Can Markets Predict Elections?
Even if markets can somehow better anticipate the outcome, writes B.K. Marcus, prediction markets will not achieve their full potential until they incorporate the power of profit-seeking self-interest, which can only be found in the private world of risk and reward. Government-developed "planned markets" such as the Pentagon's PAM, "virtual markets" such as the Hollywood Stock Exchange, and low-risk hampered markets such as the IEM (which only allows $500 trading accounts) will not be able to operate as efficiently or accurately as would true capitalist markets, which allow for real profit and real loss. [Full article]
Comments (9)
When I read the article, it made me to think about how the businesses can take advantage of the information of "so-called stock markets" of terrorism and election.
Let play a game. Let say business X is specialized in defence weapon production. As economic nature of humans dictates it, business X will seek out information that will enhance the profit of itself. The stock markets on possbility of specific terrorist action and outcome of the election exist and open for trading. The business X can use the information from both stock markets to predict the possible increase in demand for weapons to response to terrorist actions, to determine if it should produce more weapons to be prepare for the demand, or to persuade the government to buy weapons to protect itself from potential forecasted terrorist action. Also the market on president election can assist the business X to determine which candidate will win and to persuade that candidate to be favorable to the business X's interest and assists the business X to shape the policies to response the possible candidate's policies and agenda.
I know I am speaking not like a Libertarian but I am giving an example of how those two types of markets can really assist the businesses.
Published: August 4, 2004 9:19 AM
This is a very promising first outing (by Marcus). I don't watch election-night coverage on TV, but I've heard that outlets for such material give much emphasis to "their" predictions of the outcome, which I suppose rely on more-or-less-conventional polling techniques.
I wonder if any of these outlets may this year avail themselves of any of these superior "market-based" predictive devices? And if they do, what sort of disclosure/billing will they give this fact in presenting their predictions?
Almost makes me want to tune in! But like the government, which television and so much else is coming more and more to resemble, they seem most likely to just keep doing it the way they did it last time.
So, just like last time, I won't watch.
Published: August 4, 2004 12:26 PM
Bkmarcus runs that Blackcrayon site, right? Its a good site
Published: August 4, 2004 12:30 PM
As a founder of the "play-money" prediction market NewsFutures.com, I am of course keenly interested in the issue of prediction accuracy, and in particular in how it compares to that of "real-money" markets. Marcus' article provides illuminating historical context for this issue.
He argues that markets based on real money are likely to yield better predictions than play-money markets. It's the classic hunch that markets work because participants must put their money where their mouth is. As an example, he points to the significant price difference on NewsFutures (play-money) and Tradesports (real-money) regarding the Bush re-election contracts, and wishes for an experiment that would help decide which is more accurate...
It turns out that we recently performed just such experiment. Together with economists Justin Wolfers (of the Wharton School, University of Pennsylvania) and David Pennock (Yahoo! Research Laboratories), we systematically compared the Football game predictions of NewsFutures and Tradesports throughout the entire 2003-2004 NFL season. Overall, both markets were highly accurate, especially compared to individual "experts". But we found no systematic accuracy advantage whatsoever for real-money predictions over play-money predictions. In fact NewsFutures just edged out Tradesports in a prediction competition. This research will shortly be published in the forthcoming issue of Electronic Markets, 14-3 (Fall 2004), in an article entitled: "Prediction Markets: Does Money Matter?"
One important thing to remember, and which Marcus fails to mention, is that most play-money prediction markets, like NewsFutures and the Hollywood Stock Exchange, do offer some significant prizes to the most successful traders in addition to substantial non-monetary rewards like peer-recognition and self-satisfaction. They only eliminate the risk of participating, not the rewards.
Our data suggests that the reason prediction markets work so well is not so much because participants put their money where their mouth is, but rather because the better informed will come out on top.
This fits well with the hypothesis (familiar to entrepreneurs) that human activity is in general significatnly more driven by the pursuit of rewards than by the avoidance of risk.
Published: August 5, 2004 6:26 AM
Responding to Feedback
In the day and half since the publication of "Can Markets Predict Elections?" I have received some very supportive email as well as questions and challenges in both email and the Mises blog.
Rather than answering in several messages and both media, I'll try to reply to everyone here.
The Virtues of Virtual Markets
One reader appreciated my summary history of the calculation debate and the distinction between Hayek's thesis and Mises's, but said he has a hard time believing "that an exchange in election futures will reflect back anything more than the average of peoples' hunches."
It may well be true that these markets allow people to trade on their hunches, but the resulting price will not be an average of everyone's guesses; rather it will be the consensus of best guesses. As Emile Servan-Schreiber says in his blog comment, "the reason prediction markets work so well is ... because the better informed will come out on top." The market provides not an averaging mechanism but a quality selection mechanism.
This may still come down to certain people's hunches -- and predicting the future can only ever be a form of guessing -- but what we have are the most informed hunches and the best educated guesses, selected for by the market itself. This is true whether we're dealing with the Hayekian knowledge markets or the Misesian markets of property, profit and loss.
The Virtues of Real Markets
Another reader asked if the difference between the Hayekian and Misesian theses reflects "the difference between the theoretician and the practitioner? ... whereas both want to be correct, the former is primarily guided by his beliefs, while the latter is governed by an intuition ... driven by outcomes?"
I think there is something to this, but I also think it's possible for a good theoretician to know his subject better than many practitioners. What the market selects for is two-fold:
(1) not just our best guesses -- either from theory or practice -- but our level of confidence in our best guesses (and this is where I think real risk will prove to be an inextricable part of the mechanism: confidence is tested by risk-tolerance);
(2) not just our best guesses but our best guessers! In a profit-and-loss market, the best guessers will, over time, increase their influence over prices. Consistent profits will reward the best bidders and consistent losses will weed out the worst.
While I'm confident that the prizes the virtual sites offer to their winners increase the incentives that draw their best participants -- and no doubt contribute to the impressive track record these virtual markets have already shown -- the effect of such prizes on future iterations of the market will be smaller than would be the real profits one could reinvest in future predictions.
What Do Statistics Show Us?
The NewsFutures study that Mr. Servan-Schreiber brings to our attention does tell us something, but what exactly does it tell us?
The statistical challenge to the Misesian thesis (or rather, to my interpretation of that thesis and its application to the question of prediction) offers an opportunity to apply the Austrian School's praxeological insight: it is not that our understanding of economic law is a hypothesis to be tested by empirical study; rather it is economic law that helps us understand and interpret the results of such studies. For instance, if the price of a good is raised, and we find that demand for that good increases, this cannot be a disproof of the law of supply and demand, but rather a reason for those who understand the law to look for other contributing or distorting factors.
If it's true that the track records of NewsFutures and TradeSports are statistically similar, that doesn't refute the claim that the pursuit of real profits will increase the quality of prediction. What it shows us is that the profit-and-loss market isn't yet big enough to draw the attention of the biggest profit-seekers. People are making real money at TradeSports.com, but I doubt there are many futures traders who can make a living there yet. And I'm even more skeptical that anyone has made a fortune in their exchange (with the worthy exception of the entrepreneurs who created the site in the first place).
As I stressed in my original article, (a) an incentive for real research is likely to contribute to the quality of bids for contingent contracts, and (b) arbitrage is well-known to improve the efficiency of real markets (where efficiency refers to the dissemination of information and the reflection of that information in prices) and only the pursuit of real profit will draw the attention of arbitrageurs.
When I was first contemplating this subject in the spring, I asked a professional arbitrageur to comment on the discrepancy between prices at NewsFutures and TradeSports. He said that NewsFutures wasn't a real market as far as he was concerned, and that TradeSports "probably isn't liquid enough to put together a good arb."
Like everyone else, arbitrageurs face opportunity costs. Until prediction markets can offer big enough profits to bring them into the game, they'll continue to focus their attention on the opportunities that still exist in traditional capitalist markets.
That doesn't keep the current prediction markets from being the best game in town.
BlackCrayon.com
Finally, thanks to Walt Byars for recognizing my name and commenting favorably on the BlackCrayon website. BlackCrayon.com is about philosophical anarchism and the ethics of free markets and property. BlackCrayon led me to Rothbard who led me to Mises.
B.K. Marcus
Charlottesville, Virginia
5 August 2004
Published: August 6, 2004 12:36 AM
What I think is missing from all this is a way to account for the expected harm or benefit from the election outcome itself. If a rational agent were to participate in such a market, he would maximize his expected return which is = (odds of candidate A winning * financial harm expected if candidate A wins) + (odds of candidate B wining * financial harm expected if candidate B wins). This means two things. First, the candidate who is expected to be more of a financial burden, if elected, will be bid on more than would be expected if participants in the market would otherwise be unaffected by by the election outcome. This also means that the price of election futures is more like a market clearing price in the insurance market than the price of a horse race ticket. If, for example, I buy fire insurance on my home (essentially, a bet that there will be a fire) and I'm willing to pay $100/month, that isn't enough to determine how likely I perceive a fire to be. You'd also need to know how big the payout is (possible with event futures) and how much I expect the outcome of a fire (or an election) to cost me. In the case of event futures, you'd need to know what users expect to be their loss or gain from the event aside from profit or loss of the trade. I'm not certain and definitely not confident that one could obtain this piece of data without resorting to a conventional poll, reintroducing all of the problems that Marcus points out.
Published: August 8, 2004 9:32 PM
hi!
with regards to this article "Can Markets Predict Elections?", I don't quite understand the parts on arbitrage opportunities.
My question is whether there is arbitrage opportunity in the IEM winner-takes-all market? If so, would these arbitrage opportunites affect the accuracy of this winner-takes-all market in predicting the presidential election results?
Published: October 15, 2004 6:42 AM
I am impressed with this page...setup really nice. Doesn't take forever to load pics, like mine... Very impressive.. http://lincoln.ziarc.com
Published: January 31, 2006 11:19 AM
I am impressed with this page...setup really nice. Doesn't take forever to load pics, like mine... Very impressive.. http://lincoln.ziarc.com
Published: January 31, 2006 11:44 AM