Tacit knowledge

by Henry on July 29, 2003

There’s a lot of buzz in the blogosphere about a DARPA project which aims to predict terrorist attacks, assassinations and coups, through creating a futures market, in which traders can speculate on the possibility of attacks; the NYT picks up on it too. Most of the commentary is negative, but Josh Chafetz likes the idea, and invokes Hayek.

As I explained at length in a post on Hayek last year, complex systems function by finding ways to aggregate diffuse knowledge into simple indices, which then allow actors in the system to take advantage of knowledge that they don’t actually have (e.g., no one knows exactly what Americans’ breakfast cereal preference orderings are, but by watching the information-aggregating index that we call “price,” producers can generally ensure that, when you go to the supermarket, you’ll find the brand you want. Compare that to the shortages of some items and overproduction of others that centrally planned economies have produced). A futures market in terrorist attacks, while it sounds grisly, may help us to aggregate diffuse knowledge in a way that will prove superior to expert knowledge.

Seems to me though that Chafetz is wrong. As Chafetz suggests, Hayek makes some rather interesting arguments about the ability of markets to pick up on diffuse, tacit knowledge, and make it usable. And Hayek’s not the only one saying this; Michael Polanyi and GLS Shackle develop roughly similar ideas. But the key point is that Hayekian markets aggregate knowledge. They don’t create it. People tend to be tolerably well informed about their own tastes, and buying habits. Markets will do a good job of taking this diffuse knowledge and communicating it to producers. The general public is likely to be rather less well informed about the likelihood of coups, assassinations and general alarums, and thus the sum total of their tacit knowledge is likely to be an incoherent mess, or a product of shared cognitive biases, rather than a useful index of information. And indeed, DARPA’s “market” is aimed at the general public; it seems that random punters can sign up to participate on a first-come first-served basis. Whatever minimal amount of useful information is in there will almost certainly be drowned out by the noise.

This isn’t to say that information markets of this sort can’t be useful – but they need to involve people who have useful tacit knowledge to begin with. One of the problems with hierarchy is that valuable information sometimes doesn’t make it from the bottom of the organization to the top, because middle management blocks it, or because the boss doesn’t want to know. Anonymous information markets can potentially solve this problem. They might allow the people at the bottom of the ladder, who often have the best sense of what is actually going on, to share their information anonymously. Assuming that their decisions to buy and sell are kept confidential, management can’t punish them for not sticking to the corporate line. For example, one could create an information market that would allow anonymous CIA analysts to express their skepticism about Iraqi WMDs by shorting WMD “stocks” without fearing reprisal from on high. This would actually be a rather useful exercise. I wonder why DARPA isn’t funding it?



pathos 07.29.03 at 5:58 am

Talk about quick response time! I just requested a discussion of this issue in the “yuan” thread.

Henry wrote:

“The general public is likely to be rather less well informed about the likelihood of coups [than breakfast cereals], assassinations and general alarums, and thus the sum total of their tacit knowledge is likely to be an incoherent mess, or a product of shared cognitive biases, rather than a useful index of information.”

The question, though, isn’t whether the general public is less well informed about coups than about breakfast cereals. It is whether than are less well informed than the current crop of experts, politicians, think tanks, and Middle Eastern Studies departments. The answer to THAT question — considering the “incoherent mess” produced by the above groups — is not obviously, “No.”

Assumedly, someone like me who knows nothing about the chances of a coup in Turkey will not be buying futures contracts. Only those who think they will profit will do so (those guessing randomly should split down the middle, or at least show a consistent bias from which sudden swings can be monitored).

Particularly unconvincing were the critics arguments:

“This appears to encourage terrorists to participate, either to profit from their terrorist activities or to bet against them in order to mislead U.S. intelligence authorities,”

First, assumedly someone would be monitoring whether a large purchase of a “bomb Israel” contract was coming in from Saudi Arabia (or anywhere else, for that matter). The same concern would of course be raised if, out of the blue, a large investment was made on, “Nope. No one’s going to bomb Israel this week.”

Unlike stocks, where buying on insider knowledge is a crime, here buying on insider knowledge in the whole point: the purpose of the exercise (from the gov’t’s point of view) is not to insure a fair investing environment but to get the “price” exactly right.


pathos 07.29.03 at 5:59 am

In case it wasn’t clear, my response is not “This is the best idea since sliced stock options,” but rather — “Hey, something knew that might be useful. Let’s give it a chance to fail on its own merits.”


Dick Durata 07.29.03 at 6:04 am

This idea goes well with that other DARPA program to assassinate evil foreign leaders and other terrorists by sticking pins into dolls.
First you bet on Arafat dying of colon cancer, then grab the ugly doll with the Palestinian ghutra and slowly insert the silver pin in the proper location. Free money!


John 07.29.03 at 6:08 am

It is a stupid idea. Henry’s suggestions of modifications would be smart, but they aren’t doing that. Noise, noise, noise. I guess the DARPA people believe in some mysterious effect like a phase locked loop which will detect and amplify the signal with its unseen hand. Man, I can just see Gingrich doing the elevator speech that launched this. I’d love to get my hands on the white papers for this.


City Comforts 07.29.03 at 6:18 am

Pathos’ point that someone would assumedly be on the alert for bad guys seems to contradict the statement as reported in the NYT that buyers/sellers would be anonymous. Moreover, who else would have better information about impending activities than a bad guy who wanted to make a buck on his own future misdeed?

The whole thing is like something from science fiction. But then again: “life imitates art.”


Kieran Healy 07.29.03 at 6:21 am

Now we just need an efficient markets hypothesis for terrorist futures. The more DARPA’s futures market in terrorism resembles the Hayekian ideal that Chafetz aspires to, the less likely it is that the prices it produces could be used to predict anything. Whoops.


dsquared 07.29.03 at 6:48 am

There used to be exchange-traded contracts like this on the weather, on the basis that futures markets would efficiently aggregate all manner of weather forecasting knowledge. I seem to remember that the contracts never developed enough liquidity to really take off, but they performed tolerably well as a summary of existing weather forecasts.


back40 07.29.03 at 7:30 am

The program website may answer some questions.


Chirag Kasbekar 07.29.03 at 8:32 am

One of the brains behind this stuff is Robin Hanson. A look at his website would be quite worth the look:


Hal Varian had a relevant article a little while ago on Hanson’s ideas:



Chun the Unavoidable 07.29.03 at 8:49 am

It’s interesting that you mention Hansen, whose “How to Live in a Simulation” article auto-obsoleted many potential ideological critiques of transhumanism/simulation hypothesis by negative projection.

Don’t the senators raise the idea that such a market, on a large enough scale to be useful, would influence (perhaps cause) the events it was designed to predict? Is there some reason why this wouldn’t be immediately apparent?


dsquared 07.29.03 at 8:57 am

On a small practical note, I see no evidence on the website that the people involved have given any thought at all to the operational and regulatory aspects of running a securities market (and previous Robin Hanson projects don’t give me any reason to be confident that they ever will). It is, surprisingly enough, not a particularly easy job to create a deep, liquid and honest market, or to process trades in an efficient manner. I don’t see any financial guarantor here, either. If they’re really planning on opening up for trading in October, they need to get their skates on.

I also question whether the empirical evidence on the performance of futures markets is actually all that good. Looking at the Iowa Electronic Markets price charts, there appear to be quite a lot of speculative episodes in not a lot of time.




dsquared 07.29.03 at 9:25 am

Though having said that, the tradesports.com “Weapons of Mass Destruction found in Iraq before July 31” contract actually looks quite good; it was giving a 20% chance as of 23 June, which seems alright. Though the “30 September WMD” contract looks a bit sorry for itself; a seller at 8% chance, a seller at 10% chance and nothing on the bid at all!



dsquared 07.29.03 at 9:30 am

Oooh. just looked at the “lifetime” chart of that contract. 75% chance as of 23 April, rising to 80% shortly after. Better than evens as late as end-May. So actually, this contract appears to have lagged the news, not led it.


dsquared 07.29.03 at 10:32 am

Erratum: sorry, tradesports’ javascript navigation confused me. The July contract is the one with no bids; the September contract has buyers for 4 contracts@20 and 500 contracts@18.


Chirag Kasbekar 07.29.03 at 10:45 am

There’s also futarchy (‘Vote Values, But Bet Beliefs’):


from Robin Hansom’s home page


Chirag Kasbekar 07.29.03 at 10:45 am

There’s also futarchy (‘Vote Values, But Bet Beliefs’):


from Robin Hansom’s home page


Chirag Kasbekar 07.29.03 at 10:50 am

Sorry for the shabby double posting.

That should be Robin Hanson, not Hansom, of course…


Matthew 07.29.03 at 11:32 am

Newsfutures is another one. I quite like the Tony Blair will remain PM until 2004 — it currently is at $75, ie a 75% forecast of his still being there on Jan 1 2004, having risen from just over 60% after the david Kelly death, but down from a peak of 90%+ in mid June.

This also has some liqudity with 13,998 contracts outstanding. If you are pessimistic for his chances you can sell ‘muttley’ 105 contracts at $73 (the point being that if he resigns the contract becomes worthless so you make rather a lot of money, but if he stays on the contract will cost you $100 to buy). If you are optimistic you can buy 50 contracts off ‘guillespie’ at $75.


dsquared 07.29.03 at 12:21 pm

Of course, reading off from these prices of futures to percentage chances is highly problematic, a subject on which I will at some point write at length.


Tom 07.29.03 at 8:23 pm

Chun the Unavoidable posted”:

“It’s interesting that you mention Hansen, whose “How to Live in a Simulation” article auto-obsoleted many potential ideological critiques of transhumanism/simulation hypothesis by negative projection.”

That’s easy for you to say, what with your robe of eyeballs and all.

(sorry – couldn’t resist the jack Vance reference)


Dan Hardie 07.30.03 at 4:45 pm

Work out the moral hazard problem. If this market did get up and running, and if it did start to predict ‘a coup in Turkey’ then if its predictions had any credibility at all there would be serious problems for the Turks (fall in share values, downward revision of their S&P’s and Moodie’s ratings, etc). This would make it well worth the Turkish govt’s while to manipulate the market in the other direction, which it could quite easily do. And people like the Saudis, with their huge resources, would have to be assumed to be manipulating the market in whichever direction most suited them.

I can’t see the big institutional funds getting involved in such a whacko idea. If the market does develop credibility, the Saudis or the big oil firms, or other players with plenty of currency, will want to finance big shifts in value to influence the Pentagon in their preferred direction. If it doesn’t develop credibility, the market will just become an another forum for the fantasies of Den Beste and LGF fans, and suchlike ignorant loons.

Finally, the point of this is presumably to provide intelligence analysis in time for policy to be changed. You didn’t get any kudos for predicting an Islamicist attack on the US on September 12th, 2001. What the US wants is better advance warning than it got last time.

So the underlying logic of this market is: markets process information better than bureaucracies; we set up a market to predict coups, wars, or terrorist attacks; when the market ‘predicts’ a coup, attack, etc, but *before that event has taken place* the government should change policy to forestall the event.
So- given the fact that markets do develop bubbles- we’re going to see, for example, air strikes on the Bekaa Valley because everyone in the Darpa Terrorist futures market has decided to buy the ‘Hezbollah to strike Chicago’ shares. No doubt on the basis of hot tips from the rightwing blogs.



Dan Hardie 07.30.03 at 4:59 pm

Just as well these people weren’t around in World War 2, one feels. Would they have been issuing futures in ‘Bletchley Park decrypts all U-Boat codes by April 1943’?


dsquared 07.30.03 at 10:00 pm

>>So- given the fact that markets do develop bubbles- we’re going to see, for example, air strikes on the Bekaa Valley because everyone in the Darpa Terrorist futures market has decided to buy the ‘Hezbollah to strike Chicago’ shares.

Dan has hit on the real weakness of the model here; I’ll be posting something which expands on this point soon.


Dan Hardie 07.31.03 at 11:36 am

>Dan has hit on the real weakness of the model here< D-squared, I am moved. If peace has broken out between you and me, can Israel and the Palestinians be far behind? In the new spirit of amity, I wish to state that the only decent jokes I have ever seen in the blogosphere are a)on Brad Delong's website concerning this DARPA futures market, and b) D-squared's take on haikus, especially the bit about 'Agadoo'.


Dan Hardie 07.31.03 at 11:39 am

Just to spell out some of this model’s absurd problems:

One is that a futures market is subject to distortions due to moral hazard, herd behaviour etc, and is thus a lousy basis for prediction.
Leading on from that, this particular market would be especially subject to moral hazard (all those spooks and oil companies enabled to quite legally buy the opinion of the US govt), and even more so to asymmetric information (since good intelligence is by its nature kept secret).

But I think the most fundamental error of all is that a futures market works by rewarding investors who make the correct guesses about future events. If the US Government does use this market as a source of warnings about the future, it will do precisely the opposite.

Very simply: in a normal futures market I put money down to say that the price of copper per tonner will be x in August 2003, and if it does turn out to be x I make money, whereas if it is >x or


jimbo 07.31.03 at 8:58 pm

Even worse – what futures markets really measure is what Keynes called “the average opinion of average opinion” – in buying a contract, I am hoping, not that the contracted event will actually take place, but that more people will come to think that other people will come to think it will take place (and so on, ad infinitum) It doesn’t really “aggregate information” – it just codifies conventional wisdom…


logo 01.13.04 at 9:57 pm

Mein Hobby ist es Gästebücher zu besuchen. Das ist immer ganz interessant und widerspiegelt so, was die Leute im Internet wirklich denken. War auch interessant bei Dir ! Bis zum nächsten Mal. All The Best OfNew Year. Sorry for my english i’am from Germany.

Comments on this entry are closed.