Some comments on Andrew's reply:
Andrew usefully quotes the BEA explanation of the Inventory Valuation
Adjustment, which seems to me quite clear.
>
>What is the justification for this? Apparently, inflation.
Well, it might be general inflation, but it might also be relative price
changes. The point is that the revaluation of inventories due to price
changes during the accounting (or production) period is not part of the
value added to the inputs by production itself (or as Marx would say, by
the expenditure of living labor), so it shouldn't be counted in the value
added.
>I took this for
>granted in my post of the other day, but now I think I was wrong. If the
>change in the value of inventories is removed, then so should inflation of the
>rest of profits and the rest of incomes be removed, if one is doing current
>dollar accounting. If one is doing constant dollar accounting, then also to
>factor in the IVA (i.e., lower profits in times of inflation) is "double
>uncounting."
This is a tricky point, but I don't think it's double counting. Whatever
the valuation system you use, current prices, constant prices, labor time,
there may be changes in the value of the inventories of goods in process
through the production period, and you have to decide whether or not to
attribute that change in value to production itself. I would argue that it
should not be attributed to production (or the expenditure of living
labor).
>
>What other justification besides inflation might there be? What the BEA is
>trying to do with NNP and NI is not obtain a measure useful for planning and
>investment purposes, but obtain an accurate measure of what has taken place,
>how much value has been added. Do they do so? I think it all depends on how
>the PBT is calculated.
>
>Assume a year-long general strike, so that the amount of production that takes
>place during the year is 0. Accurate measures of value added and profit from
>current production (if wages, etc. = 0) should show that they both equal $0.
>Assume also that the inventory withdrawals valued according to a mix of
>historical and replacement costs is $1 billion. (How can there be inventory
>withdrawals without production? Answer: all the inventories were
>perishables.) And assume that the replacement cost of the inventory
>withdrawals has plummeted to $0. The IVA will equal $1 billion > 0.
I have some trouble understanding your language here. Inventory withdrawals
during the accounting period are counted as part of the cost of goods sold.
(In your example the sales revenue is zero.) This takes place before the
IVA. The IVA refers to the change in the value of the inventories that
remain as a result of changes in valuation during the production period. In
your example I have trouble distinguishing the two issues. In order to
compute the IVA you would have to specify the value of the inventory that
was not withdrawn and its change over the period.
>
>Now there are two possibilities. (1) PBT is measured as $0. Then the BEA's
>adjusted measure of corporate profits (and, ceteris paribus, NI and NNP as
>well) will equal $1 billion. This, I think, overstates the value added from
>current production by $1 billion. (2) PBT is measured as - $1 billion. Then
>the adjustments give accurate measures of $0 for corporate profits, NI, and
>NNP. For PBT to be measured as - $1 billion, the accountants must be figuring
>profits by taking the difference between end-of-year and beginning-of-year
>assets without making price adjustment. They began with $1 billion and ended
>with 0. Is this what PBT does?
Because of my uncertainty about the assumptions you're making, it's hard
for me to be sure how to calculate PBT in this example.
>
>Whichever is the case (I suspect (2) is), my equations do not compute the
>difference between end-of-year and beginning-of-year assets, but work directly
>with "data" from current production. Thus, they conform to (1), and any
>adjustment to them produces wrong results.
Well, everyone is free, I suppose to redefine concepts as they choose as
long as they are explicit about it. I think we reached strong consensus on
the desirability of making all the assumptions in arguments explicit in
earlier exchanges around the Okishio and Roemer papers. But as the passage
you quoted from the BEA indicates, your computations use a definition of
value added different from the national income accounts.
>The computation of total value
>(which equals total money price in current dollars after adjusting for the
>possible change in the monetary expression of value) is
>
>TV(t) = p(t)*X(t) = p(t-1)*a(t)*X(t) + l(t)*X(t)
>
>and value added is
>
>VA(t) = p(t)*X(t) - p(t-1)*a(t)*X(t) = l(t)*X(t)
These are exactly the expressions I attributed to the TSS examples, so I'm
glad we're in agreement.
The problem is that this apparently minor change in the definition of value
added, in the context of the labor theory of value principle that the value
added in the period is attributable to the living labor expended in the
period, has quite a substantial impact on the theory's predictions about
prices and profit rates, as John Ernst's example shows. This shows,
incidentally, that the labor theory of value has explanatory content and is
not just a set of accounting conventions (a critique some people have made
of the "New Interpretation"), since changes in the definitions give rise to
operationally different predictions about prices and profit rates.
>
>for one or many sectors. If production of the "year" is X(t) = 0, total value
>and value added both equal 0. Profits are computed by subtracting wages >= 0
>from value added, so if wages = 0, so do profits. Only stocks *used in
>production* enter into the equations, not all "inventories."
But in a circulating capital model, we always assume (as I did explicitly)
that the inputs have to be purchased the period before the output is sold.
This means that the inputs are inventories through the period of production
and as a result can be revalued, and it also means you have to decide how
to take account of that revaluation. If you move to the view where inputs
and outputs are simultaneously present (as in most applications of the
neoclassical production function), then I don't see where p(t-1) comes into
it at all. As I understand you, however, you make the standard assumption
that the inputs are purchased at the end of the last period, at last period
prices, which gives them a historical cost of p(t-1)a, and inevitably
subjects them to a revaluation of a(p(t)-p(t-1)) during the production
period, unless prices are stationary. Thus we have to take your second
equation as your definition of value added, which is different from the
BEA:
VA(BEA) = p(t)X(t) - p(t)aX(t)
VA(TSS) = p(t)X(t) - p(t-1)aX(t) = p(t)X(t) - p(t)aX(t) + aX(t)(p(t)-p(t-1))
= VA(BEA) + IVA(BEA)
In search (in John's words) of clarity as a step toward agreement,
Duncan
Duncan K. Foley
Department of Economics
Barnard College
New York, NY 10027
(212)-854-3790
fax: (212)-854-8947
e-mail: dkf2@columbia.edu