August 17, 2003
What Makes Financial Markets Work?
I've been thinking about this argument by Daniel Davies for nearly two weeks now. I've concluded that it is either a brilliant innovative analysis, or utterly barking mad. Unfortunately, I don't know which:
Posted by DeLong at August 17, 2003 05:19 PM
Crooked Timber: Hayekian markets reconsidered : A week late and a couple of dollars short, here are my thoughts on the now defunct Policy Analysis Market.... [N]obody seems to be at all clear on the details of what this market was meant to achieve (was it open to the general public? Only to specialists? Was it going to trade “assassination futures”? Or just derivatives on the EIU political stability indices?), let alone on its clearing arrangements, confidentiality clauses, etc.... However, I do want to comment on the fact that a number of bloggers analysed it in terms of Hayek’s concept of tacit knowledge and markets as information-creating social entities.... I'm not inclined to take seriously those critiques based on bubbles or based on supposed inefficiencies of market behaviour, at least not unless they have some explanation of why these are particular flaws of market behaviour, rather than general organisational pathologies of groups of homo sapiens. I am always in favour of people being made to put their money where their mouth is when they’re shouting off about questions of global importance... one of the potential functions that the PAM might have served would be as a means for low-ranking CIA officials to signal their genuine opinions, rather than those which were politically correct at the CIA. This is obviously not a first-best solution; first best would be a culture of honesty at the CIA. But a working market would be a benefit if it had the effect of creating a "back channel" of honest communication...
[A] wider critique of the unthinking application of "market solutions" to problems of this sort, on Hayekian grounds.... I don't believe that Hayek’s discussion of "tacit knowledge" is relevant to the question of a Policy Analysis Market. The defining characteristics of Hayekian tacit knowledge is that it’s practical, non-propositional and local in time and space. I actually think it’s something approaching a category-mistake to suppose that anyone could be in a position to have tacit knowledge relevant to the question "Is the chance greater than 22% that the government of Saudi Arabia face a coup attempt this year?". It’s a question which demands an answer in the form of a proposition, not like the question "Is orange juice for September delivery too dear at $5 8/16 a contract?", which demands an answer in the form of an action.
And this matters. Hayek's view of a market as a knowledge-creating entity is one which actually sits pretty uneasily with such things as the efficient markets theory, arbitrage pricing and other strands of thinking about markets and information which rely on reading off the closing prices and using them as if they were propositional information about something else. In the sense in which Hayek uses it, a market is an information processing system because it takes tacit knowledge as inputs and has the co-ordination of human activity as an output. The actual prices and volumes traded are epiphenomena; in general, they match up reasonably well to events in the real world (which is how the Soviets were able to use price data from Western markets as an essential input into the planning process), but that's not the point of a market, any more than it's the point of a kettle to increase the humidity of your kitchen.
Why does this matter? Well, it suggests that the prices struck in a market will be informative only if the market is well stocked with buyers and sellers operating on the basis of their own tacit knowledge. And this is not just an obscure point of Hayek scholarship; it’s actually written into the rules of the Chicago Mercantile Exchange.
The Merc, and several other commodities exchanges, makes a distinction between "hedgers" and "speculators". They do this for practical reasons; hedgers are allowed to run larger positions, because it is assumed that they are willing and able to take or make physical delivery of their contracts if necessary, and because they need to. But the distinction can also be made on sound grounds of Austrian economics. Consider the following sketch of a theory of the commodities market (to make it concrete, we’ll consider the grain futures market):
The market exists because of the hedgers; farmers are structural sellers of wheat contracts and bakers (etc …) are structural buyers. Both farmers and bakers are in the market because they want to fix their prices ahead of time in order to be able to make long-term plans about growing wheat or baking bread, and the futures market allows them to make these plans in the knowledge that they won’t be rendered unable to pay their debts because of sudden price movements in the spot market. Both the farmers and the bakers have plenty of (practical, unverbalised) tacit knowledge about grain, and the market price converts this tacit knowledge into a plan for co-ordinated action.
But experience has shown that it is pretty difficult to operate your market if you only have hedgers in it; in general, markets which don’t have speculators are illiquid. Speculators supply liquidity to the grain market, taking on the other side of the trades which the hedgers wish to make, in order that the market doesn’t have to wait until a hedger with an equal and opposite demand shows up. Speculators don't in general have tacit knowledge of the underlying security (in extreme cases, nor do they want to; the old proverb "the stock doesn't know you own it" comes to mind). Speculators assist the market’s functioning because they have tacit knowledge of the market; they have practical, nonpropositional information about the way in which liquidity is best provided to hedgers.
It follows from this that in order to be an efficient information-creating entity, a market has to have both hedgers or speculators. Although speculators are vital to the functioning of the market, you can’t have a market with nothing but speculators. And if you think about it, all the really successful "speculative" markets are ones in which the speculative activity clearly takes place in the context of a two-way market between hedgers. Commodities markets have structural demand from manufacturers and structural supply from primary producers. The stock market has structural demand (for stock) from people who want to save, and structural supply (of stock) from companies who want to raise money. The money market has structural demand from borrowers and structural supply from lenders.
There is nobody (to a reasonable first approximation) who has structural demand for more terrorism. The only people who have tacit knowledge of terrorists and would be considered to be on the long side of the market, are terrorists, who would presumably not be material participants. The problem is the same with respect to "weather derivatives", "catastrophe derivatives" and other such markets which (with due respect to the well-intentioned and often frighteningly intelligent people who try to make them work) have never taken off. The problem is that one side of the market has no tacit knowledge, and so the market does not really perform its function as an information processing entity.
I actually think that as a predictive tool over whether a market is going to work or not, this simple model works pretty well. Commodities exchanges - work well. Money markets - work reasonably well, but can be upset by the government coming in as a big player with no tacit knowledge. Stock market - can work just fine, but didn’t in the 1990s as the corporate sector became a net buyer of stocks, leaving no tacit knowledge on the short side. Betting spread "markets" - entirely speculative and notoriously subject to home town effects and long-odds effects. Hewlett-Packard's internal market for forecasting sales - works fine and actually quite a good way to resolve what would otherwise be an organisational problem between salesmen (systematically interested in talking projections down) and engineers (wanting to talk them up). And so on...
You say "utterly barking mad" like it's a BAD thing! ;)
Sounds like a reasonable critique to me. The Hayekian tacit knowledge idea, as far as I understand it, involves accurate aggregation of tons of little data points into a coherent top-level signal. But in the market for coca-cola, for example, everybody who operates in the coke market actually has some information to contribute -- its just that nobody is smart enough to gather and aggregegate that information as quickly or as accurately as a market.
A futures market on whether Leader X will be assassinated would have too few participants with any real data to simulate the effects of a real market.
Any way that's how it seems to me.
Yes, the spread betting market is indeed prone to get driven away by over-confident speculators (and if you are right-confident you wouldn't think you knew to what side it has drifted).
The grain-market on the other hand shold give farmers the possibility to drench the too bullish speculator in grain. Bakers have the same opportunity punish the overly bearish speculator.
Where is the barking madness in that?
This analysis seems limited to markets that are contingent on future outcomes. In today's economy, that's most markets (any that deal with uncertainty about future outcomes, including future values of assets), but markets can be just sellers and buyers meeting to set a price for something the sellers have produced and buyers wish to consume.
Excellent analysis. A uniquely wonderful synopsis of Hayek's ideas about knowledge producing markets. I will have to think some more before commenting in detail. However, a couple of fine points to correct:
- price and volume are emergent properties not epiphenomenon. It is a 'detailed point of scholarship' but much rests on the distinction.
- as I understood the design of the 'terrorism market' what was being traded is insurance against the risk of terrorism. I believe that there is tacit knowledge about the risk of terrorism. However, since much of that tacit knowledge would exist as classified information, there might exist a strong insider information liquidity problem.
Thanks for the link.
Although formalized fast-trading markets for disaster have their problems, there's always the insurance and reinsurnace industry as a counterexample (or then again, perhaps not).
Looking at tacit knowledge from the other side, another thing that distinguishes the markets you cite as anecdotal test cases is that they work best when there's a verifiable outcome that is (as far as feasible) outside the participants' ability to influence. (In the money markets, for example, the fed and the treasury have not merely tacit but explicit knowledge about their own intentions, and in the stock market the same goes for name players.)
At their best, it seems to me, terrorism futures would act like so many formerly leading indicators. Which would vitiate exactly the signal s one might be looking for.
I think there may be something to the idea that, to function well, markets need both hedgers and speculators. But DD's insistence that PAM would have had no hedgers -- that is, no people with tacit knowledge about the contracts being traded -- is simply wrong. His critique is focused on specific terrorist acts, and here it's plausible (though not certain) that tacit knowledge would have been absent. But as been said many times already, the vast majority of the contracts on PAM had nothing to do with specific terrorist acts. Instead, they had to do with broader questions about conditions (civil stability, economic health, military preparedness) in the Middle East. And these are precisely the kinds of phenomena that people have tacit knowledge of, because they are phenomena that emerge out of people's everyday experience.
If you want to know how likely it is that Jordan's economy will be stronger six months from now, or how stable the government will be, the people who have the best chance of answering that question are people who live and do business in Jordan. But these people would rather not talk to U.S. intelligence (even if CIA agents were interested in having a conversation about everyday life in Jordan), either because of patriotism or fear or whatever. The idea behind PAM was that it might be a way of getting those people to talk, as it were, by making it possible for Israeli engineers, Palestinian academics, and Jordanian businessmen to speculate (or, in DD's terms, hedge) precisely on the basis of their tacit knowledge.
DD suggests that no one has tacit knowledge relevant to the question "Is the chance greater than 22% that the government of Saudi Arabia face a coup this year?" In Ryszard Kapuscinski's book about the Iranian revolution, "Shah of Shahs," a man named Mahmud says, "And yet early in the summer [of 1978] I myself began to feel something changing, something reviving in people, something in the air. The atmosphere was indefinable, a little like the first glimmer of consciousness after a tormenting dream. . . . The first crack, the first little gap, appeared in the system." That is exactly the kind of tacit, local in time and space knowledge Hayek was talking about, and it was unquestionably relevant to the question of how likely it was that the Shah would be overthrown (as he was in early 1979). Of course, the knowledge was decidedly imperfect, but so is all the knowledge that individual market participants have. If aggregating imperfect knowledge produces good predictions of future orange-juice prices, it absolutely could have produced good predictions of civil unrest.
Please forgive my likely ignorance, but:
It would seem that the essential difference between the tacit knowledge of hedgers and speculators is that the former (on both sides) have a direct economic interest in rational risk pricing, whereas speculators do not. Speculators might make 10 or 50 percent on a PAM investment, but a hedger with a factory in Jordan that faces possible cessation of production or expropriation is going to balance gain and loss in either contingency so that he comes out ahead. Without a similarly-motivated hedge seller on the other side of the transaction, no rational price can be set, as such information is only discoverable after the fact.
If such things as terrorism risk are to be monetized, the price of derivatives based upon them should bear some relation to the value of the underlying "assets," no?
Just a small dsquared-type quibble. Wouldn't it be more accurate to talk about "Hayek's view of a market as a knowledge-creating entity" and Michael Polanyi's (we'll leave Wittgenstien on one side for the moment) concept of tacit knowledge.
“That is exactly the kind of tacit, local in time and space knowledge Hayek was talking about, and it was unquestionably relevant to the question of how likely it was that the Shah would be overthrown (as he was in early 1979). Of course, the knowledge was decidedly imperfect, but so is all the knowledge that individual market participants have.”
Yup, that's right. I must be really slow because the idea of insurance policies based on the possibility of violence erupting somewhere in the world seem entirely logical? What’s the bid deal? Aren’t a few people really irritated that the Bush administration came up with initial idea? Every single human being employs tacit information when making even a purchase of a candy bar? I could torture you to death---and you would still be unable to completely explain why you bought this particular piece of candy.
Insurance policies are written everyday on all sorts of possible disasters. In regards to terrorism, one would simply have to do a tremendous amount of reading and studying materials previously ignored. Moreover, was there ever a person who was born with the information to invest in pork bellies or corn futures? Of course not!
"Every single human being employs tacit information when making even a purchase of a candy bar?"
Sorry about that. There should be no question mark at the end of this sentence.
re James Surowiecki's example:
"And yet early in the summer [of 1978] I myself began to feel something changing, something reviving in people, something in the air." [...]
This could also be an example of post hoc confirmation bias. It's easy to "identify" reasons and causes after the fact. What did people write before event x happened and does the aggregate or the conjunction of feelings that there's "something in the air" add up to something reliable? What kind of analysis or algorithms will be employed? What kind of actions would it inform? (I don't know)
D-Squared is wrong to assume that only hedgers have tacit knowledge. There were no hedgers in the HP internal market but plenty of people with tacit knowledge that the market brought to light in its price forecasts. The same is true of the Iowa markets and the Hollywood Stock Exchange. D-Squared may be correct, however, that purely speculative markets have difficulty attracting liquidity. Indeed, the HP market was subsidized for this reason.
Although hedgers provide liquidity, precisely because they have a non-speculative reason for being in the market they can push prices away from risk-neutral probabilities. (See the debate over normal backwardization and contago effects in futures markets.)
Thus for forecasting it may be best NOT to have hedgers in the market but then you will have to subsidize the market in some way - perhaps directly as HP did or by making the process fun as does the Hollywood Stock Exchange.
...you can’t have a market with nothing but speculators. And if you think about it, all the really successful "speculative" markets are ones in which the speculative activity clearly takes place in the context of a two-way market between hedgers.
If I am following the logic of this post, there is no sports betting, since the natural longs and shorts (players, managers, owners) are barred from participating (although we may soon see a "Pete Rose exception").
That suggests two purely speculative propositions:
I bet there is active, speculative sports betting.
I also bet (with less assurance) that some eager student has tested the usefullness of sports lines in predicting final outcomes.
Yes, yes, but "hedgers" can control the outcome in a stablizing way, i.e. producing(consuming) more when prices are high(low). Sports bettors cannot, unless they pay a player to lose, which is neither stablizing nor constructive.
Isn't that the point of D^2's post, and could one really argue against it?
Yes, yes, but "hedgers" can control the outcome in a stabilizing way, i.e. producers can produce more when prices are high. "Sports bettors" can't do that, at least not in any fair and stabilizing way. Isn't that the simple difference that D^2's post is about?
Actually, I don't think DD's point was that the efficiency of a market depends on the ability of hedgers to control outcomes. Or, if it was, then he didn't get that idea from Hayek.
I think the word "hedgers" leads us astray. DD took the phrase from the Chicago Merc, but what he seems to mean (appropriately, I think) by
"hedgers" are people who have tacit knowledge about the underlying reality that market prices are meant to reflect. As he puts it, "the prices struck in a market will be informative only if the market is well stocked with buyers and sellers operating on the basis of their own tacit knowledge." Whether they're literally hedging or not doesn't seem all that relevant.
The mistake DD makes, I think, is assuming that there can be no tacit knowledge about things like the future of the Jordanian economy or the stability of the Palestinian Authority.