What if the problem with the Soviet Union was that it was too early? What if our computer processing power and behavioral data are developed enough now that central planning could outperform the market in the distribution of goods and services?
Pointer from Tyler Cowen.
This is just sad. The socialist calculation problem is not one of processing power. In The Book of Arnold, I try to explain it thusly:
In a command economy, central planners give each family ration coupons for corn flakes and wheat bread. If you were to give a family more ration coupons for corn flakes, then the family would gladly use them. If you were to give a family more ration coupons for wheat bread, then the family would gladly use those. Consumers have no way of signaling to planners that in relative terms they would prefer more corn flakes.
Similarly, workers have no way of signaling to central planners that they have untapped skills. Managers have no way of signaling that there are alternative production methods that might use a lower-cost combination of inputs. Indeed, the central planners do not know the cost of inputs. They only know the shadow prices that they have imputed to those inputs.
It’s also sad how this same idea keeps being presented without any reference to prior discussions of it. As if no one ever had this idea before, or responded to it.
Not much practical use has been made of it, but Tirole’s work on regulation strikes me as the right way to think about a sensible middle ground. Taken as given that the market has certain inefficiencies (i.e. monopolists set too high of a price, banks with deposit insurance take too much risk), good regulation constrains inefficient behavior while still allowing people to act on their private information in an incentive compatible way.
Regulators are overconfident in practice of course and are likely to micromanage things they have no business touching (and often in equilibrium they do so because it benefits big incumbents at the expense of small guys who can’t comply with the regulation), but the optimal degree of central planning must be neither 0% nor 100%.
Thanks for your inspiring thoughts. In order to better understand the (probably moving and context-varying) point or extension of the non-0%- and non-100% range, we need to look more deeply into the details of regulation and its outcomes.
High levels of abstraction in the analysis are of use too, but we are likely to get things wrong unless this perspective is supplemented by case studies and acquaintance with the concrete minutiae of the respective policies.
Of course, it is, for most of us (lacking time or expertise), hard to acquaint oneself with these vital details and, even if one has the knowledge, it is hard to convey these aspects and their significance in general discussions that are normally dominated by ideological ambition, which in turn is situated on a fairly high level of abstraction. We tend to organise our intellectual competition around ideologically preconceived, highly abstract patterns of interpretation rather than around common efforts at fact mining. Almost inevitably.
I would propose tentatively that regulatory success is often underestimated, especially from a libertarian perspective, and it is often insufficiently appreciated as a requirement generated in a free society, where people are enabled to compete for political impact.
Regulation may be a cost of liberty, which like everything is not a free lunch.
The Socialist calculation problem sounds a lot like the Econometrician calculation problem.
Don’t we go back to Hayek (and Mises) with the question: where do we get the information to make the “computer entries?” How do we “weight” (as per Arnold) the information in a place in an algorithm?
Basically, how do we (and for what purposes, how chosen) determine the structure of the algorithms?
Garbage in, garbage out. Next question?
There isn’t just a calculation problem. There’s a sensing and measurement problem: billions of prices and transactions that can’t be observed from a distance.
Surprise! The answer to these problems was invented long ago: Capitalism is the parallel multi-processing and sensing/measurement answer to central planning. Decisions are made on a local, decentralized way. And the net agregate processing of billions of simpler algorithms and processors is far more sophisticated than any centralized algorithm can achieve.
The “market failures” that the left is fond of pointing out are minor blemishes compared to the already seen major failures of centrally controlled economies.
yanovsky wrote a book ‘The outer limits of reason’ which lays out many problems that are not deterministic regardless of computer horsepower. Surely is such a case regardless if you could even get the inputs from everyones current subjective preferences into such a super computer?
Private actors will have computers too, which might be some version of a Lucas critique.
To be charitable, per the subtitle of this blog, assume that for some reason the Government (but not the private sector) has access to complete information about everything and everyone in the economy. (Already this requires a dubious assumption: why would the Govt have better tech than the best firms?)
In this Orwellian world, the Government knows how best to produce and how best to assign stuff to people, capital to firms, and workers to jobs. So far so good… But we run into an intolerable paternalistic conflict: a person may not wish to do the work they are best suited for, or eat what is best for them. A whole lot of misery ensues.
Ah, if only there were some way to…. Oh, right! Prices.
In bare simple terms, market orders are of NP Hard complexity, even with complete knowledge we need unlimited computing power to architect an optimal arrangement.
http://politick.me/2015/08/05/signals-and-systems/
@mseekan: actually, these problems are potentially harder than NP. Optimal allocation under both uncertainty and change is arguably isomorphic to the “restless bandit problem”, which is PSPACE hard.
http://arxiv.org/pdf/0711.3861.pdf
Thanks, I had realized that market are more like solving TSP on a constantly evolving graph. But as you stated PSPACE hard seems more appropriate.
Arnold, the post linking to your essay at EconLib seems to have vanished into the ether. Just a heads up in case you were dying to hear my feedback on it.
apparently the essay is not supposed to go up until next week, so I was told to take down the post and re-post it next week. communication mix-up.
so, Dr Kling, you write your blog posts but then (as you’ve told us) they don’t publish for a couple days, but when you write essays they actually pop up too soon.
There’s an Austrian lesson about “time preference” in there somewhere
Human beings exhibit features of irreducible personal autonomy (life stories, mental associations, inclinations and aptitudes etc) which are constitutive to the formation of their preferences. No third party has fully authentic and comprehensive access to these deeply personal dispositions. This is why computational processing power is immaterial.
Sure, government can make people act as if it new all their thoughts and desires, but it cannot truly know all their thoughts and desires.
I agree this is sad, especially in light of the tens of millions of people that Stalin killed.
But in another world (one with infinite land, say), entrepreneurs could just go try to re-create a Soviet state, armed with modern computers. Then we could see empirically who is right and who is wrong. And maybe entrepreneurs really would get better at centrally planned economies.
Unfortunately, in the real world, there is no extra land for somebody to go try this.
The best realistic alternative might be for Paul Mason or Malcolm Harris to take their ideas to Cuba and convince the Castros to give it a go.
This area of debate needs to include consideration of China’s mixed but significant record of success if it is to be useful.
What does the success result from and what does it succeed in spite of?
Until they achieve the potential of having a lot of high IQ people and a long history of relative political stability it may be too soon to talk about success.
I was rather amused by the title of the Malcom Harris article that inspired Arnold’s post here – “What if Stalin Had Computers”.
Stalin didn’t need computers. Stalin had guns – and more importantly, he had an unerring willingness to use his guns.
The use of computers implies the application of persuasion in order to induce the cooperation of others. The use of guns is an application of coercion to induce the cooperation of others. Stalin, and all like him, lack the patience (and wit) to rely on persuasion.
I recently read an excellent book on this subject, Red Plenty (http://www.amazon.com/Red-Plenty-Francis-Spufford/dp/1555976042/ref=sr_1_1?s=books&ie=UTF8&qid=1441116436&sr=1-1&keywords=red+plenty+spufford). It deals with the calculation problem in a novelistic way, which makes it very readable, but the characters are only semi-fictional and the concepts are real.