[From Bill Powers (2004.03.21.0626 MST)]
Bill Williams 20 March 2004 6:30 PM CST --
You are probably tired. Why not give it a rest for a while and rethink
the problem. Maybe you can come up with some basis for a discussion other
than one which starts with you win and I lose.
Excellent idea. Let's both do that.
I think I may have come closer to understanding how Keynes arrived at his
idea that savings equals investment. If you're still working on your
exposition, perhaps you can check this out to see if you agree.
While I'm at it, I'll discuss some general ideas about modeling, as you
previously suggested I do. In particular, I think we need to talk about the
"test bed" idea in which we separate properties of the human agents from
properties of the objective environment in which they operate, where
"objective" means only that the characteristics of this environment are
independent of the observer. It is always important in PCT modeling to have
a model of the control system AND a model of its environmment. I hope the
meaning of that will become clearer as we go along.
In Chapter 6, Keynes begins (p. 52) with some bookkeeping. "During any
period of time, an entrepreneur will have sold finished output to consumers
or other entrepreneurs for a certain sum which we will designate as A. He
will also have spent a certain sum, designated by A1, on purchasing output
from other entrepreneurs. And he will end up with a capital equipment,
which terms includes both his stocks of unfinished goods or working capital
and his stocks of finished goods, having a value G."
He then goes into the problem that part of the value of G comes from "the
capital equipment which he had at the beginning of the period." So he sees
a problem in defining income, which he proceeds to attack in a way that
seems unneccessarily complex. This is the section where he introduces the
idea of User Cost, which seems to involve estimates of how much more the
production equipment would have been worth if it had not been used for
production of output. I have finally realized what he was talking about: he
was talking about the cost of maintaining capital equipment against the
deterioration due to its use (a cost proportional to the amount of goods
produced, to a first approximation), plus other costs which are simply a
function of elapsed time whether the equipment is used or not. There are a
few savings from using the machinery, which would exist only if the
equipment were not used, such as mothballing machinery in cosmoline which
would not have to be purchased and applied if the machines were in use, and
some savings, in that not using the equipment would reduce the costs of
maintenance and change the kind of maintenance, which savings are foregone
by using the machinery. So estimating the value of the capital equipment
can get quite complicated.
The key idea that got things going again for me was the realization that
Keynes was including two different kinds of things in the term "value". For
modeling purposes, monetary value is particularly simple to calculate: it
is a quantity of money, either in cash or in some other form which has a
fixed monetary value (for example, a deposit slip from a bank indicating
that you have $250 in your account means that the monetary value of your
account is, and will remain until you change it or the bank adds interest
and subtracts fees, $250). The monetary value of money is an objective fact
-- i.e., not dependent on who is observing it. This is not the case for
other measures of value, which are dependent on who is observing, what that
observer knows, and how that observer reasons.
Since Keynes does not mention stores of money in his definition of G, I
conclude that he considers money simply another form of value comparable,
say, to possession of machinery or buildings or raw materials for
production. If everything is always reduced to an equivalent monetary
value, that would make sense.
However, for modeling it does not work. Charlotte Bruun has said
essentially the same thing. In a model of the environment, the monetary
value of a machine, a building, or a ton of sand is not known after the
moment of purchase until it is sold, and then it is computed as the number
of goods of each type bought/sold times the unit price for each type of
good. The price is set by interactions among agents as part of their
actions on the common economic environment; price is a variable of the
model. Thus the test bed requires keeping track of physical goods (or any
non-monetary assets) as goods rather than as monetary value, leaving their
money equivalents to be determined as prices are varied through time. The
"test bed" model, in other words, deals only in actualities, not in
estimates or predictions. Estimates and predictions are things that the
human agents do in their heads, things that influence their actions but
which have no other effect on what exists outside the agents until actions
take place.
Under this principle, we can legitimately deduct from income A the amount
of money actually spent on purchase, maintenance, or improvement of
equipment during the period in question. Those are objective expenses and
they actually subtract from net income. However, we can't deduct any
subjective estimate of what the cost would have been if the equipment had
not been used, because that is part of a model of an agent, a person, and
not of economic transactions per se. An entrepreneur may take that
alternative cost into account in determining what his actual profit is or
might be, but that calculation belongs in a model of the entrepreneur, not
the test bed.
So we can trivially compute the monetary values of any cash reserves
maintained by the producer, but not the values of any material goods or
other assets for which a value is not set until a price is established in a
transaction. Whether we speak of raw materials, machinery, partly finished
goods, or finished goods held in inventory prior to sale, we have to carry
them in the model simply as counts of goods of each kind. Some goods, such
as production machinery, have direct effects on productivity, which will
make productivity a function of the amount of money spent on maintaining or
improving such machinery. The machinery, even without a specific monetary
value assigned to it, will play a part in determining production quantities
and production costs. But the value of the machinery remains undetermined,
unless of course it is sold at a specific price. Then its monetary value is
known for a specific moment in time.
The "income" of the entrepreneur, Keynes says, comes down to the value of
finished output sold (actually sold at a specific price) during the period
in question, minus his prime cost, prime cost consisting of (actual) factor
costs (money actually spent on production) and User Costs, which are partly
actual and partly subjective estimates.
And this income is, according to the previous discussion, exactly the net
change in G that has occurred during the accounting period.
Skipping over numerous details, I hope safely, we come at last to saving
and investment. Keynes says, "Income we have defined above. Expenditure on
consumption during any period must mean the value of goods sold to
consumers during that period ...". Savings, he says, is the excess of
income over expenditure, and according to the previous discussion, it is
therefore equal to the change in the net value of G during the period in
question, after deduction of production costs. This shows us that savings
is measured by Keynes not in monetary terms alone, but in terms of the
change in "value" of the "capital equipment" defined at the beginning of
the chapter. This term "value" thus includes not only actual money held in
reserve, but the entrepreneur's best estimate of the market value of all
his assets.
Keynes eventually goes on: "Our definition of income also leads at once to
the definition of _current investment_. For we must mean by this the
current addition to the value of the capital equipment which has resulted
from the productive activity of the period." In other words, the same G,
less costs. He then goes on to sum up, and naturally concludes that income
equals investment, as must be the case if they are both defined as the
change in G minus costs.
However, this does not mean that the amount of money held in liquid cash
form or its equivalent is equal to the amount of money spent on the
purchase and upkeep of physical production machinery, buildings, or raw
materials -- physical assets. If we think of savings as actual cash held
back and not spent, and of investment as the actual purchase and upkeep of
capital machinery, then savings is clearly different from investment: cash
that is saved is not spent for investment; what is spent on investment is
not saved. The sum of cash held back and cash spent on investment remains
the total amount taken from income in addition to costs of production, of
course. But that, apparently, would be a nonstandard usage of these terms,
even though it is a common usage.
I hope this is an acceptable solution to the controversy over savings and
investment.
···
========================================================================
In the test-bed model, it is easy to keep track of goods and money
separately, with price being the conversion factor when actual transactions
take place. Rather than use the term "Savings", knowing that it has a
special meaning to economists or at least to Keynes, I used the term
"reserve" to mean holdings of cash or equivalent. "Investment" is not yet
part of the model, although I have provided for an "investment fund" which
is built up from current income to some desired level and maintained there.
Later, money will be withdrawn from that fund and used to maintain
machinery, or rather to maintain the productivity factor at whatever level
is desired. Some assumptions (or measurements!) will have to be made about
the effect of money on productivity by this route.
I noticed a problem in Keynes' expositions that arises from a confusion
between what Forester calls "stocks and flows." Consider the idea of
income. I think of my income in terms of a rate or a flow: so many dollars
per month, which converts to so many dollars per day, a handy number to
keep in mind. However, when Keynes speaks of income he refers to the total
quantity received during an accounting period (it's true that he says "any
period," so he is almost there). The idea of carrying over value of capital
equipment from a previous period seems to be handled awkwardly, because of
thinking only in terms of total amounts rather than stocks and flows.
In modeling, it is essential to know the difference between a variable and
its time integral, which is what we're talking about. Let's say that the
entrepreneur calculates the value of his capital equipment (given modern
computers) every day (or even every minute). The value at the start of one
day is the value at the end of the previous day plus any changes in value
due to purchases, sales, and expenses occuring during that day. By adding
up all these changes, recalculating the total value every day, the
entrepreneur can compute the cumulative value at any time after the initial
conditions were set -- say, at the end of a monthly or quarterly accounting
period. This is a simple way of handling the problem that Keynes seemed to
pose so awkwardly.
Income and expenses can fluctate rapidly and widely every day, but the
total value changes much more slowly, since it's a cumulative quantity. So
there is a built-in lag which means that not every variable in the whole
system changes simultaneously, in proportion. Keynes writes about these
lags and seems to have an intuitive idea about their consequences, one of
which is a tendency to oscillations ("cycles"). When the model is set up
properly as a simulation based on differential equations, such oscillations
and their causes will be easily understood.
I think that much of the confusion here will be resolved by going to
"accounting periods" that are short enough to make all the relationships
appear continuously variable - in the test bed, I use a period of 0.001
day. Then cumulative variables will be seen to change on a slow time scale
while rate or flow variable fluctate more rapidly. This will eventually
open the way to a hierarchical model in which higher systems operate much
more slowly than lower ones (Richard Kennaway seems to have shown that a
speed ratio of about 4 to 1 from one level to the next is optimal for
certain kinds of hierarchical control).
I look forward to seeing your discussion of income, savings, and investment,\
Best,
Bill P.