Information aggregation markets and space development
by Sam Dinkin
|Infomarts require traders to put their money where their mouth is.|
The most famous of the infomarts are the Iowa Electronic Markets (IEMs), which trade various kinds of securities. Winner-take-all markets pay $1 for the winning outcome and nothing to losers so the trade price in pennies represents the percent chance the proposition will be true. Vote share markets pay according to the vote share of two or more outcomes. These two types of securities can be used to answer most types of questions. For example, a vote-share security that pays the logarithm of the price per kilogram to orbit in 2010 could be used to see if the price will be closer to $250 or $25,000, assessing whether Elon Musk’s or Mike Griffin’s view of the world is closer to being correct. A winner-take-all security might ask if $500/kg is achieved by 2010.
If space propositions securities can be traded, valuable information can be disseminated to investors, entrepreneurs, policy makers and customers. Here are some teaser security names:
I asked George Nield, Deputy Associate Administrator for Commercial Space Transportation at FAA AST, at the Space Access ’05 conference why AST was concentrating exclusively on safety when none of the uninvolved public has been killed in rocketry accidents, while space company after space company has gone bankrupt due to poor economics. Better forecasting might result in fewer economic casualties.
Adopting informarts for space could potentially lead to space colonization years earlier at a far lower total cost than continuing with poor forecasting leading into more blind allies.
Public opinion polls do a rotten job at tabulating the strength of preference. Infomarts require traders to put their money where their mouth is. Public opinion polls treat the responses of people who make lots of good predictions the same as responses from people who make bad predictions. An infomart results in losers exiting as they run out of money. This helps the market to evolve to become smarter over time through natural selection. Public opinion polls are also a huge negative externality. The person called is typically not paid. The inconvenience of the unsolicited call and the expended time by the respondent more than double the social cost of telephone surveys.
|Better forecasting would result in fewer economic casualties.|
The most widely acclaimed infomarts are the IEMs. It shaves 0.5% off of the error associated with public opinion polls for presidential elections from 2% to 1.5%. It is also a lot cheaper and faster to run than a series of polls. A telephone survey might cost only $10,000 for a survey of 1,000 adults, but that survey would require 1,000 callers and an instant electronic analysis package in order to get within five minutes of the time it takes to tabulate the results from an electronic market which has a response rate measured in seconds or fractions of a second. Surprisingly, there is no action at the IEMs now.
If the predictions are altered by daily events, the cost of running a daily poll on one proposition alone would be $3.65 million per year. A continuous market to monitor hundreds of propositions would cost far less. A well-run infomart’s trading operation can be run at a profit. Iowa Electronic Markets do not charge commissions, but also do not provide interest for money on deposit.
For longer term bets, the interest rate paid becomes important and the market unduly thin. A small commission from traders could be taken and interest could be paid to traders who hold information securities. For example, presidential prediction market securities Democrat, Republican, and Other, which pay $10 in 2008 depending on who wins the presidency, could be purchased together in 2005 for $8.93 and the money could be invested in treasury bonds yielding 3.7% per year to make up the difference. Instead, a $0.05 commission could be charged on each trade to support the market. Without this innovation, traders would need to believe that the market price was off by more than 10% for a security in order for it to be worth trading three years in advance. For this web bet on Democrat and Republican, you need to overcome an additional 8.5% house edge. With a $0.05 commission on a $10 face value security that pays interest, the traders need to believe the market price is off by only 1% if the security is trading near $5.
Market-making costs are also important. For less interesting but more valuable topics, it is often worth it to market makers to subsidize rather than tax traders. The commission can be cut to zero and the interest rate paid on money invested in propositions increased to something approaching the stock market’s historical return (11%). Further subsidies can be used to decrease the “spread” associated with trading. At any particular moment between trades, there is a bid-ask spread. The traders that accept the market price implicitly must pay half of the spread every time they make a trade because the best prediction of the market price is typically half way between the bid and the ask. This discourages trading and reduces the accuracy of the market, especially in thinly-traded stocks.
Automatic trading programs can be used to make a market have a lower spread at the cost of a direct subsidy. For example, the presidential prediction market could start out with each of the three securities with bids of $2.50 and asks of $3.00. If the security Democrat were sold for $2.50, the automatic bid and ask would then move to $2.00 and $2.50. In following this algorithm, the market sponsor can assure that it has a prediction within 5% at all times for a cost of at most $67.50 over the life of the market. For about ten times that, markets could be open at all times with a bid-ask spread of $0.05 on a $10 face value security or a 1% spread near $5.00. For $6,750 total, ten such propositions could be continuously tracked within 1%. (Note that this prediction is still subject to error, e.g., 55.5% ± 2%.)
For $100,000 in setup costs for a high-end system and $10,000 for something quick and dirty, such a market could be up and running quickly. $10,067.50 is about what it would cost to conduct a single poll. For 1,000 propositions daily for a year, polling would cost $3.65 billion or a bit less than the annual revenues of Nielsen Media Research’s parent company. One thousand five-year propositions can be tracked continuously on an infomart for less than $1.4 million a year. That’s more than 2,000 times as productive in terms of transaction costs and don’t forget the higher accuracy.
In addition to the very successful history of the Iowa Electronic Markets, the most interesting infomart proposal was for the Policy Analysis Market. Vice Admiral Poindexter and Dr. Robin Hansen are the names most closely associated with this DARPA effort. Both are Caltech graduate school graduates, Poindexter in 1961 in nuclear physics working with Nobel laureate Rudolf Mossbauer, and Hansen in economics 1998. When the proposal became tarred with the possibility of a security paying for an assassination, it was summarily laughed out of town. The Iran-Contra scandal did not end Vice Admiral Poindexter’s career in public service, but being pilloried for the Policy Analysis Market may well have. My bet is that we have not seen the last round of Hansen or infomarts. This economics technology is worth billions and the only barrier to adoption of the technology economy-wide is education.
|This economics technology is worth billions and the only barrier to adoption of the technology economy-wide is education.|
The argument against having an assassination market is logically flawed. It is true on its face that making a market for a proposition implicitly subsidizes the proposition. Having a proposition bet on whether a streaker will show up at a sporting event all but guarantees that one will show up. The streaker can take the bet and assure the outcome of the bet. That logic does not carry over to events that strip a perpetrator of their winnings. Wouldn’t it be nice to have financial records to track of everyone who bet that an assassination would happen? We might lock up terrorists for insider trading instead of Martha Stewart.
Whether or not such a market can ever come into being sponsored by government is a debatable question. The tragic history associated with the most famous attempt to bring an infomart online will make subsequent infomarts more difficult to implement. Private infomarts have a much more certain and strong future. Infomarts may be the future of polling and the number of topics that are tracked may explode into the thousands or the millions.
There are many potential funders for such a space infomart. The general benefits to the space economy would be high. This would make it a good investment for rich space buffs, an industry association, a government space promotion authority at the federal, state or local level, a spaceport, a service provider company, or a customer company.
So what do you think a security called “space infomart 2006” would trade for?