Misconceptions on the speed of the market

Paul Cockshott wpc at clyder.gn.apc.org
Wed Nov 2 10:03:26 MST 1994

>> These corporatations make their estimations on how much previous demand
>> has emptied the shelves.  So previous demand is the ticket in our corporate
>> economy as well (This was the contention of my professor Robert Engler,
>> who has written two books on the oil corporations: we already have planning
>> in the late, corporate capitalist economy;
>Yes, yes. And markets are ex post facto systems. The point is that they
>have ex ante checks. You make a mistake, you have a surplus, so you have
>to lower your prices; you have a shortage, your competitor gets the
>business. Profit gives you an incentive not to make mistakes. This
>incentive is largely lacking in planned systems--including within big
>corporations. That's how GM got into such a mess.
The mechanism that Allin and I have been arguing for in our book would,
I think get round this problem.
1) Internally all units of production use up to date national databases of
   current labour content to cost different production alternatives.

2) The planning authority is sent by each production unit details from its
   internal spreadsheets to tell the planning authority what its current
   best production technology is.

3) The planning computers at frequent intervals, ( daily , monthly ? )
   computes from this data the required gross output to meet a given net
   output of consumer and social goods.

4) These are then transmitted as orders to the production units.
   These are similar to the orders that would come from customers
   in the market system, except that they originate from the center.

5) There is a separate distribution authority that is responsible for transmitting
   to the planners on a daily or monthly interval an updated net output
   vector. The distribution authority updates the output vector on the basis
   of the rate at which stocks are being run down. Goods are sold to consumers
   at their labour values. In cases of extreeme shortages or gluts, it might
   be necessary for goods to be sold at a premium or discount to consumers
   provided that there is no net inflation or deflation. (It is by no means
   clear that under normal conditions the system would have better dynamics
   with mobile prices than a simple stock adjustment algorithm.)

>> We have computers now that can coordinate the fit between group political
>> consciousness and individual desires.

>??? Look, Tom. Check out your Arrow Theorem. This particular
>"coordination" is no mere technical fix, even if we had big and fast
>enough computers to recompute an input-output matrix of--say--200 million
>by 200 million evevey few days (this is essentially what markets do).
>There are profound conceptual problems in solving exactly the fit you are
Whilst it would be feasible for a planning authority to recompute an input
output matrix of the required scale ( 10 to 20 million products is a more
realistic figure ), it is misleading to state that a market can do this.
The computational complexity of solving an I/O matrix by the iterative
method is
where K is the number of iterations required to converge
      m is the mean number of inputs used in each output
      n is the number of outputs
Experience shows that K is about 10. Let us assume n= 10,000,000 and
that m = 50. The number  of basic steps is then 5 billion. The time
for a computational mechanism, whether it be a computer, a single
mathematician, a million planners, or a million business men, is given
    --- x t
where P is the number of processors
      t is the cycle time
If we use a modern multiprocessor with say 10 alpha chips, each performing
20 million operations per second ( allowing 10 instructions per operation )
we get a compute time of about 25 seconds.
In the case of a market however, although there are perhaps a million
company executives with authority to alter production levels, the cycle
time is inevitably much larger because:

a) It takes days, weeks or months for an executive to make up his mind to
   alter production levels.

b) It takes months or years to adjust the actual level of production.

c) It takes a further period of tens of months for the adjusted level of
   output to feed through the wholesaling system enough to affect prices.

The combination of delays b and c are commonly used to explain the stockbuilding
and fixed capital component of business cycles, in which they are usually
assumed to have time periods of around 5 years and 18 months respectively.

Thus whilst the entire computation can be done in seconds on a computer,
in a real economy as opposed to the fantasies of the free market economists
the cycle times are such that it would take perhaps 50 years to equilibrate.
This is so slow that in practice market economies never approach
the optimum of economic theory.
Paul Cockshott , 		
Phone: 041 637 2927		wpc at clyder.gn.apc.org
				wpc at cs.strath.ac.uk


More information about the Marxism mailing list