How well did the subsidies that I provided for some Intrade contracts promote price efficiency?
For the conditional contracts DEM.PRES-GOVT.DEBT, NONDEM.PRES-GOVT.DEBT, DEM.PRES-TROOPS.IRAQ, and NONDEM.PRES-TROOPS.IRAQ the evidence is clearly disappointing. Efficient markets would never produce an implied price[1] exceeding the maximum expiry price of 100. NONDEM.PRES-GOVT.DEBT implied prices exceeded 100 about 48 percent of the time, and NONDEM.PRES-TROOPS.IRAQ implied prices exceeded 100 about 22 percent of the time. The other two contracts had implied prices that stayed below 100, partly because their prices stayed closer to 50, which meant it took less money to keep their implied prices below 100.
I also tested the efficiency of those prices by looking at the correlation between PRESIDENT.DEM2008 and the implied prices. I expect that since the majority of fluctuations in PRESIDENT.DEM2008 have little connection with troop levels and debt, those correlations ought to be closer to 0 than to 1. For most time intervals, that correlation was above 0.6, and often above 0.8. The correlations between each pair of implied prices seem like they should be closer to 1, but they were moderately lower than the first set of correlations.
Another way to look at the data is by the charts that Mike Linksvayer made. The implied values show large moves with the Democrat and Republican implied prices going in opposite directions. This could be caused by the conditional contracts moving less than is needed to reflect changes in PRESIDENT.DEM2008, and are hard to explain by any news. The financial news over the past two months ought to have caused the implied debt increase to rise in ways independent of the election result (news of a significant recession implies lower revenues and higher welfare-type spending for years to come), but I see no such effect in the graphs he presents.
I can’t find any other evidence of efficiency that would argue that those tests are misleading, so I conclude those four contracts did not provide valuable information about how the election would affect the world.
The shock response futures DEM.PRES-OIL.FUTURES and DEM.PRES-T.BONDS are harder to analyze. DEM.PRES-T.BONDS showed no expectation of an effect of the election on bond prices. Without testing many contracts of this type, it is hard to say whether that was due to inadequate incentives for traders or due to the election having no effect on bonds.
The oil contract showed some expectations of an oil price response to the election starting in May. The effect was strongest in September when the race looked close (which is when shock response futures look most likely to be profitable). During that month, the price predicted that an Obama win would cause oil futures to drop. I expect that this prediction should have been taken seriously, but it’s hard to have a lot of confidence without repeating this kind of contract a number of times.
Since the election ended up not looking close on election day, the actual change in oil futures told us little. They increased over the time period specified by the contract, but that increase mostly happened Tuesday morning when there was little new evidence of the election result. The change over the time period when the election results became certain was down as the contract predicted in September. In hindsight, I wish that I had asked Intrade to use oil futures prices from sometime on Tuesday as a starting point rather than Monday evening.
Since the information provided by this type of contract is most valuable when swing voters are making their decisions, it seems unimportant that the contract apparently stopped providing information after the election stopped looking close.
Part of the problem with the conditional contracts is that it is hard to trade them (particularly if you want to place orders that don’t get executed quickly) unless you can specify your price in terms of the implied price rather than the price of the conditional contract. Implementing that would take some nontrivial changes to Intrade’s software.
Could the subsidies have made it too expensive to move prices to the efficient price? Problems related to a software bug gave us some evidence. For most of the last week of October, the liquidity provided by the market maker was reduced by a factor of 115. The result was more volatile prices, but little sign of more efficient prices.
I would like to learn from the behavior of the contracts how to make market subsidies more effective at producing efficient prices, but so far I haven’t been able to learn anything useful about that.
[1] – The implied price means the expected price if PRESIDENT.DEM2008 is expired without causing this contract to expire at zero. I.e. what the contract price suggests the price ought to be if the election goes the way the contract says.
I wonder if part of the reason is that not enough people knew about the subsidies. From looking at the contracts online, I could not tell they were subsidized.
Pingback: The Robin Hanson-inspired, real-money, conditional prediction markets at InTrade are a world-wide disaster of thermonuclear proportion, concludes the long-time Robin Hanson fanboy who forked over real money to InTrade to have them set up and run. | Midas
Maybe you can build in some efficiency by having the decision market constantly monitor the presidential market and move the bid/ask midpoint as appropriate.
The AMM would essentially be making a market on the implied price (Mike Linksvayer’s graphs), and the prices posted in the decision market will be those AMM prices divided by the presidential market price. There are details to work out, but that’s the sort of thing I would try.
There is something to be said for the original design which does not automatically move prices based on P(B) because it should (if it’s easy to get cash into the exchange) get people involved in the market.. but this is very expensive. The subsidy should not just give money away to people who have no new information, which is what happens when the AMM doesn’t respond to price changes in related markets.
It would also save money to initialize with low risk and only tighten things up when the market tells you roughly what the implied value should be.
Thank you Jason, I should have thought of adjusting the AMM orders based on the presidential market. Although it may become hard to evaluate how much money the AMM would lose.
I’m less certain about your initialization advice. In addition to complicating the calculation of how much the AMM risks, it might cause some traders to decide early on that there isn’t enough liquidity to make the contracts worth paying attention to. That could in principle be handled by a more conspicuous description of the AMM, but Intrade hasn’t designed their system to do a good job with contracts that need more than a few words to describe them.
Yes, the market should be tightened as soon you’re comfortable that the initial estimate isn’t too horribly off, and a relatively dense/deep book of small lots could help with the perception of liquidity.
While the losses aren’t strictly bound with these approaches, with some care I don’t see how their expected value could be worse.
As someone who traded two of these contracts and followed their movements fairly closely, it seems to me that the main problem was how few traders knew about and understood the contracts and the subsidies. There was little incentive to try to figure out what the implied price should be when it was consistently possible to sell the two non-dem contracts at prices at or above the McCain or REP pres prices, witht he added plus of paying no fees.
By the time I was heavily invested in these contracts, I had come to see Obama’s victory as close to a lock, and so was usually not willing to push the price more than a point or two lower than the McCain price. This was especially so given the higher liquidity and possibility for fluctuation in the main markets. From what I observed, the apparently small number of other traders participating in these markets (I think I accounted for about a third of the total volume) were behaving similarly.
You should know that matters may been complicated by an the role of an institutional investor that, on several occasions, enterred very deep long McCain and short Obama positions, driving the price of each contract several points from its ex ante equilibrium. The opportunities for quick profits from the strikes gave traders an incentive to have significant amounts of free margin available on hand so they could take full advantage. At least in my case, that affected my thinking with regard to the contingent markets you subsidizede.
Pingback: Why Nate Silver is wrong about the usefulness of long-term prediction markets: IT DIDN’T WORK. IT IS A PATENTED FAIL. | Midas Oracle .ORG