Reading Andrews posts I am clearer as to where I disagree with him.
I was at first unclear about what he meant by prices, assuming them
to be simple weights of gold. He means amounts of labour obtained
by dividing the money prices by the MEV.
I believe his mistake is to think that one can convert prices
into socially necessary labour times by dividing by a single
scalar - the MEV or monetary equivalent of value. But this number
is just the average monetary equivalent of value. In practice one
has a vector which expresses the $ rduced per hours labour for
each industry.
When I have attempted to produce estimates of price to value dispersions
for different countries, I use an iterative procedure, that at first
assumes that this vector of specific labour coefficients is a vector
of 1's. One then makes a first estimate of the total labour content
of each industry's output by summing the direct labour input ( using
wage rate data to convert the wage bill to hours ), with the
money inputs of means of production converted into labour using the
vector of specific labour coefficients.
Since on the first step, this procedure amounts to multiplying all the
money expenditures on means of production by the same constant, this
first step is equivalent to Andrew's method of calculating values ( if we
assumed MEV to be 1).
What we obtain from this is an approximate vector of labour contents of
the outputs of each industry. From this, one can obtain a more accurate
estimate of the specific labour coefficient vector. This can then be used
to repeat the process and obtain a more accurate estimate of the vector
of industrial labour contents.
If one repeats the process a dozen times or so, one gets an answer that
is convergent to within a reasonable number of decimal places.
This leaves us not with a scalar MEV, but with a vector of specific
labourcoefficients, which can of course be used to obtain a vector of
industry specfic MEVs. The point being, that different industries have a
spread of price to value ratios. This spread is quite tightly clustered,
with the outliers being industries with a high rent content or ones with
a high degree of monopoly.
If one were to use Andrews method for estimating values one would obtain
a narrower spread of price value ratios, since, to take a pracatical
example from the US i/o tables, the fact that the output of the oil
industry sells significantly above its value would not feed through into
ones estimate of the value of output of the petrochemicals. One would
thus invalidly bias ones results on the closeness of price/value
correspondance.
---------------------------
Andrew wrote ( among other things ):
Finally, Paul commented on the need for p(t) to be socially necessary
labor-time. He said the problem does not arise if constant capital
is valued using v(t). First, where did this come from? What happened
to simulatenous valuation of input and output values all of a sudden?
Second, all I meant was that a firm might pay more or less for the means
of production than is socially necessary, but if so, that doesn't count.
p(t)*A is the amount of labor-time that is needed to *produce* A, on
average, at time t (when p(t-1)*A(t-1) is understood as one component
of that total).
There are a number of issues, important ones, that remain concerning the
value of money, etc. I don't pretend to be able to answer them all.
But the key point is that exchange can only distribute labor-time
differently, not alter the total. Changes in the MEV may or may not be
independent of how value is distributed, but as far as price and value
are concerned, their equalitiy, in money terms and in labor-time terms,
is not affected.