate is likely quite a bit too high. (Intrade prices its shares such that each “point” is worth ten cents.)

The debate market is finished, and so data on it is no longer readily available. But imagine for a moment that the strategic trader was able to purchase shares of “McCain to win the debate” at a price of 65. (The market was trading in that vicinity for much of Thursday.) Each of those shares nets the trader a gain of 35 x $0.10 = $3.50 when the market closes, meaning that the trader needs to purchase fewer than 300 shares to break even. If the volume on this secondary debate market was higher and the trader was able to purchase as many as 2000 shares in the days leading up to the debate, she could have turned a profit in the thousands of dollars.

In the grand scheme of the markets, these are small (and relatively insignificant) amounts of money. But this strategic exploitation of the information markets raises the question of whether similar manipulation could occur in the regular financial markets as well. For instance, it is possible to trade in derivatives pegged to a market’s volatility. A trader might be able to invest heavily in these securities, and then churn stocks in the reference market at small losses (possibly inducing others to do the same) in order to drive up the measures of volatility. Whether this can succeed as a money-making strategy depends on the volumes of the reference and secondary markets, as well as the costs involved in manipulating the reference market, but it would not appear to be outside the realm of possibility. The lesson is that designers of information markets and derivatives alike must be careful to isolate their reference values from manipulation, lest the derivative itself come to represent little more than manipulation and strategic behavior.

Comments

[Jason Ruspini]

More easily, two friends could conspire to trade back-and-forth in their preferred range to create very large volume with minimal p&l implications ("wash" trades) and side-payments if necessary.

http://groups.google.com/group/Prediction-Markets/browse_thread/thread/7d88abc019dc6552/249a39a22a9422ab