A response to Alejandro's PIAF (and the slew of repeats of the same that I
received):
----------
From: owner-ope-l@galaxy.csuchico.edu on behalf of Gerald Levy
Sent: Sunday, October 26, 1997 2:20 PM
To: ope-l@galaxy.csuchico.edu
Cc: multiple recipients of list; Andrew Kliman
Subject: [OPE-L:5661] [OPE-L] [ALEJANDRO R] andrew: random profits
I had written: "So, in principle, theres no problem here of using the
dependent variable to explain itself."
Ale asks: "1. Could you explain the last sentence? I think it suggests a kind
of "tautology" like "prices explain prices" ("dependent variable explain[s]
itself"). IMO this is why Allin is thinking that this is a "theory"."
Yeah, the issue is exactly whether the naive hypothesis (NH1) uses prices to
explain prices, though "explain" here is not meant in the theoretical sense,
but in the sense of accounting statistically for variations in prices.
But I don't understand Ale's last sentence.
Ale: "Isnt there a "temporal determination" in which there is no such a
"tautology"? "Dependent variable" is "output prices, period t+1" while
"independent variable" is "input prices, period t"."
Exactly. In any particular period, p(in) and p(out) are two distinct
variables. This is also why claims that the TSS interpretation has a
"circular" explanation of price determination, or is incompatible with the
determination of value by labor-time because prices are explaining prices, are
totally mistaken.
Ale: "I think the issue of "determination" is perhaps the most interesting
thing one can find in the TSS interpretation. However, unfortunately, it is
hard to have a clear and understanable account of it."
Huh?
Ale: "For example, can you really trace *temporally* the price of inputs, so
that you can distinguish them from the output prices?"
I'm not sure. Cockshott and Cottrell did something which I didn't understand,
in their paper at the last EEA, to estimate the temporal prices of production,
apparently taking turnover into account. But as I said, I didn't understand
the procedure or its relation to the raw data.
Ale: "When Allin tested NH1, did he really have *different* series of prices
for inputs and outputs? Or, did he actually work with a *unique* set of prices
corresponding to a generic "period t", lets say "year 1996"?"
He can answer that better than I, but I believe the raw data are annual
expenditures on inputs and the price of the year's output. Thus, the costs
are based on actual input prices (the price is of course the actual output
price). An estimation problem arises because some stuff produced during the
year not only counts as part of output price but also as input expenditures
due to multiple production periods during the year.
Here's an example with 3 sectors. Assume the following inputs and outputs per
period, with turnovers synchronized and 2 production periods annually in each
sector:
sector inputs from: output
A B C
------- --- --- --- ------
A 160 100 50 400
B 40 150 100 400
C 100 50 150 410
Assume that initial input prices are $1 per unit in each sector. If prices
remain constant throughout the year, we have:
sector cost price
------- ----------- ------------
A(1-6) $310 $400
B(1-6) $290 $400
C(1-6) $300 $410
A(7-12) $310 $400
B(7-12) $290 $400
C(7-12) $300 $410
The correlation between costs and prices is zero.
If, however, if output prices at the end of the first period, which are also
the input prices of the start of the second period, change to $1.20, $1.00,
$0.80 in A, B, and C, respectively, and if these prices still hold at the end
of the year, then we have:
sector cost price
------ ----------- ------------
A(1-6) $310 $480
B(1-6) $290 $400
C(1-6) $300 $328
A(7-12) $332 $480
B(7-12) $278 $400
C(7-12) $290 $328
The correlation between costs and prices is 0.63. This, IMO, is the "true"
correlation. In other words, we don't have the costs of a period "explaining"
the prices of the same period.
If, however, we are unable to get per-period figures, and have only annual
costs and prices, we see the following data:
sector annual cost annual price
------ ----------- ------------
A $642 $960
B $568 $800
C $590 $656
The correlation between costs and prices is now 0.71, not 0.63.
Hence, in this example, the fact that the change in output prices in the first
period leads to changes in costs in the second period has artificially
strengthened the relation between costs and prices a bit, if we do not or
cannot break down the information by production period.
In other examples, it could go the other way. I suspect that the main
determinant of spurious correlation here is the relative extent to which
sectors use their own inputs. Because they tend to have high input-output
coefficients for their own stuff (large main diagonal elements of the A
matrix), I think, at least for fairly aggregated data, the correlations
between cost and price and, to a lesser extent, the correlation between NH1
prices and actual prices, will tend to be a bit larger than they should be
when annual data are used.
Andrew Kliman