There is no way to predict the price of stocks and bonds over the next few days or weeks. But it is quite possible to foresee the broad course of these prices over longer periods, such as the next three to five years. These findings, which might seem both surprising and contradictory, were made and analyzed by this year’s [Economics Nobel] Laureates, Eugene Fama, Lars Peter Hansen and Robert Shiller. (
"The idea is that when the information isn't feasible to compute, then markets come into their own. The market accomplishes an aggregation no individual computing agent could--exactly because the information is so complex and dispersed."
That only works for existing financial markets because the participants each know a different piece of the puzzle: their personal preferred price and personal flexibility (the market solves a communication problem), plus the market is a self-fulfilling prophecy machine. With prediction markets you lose those properties (the event can completely ignore the prediction market and rather than having unique and sufficient pieces of the puzzle the participants all hold similar or even the same piece, because they'll rely on a handful of complicated calculation models, and even if they had unique pieces there would be no guarantee they were sufficient), the dynamics are just fundamentally different. I don't quite know why Robin Hanson doesn't see that.
"If it isn't available to anyone, then prediction markets (if they correctly combine the available information) will do better than anyother method of prediction."
The least worst option can still be so bad it's not worth it (and this is to be expected for most interesting events you'd want to predict) and as you say these prediction markets come with a cost.
No, reliability (or accuracy if you prefer) may still be very low because the information required for good predictions is not available to the investors (inside information is required or it's really complicated or not even feasible to compute).
The idea is that when the information isn't feasible to compute, then markets come into their own. The market accomplishes an aggregation no individual computing agent could--exactly because the information is so complex and dispersed.
If there's information that's not available to investors, then either it's available to other people or not. Only in the former case is it a potential argument against prediction markets. If it isn't available to anyone, then prediction markets (if they correctly combine the available information) will do better than any other method of prediction. And that's the only relevant question (besides how much do the methods cost to implement, which is the big question that Robin tends to avoid).
[Added.] But there's one important reason prediction markets make information less available than it would be without prediction markets. This does require consideration I haven't seen it given: Prediction markets will foster hoarding information that would otherwise be public. (Thus, there is conflict between prediction markets and public argument and discussion.)
Massage the price away from a truthful forecast is quite expensive (given that we don't live in a world where expected GDP growth has a very strong causal effect on realized GDP). I would worry more on massaging taking place on the measurement side.
So it may very well be that a prediction market is no more reliable than tossing a coin or if it is more reliable it would still come with such a low confidence level that it does not yield practical advice besides a long term average, except that interesting events are not common enough to repeat a large number of times before real world circumstances have changed (you might get a prediction that under current real world circumstances the US Democratic party has a 55% probability of winning the next presidential election, but the usefulness of that prediction is almost zero since there are 4 years between US presidential elections and real world circumstances will have changed a lot even 5 elections from now and 5 or less repetitions aren't really statistically significant at 55-45 odds.)
"If prices (per Fama) incorporate all the information available, then they do the most accurate job possible in predicting from the available information"
Yes.
"prima facie, supporting prediction markets."
No, reliability (or accuracy if you prefer) may still be very low because the information required for good predictions is not available to the investors (inside information is required or it's really complicated or not even feasible to compute). Also Shiller's findings should not be ignored, especially in smaller markets (small compared to the largest stock exchanges) and with timescales of less than multiple decades.
Reliability isn't the right concept here; it's accuracy. If prices (per Fama) incorporate all the information available, then they do the most accurate job possible in predicting from the available information--prima facie, supporting prediction markets.
No, it's a lot more complicated than that. Peter Hansen made techniques that measure the performance of models, so he's already out of the pircure. Fama said market prices behave according to random walk and Shiller discovered a pattern in some deviations from this random walk (bubbles, investor panic, billionaires betting against the market, etc...) What this means is that Fama believes all the information available to investors in already incorporated into the market price, but it may very well be that all the information available to the investors is not sufficient to reliably predict a future event that is primarily driven by non-market forces (such as a political election). Shiller believes investors, as a group, will not act entirely rational upon the information that is available to them and that just adds to the problem, especially since the dynamics of primarily non-market driven events are often very different from that of market driven events (for example a political hype can actually help a political candidate indefinitely, whereas a market bubble always eventually ends in a bust).
If Fama and Shiller believe that information available to investors is not enough to reliably predict primarily non-market driven events then they will not support futarchy.
Upon further thought, an alternative to customer-as-market-maker might be for the customer to stake a bunch of pundits to trade with each other instead of pontificating about their views. Winning pundits would get a share of their profits, and losing pundits' losses would be borne by customer. That might provide enough non-informed participants willing to trade (because they're betting with someone else's capital) to attract informed participants, even outsiders betting their own capital. The profit opportunity for the informed participants would still be limited by customer's willingness to lose money in trading. However, at least the pundits would be incentivized to be "right" about their expressed views.
Doug's point about attracting liquidity is a good one, and I'm not sure Robin's answer completely addresses it.
It seems to me, that most of the deep and liquid (and hence most likely to be informationally efficient) financial markets attract participants with non-informational motivations. For example, passive stock investors participate in equity markets to gain equity exposure ("beta"). Commodities producers and consumers may participate in commodities futures markets to hedge risk, as Robin points out. Even sports gambling markets provide entertainment value, which may attract participants that don't actually believe that they possess superior information.
The presence of non-informed participants may be helpful or even essential in attracting both market-makers to provide liquidity and informed participants. Market makers may believe, that while there are some participants with more information than themselves, there are enough non-informed participants that the market makers can earn enough net profit from the bid-ask spread to compensate for the losses resulting from trading with informed traders. The presence of non-informed traders may help convince informed traders that the information they think they possess truly has not yet been incorporated into market prices. In short, the presence of non-informed traders may help market makers and informed traders answer the question, "Why would I want to trade with someone that is actually willing to trade with me?"
Markets (financial and non-financial) exist to allow trading between parties with *different* needs, preferences, and characteristics. In a pure information market, what asymmetry between market participants would induce trading? Without such asymmetry, it seems like we would have to rely on participants' overconfidence in their own information. What are the most liquid pure information markets that have been established to date?
If an information market cannot attract non-informed participants, then the subsidy that Robin's hypothetical customer would have to pay might be quite higher than the cost of setting up an exchange. The customer might need to act as a market-maker, providing liquidity only to informed traders, which could result in substantial trading losses, much higher than the cost of media, pundits, etc. I agree, though, that in many cases those alternatives are established for reasons other than information gathering.
The fundamental problem with prediction markets is that the vast majority of people don't vaguely understand uncertainty at all, and the very small select few don't understand it deeply at all.
So prediction markets provide a solution to a problem that in the minds of many doesn't exist.
That is why they are not used or respected. Bottom line. End of story.
I can only speak to Prof. Schiller, who was one of the very first supporters of HedgeStreet back in the day and in fact wrote a letter to regulators in support of the Company some ten years ago. He is deeply knowledgeable of the structure of markets, and their information aggregation properties, so he definitely knows about info aggregation, but for a variety of reasons chooses not to fight this particular cause. Not sure why, can't really speculate, but in his case it is definitely a case of being aware of these properties and for some reason no deciding to focus on advocating them.
Note that the award went primarily (in so far as this can be said about a shared award) to Shiller for statistically demonstrating the strong limits of the efficient markets hypothesis. Fama got the award for his contributions in the field (basically creating it), but Shiller got the award for being the one in the field who was correct.Efficient markets are not generally that maligned, they are extremal popular among Thomas Friedman level columnists who like to think they are smarter than subject matter experts.
Consider that the non legality of prediction markets as evidence for or against the idea that those who set legal regimes think accurate information would be to their benefit or detriment. Note that if lobbies wanted these markets to exist they would.
"The idea is that when the information isn't feasible to compute, then markets come into their own. The market accomplishes an aggregation no individual computing agent could--exactly because the information is so complex and dispersed."
That only works for existing financial markets because the participants each know a different piece of the puzzle: their personal preferred price and personal flexibility (the market solves a communication problem), plus the market is a self-fulfilling prophecy machine. With prediction markets you lose those properties (the event can completely ignore the prediction market and rather than having unique and sufficient pieces of the puzzle the participants all hold similar or even the same piece, because they'll rely on a handful of complicated calculation models, and even if they had unique pieces there would be no guarantee they were sufficient), the dynamics are just fundamentally different. I don't quite know why Robin Hanson doesn't see that.
"If it isn't available to anyone, then prediction markets (if they correctly combine the available information) will do better than anyother method of prediction."
The least worst option can still be so bad it's not worth it (and this is to be expected for most interesting events you'd want to predict) and as you say these prediction markets come with a cost.
No, reliability (or accuracy if you prefer) may still be very low because the information required for good predictions is not available to the investors (inside information is required or it's really complicated or not even feasible to compute).
The idea is that when the information isn't feasible to compute, then markets come into their own. The market accomplishes an aggregation no individual computing agent could--exactly because the information is so complex and dispersed.
If there's information that's not available to investors, then either it's available to other people or not. Only in the former case is it a potential argument against prediction markets. If it isn't available to anyone, then prediction markets (if they correctly combine the available information) will do better than any other method of prediction. And that's the only relevant question (besides how much do the methods cost to implement, which is the big question that Robin tends to avoid).
[Added.] But there's one important reason prediction markets make information less available than it would be without prediction markets. This does require consideration I haven't seen it given: Prediction markets will foster hoarding information that would otherwise be public. (Thus, there is conflict between prediction markets and public argument and discussion.)
Massage the price away from a truthful forecast is quite expensive (given that we don't live in a world where expected GDP growth has a very strong causal effect on realized GDP). I would worry more on massaging taking place on the measurement side.
So it may very well be that a prediction market is no more reliable than tossing a coin or if it is more reliable it would still come with such a low confidence level that it does not yield practical advice besides a long term average, except that interesting events are not common enough to repeat a large number of times before real world circumstances have changed (you might get a prediction that under current real world circumstances the US Democratic party has a 55% probability of winning the next presidential election, but the usefulness of that prediction is almost zero since there are 4 years between US presidential elections and real world circumstances will have changed a lot even 5 elections from now and 5 or less repetitions aren't really statistically significant at 55-45 odds.)
"If prices (per Fama) incorporate all the information available, then they do the most accurate job possible in predicting from the available information"
Yes.
"prima facie, supporting prediction markets."
No, reliability (or accuracy if you prefer) may still be very low because the information required for good predictions is not available to the investors (inside information is required or it's really complicated or not even feasible to compute). Also Shiller's findings should not be ignored, especially in smaller markets (small compared to the largest stock exchanges) and with timescales of less than multiple decades.
Reliability isn't the right concept here; it's accuracy. If prices (per Fama) incorporate all the information available, then they do the most accurate job possible in predicting from the available information--prima facie, supporting prediction markets.
No, it's a lot more complicated than that. Peter Hansen made techniques that measure the performance of models, so he's already out of the pircure. Fama said market prices behave according to random walk and Shiller discovered a pattern in some deviations from this random walk (bubbles, investor panic, billionaires betting against the market, etc...) What this means is that Fama believes all the information available to investors in already incorporated into the market price, but it may very well be that all the information available to the investors is not sufficient to reliably predict a future event that is primarily driven by non-market forces (such as a political election). Shiller believes investors, as a group, will not act entirely rational upon the information that is available to them and that just adds to the problem, especially since the dynamics of primarily non-market driven events are often very different from that of market driven events (for example a political hype can actually help a political candidate indefinitely, whereas a market bubble always eventually ends in a bust).
If Fama and Shiller believe that information available to investors is not enough to reliably predict primarily non-market driven events then they will not support futarchy.
If this summary is accurate, Prof. Fama would be the one more likely to support prediction markets:
http://www.bbc.co.uk/news/m...
Upon further thought, an alternative to customer-as-market-maker might be for the customer to stake a bunch of pundits to trade with each other instead of pontificating about their views. Winning pundits would get a share of their profits, and losing pundits' losses would be borne by customer. That might provide enough non-informed participants willing to trade (because they're betting with someone else's capital) to attract informed participants, even outsiders betting their own capital. The profit opportunity for the informed participants would still be limited by customer's willingness to lose money in trading. However, at least the pundits would be incentivized to be "right" about their expressed views.
Doug's point about attracting liquidity is a good one, and I'm not sure Robin's answer completely addresses it.
It seems to me, that most of the deep and liquid (and hence most likely to be informationally efficient) financial markets attract participants with non-informational motivations. For example, passive stock investors participate in equity markets to gain equity exposure ("beta"). Commodities producers and consumers may participate in commodities futures markets to hedge risk, as Robin points out. Even sports gambling markets provide entertainment value, which may attract participants that don't actually believe that they possess superior information.
The presence of non-informed participants may be helpful or even essential in attracting both market-makers to provide liquidity and informed participants. Market makers may believe, that while there are some participants with more information than themselves, there are enough non-informed participants that the market makers can earn enough net profit from the bid-ask spread to compensate for the losses resulting from trading with informed traders. The presence of non-informed traders may help convince informed traders that the information they think they possess truly has not yet been incorporated into market prices. In short, the presence of non-informed traders may help market makers and informed traders answer the question, "Why would I want to trade with someone that is actually willing to trade with me?"
Markets (financial and non-financial) exist to allow trading between parties with *different* needs, preferences, and characteristics. In a pure information market, what asymmetry between market participants would induce trading? Without such asymmetry, it seems like we would have to rely on participants' overconfidence in their own information. What are the most liquid pure information markets that have been established to date?
If an information market cannot attract non-informed participants, then the subsidy that Robin's hypothetical customer would have to pay might be quite higher than the cost of setting up an exchange. The customer might need to act as a market-maker, providing liquidity only to informed traders, which could result in substantial trading losses, much higher than the cost of media, pundits, etc. I agree, though, that in many cases those alternatives are established for reasons other than information gathering.
The fundamental problem with prediction markets is that the vast majority of people don't vaguely understand uncertainty at all, and the very small select few don't understand it deeply at all.
So prediction markets provide a solution to a problem that in the minds of many doesn't exist.
That is why they are not used or respected. Bottom line. End of story.
I can only speak to Prof. Schiller, who was one of the very first supporters of HedgeStreet back in the day and in fact wrote a letter to regulators in support of the Company some ten years ago. He is deeply knowledgeable of the structure of markets, and their information aggregation properties, so he definitely knows about info aggregation, but for a variety of reasons chooses not to fight this particular cause. Not sure why, can't really speculate, but in his case it is definitely a case of being aware of these properties and for some reason no deciding to focus on advocating them.
Note that the award went primarily (in so far as this can be said about a shared award) to Shiller for statistically demonstrating the strong limits of the efficient markets hypothesis. Fama got the award for his contributions in the field (basically creating it), but Shiller got the award for being the one in the field who was correct.Efficient markets are not generally that maligned, they are extremal popular among Thomas Friedman level columnists who like to think they are smarter than subject matter experts.
Consider that the non legality of prediction markets as evidence for or against the idea that those who set legal regimes think accurate information would be to their benefit or detriment. Note that if lobbies wanted these markets to exist they would.
good idea, except for requiring it be linked with a google account.
The era of arrogant journalism is long gone, the rise of blogging has emerged: http://www.21stcenturynews....