[OPE-L:5510] Re: William of Ockam's Razor and Political Economy

From: Paul (clyder@gn.apc.org)
Date: Mon May 07 2001 - 16:48:36 EDT

On Sat, 05 May 2001, you wrote:
> Inspired by Nicky's [5487] -- excerpts of which
> Does Ockham's Razor then not suggest that we
> should embrace VFT in preference to Marx's
> theory? By the same token, does Ockham's Razor
> suggest that we should embrace surplus approach
> theory in preference to either VFT or Marx's
> theory?
> 1) to what extent can an appeal to empirical
> evidence 'settle' the question of the superiority
> of alternative paradigms?   

There  is an extensive literature on the  relationship
between Occams  razor and choice between theorems.
This goes under the heading of the Mininum
Description Length principle, for an exlanation see:
or do a web search on Minimum Description Length.

Basically it says that given some observed data D, and
given a choice of theorys  T:A ,B, ...  then the best
theorem is the one which minimises

I (D - P) + I (T) 

Where P is the prediction produced by theory T
I(X) is the information content of a sequence.

Thus this says that we chose the theorem for which
the information content of the sum of the difference
between prediction and obserervation plus the
length of the theorem is minimised.

The point being that one can always produce a better
estimate of a dataset ( timeseries, set of prices, image, etc)
by building additional constants into the theorem.

The application of Minimum Description Length
Theory to the labour theory of value is addressed
in my and Allin's paper 'The scientific status of the Labour
theory of Value', see:
( the html version loses some of the precision of the
mathematical typsetting in the original, if people want
a PDF version I can supply it)
It was presented to the IWGVT in 1997.

The paper attempts to measure the formula above,
or a simple derivation of it for 
1. The simple labour theory of value
2. Sraffas formulation of price of production theory
3. The TSS theory of prices

All theorems were used to predict the final price vector
in the US input output table. Our conclusion was that
MDL principle resulted in the classical labour theory of value
being the most elegant.

We did not evaluate value form theory in this comparison.
As I understand it the value form theory makes no predictions
about prices at all. At one level it is arguable that this is 
simpler than theories that try to predict prices, since given no
prediction the value form theory would have information 

i.e, it predicts nothing so that the information content of
the data itself is the length of the theory.

whereas the labour theory of value has information 

I(D-P) + I(T)+I(L)+I(A)

where I(T ) is the encoding of the matrix formula for value,
I(L) is the encoding of the vector of labour values, and
I(A) is the encoding of the input output matrix. Even allowing
for the sparseness of the matrix, the sum I(L)+I(A) is almost
bound to be greater than I(D) since A is a square matrix
and D is a vector.

Thus in one sense it is almost certainly true that it is simpler
just to take prices as given rather than to construct a theory
to predict them, since the amount of data that one has to
use as input to a price theory is almost certainly greater than
the amount of data that one gets out. But is this a fair comparison?

There are certainly fields of investigation where the observed
data can be well predicted from a very short formula, for example
lunar motion - where the data are essentially harmonic.
Such data have inherently a low information content on grounds
of Chaitin Kolmogrov complexity theory. If we look at other
data - like observed air temperatures and pressures over
central Europe on a 100 km grid over a 24 hour period, the
data is computationally irreducible and can only be approximated
by theorems which bring in a wealth of ancillary data relating
to wind velocities, temperatures and pressures over a larger
region and a finer grid.

On one level it may be said that since we need so much
additional data to predict the weather all climatic models are
less elegant than simply observing the actual weather.
But surely simple observing the weather or observing prices
has less explanatory value than a theory which produces
some reasonably accurate model of what actual weather
or prices will be even if it has to use ancillary data
to do this.

A theory of prices or a theory of climate allows us to
predict what prices or climate would be like if circumstances
changed in some way. Post hoc observations whilst
being cheaper in information content, provide no similar
way of comparing alternative possibilities.

The labour theory  of value, along with an I/O table does
enable us to say what for example the price consequences
of a 50% increase in the labour content of electricity is 
likely to be. Value form theory, as I understand it, makes
no predictions on this.


>In practice, this
> becomes a very difficult way of accessing
> alternative theories. E.g. if we had a clear
> opposition between one theory for which
> money and labour-power must be commodities
> and another theory in which they are not,
> then empirical and historical evidence (in
> addition to Ockham's razor) might suggest
> which theory is preferable. There is not such
> a 'clean' opposition, though, that can be tested
> since the theory that held that labour-power
> is a commodity also holds that it is a *unique*
> commodity and that there may no longer be
> a money commodity (pace Claus and Akira).
> Even on the question of whether gold still is the
> money commodity, we have seen ambiguous
> empirical evidence.
> 2) one could suggest that the theory which
> reconstructs (e.g. through the method of
> systematic dialectics) in thought the 'essential'
> inter-relationships that characterize the capitalist
> mode of production in the 'simplest' and
> 'clearest' way is the superior theory (this also
> assumes Ockham's razor).  Yet, this is a *very*
> difficult thing to determine in practice for an
> inherently complex system in which things are
> by no means necessarily as they seem at the
> surface level of appearances. Thus, that gets
> us back to the question: how simple is *too*
> simple?
> Thus, how *do* we have a 'cross-paradigm'
> discussion about the merits of alternative theories
> of capitalism without retrograding into a
> discussion of Marx's perspective?
> At the IWGVT, Gary asked (following in the
> tradition of Steedman) what (and here I am
> paraphrasing) one theoretical perspective
> (the TSSI) was able to tell us that surplus
> approach theory couldn't. In so doing, I
> think he (and Steedman before him) were raising
> the issue of Ockham's razor as a way of choosing
> between alternative theories of political economy.
> I don't agree that this is necessarily the best
> question to be asking (see above), but I think
> it is reasonable to ask how in the lack of clear
> empirical evidence that supports one theory and
> negates another (very rare in political economy)
> or lacking a clear logical inconsistency or
> fallacy by one theory (as is often claimed --
> but infrequently proved) how one *does*
> (without resorting to authority, i.e. without
> resorting to the assumption that in the lack
> of clear evidence to the contrary Marx was
> 'right',  or political claims, i.e. claiming that
> one theory is preferable to another because of
> supposed political advantages of one
> perspective [which runs the danger of introducing
> a 'bias' into what purports to be an analysis of
> the law(s) of motion of capitalism])  compare
> and critically evaluate alternative perspectives
> of the political economy of capitalism?
> Any thoughts? Anyone?
> In  solidarity, Jerry
> ----- Original Message -----
> From: "nicola taylor"
> Subject: [OPE-L:5487] Re: Re: Counteracting factors (snip, JL)
> > Is the conceptual problem with Jerry's reformulation, or is it with a
> > labour embodied value theory?  Suppose that money is not a commodity but
> > pure form (having no labour-value substance); i.e. money is a more
> concrete
> > determination of value, and price as it's complete and finished
> expression.
> >  Then price *is* value, at a different level of determination.  Now
> suppose
> > that labour-power is not a commodity (since it is produced *outside* of
> > capitalist relations of production, in the household); then it makes no
> > sense to think about the wage as a 'value' or a monetary equivalent to the
> > subsistence bundle of reproducing labour-power as a commodity.  Rather the
> > wage is determined by supply and demand, by workers struggles, and by
> state
> > intervention.  Now, if labour-power and money are not commodities in the
> > sense described by Marx (Marx's own condition being that commodities are
> > produced within capitalist relations of production), then the triple
> prices
> > represent a fall in the purchasing power of the wage (yes, in the
> > neoclassical sense), therefore, a redistribution from the working consumer
> > to the capitalist class.
> >
> > A theory that affords ontological primacy to what goes on in production
> > does not, imo, provide an adequate understanding of oligopolistic
> behaviour
> > or the changes in contemporary capitalism (if Jerry is saying something
> > similar, then I agree with him).  In my own view, the reason for this is
> > that money is not given any ontological significance in the orthodox
> > abstract-labour theory of value, but is opposed to value as merely
> > phenomenal form.  So the redistribution from worker to capitalist can only
> > be theorised as a redistribution of use-values.  This is wrong.  The
> > capitalist economy is clearly 'form-determined' - at very least in the
> > sense that capitalists make an 'ideal precommensuration' in *money terms*,
> > before purchasing labour-power for *money wages*, and allocating it to
> > particular industries.  What is needed to understand the complex
> > interrelation of production and consumption in the reproduction of the
> > system, then, is a systematic dialectical reconstruction of Marx's core
> > categories, answering to the question of *how* value and abstract labour
> > are determined not as substance, but as historically specific capitalist
> > *forms*.
Paul Cockshott

This archive was generated by hypermail 2b30 : Sat Jun 02 2001 - 00:00:06 EDT