A reply to Duncan's ope-l 3540.
First, let me clear up an interpretative issue. Duncan writes: "'total value
produced' includes constant capital that represents production from many past
vintages, and the labor time in the denominator would also have to weight past
labor inputs in some way or other. I think this is what the difference
equations above are telling us."
He's referring to equations such as my
[1/e(t+1)]v(t+1)Q(t) = [1/e(t)]p(t)*A(t) + N(t) (total value).
I interpret Marx as holding that the value transferred (in labor-time) by used
up constant capital is [1/e(t)]p(t)*A(t), if A is taken to include
depreciation of fixed capital, whether or not the elements of A were acquired
in period t or in period t - k. That is, the value transferred is determined
by their *current* (pre-production) prices, p(t). I think this is, for
instance, what Marx says at the end of Ch. 8 of Vol. I and in Ch. 6 of Vol.
III. Yet each increment of the money-value actually invested as capital must
be deflated by the MEV reigning at the moment the investment is made.
Duncan also writes: "I understand how Andrew uses these equations to develop
examples where the path of the mev (e(t) in the equations above) is given. But
it isn't so clear to me how one could use these equations to estimate the mev
given
data on labor time, inputs, outputs, and prices. The problem is that in order
to figure out e(t+1) you seem to have to know e(t). In the examples this is
easily dealt with by assumption, but with real data what would you do to
calculate the first e(0) for period 0?"
Duncan has asked this question several times now. It deserves an answer. I
think Alan can answer it better than I, because he's tried to estimate the MEV
with actual data, and I haven't. But I'll try to address it.
Clearly, e(0) must be estimated. One way of doing so is to use Duncan's MEV,
i.e., take some adjusted measure of nominal NNP and divide by productive labor
expended. This constrains the estimated e(0) to equal the estimated e(1).
The shorter the time span for which this estimate is made, the more correct it
will be. Because the change in the MEV is principally determined by growth of
labor productivity and inflation (money per unit of use-value), and because
the shorter the time span is, the less productivity and the price level will
change, the actual difference between e(0) and e(1) will be smaller the
shorter the time span. Perhaps the use of quarterly data isn't so bad.
There are more sophisticated methods of estimation possible, such as using the
estimated growth rates of the MEV to extrapolate back to e(0), then proceeding
recursively. I don't know that it would make much difference. Other
estimation problems are so huge - e.g., estimating the amount of productive
labor performed - that this isn't something I'd worry about particularly.
Moreover, I've run a lot of simulations, and after a few periods, the growth
rates of the actual and the estimated MEV converge, even when the difference
between the actual and the estimated e(0) is large. So if one is interested
in a particular period, just start estimating the MEV starting about 5 periods
earlier.
There are many other, much more difficult, problems involved in trying to
estimate labor-time value and "price" magnitudes, once one abandons pure
simultaneism, for which the input/output approach "works." I agree with
Michael Perelman that the problems are insuperable in practice, so that the
project of operationalizing Marx's value theory as a whole is doomed to
failure.
I have to say that it doesn't seem to me that much is thereby lost. Marx
himself was able to explain a lot, and disclose the *meaning* of a lot of
things, with very rudimentary measures in some cases and with no measures at
all in others. A lot of difficult problems in "Marxian economics" would
simply disappear if people didn't try to use the categories developed in
_Capital_ to answer questions that differ markedly from Marx's. You can use a
shoe to drive in a nail, but it works better to use shoes for walking and
hammers to drive in nails. It seems to me indeed that if one reads _Capital_
in light of Marx's concerns and Marx's project, there is hardly anything left
to "develop" theoretically. If one reads it in light of other concerns and
other projects, then the thing appears obscure, incomplete, and
self-contradictory.
This isn't surprising. Yet my rather obvious remarks make some people
anxious. Let me make clear that I'm certainly not claiming to have a "direct
line to Marx." Instead, I'm suggesting re-reading Marx in the following way:
start from the very beginning of the text and, when a passage seems to leave
the matter unresolved, try to think of another way of understanding the
matter, of interpreting the categories, and of understanding Marx's project
and method, so that the text does resolve the issue it poses.
Andrew Kliman