[OPE-L:3588] Re: Marxian empirical research topics

Duncan K. Fole (dkf2@columbia.edu)
Mon, 4 Nov 1996 06:00:27 -0800 (PST)

[ show plain text ]

>Duncan wrote in [OPE-L:3571]:
>> I've guided some very fruitful attempts to apply the circuit of capital
>> methodology empirically to the U.S. economy, though no one has ever got to
>> the point of publishing the results, to my chagrin.
>I'm guessing that you mean dissertations at Columbia. If so, what and whose
>dissertations are you referring to?

One was Andrew Senchak's 1982 Columbia dissertation. Since then a graduate
student at the New School has extended Senchak's basic methodology to a
longer time scale. There is also a graduate student at Utah who began a
similar project, but I don't know how far it has progressed. Peter
Matthews, now a faculty member at Middlebury College, also completed a
circuit of capital measurement for the U.S. as part of his Ph.D. thesis
from Yale, but as yet it has not been published. Matthews has been generous
in sharing his results with other interested researchers.

>What is the "circuit of capital methodology" as it relates specifically to
>empirical work?

These studies use the decomposition of the rate of profit I developed in
_Understanding Capital_, r = s/K = (s/v)(v/c)(c/K) = ekn where r is the
rate of profit, K is the stock of capital invested, including both fixed
capital and inventories, e is the rate of surplus value (the ratio of
surplus value to value added), k is the "composition of capital" (the ratio
of the flow of variable capital to the total flow of constant capital,
including depreciation on fixed capital) and n is the rate of turnover of
capital (the ratio of the stock of fixed capital and inventories to the
flow of constant capital).

The studies so far indicate a slowly rising rate of surplus value, e,
offset by a declining composition of capital, k, leading to a very gradual
fall in the markup on costs, q = s/(c+v) = ek. There is a sharp structural
break noticeable in the data in the early 1970s when k dropped abruptly
(corresponding, presumably to the sharp rise in energy prices). The rate of
turnover shows a substantial cyclical component, but no strong trend. This
is one aspect of the macroeconomic evidence I think broadly confirms what I
called the "Marx-biased" pattern of technical change, that is,
capital-using and labor-saving.

>Another guess: have similar attempts been made in France (other than by
>Dumenil/Levy) to apply this methodology to empirical work?

I'm not aware of any French work specifically using the circuit of capital
method. Dumenil and Levy have a very similar decomposition, but don't put
as much emphasis on estimating turnover times dynamically, which is a
central feature of the circuit of capital model. Their studies of the
evolution of the profit rate are highly compatible with the circuit of
capital results. Shaikh and Tonak also study the empirical evolution of
U.S. capitalism using an input/output framework, and with more emphasis on
measuring unproductive labor. Their results are also broadly consistent
with the circuit of capital picture, though there are some interesting
methodological questions involved in reconciling the details of the two

Senchak's dissertation is available from University Microfilms, but was
never published. (He now works for an investment bank.)

The method itself is fairly straightforward. One difficult issue involves
the measurement of constant capital, because national income accounts are
value-added accounts and net out intermediate inputs to production. The
studies I've mentioned above use the Census Bureau's Annual Survey of
Manufactures, which includes data on manufacturing purchases of inputs (as
well as depreciation), and then assume that the manufacturing composition
of capital can be extrapolated to the structure of capital in the whole
economy (which is, of course, only an assumption). The other somewhat
difficult problem lies in estimating the turnover times of inventories and
fixed capital. The studies I've described use a "FIFO" (First-In First-Out)
assumption on inventories to reconstruct a time series of turnover. If
anybody is interested, I can explain this method more carefully. The
measurement of the turnover of fixed capital, as John Ernst has been
pointing out, involves making some assumption about the time-lags
structure. Matthews, for example, assumes a "Koyck lag", exponentially
declining and estimates the turnover econometrically. Other approaches to
this problem are possible, and we just don't know very much about the
sensitivity of the measurements to this assumption.


Duncan K. Foley
Department of Economics
Barnard College
New York, NY 10027
fax: (212)-854-8947
e-mail: dkf2@columbia.edu