I read it again and I still don't understand. Maybe I should clarify my question.

I understand why price manipulators fail to make the market look the way they want, and instead make it more accurate.

What I don't understand is why insider traders, with access to true relevant information that the other traders cannot get merely by further research, aren't "neutralized" by the same mechanism.

That is: Assume a shallow consideration of a given question suggests a probability of 60%, more extensive research turns up data implying a more accurate 65%, and one insider has special data implying an even more accurate 64%. Further assume that the other traders besides the insider can't turn up the insider's special data with even extensive research -- the only way they can learn what the insider knows, if at all, is through the insider.

As I understand the article, if a manipulator tries to mess with the market, or if the insider tries to profit from the special data, in either case the market will settle on 65%.

What can the insider do to prove they're not a manipulator masquerading as an insider? If nothing, then how can the insider move the market to 64% when the manipulator can't move it to 66%?

Expand full comment

Read the post again. Ex post, it doesn't; ex ante, it does.

Expand full comment

If market forces overcome the effects of manipulative noise traders, why doesn't the same principle eliminate the effects of traders with special information about the event in question?

How does the market distinguish between a trader who's betting against market consensus because she knows something the other traders don't, versus a trader who's betting against market consensus because she wants the other traders to think she knows something they don't?

Expand full comment

Sorry - to finish my comment. "Cap and trade" is the modern equivalent of buying indulgences in the Church of Environmentalism.

Expand full comment

>whole point of pricing carbon is to create an incentive to economize

Bull. The cost of the fuel is incentive to economize on its use. There is no good evidence that atmospheric carbon dioxide is a negative externality, and thus NO REASON to "economize" on its production.

Expand full comment

I think many people don't understand prediction markets. Someone should make a YouTube video explaining this.

Expand full comment

This complaint is sad but predictable. Just as Robin says, it is the exact opposite of the truth. In my personal experience w/ prediction markets, I think one of their great weaknesses is thin markets due to adverse selection. Prediction markets work the best when there are many punters to make it worthwhile for the professionals - or at least some type of subsidy. People attempting to manipulate the markets are people subsidizing the markets, thus addressing one of the major problems!

Expand full comment


My beef is that in a response to a critique of decision markets in the context of a specific case ( global warming), you respond with a paper that makes assumptions that are not even close to the reality of the situation.

Also, I am quite familiar with academic modeling. I just have very little respect for the way it is done.

I am not against models in general. Nor am I against math, or making simplified assumptions, or for using approximations due to to time constraints. I work in software, and I often write documents to my co-workers about how I think a system should be engineered. I will use simplifications and approximations. As you say, an exact spec would be completely unwieldy. But these documents are nothing like an academic paper. Your goal is essentially an engineering goal - you are trying to examine whether a certain design will have the desired outcome. Thus I hold you to an engineering standard.

Here is how I would alter your paper to make it a convincing engineering design document:

a) First, start with the assumptions and explain and justify every non-obvious assumption you make. For instance, your assumption of "a commonly known strength of desire to manipulate" needs to be justified if your paper is to be at all convincing. Garbage in, garbage out, if your assumptions are not correct or not justified, nothing else in your model matters.

b) Second, solve for the average case and the worst case, not the best case. For a model, you do not need exact numbers. Simplifications are fine. But you should attempt to use ball park figures for the average case. An "unlimited budget" is not an average case. Pick a typical securities market, and use that as a ballpark guess. Proving that your model works in a best case is a useless exercise.

c) Third, it's much more important to have ball park figures for all the assumptions, than in depth analysis of one aspect of the system. You do the opposite. You painstakingly write out the math, using the most advanced techniques possible, for a very narrow set of assumptions. Unfortunately, because the assumptions are so narrow, the paper is not at all convincing for any real world scenario. If you are time constrained (as we all are) It's almost always better to do back of the envelope math for a very broad set of assumptions. Then if there is one particular area where precise math is needed to make an accurate judgment, you can add more advanced math as needed (although most of the time it turns out the back of the envelope estimates are good enough, your paper would be just as convincing if you simply cut out pages 4 through 11).

Your paper passes muster in academia. But if your communicating on a blog frequented by engineers, you'll need to do a lot better. Come up with a simplified model that actually makes reasonable estimations of budgets, noise in determining the asset value, knowledge about bias, etc.. Also add in a reasonable proposal for how the winner of the bet gets decided. You don't need to be "exactly realistic". Just get it in the right ballpark. As it is, the assumptions in your paper are so unrealistic as to make any further discussion impossible.

Expand full comment

You are clearly not familiar with the practice of academic modeling. Models are not expected to be exactly realistic to be persuasive; an exactly realistic model would be completely unwieldy. Issues of how to define and measure global warming are pretty independent of the issue of manipulation by traders; every post can't deal with all issues.

Expand full comment

From Robin Hanson's paper:

Of course the fact that we have a particular model illustrating these results hardly implies that these results always hold in every context. Our model assumes risk-neutrality, normally distributed values and signal errors, interior choices of information quantity, only quantal-response-type irrationality, no meta-signals about the signals of other agents, no transaction costs of trading, no budget constraints, and a single rational manipulator with quadratic manipulation preferences and a commonly known strength of desire to manipulate. However convenient these assumptions may have been for solving the model, one can reasonably question the empirical relevance of models based on them.

No budget constraints? Commonly known strength of desire to manipulate?

the theory makes some assumptions and our lab and field data do not cover every possible contingency

No, it doesn't cover the contingencies that are actually likely to exist in the real world. It boggles my mind that you keep citing this paper, when this paper very clearly makes so many dubious assumptions as to be useless for actually showing that decision markets would work in the real world.

Furthermore, your paper glosses over two points which are vitally important: How does the question get defined? How is the outcome measured?

If the measurement of the outcome has a lot of noise, then the impact of the manipulators becomes even more pronounced.

If you actually care about convincing people, you should work through a real world example of how decision markets might work. Use global warming as an example. What is the question/policy posed to the market? How would the result be measured? We can then work through the logic of how actual, real world, players would respond to the market.

Expand full comment

Reply to Peter,

Well, if there is free rider, and if I am worse off with others joining me, then it does alter incentives on one side of the prediction market, doesn't it. And, don't forget, as a consumer and a shareholder of Exxon, I pay for that battle on that side too. Sounds like if you believe global warming is caused by CO2 emissions, you pay a higher cost to bid your beliefs than if you do not. Do not like to play weighing games with someone who puts their thumb on the scale. Do you?

Expand full comment

No one's saying you can't do that if you want.

Are you trying to imply that it's a bad thing that people have individual incentives to fight the API? That there should be a free rider problem for people to gloriously overcome?

Expand full comment

Reply to Peter Twieg,

Oh, great, what you're saying is that I have no incentive to get others to join me to oppose global warming by bidding with me in the prediction market. I don't have to share the pot with them. Sad.

No wonder economics is called the dismal science.

Expand full comment

You only need a 51% advantage to take a bet. If you want a 30 year bet and can only get a 10 year bet you plan to bet a certain amount each year.

Expand full comment

But, although others bid with me, there is a free rider problem…some people do not bid or contribute, figuring, well, hell, if Bill is willing to bid, I won’t have to.

What? You stand to receive more money because Bill doesn't bid. The benefits of your bidding are almost wholly internalized.

Expand full comment

How much capital do the large players have relative to the rest of the market? If other participants think large players are distorting the market rather than betting accurately, they will be willing to put up more of their capital (they think they are selling a dollar for more than a dollar). If there are no significant barriers to entry, the potential market size is huge.

Expand full comment