Saturday, March 29, 2008

There's no Debatin' these Socialist Calculations!!

Apparently Salvadore Allende didn't get the memo that Lange & Lerner lost the calculation debates. Allende's government (while it lasted) was attempting to implement a plan to "manage Chile’s economy using a clunky mainframe computer and a network of telex machines."

The project, called Cybersyn, was the brainchild of A. Stafford Beer...who employed his “cybernetic” concepts to help Mr. Allende find an alternative to the planned economies of Cuba and the Soviet Union.....

A Star Trek-like chair with controls in the armrests was a replica of those in a prototype operations room. Mr. Beer planned for the room to receive computer reports based on data flowing from telex machines connected to factories up and down this 2,700-mile-long country. Managers were to sit in seven of the contoured chairs and make critical decisions about the reports displayed on projection screens.

While the operations room never became fully operational, Cybersyn gained stature within the Allende government for helping to outmaneuver striking workers in October 1972. That helped planners realize — as the pioneers of the modern-day Internet did — that the communications network was more important than computing power, which Chile did not have much of, anyway. A single I.B.M. 360/50 mainframe, which had less storage capacity than most flash drives today, processed the factories’ data, with a Burroughs 3500 later filling in.

(the program was intended) to use the telex communications system — a network of teletypewriters — to gather data from factories on variables like daily output, energy use and labor “in real time,” and then use a computer to filter out the important pieces of economic information the government needed to make decisions.

Sweet! Here's more from Wikipedia:

There were 500 unused telex machines bought by the previous government, each was put into one factory. In the control center in Santiago, each day data coming from each factory (several numbers, such as raw material input, production output and number of absentees) were put into a computer, which made short-term predictions and necessary adjustments. There were four levels of control (firm, branch, sector, total), with
algedonic feedback (if lower level of control didn't remedy a problem in a certain interval, the higher level was notified). The results were discussed in the operations room and the top-level plan was made. The software for Cybersyn was called Cyberstride, and it used Bayesian filtering and Bayesian control. It was written by Chilean engineers in consultation with a team of 12 British programmers.

Move over Von Mises! It looks like maybe it was Augustin Pinochet who actually won the calculation debate!

Hat tip to Mrs. Angus


Gabriel M said...

The page you link on the calculation debates doesn't support your claim that Lange & Lerner lost and neither does history, I think.

By their standard, they won. I wrote a short entry on it.

Gabriel M said...

P.S. Pinochet was not big on debates, if you know what I mean, so your remark is a bit unfortunate.

I'd hate it if the casual reader would wrongly infer that torture camps and secret police abductions are a valid way to make a point.

Anonymous said...

He won, with help from the CIA

Angus said...

let me just say two things.

1: Hayek beat down Lange&Lerner 1-2-3 in the center of the ring. He came off the top rope with a bionic elbow and then hit them with the leg drop.

2. My punch line is a JOKE. At KPC we only endorse torture for grad students.

Anonymous said...

Ah the intellectual arrogance that leads people to believe they can direct trillion dollar economies with millions of participants from silly little Captain's chairs.

And brilliant imagery of Hayek on the top rope.

Gabriel M said...

Angus, sorry for not getting the joke. As for who won, I guess it's sterile. We could focus instead on what claims where proved/disproved. If the issue was that of a planner replicating competitive equilibria by playing the auctioneer, then it should be clear.

Anonymoous @ 11:19 PM... but isn't exactly that, "direct[ing] from silly little Captain's chairs", which is happening in US corporations which are larger than some nations?

Grant said...

True Gabriel, Lange showed that equilibria could be achieved (albeit by a very imperfect mechanism). But reaching equilibrium "is emphatically not the economic problem which society faces". I really don't think anyone who has read and understood The Use of Knowledge in Society or Human Action (although I think Mises' points about some things were less clear) can think socialism is a viable alternative to capitalism. Thankfully, no one is even trying to implement pure socialism anymore.

No amount of powerful computers will ever achieve the efficiency of a market, because markets can take advantage of knowledge that the planner does not have and cannot understand. Any equivalent market would necessarily have greater numbers of more specialized computers at its disposal.

Re: large multinationals, yes they do face calculation problems when allocating resources within themselves. But they exist in markets, and so have huge amounts of information available to them in the form of market prices in labor, commodities, etc. I'd imagine this is similar to semi-nationalized industries, such as health care, banking and education (no industry in any nation I'm aware of is totally vertically nationalized, nor ever has been).

Gabriel M said...


I respectfully think you're short-changing these results. In contrast, in the year 2008 the question of dynamics in decentralized economies is still open. (There exist counterexamples to tatonnement, trading at non-clearing prices change the equilibrium, etc.)

Much too much about markets qua mechanisms is glossed over, hand waved if you will, including (if not especially) in Hayek.

I'm not sure how to interpret your claim re: having or not having information available, and by what standard should I judge when enough is enough and why it is relevant.

Finally, the issues of these debates was narrower than the viability of socialism, or lack thereof, but rather the computation of allocations.

The paper pledge mechanism makes it theoretically and intuitively clear than a centralized market can generate efficient (and socially preferable?) allocation. Where do we disagree?

Grant said...

On the socialist side the debates might have been narrower for Lange, but it wasn't for Hayek and especially Mises (who were likely rather angry at the German 'National Socialist' party). To them, the division of labor (or 'division of knowledge' with Hayek) can only exist with markets. Mises thought pure socialism would cause the breakdown of all decent ethics and morals of society.

I don't disagree that socialist equilibrium is possible if the planner has full knowledge of what he is distributing. Hayek tries to point out this isn't the problem society faces. He believes the problem is how to make better use of available resources than is currently being done, a situation which necessarily requires people to do things without others telling them exactly what to do (after all, if the current utilizers of resources knew how to make better use of them, they'd be doing so).

As for how much information is needed, I think that kind of misses the point. You want the individual with specialized knowledge to be able to act on his knowledge, because that is the only way such knowledge is going to be utilized. We can't know how much knowledge is 'enough' because we don't know what is possible. We can't even imagine the span of all knowledge of how to use a given resource, let alone actually know all of it.

I think a good example is Warren Buffet. He's been very vocal and honest about the investing techniques he uses, but no one has been able to duplicate his success. Its likely much of his knowledge is tacit, and he may not even know what he has. How would he, as a normal socialist citizen, go about communicating his knowledge to a planner of capital?

I believe Lange conceded the debate for these reasons, and admitted socialist investment in capital goods would basically have to be random.

Gabriel M said...

Hayek's or Mises' motivations, their anger at the nazis etc., are beside the point, I think. I'm not happy for having been born in an ex-communist country, but that doesn't change the facts.

The planner needs to know nothing besides the "pledges" made by agents, i.e. how much each agent commits to buy/sell if the price would take a particular value. By iterative process of quoting potential prices, the planner finds clearing prices.

The people on the socialist side were not advocating "command and control economies".

Finally, could you please reference the paper or book where Lange "concedes"?

Grant said...

Well, as I've read it, the Lange model never included a pledge-based Walrasian auction. What I've read indicates he favored what was essentially trial-and-error to achieve equilibrium, which Mises dismissed as being a useless solution.

I don't think a pledge-based system to arrive at prices would work in a socialist system, because the participants don't have any incentive to report their true valuations. To be honest I haven't studied Lange enough to know what he proposed the system he did.

Trial and error can of course somewhat work for setting prices of consumer goods, but its absolutely horrid for goods which require time to mature (i.e., capital goods). This is where 'Hayekian' knowledge is needed most. Setting prices in consumer goods is relatively trivial, after all (e.g. no one trades 'X-Box futures').

The citation I have for Lange admitting that capital investment would be arbitrary is page 65 of On the Economic Theory of Socialism.

Of course, all this ignores the incentive problems of socialism, but I think studying those has been a more modern development of public choice economics.

Gabriel M said...

Thanks! Will check it out, but I don't expect to be able to return here soon.

My main point was, to reiterate, that there are ways for a centralized authority to compute efficient allocation. The many "market socialism" theories and models are a testament to that.

Grant said...

The above should read "To be honest I haven't studied Lange enough to know what he proposed the system he did."

I have read that Lange argued capitalism also produces arbitrary investment, but I'm not sure if anyone really took this seriously. Clusters of errors (e.g., the housing bubble) certainly do exist, but are obviously the exception and not the rule.

My main point was that you are correct to say that that a central authority can compute 'efficient' allocation, but its measure of what is 'efficient' is based on its own limited knowledge. So its idea of efficiency is less than what a market would create.

I think the Austrian tradition is much better suited to that sort of analysis than equations and equilibrium theory. To even suppose that one can compute economic equilibria is to assume one has the knowledge of all the actors in the market.

Gabriel M said...
This comment has been removed by the author.
Gabriel M said...


With the risk of repeating myself and dragging this on to the displeasure of our hosts, I don't think that what you're saying follows.

All the information one needs is the information an auctioneer gets by quoting prices: i.e. "options"/firm offers to buy/sell at those prices.

The auctioneer need not know the preferences or technology available to market participants, only their net supply for a range of prices.

It's supply and demand, and it would work the same for investment goods.

I haven't found your Lange reference so far, so I can't comment on that yet.

Tracy W said...

Gabriel - you say "The planner needs to know nothing besides the "pledges" made by agents, i.e. how much each agent commits to buy/sell if the price would take a particular value. By iterative process of quoting potential prices, the planner finds clearing prices."

Why in this case have a planner at all then?

It seems like a really slow way of replicating what a market does - since there are billions of people and billions of products in the world. Can you imagine spending hours each day sitting committing to buying/selling x for every single product you buy or sell?

Gabriel M said...

I did not say that socialism was a good idea, my comments were about the calculation of allocations issue from the "socialist *calculation* debates".

But to try to answer your question, one answer would be that you don't know that a market does/approximates that. You have a few clues and suggestions in that direction, but the theory still has problems (I commented above on some).

Grant said...


I'm really not sure what sort of system you are talking about. Like Tracy indicates, it sounds like a market and not any sort of socialism to me. After all, if pledges from market actors determine prices, we aren't talking about socialist calculation; thats market calculation, similar to eBay.

Of course, in reality things aren't as simple. There are many different auction designs, and I think if the Walrasian auction produced a result closer to equilibrium than other sorts of markets, people would be using it (although I don't think its very dissimilar to other markets in existence). There are real-life markets where participants purchase goods by pledging their worth, of course (InklingMarkets works this way, I think). The Internet may allow more centralized exchanges to to emerge, but I still think we will see quite a number of different market designs used.

Lange's proposed model involved using trial-and-error to find an equilibrium, not pledges. That is what Mises and Hayek were debating.

Gabriel M said...

Grant, I won't comment on what exactly Lange had in mind because I'm not qualified. I'm searching for your reference and that should help me with this.

I will make the following claims, though, and hope that I've supported them with contemporary research/theory:

1. Public ownership of the means of production (firms) is not an impediment to efficient allocations. There is no "problem of calculation". (They can just "play market" in a centralized/paper pledge fashion.)

2. Tatonnement will actual trading (at non-clearing prices) rather than pledges, if it works at all (e.g. counterexamples to tatonnement), will still get you an efficient allocation, although not the one that would prevail with pledges.

3. All this is irrelevant, even for contemporary socialists, since some form of shadow values or prices or whatever you want to call them will show up in your planning problem anyways.

In any case, my problem is with libertarians that invoke the "socialist calculation debates" as an argument against (quasi-)socialist, contemporary proposals. I find their argument weak and conclusions unsupported by history or theory.

Grant said...

1) I don't disagree. To the extent that the state "plays market", there is market calculation, and pledgers have real market incentives to pledge as honestly as they would in a market. But doesn't that sort of defeats the claimed purposes of socialism?

The problem arises when the state stops playing markets and starts exercising some form of command and control. You mentioned contemporary proposals (I assume for market intervention), so let me give an example:

Currently, the SEC regulates public stock exchanges in the USA; the state is exercising some form of command and control over trading stocks. What is the economic value of those regulations? There is no way I know of to compute this.

Compare that to a situation more similar to online markets, which regulate themselves. Each stock market would have its own set of regulations in order to attract both investors and firms, and the value of these regulations (among many other things) would be reflected by the revenue of the stock market.

Technically, any market (such as eBay) is more successful (at least in the long run) if it produces something closer to equilibrium. I'd look at real competitive markets to understand what methods work best.

You can certainly argue that we don't need market prices to calculate the value of some things (such as breathable air), but I don't think its accurate to suggest that contemporary socialistic policies don't run afoul of the calculation and knowledge problems.