Re: [OPE-L] measuring causes

From: Paul Cockshott (wpc@DCS.GLA.AC.UK)
Date: Fri Dec 16 2005 - 13:54:09 EST

Mike wrote:
         This is a great opportunity for a systems-illiterate person
like me to get an answer from a specialist. Assume that we have what
Marx called a 'structure of society, in which all relations coexist
simultaneously and support one another.' If you have a system with
substantial interaction, feedback, etc (eg. change in X produces
change in Y which produces change in Z which produces change in A
which produces major change in Y), then how can we reason from the
quantitative results of change to the weight to be assigned to
causes? The unambiguous results you suggest would certainly be true
if we assume no interaction effects (ie., that we can assign weights
to factor A and to factor B but the value of AB approaches zero), but
then the conclusion is embodied in the assumption.

I cited that example because it ( with slight changes to the figures )
came out of a conjectural analysis of the UK that Allin and I did 
back in 1976. The interaction effects here are negligible because
Marxist economics gives us clear formulae for the rate of profit
as a function of the other variables. One can take the partial
of the rate of profit with respect to them along with the known
data on the change in each of the individual variables and determine
from this how much the change in c/v, s/v and p/s would, taken by
themselves have produced as a change in the rate of profit.

The point is that the whole formula captures the interaction between
them, and since it is an analytic formula, you can compute its

 On the other hand,
how can we come to a weighting of causes from the empirical results?
         in solidarity,
Michael A. Lebowitz
Professor Emeritus
Economics Department
Simon Fraser University
Burnaby, B.C., Canada V5A 1S6

Currently based in Venezuela. Can be reached at
Residencias Anauco Suites
Departamento 601
Parque Central, Zona Postal 1010, Oficina 1
Caracas, Venezuela
(58-212) 573-4111
fax: (58-212) 573-7724

This archive was generated by hypermail 2.1.5 : Sun Dec 18 2005 - 00:00:01 EST