[OPE] Kliman's 'uncontested proof'

From: Dave Zachariah <davez@kth.se>
Date: Wed Jun 16 2010 - 15:29:28 EDT

I was recently prompted by Andrew Kliman about subjecting the labour
theory of value to statistical tests. Kliman objects to measuring the
correlation between industry sector outputs in terms of price P and
labour-value V. Instead he suggest one must 'deflate' the outputs in
some way.

He seems to miss the fundamental point that given the constraints on the
available data the specific test is precisely trying to quantify *how
well* size in terms of price correlates with size in terms of
labour-value. If they correlate weakly the labour theory of value would
be a relatively weak theory of market exchange. Hence, deflating the
output price and output labour-value with a third output size variable
would make little sense for this test.

Nevertheless Kliman insists that one must deflate price P and
labour-value V with output costs C. In his paper in Cambridge Journal of
Economics from 2005, he reports that when doing this the strong
correlations between P and V reported in the literature is destroyed,
indicating that they are all spurious. He shows this by means of a
computer simulation but motivates this by a 'deductive proof'.

    "For some reason, C[ockshott] &C[ottrell] simply ignore this proof.
    I am at a loss to explain why they fail to discuss (or even mention)
    it and why they seem not to recognize its validity. The proof is
    straightforward and mathematically trivial." (p.318)

His reasoning is as follows: If

    ln P_i = ln V_i + e_i

holds then it also holds that

    ln P_i - ln C_i = ln V_i - ln C_i + e_i

and hence correlations would be preserved. But since he finds that the
latter does not hold he concludes must be that the first relationship
does not hold in reality and "[a]s long as it remains unrefuted---and I
do not see how it could possibly be refuted---my regression results stand."

Well, it seems like the reviewers failed to notify him that P, V and C
are *statistical* quantities. Effectively he is saying that if rho(P,V)
is high then rho(P/C,V/C) is necessarily high or else the correlation
between P and V is spurious. But is should be clear that if C is
correlated with P and V this does not hold. And this is precisely the
case if one lets C be output costs.

This is the basis of what the author has labeled "the uncontested proof".

//Dave Z
ope mailing list
Received on Wed Jun 16 16:31:14 2010

This archive was generated by hypermail 2.1.8 : Wed Jun 30 2010 - 00:00:03 EDT