Manipulation can affect prices

For the last two weeks a very interesting manipulation has been going on in Intrade’s “Hillary Clinton for President” contract.

1. The contract had been trading between 23 and 26 all year. It has consistently been about half the price of the “Hillary to get nominated” contract price. This ratio implies that, conditional on nominating Hillary, the Democrats have a 50 percent chance of winning the Presidency.

2. Comparing this with the unconditional probability of a Democratic victory (about 56 percent throughout 2007) suggests that Hillary is a slightly weaker general election candidate than the alternatives (Obama, mainly). [Note I say “suggests” because the comparison of conditional probabilities implies a correlation, but not necessarily that a Clinton nomination would cause a better outcome for the GOP. For more see the fifth question in this paper].

3. Around May 12, someone started buying “Hillary for President” pretty heavily, driving the price up to 40. This price is clearly ridiculous for two reasons:

3a. You could sell the President contracts of Hillary, Obama, Gore, and Edwards for a combined 69 (40+17+8+4) and buy the “Democrat to win” contract for 56.

3b. Since there was no movement in the nomination contract, the conditional probability of Hillary was now a ridiculous 40/52 = 77%, while the conditional probability of “Not Hillary” was 16/48 = 33%.

4. Unlike past manipulation attempts, this manipulator isn’t just dumping a ton of money in to move the price once. He (or she) is moving the price, and then providing support to keep the price high. Note that the price stayed at 40 for about a week (on higher than normal volume).

5. I mentioned the manipulation at the end of my talk at the Palm Desert prediction markets conference, figuring that there was no surer way to get a $100 bill picked up than to tell that crowd about it. Someone emailed Greg Mankiw and he blogged about it the next day. (Justin and I also just tipped off Tyler Cowen). Since then there has been some downward price pressure, but the manipulator isn’t throwing in the towel. He/she keeps replenishing the bid side of the order book, albeit giving ground in the process.

6. By my calculation, the manipulator has spent about $10k to push the Hillary contract up around 12 pts on average for 2 weeks, buying about 8,500 contracts in the process. [I’m assuming 26 is fair value and just summing up volume*(price - 26)].

7. So what do we learn from this?

7a. Manipulation doesn’t have to be as ham fisted as the 2004 Bush reelection contract manipulation.

7b. The manipulators are getting smarter. This manipulator was smarter in one sense by providing price support after the fact. But of course, he/she shouldn’t have pushed the price up to such an obviously ridiculous level (and should have bought and sold other contracts to keep the pricing relationships consistent). The same mistake probably won’t be made next time.

7c. By prediction markets standards, manipulation is expensive. But by political spending standards, manipulation could be reasonably cheap. That said, I can find only one media mention of the inflated Hillary price. $10k for one blog mention probably isn’t great value for money, but the Intrade prices get cited a lot these days, so the manipulator may just have been unlucky.

7d. None of this disputes Hanson and Oprea’s point that, if anticipated, manipulation could increase average prediction market accuracy. In their model, traders all have rational expectations about how much manipulation to expect. In the real world, they may need some help (hence this post).

7e. Although the Hillary price is down to 34.5 (bid-ask midpoint at time of writing), there are about 500 contracts bid above 33, so there is still plenty of free money there if you want it.

Hillary Clinton Chart 2007 Manipulations EZ

Related Posts

Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>