[OPE-L:5516] Re: Humbug Aggregate Price-Value Correlations

Paul Cockshott (wpc@cs.strath.ac.uk)
Wed, 24 Sep 1997 08:11:15 -0700 (PDT)

[ show plain text ]

Alejandro Valle
> The correlataion coeficient for Mexico in 1980 is about 0.83 without
> reduction from complex to simple labor using wages and about 0.95 with
> such reduction. The MAD for the same year is about 80%. The previous
> figures does not consider productive labor and fixed capital problems.
> This results, maybe, point out that correlation coeficient is not good
> because of spurious correlation. I am not sure that 80% means that MTV
> does not hold because there are many theoretical problems still unsolved,
> as Lefteris pointed out. One important explanation for the impresive
> magnitude of the MAD in Mexico is oil rent: the production cost is 2-3
dol.
> and the selling price is 15-16 $ per barrel. Hence there is a good reason

> for high value price deviation in the Mexican economy. In fact, this is
> the highest value-price deviation. I am sitll working on this.
>
Paul C

The price value deviation for the oil sector is very high for Britain
as well. Model A shows the measures for regression of labour values
of industries against their prices including the oil industry, Model
B excludes oil.

Regression of prices against values for the UK in 1984
------------------------------------------------------
measure Model A Model B
2
R 0.955 0.976
MeanAbsErr 13.5 %
Max Error 157 %

I do not consider that there exists a real problem of spurious
correlation, for the reasons outlined by Allin, though I can
accept that correlations, based as they are on an 'energy' or
squared metric may be less appropriate for economic measures
than a metric based on absolute deviation.

By the way, I have just read your paper in Reasearch in Political
Economy and find that I agree almost 100% with your approach to
what value theory is about.