computers and the economic calculation debate

04/22/23 Paul Cockshott 1

Computers and the economic calculation debate


Introduction

I will be looking at the extent to which computing technology has improved the possibilities for planned economies.

Web site discussing these issues http://reality.gn.apc.org


Topics of Discussion

Plans and computers Value and prices under Socialism Payment


Historical Background

Immediate - the work of Prof Nove of Glasgow University and its impact in Britain

Long term - the work of the Austrian school of economics, particularly von Mises and Hayek

Current relevance - application of Hayekian economics to formerly planned economies

Collapse of of production Drastic fall in living standards and life expectancy


Free market deaths 7.7 million Excess Russian deaths 1991-

2001

0.00

500.00

1,000.00

1,500.00

2,000.00

2,500.00

1985 1990 1995 2000

deaths


Plans and computers

Starting with Von Mises, conservative economists argued that effective socialist planning was impossible for 3 reasons:

1. No effective cost metric in absence of market

2. Complexity too great – millions of equations argument.

3. Impossibility of capturing tacit knowledge


No Cost Metric

Von Mises argued that without a market one could not cost things and thus had no rational basis for deciding between production alternatives.

One exception he allowed was the use of Labour Values – we will return to this


Lack of Metric continued Suppose you have to select one of two

techniques of producing for example polyethelene – each is technically feasible but which would be the best one to chose from the standpoint of the economy as a whole.

In a market economy you cost the two techniques in money terms and select the cheapest.

If money and prices did not exist how could you do it?


Marxian response – use labour time as the metric The Labour theory of value provides us

with an immediate response here – you select the technique which minimises the total expenditure of labour.

Von Mises replies that the use of labour values is impractical for two reasons

1. The computational complexity of estimating labour values is simply to great

2. Reduction problem – how to reduce complex to simple labour


Millions of equations Von Mises asserted that one would

need to solve millions of equations to come up with the answer.

Computers obviously change this as they can solve millions of equations

Need to be quite precise about how many million equations and just how hard they are to solve

This is a branch of complexity theory


Complexity

The complexity of an algorithm is measured by the number of instructions used to compute it as the size of a problem grows.

We will look at a simple example before going on to economic planning


Searching Suppose that I have a telephone

directory for Berlin and a phone number.

It is clearly possible in principle to look at every number in the directory until I find who the number belonged to.

The task would probably take several days.


Example

Suppose I have 2 directories1. Has 1000 entries 2. Has 1,000,000 entriesTo look up a name will take 1000 times

as long in the second directory, but to look up a number – given the name will only take twice as long.


Indexing If I have a name on the other hand, I can

probably look up the phone number in less than 60 seconds.

The complexity of looking up by name is proportional to the logarithm of the number of people in the town.

The complexity of lookup by number is proportional to the number of people in the town.

The key is to select methods of low complexity.


Use of Input Output table

From the I/O table one can compute how much of each intermediate product required to produce each final product.

In particular we can compute the labour content of each output.


Part of the USA Input Output table

Table 2.--The Use of Commodities by Industries, 1987 BenchmarkAgricultural,

Livestock Other Forestry forestry,

and agri- and and Metalic

livestock cultural fishery fishery ores

products products products services mining

Commodity \ Industry ( 1) ( 2) ( 3) ( 4) ( 5+6)0 1 2 3 4

1 Livestock and livestock products 0 16817.9 1583.6 26.5 1250.5 0.0 2 Other agricultural products 1 23777.7 3855.2 0.0 2088.9 0.0 3 Forestry and fishery products 2 0.0 0.0 167.8 32.0 0.0 4 Agricultural, forestry, and fishery services 3 4003.2 6541.7 1288.5 8.2 0.4 5+6 Metallic ores mining 4 0.0 0.0 0.0 0.0 518.8 7 Coal mining 5 0.0 0.0 0.0 0.0 10.5 8 Crude petroleum and natural gas 6 0.0 0.0 0.0 0.0 0.0 9+10 Nonmetallic minerals mining 7 6.4 253.5 0.0 2.3 7.211+12 Construction 8 458.4 710.1 82.8 288.4 87.5 13 Ordnance and accessories 9 0.0 0.0 29.4 0.0 0.0 14 Food and kindred products 10 11565.6 0.0 305.0 33.4 1.2 15 Tobacco products 11 0.0 0.0 0.0 0.0 0.0 16 Broad and narrow fabrics, yarn and thread mills 12 0.0 44.1 0.0 0.0 0.6 17 Miscellaneous textile goods and floor coverings 13 25.4 25.7 71.6 114.4 0.0


Computability of labour content

Suppose we have 10,000,000 different types of goods produced in an economy (Nove quotes this)

Labour content given by a simple equation =A+l where is a vector of labour contents,l a vector

of direct labour inputs and A an input output matrix Clearly too big to invert, matrix is even too big to

store in a computer containing : 1014 cells.


Exact solution impossible

ProductsIn economy

multiplications Seconds taken

uniprocessor multiprocessor

1000 1,000,000,000 10 0.1

100,000 1015 107 100,000

10,000,000 1021 1013 1011=3150 years


Simplification

Matrix is sparse, most elements are zero Replace by linked list representation, we

estimate the number of inputs directly used in a product is logarithmic in the size of the economy.

Solve iteratively - use about 10 iterations, Complexity of order nLogn in number of

products. We estimate that it takes a few minutes on a modern machine.


Solution We only need to know labour values to

about 3 significant figures. Initially just include direct labour inputs. The produce second estimate taking into

account indirect inputs. Repeat this step about 10 times.

You end up with a figure accurate to about 3 digits.

This is accurate as our knowledge of prices – which are rarely accurate to more than 3 figures.


Approximate solution is feasible

products multiplications Seconds taken

uniprocessor multiprocessor

1000 150,000 0.0016 0.000016

100,000 100,000,000 1 0.01

10,000,000 6x1010 600 6 seconds


Feedback mechanism

We assume a real time feedback mechanism which uses sales of products along with democratically determined general goals to set net output targets for all goods. The planning computers must derive the gross outputs required to meed these net outputs.


Model we propose

Drawn on the principles of Robert Owen (of New Lanark), and Karl Marx

New Lanark Robert Owen


Payment in labour

Workers paid in labour tokens, 1 per hour. Goods priced in labour tokens proportional to the labour required to make them. (some discounting possible )

Industry publicly owned and planned in physical units.


Owenite Labour Note


Labour notes not money

Marx points out that labour notes are no more money than a ‘theatre ticket’ is.

They presuppose not commodity exchange but the direct socialisation of production


Market clearing prices used for finished goods

If stocks of unsold goods grow – then reduce selling price

If stocks fall – then increase selling price

If price above labour value - then increase output

If price below labour value – then reduce output


How close are prices to labour values?


International correlations of prices to labour values


Comparison with today Today market prices are an imprecise estimate

of the labour cost of producing a commodity. True labour values more accurate estimate of

costs Capitalism only accounts for the paid portion of

the working day. As a result it systematically underestimates the costs of labour as compared to machinery – whose cost it pays in full.

This encourages the wasteful use of labour and the under-use of machinery in capitalist economies.


The reduction problem

How do we reduce complex labour to simple labour – the work of an airline pilot to the work of a cook?

In principle it is simple – we add up the labour cost of training a person and divide it by the number of hours they will work during their life.


Why the fuss?

Behind this ‘technical’ objection by Mises hides class prejudice.

How, the upper class intellectual thinks, can my work possibly be compared to that of an ordinary worker.


Why computers better than markets

The market can be viewed as computing engine - this is explicit in Hayek.

Cycle time is slow, measured in months or years.

Arrives at answer by physically adjusting production up or down.

Constantly tends to overshoot in an unstable way.

Human costs to these adjustments


Computers are faster

Computers can predict where an ideal market economy would get to if it ever had the chance.

Production can then be adjusted directly to this target.

Cycle time for computation is in the order of hours not years or months.


Tacit Knowledge Hayek argued that socialism could never

handle the tacit dispersed knowledge that enables an economy to function. The price mechanism was a cybernetic control system that transmitted private information to where it was needed.

Example he gave was of a shipping clerk who has private expert knowledge of the sailings and arrivals at various ports.


Boadicea Paradoxically transport –

air transport at least was the first industry to be subjected to comprehensive computerised planning. The Boadicea airline booking system opened in the 1960s

Now all airline booking is computerised and shipping clerks are a thing of the past. Boadicea – early

anti-imperialist


Boadicea computer of 1960s

I have an affection for Boadicea, this B5700 buffer processor was my first personal computer in the 1970s, when Greg Michaelson and I salvaged it from scrap.


Objectivist tacit knowledge Clearly it is the Airbus factories

that have the information about what parts are used to make an A340. This information corresponds to what Hayek called tacit knowledge---but it is of course no longer human knowledge.

Literally nobody knows what parts go into an A340. The information, too vast for a human to handle, is stored in a relational database.


Industrial records At an earlier stage of industrial development it would

have been dealt with by a complex system of paper records.

Again the knowledge would have been objective, residing in objects rather than in human brains.

The very possibility of large scale, coordinated industrial activity rests upon the existence of such objectivised information.

Hayeks subjectivism makes him misunderstand the objectivity of industrial information.


Computers and democratic control

We propose system of online electronic voting on key issues like the proportion of national income to be allocated to health, eduction, research etc.

This done in terms of the fraction of the working week in labour units that is to go on it.

Taxes automatically adjusted to the democratic vote on social labour allocation.


Payment Payment assumed to be 1 hour per hour

worked minus taxes. No differentials for different grades of

labour. Enterprises charged more by the state for

skilled labour since this costs more to educate.

Prevent accumulation of human capital but ensures efficient use of scarce labour.


Incentives

Would there still be an incentive to aquire skills

Yes – because skilled work is more interesting and enjoyable than unskilled work even aside from payment questions.

Equal pay is fundamentally democratic.


References

“Alternativen aus dem Rechner” Cockshott and Cottrell,

A number of related papers from web pages. http://reality.gn.apc.org http://www.dcs.gla.ac.uk/~wpc/reports/ http://

ricardo.ecn.wfu.edu/~cottrell/socialism_book/ Book now available in English, Swedish,

German, Czech. Bengali and Spanish translations in progress

computers and the economic calculation debate

Documents

use of labour values

mises replies

paul cockshottplans

paul cockshottexamplesuppose

paul cockshottindexingif

phone number

number of instructions

total expenditure of