Dialog:
* Port:
o Prediction markets are crazy cool.
o If you had to design a method of prediction that resists being tested, this would be it.
o Prediction markets only work with groups of people. Yet, the group knowing the prediction could change the outcome.
o For all we know it predicts perfectly by destroying the perfection of its own prediction.
o It is like only being able to find gold by turning it into lead.
This is commentary from someone who knows next to nothing about prediction markets. Prediction markets don’t resist being tested. When they are tested, their proponents tend to take the rosy view, and more often than not, their conclusions are incorrect.
In some markets, the participants may have some (usually quite small) impact on the outcome. However, in the vast majority of the cases, no PM participant would have the power to influence the outcome in any signifcant way.
The last two points are gobbledegook. Meaningless statements.
Chris, maybe you should stick to legitimate reporting of prediction market issues.
I saw that. Maybe my sense of humour is off today!
Paul, that was filed into “humor”.
Paul,
Generally speaking, I will not “stick” to prediction markets, because posts about PMs are the least popular. People are not interested in PMs.
Assuming that the cartoonist refers to the presidential prediction markets, I can understand the perception.
Given the visibility of prediction markets during the presidential election, it may be the case that the PM numbers actually *influence* the voters and create a perception that the winner is established. Voters want to vote for the winner, so they vote for the leading candidate, so the PM results end up being a self-fulfilling prophecy.
Not a likely scenario, but not completely implausible either.
That is why I republished this cartoon. To spot what others are saying about prediction markets.
Panos said:
“Given the visibility of prediction markets during the presidential election, it may be the case that the PM numbers actually *influence* the voters and create a perception that the winner is established. Voters want to vote for the winner, so they vote for the leading candidate, so the PM results end up being a self-fulfilling prophecy. ”
But, don’t you think this type of PM indicates the participants have a complete lack of information about the outcome? They cease to have independent thoughts and opinions, too. Another first principle of PMs. If the basic principles (conditions) are not met for a prediction market, it is hardly fair to comment, as the cartoonist did.
Personally, I don’t think the cartoonists’s views were fair comment on the topic, and I am still trying to “get” the humour part!
Paul
Hi Paul,
“don’t you think this type of PM indicates the participants have a complete lack of information about the outcome”
No, the participants in the PM *do* have information about the outcome. The PM influence on the voters is also part of the equation. It is not the only factor, may be not even a major factor. But you cannot deny that it is part of the information available to the voters.
“They cease to have independent thoughts and opinions, too.”
I wanted for long to blog in detail about this, as this indeed puzzles me. Let me give below some incomplete thoughts.
When analyzing PMs, we assume that the traders are independent of each other. However this assumption is not true: All traders see the overall price, so in a sense they know what all the other participants think. So, there is no real independence.
Still PMs work. Therefore, independence does not seem to be a prerequisite for a PM to work. One reason is that, even if the traders can see the aggregate price, they still have the incentive to deviate from the crowd, in order to profit. So, it seems that (a) the incentives to be truthful and (b) the inability to influence the actual outcome are more important than independence.
The “inability to influence the actual outcome ” of course assumes that PM’s are not a major influence of the voting public. If PM’s become major factors in deciding the election, then they will degenerate into a self-fulfilling prophecy, aka “Keynesian beauty contest”
Hi Panos…
In my comment, I didn’t mean to imply that *all* participants had a “complete lack of information”, just some of them. The ones that delegate their opinion to the group’s consensus. Certainly, every participant has *some* information about the outcome, but that does not mean the information is good or useful.
Those with inferior information (and presumably, they would know this) would be more likely to act on information derived from the other participants (i.e. the current market prediction). Also, those who are more risk averse will tend to consider the market consensus more than those willing to take more risk.
I agree with you that independence is almost never met. I disagree with your comment that independence “does not seem to be a prerequisite for a PM to work”, because “PMs work”. Who says they “work”?
I fail to see a level of accuracy and consistency in PMs that would allow me to conclude that they “work” (based on the research so far). You may recall that when I looked at the HP prediction markets, they only “worked” 6 out of 8 times, the degree that they “worked” was very marginal, and in many of the markets the error was greater than 25%. There are numerous examples of public and enterprise PMs failing. So, I don’t think we should be assuming that PMs “work”. Maybe if there were greater independence, the markets would be more accurate and more consistently so.
I will note that there is very little research on the topic of participant independence (and there should be). When I first got started in prediction markets, one of the issues in my “bucket list” was to determine the effect on independence of providing information to the market participants. That is, if the information available to the participants is somehow controlled or filtered, what effect will this have on the prediction and independence?
Another issue related to independence arises when the participant pool is made up of members of a competing pool of forecasters. For example when EPMs are compared with internal corporate forecasts, the EPMs invariably use the same people for both tasks. How independent is that? I was also interested in pursuing the effect of the current aggregate price on participant trading behaviour.
I agree with your last comment, but think that it is a very rare phenomenon.