Justin Wolfers wants to know

And can you think of a better reason than that?

There is currently some very unusual activity going on in the election markets at
InTrade: Over the past two weeks, there appears to have been a concerted effort
to bid up the price of the contract tied to Hillary winning the Presidency, and
on the back of essentially no news, the price climbed from about 25, up to 40,
and now is at about 35.  My usual coauthor, Eric Zitzewitz, suggested that
this looks a lot like manipulation, and the best evidence of this (apart from
the absence of any news) is that over the same period, the odds on Hillary
winning the Democratic nomination are essentially unchanged at 50.  It
just can’t make any sense to buy PRES.CLINTON at 40, when
2008DEM.NOM.CLINTON is at 50. 

All told though, this manipulator has been
surprisingly successful.  And this is where our experiment comes in.
I wonder whether the manipulator will be just as successful when a broader
audience becomes aware of this possibility?  We tried a first experiment a
week ago – Eric mentioned this anomaly while at a prediction markets
conference in Palm Desert (Robin was with
us).  In turn, this led to a bit of pressure on the price, but the
manipulator held firm.  We were wondering whether a simple post on
Marginal Revolution (tipped off by a loyal reader) pointing to this mis-pricing
might lead to further price pressure?  As such, if you were to blog on the
mis-pricing, that would serve a second purpose, of giving us truly experimental
variation in public attention.  And all of this holds the promise of
learning a bit more about manipulation in prediction markets.

OK people, trade away!

Addendum: Here is Eric Zitzewitz, over at Chris Masse, on same.  And Koleman Strumpf says no manipulation.


Greg Mankiw has blogged on this before. It appears that transaction costs and contract technicalities make arbitrage impossible, since there is a non-trivial chance that the initial Democrat nominee will step down.

Maybe, but it could also reflect the markets saying that the probability of any likely democratic candidate winning are improving. The news that would change this probability is the continued poor performance of all the likely republican candidates.

What is happening to the contract for the other democratic candidates? If they are also improving like the Hillary contract it would tend to support my thesis.

What possible incentive exists to manipulate the price of such a contract?

I can think of none.

The truth is, polls are very very important. They are what the moderately informed use to gauge their position among others. Say you know you are more conservative than "most" people. You can use this information along with another internal gauge of the candidate's positions along a liberal-conservative line to re-assess if you are supporting the right candidate. This is actually a pretty good model of what people do subconsciously.

I had a colleague who went to Russia to help advise the new republics on their democratic processes. Rival candidates were publishing wildly varying polls suggesting support for themselves, in an attempt to sway the populace in this way. It was solved not by outlawing the publication of false polls (a tricky thing to determine), but by banning the publication of *any* poll in the last week or so before the election.

If the contract is being manipulated, it is so that Joe Reporter, when he decides to write a balanced article on the runners in the field, will either give out misleading data or weight his treatment inappropriately.

It would be interesting to know how much money has actually been risked (given a... 1/3 chance of primary selection). Then again, there is always the chance that someone has decided to stake money on an independent piece of information--isn't that what the market is supposed to be all about?

First, a lot of liquidity is being artificially blocked from Intrade because of the legality situation. I for one maintain only "beer money" there because of this.

Second, right now it looks like you could sell PRES(Clinton|Obama|Edwards|Gore) for 62.3 and buy PRES(Dem) for 57. I would guess the required margin on this would be 37.7 (worst case scenario for the individual basket sale) + 57 (worst case scenario for PRES(Dem)), so even before transaction fees, you are looking at a 5.3 gain on 94.7 total margin over one year out?? That could just be a result of discounting. You can make 5% risk-free in that time.

Plenty of this going on in Tradesport, as people note due to transaction cost
(and cost of capital).

A few examples:

I noticed a week or so ago that the combined chance of Hillary or Obama becoming President was almost as high as the chance of ANY Democrat becoming president. Non of those two will run as an independent, so this is just mispricing.

As I am writing this Gore has a 10,4% or so chance of winning the nomination, and a 8% chance of winning the presidency.

Really? Does the market really believe Gore has an 80% chance of winning if he becomes the nominee? Given that other similar candidates (say Edwards) have about 50%?

Don’t ignore the cost of capital here. I would go and trade if the election was tomorrow. But if you want to put enough money to make it worth the effort you have to bind up cash for a long time (assuming the market does not immediately adjust, which it does not seem to do).


I am not convinced the Hillary President stuff is manipulation, since many other candidates have these dichotomy between winning and becoming President. However it should be clear to anyone that the Ron Paul 2-3% odds of becoming the nominee is a free lunch, due to his cyber-army. This is higher than the *combined* odds of other GOP outsiders (Brownback, Hunter, Tancredo, Gilmore, F. Thomson), for a candidate that get 0% in the polls of likely GOP voters†¦.

Real world examples keep hurting Robin's theories. If we can't get a liquid market for something with as wide appeal as this, how do we expect to be able to use markets to do stuff like decide who should be CEO of a company, etc.?

so wolfers is manipulating his "experiment"?

doesn't sounds scientifically sound to me...

As with any market, if you think the conditional probabilities are off then trade against it. you think gore at 80% prez conditional on winning dem nom? then sell gore prez and buy gore dem nom. Same thing for the Hillary contract

I for one think 80% might be right. It implies that he's coming with a full head of steam on the environment issue and has saved boatloads of cash by not spending against hillary and obama early.

In any case, I'm not sure why it would be called "manipulation." Even if one buyer was out lifting all the Hillary prex contracts he could, it's just taking a market position, right? Well, unless a Hillary donor has put aside a couple million bucks to bid up the price, making the prediction market polling look favorable -- but in that case they would be wise to buy across the board for prez and for the nomination.

There's a simpler arbitrage possibility. This morning when I was placing sell orders on 2008.PRES.CLINTON(H) and 2008.PRES.GORE, I noticed that the bid prices for the top eight contenders added up to 110.6%.
It's not unusual for Intrade probabilities to add up to slightly more than 100% (probably due to their interface making selling look like unnatural or risky short-selling), but the size of the arbitrage opportunity seems unusually large.

Comments for this post are closed