Evangelical Control Theory

[Bill_Williams1]

From[Bill Williams 21 May 2004 2:OO PM CST]

[From Rick Marken (2004.05.21.1100)]

Bill Williams (21 May 2004 12:20 PM CST]

I still don't follow why income would equal expenditure, even at the
aggregate level. This would be true, I think, only if you define income as
equivalent to S (money received by a seller) and expenditure as equivalent
to P (money paid by a purchaser). Is this what you are saying?

This is a part of what the usual aggregate income analysis amounts to.

How else are you going to consider this in an aggregate context?

However, there is also the real side of the transaction in which the
commodity sold (income) equals the commodity purchased (expenditure).

Bill Williams

[Bill_Williams1]

From[Bill Williams 21 May 2004 3:30 PM CST]

F[From Rick Marken (2004.05.19.2100)]

Bill Williams (19 May 2004 7:20 PM CST)

Me:

I think this because, like all other
behavioral scientists, they never talk about possible _controlled
variables_.

You say "like all other
behavioral scientists, they never talk about possible _controlled
variables_." However, I had a professor who discussed at length the
question of the control of behavior in what he called difficult
circumstances. He was for this reason a critic of behaviorism.

Controlled variables are controlled _by_ behavior (actions). Your
professor was not talking about controlled variables as we understand
them in PCT.

Sorry Rick, I guess I will have to update you score card again. As best I
can tell the professor was talking about the purposeful control of behavior
in the face of a disturbance.

Minus 1

I believe that you will not be able to find an economist
who talks about _controlled variables_. But if you do, let me know.

See Herbert A. Simon 1952 "On the application of Servo Mechanism theory in
the Study of Production Control." Econometrica Vol 20 # 2 April p.
247-68.

The use of control theory-- the Bellman equations is a standard feature of
contemporary orthodox economics.
Unfortunately, orthodox theorists haven't yet gotten the idea that control
theory could be used to replace a value theory based upon maximization.

minus 1

You haven't learned that you would be better off not to make claims about
questions regarding which you are not informed.

Bill Williams

[Bill_Powers1]

[From Bill Powers (2004.05.21.1554 MDT)]

Bill Williams 21 May 2004 3:30 PM CST --

As best I can tell the professor was talking about the purposeful control
of behavior in the face of a disturbance.

That's a common mistake in talking about negative feedback control systems.
Of course control systems actually _vary_ (not control) their behavior as a
means of controlling perceived controlled variables.. Is that what your
mentor actually said? Did he speak of purposeful control of perceptions by
means of varying actions?

I should also add that lots of people have noticed control phenomena
(William James is just one example). That's a bit different from having a
theory that will explain them.

I think perhaps that the scoring should be left to a disinterested party.

Best,

Bill P.

[Richard_Marken2]

[From Rick Marken (2004.05.21.1520)]

Bill Williams (21 May 2004 2:OO PM CST)

Rick Marken (2004.05.21.1100)]

I still don't follow why income would equal expenditure, even at the
aggregate level. This would be true, I think, only if you define income as
equivalent to S (money received by a seller) and expenditure as equivalent
to P (money paid by a purchaser). Is this what you are saying?

This is a part of what the usual aggregate income analysis amounts to.

I think I see the problem. In any transaction, one person (the "seller")
gets "income" (as you define it) and the other (the "buyer") makes an
"expenditure" that is precisely equal to what the seller gets as income.
So in this sense, income (what the seller of goods or services gets) always
equals expenditure (what the buyer gives to the seller). This equality of
income and expenditure is maintained in H. economicus and Econ004.

When I say that income can exceed expenditures I am talking about income
that is already "on hand", not the income that is another name for the
expenditure of a buyer. I think a better word for what I have been calling
"income" is "income reserve". What you call income increases the income
reserve of the "seller" and decreases (as an expenditure) the income reserve
of the "buyer". When an economic agent buys goods or services, their income
reserve is depleted by an amount proportional to the expenditure and, at the
same time, the income reserve of the agent from whom the purchase was made
is increased by an amount proportional to the expenditure. When an economic
agent sells goods or services (including their own labor) their income
reserve is increased by an amount proportional to the income received and,
at the same time, the income reserve of the agent to whom the goods or
services were sold is decreased by an amount proportional to the
expenditure.

At the aggregate or individual level, there is a mathematical requirement
that income as you define it (the money received from a transaction) equal
expenditure as you define it (the money paid in a transaction) because it is
_the same money_. But there is no such requirement regarding income
reserve and expenditure because they are not necessarily the same money. For
example, when I sell a widget for $100 my income is $100 and the buyer's
expenditure is $100, because the $100 I get is the $100 the buyer gave me.
The income I received is now added to my income reserve. If my income
reserve were $0 before the sale then it is now $100. I can now use all, part
or none of my income reserve to buy something for myself (and provide income
to the seller). If my expenditure is less than my income reserve then my
income reserve is greater than my expenditure. So what I have been saying is
that it is possible for the income reserve of the aggregate consumer or
producer to be greater than the expenditures made by these entities over a
particular time period. Leakage occurs when the income reserve of the
aggregate consumer is positive during any fixed time period. A positive
income reserve simply means that the aggregate consumer has not spent all of
its income reserve during a particular time period.

Regards

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

[Bill_Williams1]

From[Bill Willliams 21 May 2004 8:45 PM CST]

[From Rick Marken (2004.05.21.1520)]

>Bill Williams (21 May 2004 2:OO PM CST)
>
>> Rick Marken (2004.05.21.1100)]
>>

. I think a better word for what I have been calling

"income" is "income reserve".

I do not see the point of continuing this discussion.

What you call income

Rick, I am not a Keynesian. And, this isn't merely what _I call income_.

I am not interested in continuing this discussion. So, I am deleting the
remainder of what you have to say.

Bill Williams

[Bill_Williams1]

From[Bill Williams 21 May 2004 8:50 PM CST]

[From Bill Powers (2004.05.21.1218 MST)]

There's no point in complaining about this usage or saying it's wrong.
Economists got there first and they can define their terms any way they
want. I think we can say "composite" instead of "aggregate" when we want

to

mean that more than one person is included in a model entity.

I don't see any problem with the term "composite." The crucial point as far
I am concerned is that if one is talking about the economic as a whole is
that the definitions being used are consistent so that we can start from a
first principles position. The most basic starting point, and the only
starting point that I know of, is a transaction. As long as consistency is
maintained, the terms used are of comparatively little concern. For me it
is an advantage if standard terminology is used. When it is not used I find
the discussion somewhat like a foreign language that uses terms I know but
assigns them different meanings. Or, what is worse, shifts the definition of
terms so that there is an equivocation going on as the argument proceeds.

In the modeling I don't expect that this should be a problem.

Bill Williams

[Bill_Williams1]

From[Bill Williams 21 May 2004 11:10 PM CST]

[From Bill Powers (2004.05.21.1234 MST)]

Bill Williams 21 May 2004 11:50 AM CST--

>Reading through Bill Powers' description of his approach, so far, to
>modeling the economy I don't see of substance that I would object to.
>However, I think the use of the term ["deaggregation"] may be a bit
>misleading.
> Powers is using a verb to describe moving from one big box in which

all

> the economic agents are treated without distinctions, to a

"multi-sector"

> analysis in which agents are assigned to particular boxes based on some
> trait-- such as ownership of capital or not capital owning, or worker or
> management.

Correct.

>I think it imay be confusing to describe the analytical division of an
>economy into sectors as a process of ["disaggregation"] when
>the intent to is create an aggregate model of the economy.

_My_ intent is to create a model that works reasonably like the real
system,

Describing your intent in terms of "a model that works reasonably like
the real system" may have the unfortunate effect of creating an
impression that you believe that this is not the intent of other people
who build economic models.

Bill Williams

[Bill_Williams1]

From[Bill Williams 21 May 2004 12:20 AM CST]

[From Bill Powers (2004.05.20.1412 MST)]

Bill Williams 19 May 2004 10:00 PM CST--

>However, there is nothing in such an approach that necessarily requires

that

>one refer to this or that economist as an asshole or a tax cheat.

I have done neither.

But, you have.

I have never, to my present recollection, called any

economist an asshole (though I did, correctly I believe, refer to Mises as
an old fascist),

Do you remember saying,

  I have a few of Mises' writings on my hard disc, including that one, I
  think. A man for whom the term "self-centered idiot" was surely invented
  (that was originally a shorter, 7-letter, term).

It took me a moment to figure out that yes, "asshole " is a 7-letter term.

and I never accused anyone of being a tax cheat.

Right. And, Rick never called Michelle "an ignorant slut." Of course not
who would ever think such a thing.

I did say

that Keynes'definition of user costs sounded like an attempt to reduce
income for tax purposes,

And, wondered how he managed to get away with it.

>Not that the work of these heterodox economists is free from deficiency.
>Not at all. However, an approach that lumps the whole of the profession
>into one benighted, malicious, elitist conspiracy involves some rather
>obvious difficulties.

Let's not blow a few rather restrained criticisms

Dignifying what you said as "restrained criticism" won't wash.

up into a different

malicious elitist conspiracy.

> One of which is the difficulty of imposing upon ones
>self the task recreating from scratch, in intellectual isolation the

whole

>history of the effort to understand how the economy functions.

I wouldn't say that I have lived in intellectual isolation.

To characterize Keynes as a guy who created a tax scam, and somehow
as you said, "got away with it." is an indication of severe intellectual
issolation.

My reading has

been limited and present access to libraries is also limited, so in a
practical sense recreating a system from scratch is probably the best

thing

I can do. Anyway, it's not necessary to reconstruct the whole history of
the economics effort: all that's necessary is to get the model right. If
the right variables are included in the model, terminology can always be
adjusted to fit what has gone before.

>As I have repeated said, modeling the economy from the standpoint of a

first

>principles approach has the potentiality of creating, or at least

clarifying

>an understanding of how the economy functions. However, there is nothing

in

>such an aspiration that requires the use of terms such as asshole, or
>accusations about tax cheats.

Correct. But isn't it almost time to get over that misinterpretation?

Get over? Misintrepreation? I am afraid not. Let's take the accusation
that
von Mises is a facist. This is another example of just what I am talking
about.
Obviously von Misesis is not one of my favorate characters. But, he wasn't
a facist. von Misses was an indivigualist. Not even close.

> Nor, does such an approach necessarily require discarding, or ignoring
> every understanding that has been ever achieved regarding how the

economy

> functions.

That's different. Actually, what could be a stronger validation of

previous

understandings than to have the same ideas developed from scratch without
reference to the previous work? I much prefer to work from scratch when
that is possible, although some degree of borrowing is inevitable no

matter

what theory is under examination.

Even the most radical revolutionaries develop their own jargon, their own
conventions about how to represent the system, their own set of ideas that
are taken as unquestionable, and their own sets of heroes. This always
means that there is a protected set of assumptions which nobody examines,
either because they are agreed to be self-evident, or because they were
first said by someone who is agreed to be above criticism. Working from
scratch to the extent that is actually possible, one is at least not
subject to those influences.

>Nor is there any justification for Bill Powers to claim that I have never
>made any contribution to efforts to apply control theory to problems
>involving human behavior or at least economic behavior.

I would be astonished to discover that I had ever said any such thing.

But, you did.

The

problem is that you suffer from a severe case of insult inflation,

I don't think so. For you to say that I have never contributed anything to
CSG is so wide of the mark that it isn't the sort of insult that causes me
any distress. But, it does indicate a willingness on your part when your
emotions are engaged to say stuff that isn't remotely true.

you were intolerant of any degree of criticism whatsoever,

When you say, and I repeat absurd things such as claiming that I have
never contributed anything to CSG it isn't a matter of my being overly
sensitive to criticism-- it is a matter of you being out of your mind.

but felt it

necessary to make the criticism seem to others heinous enough to warrant a
full-fledged counterattack.

When, you claim that I have never contributed anything to CSG, I don't feel
any
constraint is required in opposing what is in my opinion an effort to do
anything
you can to descredit me.

>But, I've known Bill Powers now for quite some time and I recognize that
>when he gets caught up in an argument he sometimes says things that he
>knows are not true.

No, Bill, you say I said things that I know are not true, which is quite
different from what I actually said. You used this technique in reporting
to the world that Rick Marken proposed that we put child molesters in
charge of law enforcement. If Rick denied, correctly, that he ever said
such a thing, you would accuse him of having said things he knows are not
true and of now trying to wiggle out of it. What you accuse me of saying
and knowing is not true is not what I said, in fact. What you said I said
is not true, but I never said it. You are very poor at accurate quotation
of what people say.

You can say this but you are wrong.

>I still find this an incontinent trait, but it doesn't really bother me

that

>much anymore.

It's amazing how much time you devote to criticizing me for something that
doesn't bother you.

You see. You are at it again. You are misquoting me. I said, "doesn't
really
bother me _that much_. This is quite different from your misquotation in
which you say, "something that doesn't bother" me. All the difference in
the
world.

>And, by a practical application of control theory I have found that I
>can cope with the problems that Bill Powers struggle of love and hate

with

>his dad ...

This amateur psychologizing may make you feel better but it has little to
do with me.

I think it has a lot to do with this stuff.

From remarks you have dropped I can't help wondering how much

all this has to do with _your_ father.

At least my dad didn't burden me with views on the economy.

> From my standpoint, it has been Bill's assumption that his dad had it
> right ( see the tape from Boston where Bill is defending his dad's

work )

> that has
>prevented him from understanding the Keynesian system.

You seem to assume that if I object to a misinterpretation or a
misrepresentation of what my father said, I support what he actually said.
It doesn't seem to be part of your world view to insist that a person be
treated fairly even if you don't agree with him, or that if you criticize

a

person, it should be for what he did or said instead of something he never
did and never said..

As if treating people fairly is part of your world view?

For me those are separate matters -- whether TCP said

what you report him as saying (not having read his book), and whether what
he did say is acceptable. I think you grossly misrepresent him, yet in

many

of the same matters I simply don't agree with his approach or his

opinions.

I can therefore say you are wrong about him, without turning myself into a
supporter of his actual ideas.

This is really rich-- you actually aren't a "supporter of his actual ideas."
Take a
look at the Boston tape.

There are some things in his book that I find reasonable, and others I do
not. On the whole I don't think he took us a very long way toward a
workable theory of economics, but he did do a few laudable things, such as
pay attention to data rather than arguing from abstract principles.

I noticed just how consistent he was in this regard, remember that big hole
in
his data-- the hole from the great depression and WWII.

> From my point of view, if I am to participate on the CSGnet I have a

choice.

>When Bill Powers makes a completely absurd charge that Keynes chapter in

the

>General Theory is an intellectual dodge to justify a tax cheat scheme, I
>can ignore it.

No you can't.

How much would you like to bet?

The evidence is that you keep returning to probe this sore
tooth and have been totally unsuccessful at ignoring it.

I wouldn't call ignoring what you said about Keynes in terms of "success."
And,
if I wish I could, and perhaps may choose to ignore future absurdities.

I disapprove of

accounting methods that rely on things that didn't happen,

You are mistaken here, obviously things _did_ and _do_ happen. Perhaps what
bothers you is that the things that do happen do not support your dad's
work.

and suspect that

they may have been invented to reduce tax liabilities, but since this is
done all the time it can hardly be called tax cheating -- especially if

the

IRS or the Inland Revenue allow it. Your exaggeration reveals your opinion
of it, not mine.

Sorry Bill, you are not being truthful.

I see no place for things that did not happen in a model

of what does happen,

Not being truthful again.

so whether I approve of it as a social matter is

irrelevant; it is of no use for modeling (except, as I have said
repeatedly, as an aspect of a psychological model of the entrepreneur)..

>It is more a matter of my irritation as a result of being confronted by,

as

>Bill Powers might phrase it, such "shocking stupidity." It really is in
>quite a stark contrast to the quality of the work that Bill usually does.

I have a vague recollection of the phrase "shocking stupidity," but doubt
very strongly that you are presenting it in context.

You used "shocking stupidity" as I remember it to describe Tom Bourbon.

Is this something like

your interpretation of Rick's "bitch" comment about Michelle, which he did
not actually make? (He said, "If I were to call Michelle ..." whatever it
was, a sentence construction that seems to be over your head.)

This is a choice example of a severe deficiency of moral judgement on your
part.
Pasting the "If I were" onto his calling Michelle "an ignorant slut."
doesn't mean
that that he didn't do it. I've explain this before. There is a criteria
known as
the "reasonable man." You say, that the sentence construction is "over my
head."
Would any "reasonable man" come to this conclusion? No. Or, for that
matter what
would a reasonable man think, or for that matter Michelle's husband, was
the
purpose for Rick's calling Michelle an "ignorant slut?" According to a
reasonable
man standard, Rick pasted this "If I were" onto calling Michelle an
"ignorant slut."
for the purpose of denying what he very plainly went on to do. Obviously it
suits you, you can call people you don't even know, fascists and assholes,
and
defend Rick when he uses a dishonest dodge so that he can call Michelle an
"ignorant slut" and then deny that he did so.

I think your phrase "shocking stupidity" fits your arguement as well as
anything I
can think of.

Bill Williams

[Bill_Powers1]

[From Bill Powers (2004.05.22.0852 MDT)]

Bill Williams 21 May 2004 12:20 AM CST--

I think your phrase "shocking stupidity" fits your arguement as well as
anything I can think of.

Well, then, it's about time to get back to more interesting topics, isn't
it? I will try not to be shockingly stupid about building economic models.

Best,

Bill P.

[Bill_Williams1]

From[Bill Williams 22 May 2004 11:35 AM CST]

[From Bill Powers (2004.05.22.0852 MDT)]

Bill Williams 21 May 2004 12:20 AM CST--

>I think your phrase "shocking stupidity" fits your arguement as well as
>anything I can think of.

Well, then, it's about time to get back to more interesting topics, isn't
it? I will try not to be shockingly stupid about building economic models.

As far as I am concerned, it is always time to be building economic models.

And, I don't think I have ever said that your modeling has been "stupid, "
let alone "shockingly stupid."

I know you have been busy at other worthwhile tasks, the Necker cube was
really pretty neat.

However, I would be interested in seeing what happens when you model a
consumer who has a reference level for consumption that is well beyond their
income. So, they have a control loop for consumption. They also have a
control loop for staying within their budget. If the two loops are the same
the consumer ends up somewhere way over their budget but not consuming all
that they might wish. Since they can not consume, not at least for very
long, at a rate that is above their budget, there needs to be some way of
reducing their expenditure over and above the basic budget control loop.
I've modeled this by integrating the budget error. There are probably other
ways that didn't occur to me.

I would be interested in see how you model such a consumer. I've attached
my version, in which such a consumer experiences first a squared impulse in
income upward and then a downward impulse. If this seems appealing we could
compare notes. I've gotten econometric results that seem to be consistent
with a consumer finding it easy to increase expenditure when income is
increasing. They seem to find it harder to decrease expenditures when income
is falling. According to orthodox theory there shouldn't be any difference
between adjusting to an increase in income and adjusting to a decrease in
income. But, the data seems to be very noisy and the empirical evidence for
the effect is so far rather weak. I have an idea that may improve model so
that I can get better results, but that work is still in process.

Bill Williams

[Richard_Marken2]

[From Rick Marken (2004.05.22.1050)]

Bill Williams ( 22 May 2004 11:35 AM CST) --

However, I would be interested in seeing what happens when you model a
consumer who has a reference level for consumption that is well beyond
their
income. So, they have a control loop for consumption. They also have a
control loop for staying within their budget. If the two loops are
the same

How could two different loops be the same?

the consumer ends up somewhere way over their budget but not consuming
all
that they might wish. Since they can not consume, not at least for
very
long, at a rate that is above their budget, there needs to be some
way of
reducing their expenditure over and above the basic budget control
loop.

Isn't that what the control loop for staying on budget is for?

I've modeled this by integrating the budget error.

It's not clear how this would solve what you describe as the problem.

I've attached my version,

There was nothing attached to your post. Make sure you send the source
code when you do attach your model.

Regards

Rick

···

----
Richard S. Marken
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

[Bill_Williams1]

From[Bill Williams 22 May 2004 1:10 PM CST]

[From Rick Marken (2004.05.22.1050)]

> Bill Williams ( 22 May 2004 11:35 AM CST) --

> However, I would be interested in seeing what happens when you model a
> consumer who has a reference level for consumption that is well beyond
> their
> income. So, they have a control loop for consumption. They also have a
> control loop for staying within their budget. If the two loops are
> the same

How could two different loops be the same?

Same in the sense that they have the same gain, slowing and so forth... One
loop "pulls" the consumer toward a level of consumption way above the
consumer's budget, the other loop pulls the consumer toward a level of
spending consistent with the budget.

> the consumer ends up somewhere way over their budget but not consuming
> all
> that they might wish. Since they can not consume, not at least for
> very
> long, at a rate that is above their budget, there needs to be some
> way of
> reducing their expenditure over and above the basic budget control
> loop.

Isn't that what the control loop for staying on budget is for?

Sure. But the "tension" between the two loops results in the consumer
spending more than the budget can provide. To start with I don't accumulate
budget errors. I have more complex models which include consumer savings and
consumer debt, and interest income or interest payments.

> I've modeled this by integrating the budget error.

It's not clear how this would solve what you describe as the problem.

As the budget error is integrated the "power" of the budget loop is
increased and this reduces the level of expenditure. I am sure there must
be other ways of doing this, but integrating the budget error seems to work
as a starting point.

> I've attached my version,

My mistake. This time should do it.

There was nothing attached to your post. Make sure you send the source
code when you do attach your model.

Given the universities E-mail policy all I _can_ send is source code.

Bill Williams

vd51.pas (7.36 KB)

[Bill_Powers1]

[From Bill Powers (2004.05.22.1614 MDT)]

Bill Williams 22 May 2004 1:10 PM CST --

As the budget error is integrated the "power" of the budget loop is
increased and this reduces the level of expenditure. I am sure there must
be other ways of doing this, but integrating the budget error seems to work
as a starting point.

>
> > I've attached my version,

I find that I can run your programs under Windows XP, but only at 640 x 480
resolution and using standard Borland bgi files. So we can go on for a
while longer with Turbo Pascal. However, I do recommend getting Delphi 7.0
-- you will get a hefty university discount. I can make it relatively easy
for you with some basic setups that will make programming work almost the
same way it does with TP7.

I think that before we get more deeply into your program we need to
simplify some things. I find it pretty hard to understand what you're doing
-- explanations as comments in the source code would be very helpful. We
can discuss the basic control system first, if that's OK with you. And I'd
like to suggest some principles for modeling control systems that will make
all models easier to understand. Here is the first controller:

  procedure controller1;
     begin
       e := r - o;
       be := b - o;
       if be > 0 then
         begin
           be := 0;
           bi:= bi - bi/ ( damping ) ;
          end
        else
          begin
            bi := bi + be/( damping );
           end;
        cl := ( gain * e - o)/slow;
        bl := ( bg * bi - o)/slow;
        o := o + cl + bl;
       end;

There seem to be two control systems merged together here. However,there is
no environment: the perceptual signal is simply the output o. The first
principle I'd like to suggest is that we keep control systems separated,
and the second is that we always include a model of the environment.

As I understand your code, there are really only two control systems here,
but you have repeated the code five times to allow plotting five conditions
in one pass. That's OK, but there's an easier way to do that, which we can
discuss later. Let's focus for now on the two control systems.

One control system, if I understand correctly, is for consumption. This one
would involve the program lines

        e = r - o;
        cl := ( gain * e - o)/slow;
        o := o + cl;

First, let's make the notation consistent, using c to indicate consumption:

        ec := rc - oc
        oc := oc + (gc*ec - oc)/sc

Second, note that I have omitted one step. The way the original program is
written, there are two integrations in the loop, one leaky (the original
second line) and one pure (the original third line). This will make the
loop show over- and undershoots, and if the gain is high enough, runaway
oscillations. I don't think you meant to get into such complications, so I
have written the loop with only a single leaky integrator. This can be
stabilized for any gain simply by adjusting the slowing factor.

The other control system is represented by

       be := b - o;
       if be > 0 then
         begin
           be := 0;
           bi:= bi - bi/ ( damping ) ;
          end
        else
          begin
            bi := bi + be/( damping );
           end;
        bl := ( bg * bi - o)/slow;
        o := o + bl;

Again, we have two integrations in the last two lines, but we also have a
decay factor of somewhat puzzling construction in the compound if-then
statement. As far as I can see, there is no control loop here becaused bi
is not a function of the output. Whatever positive or negative value it
starts with, that value simply declines or increases toward zero regardless
of the output. Have I missed something here?

What we need is a second control system independent of the first one that
is concerned in some way with the budget -- details to be worked out. It
should work like the first one, except that I suppose we could say that
it's a one-way system: there is an error only if expenditures exceed the
budget, other wise the error is zero.

And now we come to the environment. The best way to analyze a control
process is to define a controlled variable that is affected by the system's
output and possible disturbances, and which is sensed as a perceptual
signal. True, one can often simplify the equations to eliminate seemingly
unnecessary intermediate variables, but that seldom pays off in
comprehensibility. And in the present case, it can conceal environmental
interactions that are important to getting a correct model.

So, what variable does the consumption control system control and perceive?
That seems obvious, but consumption can be thought of either as a rate
variable or as a cumulative variable, and we have to decide which is meant,
and stick to the decision throughout the rest of the model. Let's say that
we will measure consumption in goods per unit time -- a rate variable. This
means that the perceptual signal (which we will call pc) will stand for the
rate of consumption, and the reference signal rc will specify a preferred
rate of consumption. The error signal (rc - pc), or ec, if positive, will
indicate that the rate of consumption is too low, not that the total amount
consumed is too low. If you don't want to call too high a rate of
consumption an error, you can specify that negative errors are shown as zero.

If there is an error, the resulting output should increase or decrease the
consumption rate as appropriate. If we call the actual rate of consumption
qc (where q by convention refers to an environmental quantity), and if we
propose that there are possible disturbances of this rate, dc, we can say that

qc = oc + dc;

That completes the rate-of-consumption control loop. There is nothing
remarkable about it.

Now we have to devise a budget-related control loop. What is the controlled
quantity to be? Just calling it "budget" is not sufficient to define it. We
could, as an alternative, call it "net rate of expenditure". A control loop
similar to the first one could immediately be constructed, which will make
the perceived rate of net expenditure equal the reference rate, whatever it
is. There will be two influences on rate of expenditure: purchases, and
income. Income, measured as a rate (dollars per unit time) disturbs the
actual rate of expenditure in the negative direction, and purchases disturb
the rate of expenditure at a positive rate measure as dollars per good
times goods per unit time, which comes out to dollars per unit time and can
be added to the controlled variable. The net rate of expenditure is the
rate of purchases times average price minus rate of income. That is what
the perceptual signal would represent, and we can assume some reasonable
reference level for it: slightly negative, implying some rate of building
up savings, or zero, implying that the person wants expenditure to be no
greater than income (a one-way control system). An output function with a
single leaky integrator will allow this control system to be stabilized for
any gain by adjusting the slowing factor. The output could be used to
restrain or allow spending by the other control system, or could be used to
vary income up to the limit possible. Fill in your theory of consumption here.

Notice that the consumption rate disturbs the expenditure rate since money
must be spent to make purchases to consume. Income is a disturbance in the
favorable direction. The environment contains these links -- they are not
part of either control system, and would not change if the two control
systems were redesigned.

You could also design the second controller as a quantity controller, with
income adding to the cash reserve and purchases subtracting from it. The
perception would then be of the quantity of money in the cash reserve and
the reference signal would specify the level to be maintained. The output
could still vary the income rate or the expenditure rate, but now could be
proportional: the cash reserve variable inserts the single integrator
required for loop stability.

I think it would be most useful now to rewrite the two control systems as
separate systems and include a separate environment. Try applying the above
principles and suggestions -- I think you'll find they make the modeling
easier to do and for others to understand. For now you need only one pair
of controllers; we can simply set the initial conditions and run the
program over and over with different parameter settings each time, to show
plots of different conditions. This will considerably simplify the program.

Best,

Bill P.

[Bill_Powers1]

[From Bill Powers (2004.05.22.2325 MDT)]

Bill Williams 22 May 2004 1:10 PM CST --

Looking over your program again I see that I misread the last two lines of
each controller, though they were reproduced correctly in my post. There is
a single leaky integrator, but it is a composite one which is hard to
interpret:

        cl := ( gain * e - o)/slow;
        bl := ( bg * bi - o)/slow;
        o := o + cl + bl;

By substitution, this is equivalent to

        o := o + (gain * e - o)/slow + (bg * bi - o)/slow

  The above line is equal to

        o := o + (gain/slow)*e - 2*o/slow + bg*bi/slow

As I pointed out, bi is not affected by o, but simply decays exponentially
to zero. Therefore, bg*bi/slow also decays exponentially to zero.The second
term on the right is the pure integration component, the next term is the
decay per iteration (the "leak"), and the last term is an additive
disturbance that begins high and decays to zero quite rapidly. The last
term begins at 100, and is multiplied by 7/60 on each iteration, so
successive values are 100, 11,67, 1,36, 0.158.-- after only four iterations
the effect of bi is no longer visible in the traces. By the way, you bi1
through bi5 were not initialized, so this term remains zero in the other
four control systems and has no effect at all.

So -- all the more reason to go to a second version. The reason I point
this out is that it's a warning that every modeler should heed. Every model
will do _something_, but it's not wise just to assume that it's doing what
you meant for it to do, and rush on to interpret its behavior. Making sure
you know what your model actually is and what its behavior actually means
is by no means simple. I don't really know how to describe the process of
verification, except to say it demands a very skeptical examination of
every line of code in your program, with the question in mind, "What does
this line of code actually accomplish, as opposed to what I wanted it to
accomplish?" This is hard to do since you just wrote that code convinced of
what it would do. It's much better to get a stranger to examine the code,
but where can you find one of those when you need the scrutiny? You just
have to practice doing it yourself. This is very good for the character.

So let's press on to VD5a.pas.

Best,

Bill P.

[Bill_Williams1]

From[Bill Williams 22 May 2004 1:15 AM CST]

[From Bill Powers (2004.05.22.2325 MDT)]

Bill Williams 22 May 2004 1:10 PM CST --

So let's press on to VD5a.pas.

By all means.

I've read through your comments but not studied or considered as yet what
modifications to make in the code. As the date 2001 indicates the program
was written quite some time ago, so I have to study it to see what is going
on.

First:

Yes, the code could and should be rewritten as an indexed version if the
goal is to model a variety of consumers.

Copying the code and creating five consumers was quick and dirty. I didn't
notice that as a result of copying only the first consumer's variables were
initialized.

And, I in general approve of standardization so that communication costs are
minimized.

Clarity in coding is an obvious goal, however, when the only time that I
do any programming is when I have an idea about something to model-- and
this comes around about twice a year-- so you can see the problem.

I'll have to study what you say about the environment. I guess I assumed
that the real numbers _are_ the environment.

And, about the dimensions which I didn't define. You are saying the same
thing my highschool chemistry teach said-- it doesn't mean anything until
you attach explicit dimensions to the numbers. I agree, but as you see I
often neglect this point.

End of responses for now to your remarks.

Second:

I should have written up a "motivation" section explaining what the program
is intended to demonstrate.
Basically, the program is intended to show that using control theory it is
possible to model a typical consumer using a method other than the familiar
one that uses the principle of maximization.

However, the model of the consumer based upon maximization works fairly well
in the economists view. Despite the someone peculiar assumption that the
consumer's wants are infinite the model of the consumer based upon
maximization is a generative model that explains why a consumer decreases
the purchases of a commodity when its price goes up. No, alternative
generative model to that provided by maximization has been available to the
critics of a consumer modeled by maximization.

Using control theory an alternative scheme can be implemented in which the
consumer's wants are more plausibly, not infinite, but quite a bit beyond
the rate of expenditure which the consumer's budget will allow.

The control theory model the consumer has the advantage over orthodox model
in that if the price of the good goes to zero the consumer doesn't want an
infinite quantity of the good.

And, in particular the VD5.PAS program generates a behavior through time
that is different than the one predicted by the orthodox model of
consumption. According to the orthodox model, a consumer's behavior is
ergodic with respect to time. An adjustment to an increase in the consumers
income is assumed to be different in no way different than a consumer's
response to a decrease in income. But, there has for some time been an
opinion that it is easier for a consumer to adjust to an increase in income
than it is for the consumer to adjust to a decrease in income. ( Veblen
1898, Duessenberry 1947 ) I am assuming that if it is easier for the
consumer to adjust to an increase in the budget that they will increase
their expenditure more rapidly when income increase and reduce their
expenditure more slowly when income decreases. There is some empirical
support for this belief.

The model generates a rapid adjustment to an increase in income-- a square
pulse in VD5.PAS and a slow decrease of expenditure when income decrease.
There are a number of additional features of the VD5.PAS model of consumer
behavior that amount to predictions-- such as the overshoot of expenditure
on the upward side of the adjustment, and the lack of an overshoot on the
downward adjustment. But, the difference in the speed of adjustment
depending upon the direction of the adjustment is the primary difference
between this control theory model and the orthodox model of the consumer
based upon maximization.

This effect is generated because-- in the adjustment to an increase in
income, there is for the time being no conflict between the consumers
control loop for consumption and the budget control loop. However, when
income is decreasing their is a conflict between the two loops, and the
decrease in consumption comes about in part because of the budget loop
including an integration term.

End of motivation text.

Comment on the model:

One issue that has bothered me is whether it is vulnerable to the criticism
that all that I have done is pick out some coefficients that generate the
behavior that I was looking for. And, maybe this is true. However, the
maximization based model of a consumer doesn't allow one to account for the
difference between a consumer's behavior in reaction to an increase or a
decrease in income.

My main concern has been that the empirical evidence for the difference in a
consumer's reaction to an increase as opposed to a decrease in income isn't
very strong-- not at least the what I could come up with. However, after
thinking about this for the last three years I finally came up with an idea
that may change this-- I should find out this week. If I could get some
reasonably strong empirical evidence for the effect it would provide a much
better basis for arguing that the use of control theory makes a significant
contribution.

However, The model purely on a theoretical basis has some appealing
features-- it handles time relationships in a consistent way-- where the
existing methods contain internal contradictions.

Bill Williams

[Bill_Powers1]

[From Bill Powers (2004.05.23.0737 MDT)]

Bill Williams 22 May 2004 1:15 AM CST --

I've read through your comments but not studied or considered as yet what
modifications to make in the code. As the date 2001 indicates the program
was written quite some time ago, so I have to study it to see what is going
on.

That solves one mystery -- why, when I saved the program into the Williams
directory in the TP directory, I was asked, "overwrite existing file?" At
least this time I looked at it more closely.

Clarity in coding is an obvious goal, however, when the only time that I
do any programming is when I have an idea about something to model-- and
this comes around about twice a year-- so you can see the problem.

Yes. This time let's stick with it until we're both sure of what we have.
Discussing code is a good way to debug it and understand what it's doing.

I'll have to study what you say about the environment. I guess I assumed
that the real numbers _are_ the environment.

There has to be some concept of what is physically going on behind the
numbers. What does the output number from the consumption control system,
o, mean? In other words, what do you visualize this control system doing to
its environment? What I see in my mind's eye is a person handing money over
to a store (at some rate, so many dollars per day or minute) and receiving
goods back at another rate, with the price as the factor converting from
dollars per day to goods per day. If the goods are consumed on the spot,
then the controlled variable would be another rate variable, the rate of
purchase minus the rate of consumption. The perceptual signal would
represent by its magnitude the rate at which goods are being consumed. The
reference signal would specify the intended rate of consumption. The error
signal would drive a leaky-integrator output function, to insert the
required single integration. This representation, however, leaves some
loose ends, like the question of what happens to goods if more are
purchased than consumed (in technical economic language, I suppose
"consumed" is not the right word here -- say "used.")

As I said, there is an alternative way to model this, and I'm going to
recommend that you try it.. You could say that the person has an inventory
of goods, which is being depleted by use and by depreciation. This
inventory increases at a rate equal to the rate at which goods are
purchased, and decreases at a rate determined by how many goods are
consumed per day (or minute or microsecond) and what fraction of the
remaining goods is lost to depreciation per unit time. Thye actual
inventory would be the sum of all increases and decreases up to the current
moment. The perceptual signal would now represent the quantity of goods on
hand at present, and the reference signal would specify the desired
quantity (any amount from zero on up, but not negative) The error would
drive the output proportionally (not through any kind of integrator)
because the cumulative inventory now plays the part of the one integral we
allow in the loop. Of course later you can play with added lags, but for
the initial design we should keep everything as simple as possible (but as
Einstein said, no simpler than that).

The controlled variable we call inventory, then, would be written like this
as a program step in a model of the environment of the control system:

inventory := inventory +
            (
               rate_of_purchase
             - rate_of_use
             - depreciation_factor_per_unit_time* inventory
            ) * dt

There are three rate variables inside the parentheses, each representing an
increase or decrease in the amount of goods in inventory per unit time. To
make sure of capturing all significant changes, we might choose hours as
the basic time unit. To make sure of smooth variations, however, we might
then make a single iteration of the progrem represent one minute, which
means that the rate per hour has to be divided by 60 to get the _amount_ of
change that occurs in one iteration. That's where dt comes in; we set it to
1/60. If you have goods coming in at the rate of 6 per hour, then to
compute the amount accumulating in one iteration you have to multiply the
rate by the time interval: 6 * dt, or 6*(1/60), or 0.1 units of goods in
one iteration. The inventory will increase by 0.1 units in this iteration
due to the influx of goods.

Similarly, if goods are being used as the rate of 5 per hour, in one
iteration we would find that 5/60 or 0.08333 units of goods are used up in
one iteration (one minute, 1/60 hour). The inventory would decrease by
0.08333 units in that iteration, for a net rate of change so far of +0.1 -
0.08333 = +0.0167 units per iteration.

Finally, the depreciation rate is proportional to the amount in inventory
-- a certain fraction of current inventory is lost per hour -- say it's
0.0001 per hour. If inventory is only 100 units, the loss is 0.01 unit per
hour and 0.000167 per iteration.. If inventory is 10,000 units, the loss is
10 per hour or 0.0167 per iteration. In fact, if the inventory had
increased to 10,000 units, net change of inventory per iteration would now be

    + 0.1 (from the purchase rate)
    - 0.08333 (from the usage rate)
    - 0.0167 (from the depreciation factor and inventory)

···

____________
    = 0 units of goods per iteration.

The inventory would be constant, losing as much per iteration as is gained.
If the inventory fell, the depreciation rate would fall and there would be
a small surplus of intake. The inventory would rise toward 10,000; If the
inventory increased beyond 10,000, the depreciation rate would also
increase and the net change would be negative. Inventory would fall toward
10,000. So 10,000 is a stable equilibrium point as along as the purchase
rate and the usage rate remain the same.

HOWEVER, we're talking about a control system, not an equilibrium system.
If the reference level of this control system for inventory is 2000 units,
purchases will proceed at a high but rapidly slowing rate until the
inventory reaches 2000. At that point, the output, the purchase rate, will
drop to the level required to make up for usage and depreciation (without
having to know about either of those rates).

I suggest modeling just this one control system to see how these
relationships come about. If you like I'll lay out the basic computations,
but you may prefer to work that out for yourself, to make it stick better.
Here's an outline as a starter:

=================================================================
Controlled variable: Inventory of goods.

Reference level: desired size of inventory of goods

Output: rate of purchase of goods (positive only). Ignore source of money.

Disturbance 1: rate of usage of goods

Disturbance 2: rate of depreciation of goods.

Unit of time: 1 hour

Duration of one iteration: 1/60 hour

From that skeleton you should be able to construct the code for the whole
control system. Note that we don't need a slowing factor because of the
integration inherent in the inventory.

When we're sure that this control system is working right, we can turn to
the budget control system and make it work right by itself. Then we can add
the interaction between the systems: making purchases cost money. We may
then decide to add a second level, a higher system that considers the
realities of life, or a few of them, and adjusts the reference levels for
working to earn money, acquisition of goods, and maintenance of prudent
savings. We can add credit by saying you can spend a little more money than
you have up to a point, and whatever other details look useful or interesting.

I should have written up a "motivation" section explaining what the program
is intended to demonstrate.
Basically, the program is intended to show that using control theory it is
possible to model a typical consumer using a method other than the familiar
one that uses the principle of maximization.

I agree that this is a good aim. However, don't forget my warning to
modelers: Before you start using the model to prove anything, make sure
it's the model you had in mind, The mere fact that its behavior looks like
what you want is no guarantee that you've produced the model you thought
you wrote. Even if it seems to prove your point, it could be a hidden
iceberg that will sink your ship later when you -- or worse, someone else
-- discovers that the model, under new conditions, does not do at all what
you say it does .

In the present case, you thought you had written a consumption control
model and a budget control model -- but you hadn't. There is no budget
control, because the budget loop is not closed. There is only a budget
disturbance that decays to a negligible amount after only four iterations
(out of the 1280 iterations in a complete run). So this model, despite the
way it behaves, does not illustrate an interaction between a consumption
control system and a budget control system. I suspect that you had to do a
good deal of tweaking of parameters to get the curves to look the way you
expected them to look. However, that only goes to show how we can force a
model to work when it really shouldn't work. Your model is basically a
consumption control system, with no influence from budget to speak of. It
shouldn't have worked right. But as Jiminy Cricket sang, "Tweaking
parameters (wishing) will make it so."

You shouldn't berate yourself for these slips. Every modeler makes them,
and making them should be one of the most valuable lessons on the way to
accurate modeling. I certainly made enough of them -- how do you think I
learned how to talk about them?

Using control theory an alternative scheme can be implemented in which the
consumer's wants are more plausibly, not infinite, but quite a bit beyond
the rate of expenditure which the consumer's budget will allow.

Yes, this can be done -- when we have a model that actually shows the
relationship between consumption and budget control. We don't have that
yet, so it would be best to postpone elaborations.

And, in particular the VD5.PAS program generates a behavior through time
that is different than the one predicted by the orthodox model of
consumption.

Yes, but only because the parameters were adjusted to make the consumption
control system produce curves of the sort you thought the combined system
should produce. If you made the above claim with the model in its present
state, you would eventually regret doing so.

The model generates a rapid adjustment to an increase in income-- a square
pulse in VD5.PAS and a slow decrease of expenditure when income decrease.
There are a number of additional features of the VD5.PAS model of consumer
behavior that amount to predictions-- such as the overshoot of expenditure
on the upward side of the adjustment, and the lack of an overshoot on the
downward adjustment.

But all that doesn't happen for the right reasons, so it would be best to
lay this interpretation quietly aside until your grounds are sure. This is
the time during model building for moving slowly and making sure of
handholds and footholds, as well as pitons, ropes, and anchors. We can
arrive at a point where every detail of the model is defensible and
reasonable, and whatever it does grows from its basic organization rather
than only curve-fitting. Then you can draw conclusions and make
interpretations with confidence because you have understood everything
there is to understand about the model. And on the way you will probably
have improved your understanding of the system you're trying to model.

Best,

Bill P.

[Bill_Williams1]

Bill Williams 23 May 2004 2:20 PM CST]

To avoid the possible confusion that might be generated by considering

more than one issue at a time, proceeding "step-by-step, I have limited

myself to considering one area of disagreement-- that is the

"substansiveness" of the economic theory of valuation.

[From Bill Powers (2004.05.23.0737 MDT)]

Bill Williams 22 May 2004 1:15 AM CST --

>I've read through your comments but not studied or considered as yet what

>modifications to make in the code. As the date 2001 indicates the

program

>was written quite some time ago, so I have to study it to see what is

going

>on.

That solves one mystery -- why, when I saved the program into the Williams

directory in the TP directory, I was asked, "overwrite existing file?" At

least this time I looked at it more closely.

>Clarity in coding is an obvious goal, however, when the only time that

I

>do any programming is when I have an idea about something to model-- and

>this comes around about twice a year-- so you can see the problem.

Yes. This time let's stick with it until we're both sure of what we have.

Discussing code is a good way to debug it and understand what it's doing.

>I'll have to study what you say about the environment. I guess I assumed

>that the real numbers _are_ the environment.

There has to be some concept of what is physically going on behind the

numbers.

Here is point where we part company-- that is "What is it that is going

on behind the numbers that are used?" Is there a substances? Or, is

there anything that can be said to be "genuinely real" involved. My

answer would be that there is no "substance" but there might be a reality.

In economics there is nothing going on that _is_ physical. Now, what

is going on is _connected_ with the physical and biological world.

But, there is no physical metric involved. The commodities in economics

may be counted and measured in physical units quarts of milk, of whisky

or lubricating oil. However, the goodness or economic value of these

different quarts can not be added up by adding their _thingness_.

Using orthodoxy's conception of maximization what economists say they

are doing is adding up how much utility is contained in each good.

This isn't in my view an adequate approach to measuring economic value.

For one thing it generates paradoxes such as the Giffen case.

The Giffen case has some interesting things to say about economic valuation.

There is of course in the Giffen case a physically measurable quantity

involved-- the calorie. In thinking about economic value everyone is

constructed so that they need a certain number of calories to survive.

But there is more to life than just calories. When we modeled the Giffen

case, I at least was assuming that the model included two very different

kinds of good. One was the calorie, but there was also the notion of good

or at least better taste. As I talk about the Giffen case now, I introduce

two goods, one a cheap and bitter tasting gruel, and also a more expensive

but sweet tasting gruel. So, there is a problem, how for the consumer to

first, or at least more urgently, obtain sufficient but not an excess of

calories. An excess might seem to be OK, but over time a constant excess

would result in the consumer being crushed under a pile of fat. Now,

distinct from this physiological requirement, for calories there is the

question of taste. Tasting good may have some physiological implications.

Our taste for salty foods, for sugar, and for fats would seem to have

physiological connection to an evolutionary past in which salt could be

difficult to obtain, where sugar was connected not only to the caloric

density of a fruit, but also to the vitamin content of a mature fruit,

and fats required the risky killing of animals whose meat was fatty.

But, these requirements or at least the desirability of salt, and sugar

and fat do not necessarily have a common metric with the calorie. There

is a connection here to recent work in biology concerning multi-level

evolution. Biologists in the not too distant past used to borrow the

economists notion of maximization as an principle in evolutionary

theory. And, the orthodox economists were happy to point to biology

to show that maximization was "science." Now, or at least more recently

both in economics and evolutionary theory some people are thinking that

economic value and evolutionary fitness should be thought of in terms

of a multi-level analysis. And, by multi-level one of the things they

mean is that there can be no commensurable unit that makes a connection

between the levels.

It might be possible to construct a rigorous proof that the way we

modeled the Giffen paradox can not be handled in terms of a common

commensurable unit of value. If, and I am assuming this it is true,

then value theory in economics will have to undergo a drastic change.

I don't at all think that this means that a value theory that takes

into account that there are values of different kinds can not be

constructed. But, it does take the analysis to a place that is very

different one where it is assumed that economic value is a substance.

According to my understanding there are decisive reasons and even

empirical evidence why economic value must be _non-substantive_.

This answer to the question, which begins here with your assertion

that

There has to be some concept of what is physically going on

behind the numbers.

is that economics is not physics. Economics is closer to biology,

but it really isn't biology either. The Giffen case appears as if

it might be closely connected to biology, but the phenomena involved

can be observed where the commodities involved are markers for social

standing rather than physiological issues.

I take your point about the desirability of attaching "something real"

as a dimension to represent consumption. Eventually I think it will

be possible to construct a value and price index using control theory

that will have a "reality" in the sense that the indexes will pertain

to economic valuations in a way that is consistent with the ways in

which people value services and commodities. To construct such an

index will require an understanding of human behavior in an economic

context of the sort revealed by the Giffen effect, and many other

behaviors that are paradoxical when viewed from the standpoint of

orthodox neo-classical economics-- and its assumption that economic

value is a substance.

A control theory approach to economic behavior makes possible an

analysis of economic behavior which avoids the many paradoxes that

pop-up when it is assumed that valuation, all valuation, can be

measured in terms of some commensurable substantive unit.

The difficulty I had with understanding what you meant by the

"environment" had its source in my not being aware that you were

thinking about the economic environment in physical terms.

Bill Williams

[Bill_Powers1]

[From Bill Powers (2004.05.23.1510 MDT)]

Bill Williams 23 May 2004 2:20 PM CST--

To avoid the possible confusion that might be generated by considering
more than one issue at a time, proceeding "step-by-step, I have limited
myself to considering one area of disagreement-- that is the
"substansiveness" of the economic theory of valuation.

There is, as you are aware, also a PCT theory of values.I don't know if
it's the one you're searching for.

There are two ways to consider value. One is to say that something is
valuable if an organism adjusts its reference level for that thing to a
high setting. The other is to say something is valuable if it objectively
promotes the survival of the organism. Obviously, organisms can set high
reference levels for things that are objectively bad for their survival,
and fail to set high reference levels for things that are good for their
survival. But in most cases, for organisms that manage to survive, the
organisms set reference levels in the appropriate way.

I think the general evolutionary position is the one to use here. We can
say that those organisms that have survived are those that learned to set
high reference levels for things that are actually good for them, and low
reference levels for things that are actually bad for them, where good and
bad in the sense intended are defined by survival of a species and failure
to survive. the reference signals specify the values that direct the
organism's behavior.

This allows us to use reference levels as the operating definition of
value, with the understanding that organisms generally set reference levels
as appropriate for survival.

Here is point where we part company-- that is "What is it that is going
on behind the numbers that are used?" Is there a substances? Or, is
there anything that can be said to be "genuinely real" involved. My
answer would be that there is no "substance" but there might be a reality.

I think you are reading more philosophical depth into my statement than I
intended. When I say there has to be something going on beneath the
numbers, I mean that the number representing the output of the control
system model has to stand for some actual process by which (in this case)
the consumption control system affects the environment so as to maintain or
change the current rate of consumption. The output action of the system
resulots in executing transactions at a certain rate, entailing spending
money at a certain rate and receiving goods are the corresponding rate
(spending rate divided by average price per good). The actual output
variable can thus be identified as the rate of spending money, which takes
place in the environment of the control system. This action affects the
environment so as to cause goods to appear in the consumer's inventory.

So far this has nothing to do with value. The subject of value comes up
when we ask why this entity is acquiring goods. In PCT, the simplest
one-level answer is simply "Because this control system has a reference
level for possessing a certain inventory of goods." That completely
explains the actions and the fact that the actual inventory will approach a
particular value and stay there (if possible). Of course it does not
explain why the reference level for these goods has been set to a
particular level. That question opens up the whole hierarchy of control
systems: the ends toward which this control process is only a means, and
the even higher levels for which even those higher ends are only means.

And the hierarchy itself, however many levels it may contain, is finite, so
we eventually come to a highest level of reference signal, and no longer
can explain its setting by referring to still higher systems. In my scheme
this is where the reorganizating system comes in and connects to what I
call "intrinsic variables." That is also where the evolution of species
comes in.

My hypothesis (adopted from Ashby) is that there is a set of basic
variables which serve as markers for the life support systems. There are
enough of them that if they are all kept close to particular values,
chances are that the physical organism is working properly in all important
regards. More to the point, any deviations of these basic variables from
what we may as well call their "intrinsic reference levels" indicates that
something is going wrong that needs to be fixed. I have not tried to
propose every possible way in which "intrinsic errors" of this kind might
be corrected, but have focused on one basic and especially powerful method
that I call "e. coli reorganization." This process alters the organization
of the brain at random, and ceases to alter it only when all intrinsic
variables are once again at their respective intrinsic reference levels.
Those intrinsic reference levels, of course, are a product of the evolution
of the species.

So I have proposed a mechanism by which the actual, objective needs of the
organism at a very basic level can end up establishing reference signals,
or relationships among reference signals, in the brain -- establishing the
values that the control systems in the brain seek to achieve via actions
on the environment.

In economics there is nothing going on that _is_ physical. Now, what
is going on is _connected_ with the physical and biological world.
But, there is no physical metric involved. The commodities in economics
may be counted and measured in physical units quarts of milk, of whisky
or lubricating oil. However, the goodness or economic value of these
different quarts can not be added up by adding their _thingness_.
Using orthodoxy's conception of maximization what economists say they
are doing is adding up how much utility is contained in each good.
This isn't in my view an adequate approach to measuring economic value.

I agree, because it doesn't make the connection between the objective needs
of the organism and the detailed reference conditions that the organism
pursues.
....

A control theory approach to economic behavior makes possible an
analysis of economic behavior which avoids the many paradoxes that
pop-up when it is assumed that valuation, all valuation, can be
measured in terms of some commensurable substantive unit.

Yes, I agree that searching for some common standard of value is futile.
Values do not inhere in the thing valued, but in the system that is able to
specify values for different things. There is no way to compare the value
of a symphony with the value of a chocolate soda. "Indifference curves" are
not up to the job. And a common metric would deny the fact that reference
signals are always being adjusted by higher systems: values are not fixed.

But let's not confuse that with the physical nature of PCT. A reference
signal is (if the model is correct) a real physical thing inside a brain,
not an abstract concept. The actions by which we make the environment
change to move our perceptions closer to their reference levels are not
abstract, but ordinary physical actions involving forces and masses and the
expenditure of energy. They consist of such simple processes as getting out
a wallet and passing money across a counter, and reaching out to accept a
package to take home. PCT applies to all levels of organization, not just
the higher ones. It forms a bridge between thought and action and it gives
abstract-seeming ideas like wants and intentions (and values) actual
physical effects on the world.

The difficulty I had with understanding what you meant by the
"environment" had its source in my not being aware that you were
thinking about the economic environment in physical terms.

To me, there is no difference in kind between the organization of control
systems in the brain and the physical entities in the world outside the
body, the environment; they all coexist in the only world there is. Inside
my models, there are "mental" processes like wanting and comparing and
perceiving (and valuing), but just outside them, the outputs of the model
enter the ordinary physical world where they cause things to happen, and
among those things are the happenings that are represented in our
perceptions, back inside the model. This keeps the model from floating off
into imaginary conceptual spaces like a hot-air balloon, georgeous but
insubstantial: it is anchored in the world of experience at both ends: in
the experiencer, the brsin, and in the world outside the brain.

So you needn't worry that I'm leading you down a garden path; my remarks
concerning numbers had merely to do with the physical process by which we
take possession of goods by handing money to the seller. The money and the
goods are aspects of the physical environment (even if the money is only a
mark in a ledger or a magnetized domain on a disk drive). Our perceptions
of the goods, the reference signals that give them value, and the
comparison that leads to output action, are part of the brain's operation
-- which is part of the physical world, but a rather specialized part.

Best,

Bill P.

[Richard_Marken2]

[From Rick Marken (2004.05.23.1700)]

Well, I spent almost my whole afternoon doing this so I thought I would
post it to show how one could go about modeling a budget and
consumption controller using a spreadsheet.

The spreadsheet contains a model of two control systems, one
controlling cash on hand ("Budget Reserve" or B measured in $) and one
controlling consumption of goods ("Consumption" or C measured in Lbs).
The blue cells are for setting parameters. You can enter numbers into
these cells to see how changes in parameters (identified in the
adjacent cells) affect the behavior of the model. Among the parameters
you can change are the references for B and C. You can also change the
gain of the C control system (a proportional controller). You can also
change Income/cycle, i, which is the amount added on each cycle to B.
You can also change the amount of goods that can be purchased/$ of
spending (c/$) and the amount by which the goods depreciate, dc (which
is the loss of consumed goods, C, per cycle, measured in Lbs).

The parameters are currently set so that the budget will steadily grow
and consumption can be controlled (C can be kept close to the reference
for C). If the parameters are changed so that the budget, B, declines
(one way to do this by reducing Income/Cycle, i) then the system loses
control of C.

The budget controller is not a very good controller because it can't
adjust income, i, to make up for loss of budget, B, due to spending on
consumption. Income is fixed. The budget controller will stop spending
when B goes below the reference for B so there is some minimal control.
But as soon as the budget starts increasing again it is quickly spent
by the consumption controller. So things go poorly when income is not
sufficient to keep the budget nearly constant or increasing. A more
interesting budget controller would be one that increases it's efforts
(output) to increase it's income when the budget starts going below the
reference. The effort required to increase income could be added to
the depreciation on consumption, dc. This could make for interesting
demonstrations of the trade off between working harder to allow control
of budget and the need for increased consumption to make up for the
depletion of resources involved in working harder.

The "Run" button takes the model through 100 cycles (iterations) of the
control loops. The states of the two controlled variables (B and C) are
plotted after the run. So you can see how changes in parameters affect
the behavior of the controlled variables by changing parameters and
pressing "Run".

I written the equations for the cells next to the cells in an effort to
make it easier to see what is happening in the control systems and in
the environment. Environmental variables are below the double
horizontal line; control systems signals are about this line. I've
tried to use capital letters for "stock" (accumulated) variables (B, C)
and lower case for "flow" variables (i, dc).

Best regards

Rick

ConsumeBudget.xls (40 KB)

···

---
Richard S. Marken
marken@mindreadings.com
Home 310 474-0313
Cell 310 729-1400

[Bill_Williams1]

Bill Williams 23 May 2004 5:45 PM CST]

[From Bill Powers (2004.05.23.1510 MDT)]

Bill Williams 23 May 2004 2:20 PM CST--

I hope you won't take offense that while I am sure you consisder the
argument that you present {1510 MDT) is correct and adaquate, and that the
argument suits you, it doesn't suit me. My principle objection is remains
that value is not a substance. I don't find a materialist argument
convincing. Materialism as, I see it, is perhaps the most , or at least one
of the more naive forms of idealism. Since you define PCT in terms of a
materialist conception of the world, I don't find PCT convincing.

I do, of course, recognize the value especially in combating behaviorism of
emphasizing the control of perception. And, I don't recall having found
fault with the specific features of HPCT, or perhaps any of your work
concerning behavior as a whole, aside perhaps from the idea of the
univerial error curve. I am to some extent will to listen to alternative
views, but basically control theory provides me with the tools I need to do
some useful stuff in economics. I do not, however, find some of your
philosophic interpretations appealing. And, when you venture into economics,
where these interpretations become norms specifying what is meaningful and
what is not, we part company.

I simply don't find your materialist preconceptions helpful. In my opinion,
an adaquate economic theory, or any theory of the social process requires a
pluralist rather than a monistic conception of the process of valuation. As
best I can judge a monistic conception of value theory generates familiar
and characteristic mistakes in application.

>To avoid the possible confusion that might be generated by considering
>more than one issue at a time, proceeding "step-by-step, I have limited
>myself to considering one area of disagreement-- that is the
>"substansiveness" of the economic theory of valuation.

There is, as you are aware, also a PCT theory of values.

Of course. Not only am I aware that there is such a thing, but I also am to
some extent familar with the errors which it contains. The PCT theory of
values, isn't one that I need to consider adopting.

I don't know if it's the one you're searching for.

Who said anything about me being " in search " in the sense you mean of
anything? This very sentence is I think an indication of what goes wrong
with behavior that presupposes a monistic theory of value.

However, to correct a possible misapprehension on your part-- No, a
monistic theory of value is not what I "am searching for" if, that is, I was
in the process of searching . The way you treat values is so inchooherent
that it really isn't worth my thinking about adopting.

There are two ways to consider value.

Actually there are lots more ways of thinking about this than you might
suppose.

One is to say that something is

valuable if an organism adjusts its reference level for that thing to a
high setting. The other is to say something is valuable if it objectively

Notice the influence of positivism here. What function "objective function"
does the use of "objectively" serve in this sentence? None. The sentence
would functionally mean the same if the word is removed. So, what function
does the word "objectively" serve? It serves an ideological function of
announcing that the discussion is being held in terms of an idealistic
philosophic conception known as "materialism." I can do without
materialism.

promotes the survival of the organism. Obviously, organisms can set high
reference levels for things that are objectively

Notice the influence of positivism here. What function "objective function"
does the use of "objectively" serve in this sentence? None. The sentence
would functionally mean the same if the word is removed. So, what function
does the word "objectively" serve? It serves an ideological function of
announcing that the discussion is being held in terms of an idealistic
philosophic conception known as "materialism." I can do without
materialism.

bad for their survival,

and fail to set high reference levels for things that are good for their
survival. But in most cases, for organisms that manage to survive, the
organisms set reference levels in the appropriate way.

I think the general evolutionary position is the one to use here.

I agree, as long as objectivism, materialism, and physicalism are not
introduced as implicit assumptions.

We can

say that those organisms that have survived are those that learned to set
high reference levels for things that are actually good for them, and low
reference levels for things that are actually bad for them, where good and
bad in the sense intended are defined by survival of a species and failure
to survive. the reference signals specify the values that direct the
organism's behavior.

This allows us to use reference levels as the operating definition of
value,

We part company here. I would define "value" in terms of consequences. Any
particular reference level may have good or bad consequences. So, I do not
agree with your definition of value.

with the understanding that organisms generally set reference levels

as appropriate for survival.

I would agree that typically this is what organisms in fact do. However,
typically, isn't good enough to serve as a foundation definition.

>Here is point where we part company-- that is "What is it that is going
>on behind the numbers that are used?" Is there a substance? Or, is
>there anything that can be said to be "genuinely real" involved. My
>answer would be that there is no "substance" but there might be a

reality.

I think you are reading more philosophical depth into my statement than I
intended.

Perhaps, however, as I percieve it , you are expounding a view that is
lacking in sufficient depth to serve as an adaquate basis for a viable
theory of value.

When I say there has to be something going on beneath the

numbers, I mean that the number representing the output of the control
system model has to stand for some actual process by which (in this case)
the consumption control system affects the environment so as to maintain

or

change the current rate of consumption.

Where we seem to disagree is in part located in the above in which you say,
"some actual process." I would agree that there being "some actual
process." Where I would disagree is a matter of what and how that "actual
process" is to be specified. There are I believe good reasons to reject a
substance theory of value, and a monistic conception of reality. I will
return this point below when you agrument touches on this point again.

The output action of the system

resulots in executing transactions at a certain rate, entailing spending
money at a certain rate and receiving goods are the corresponding rate
(spending rate divided by average price per good). The actual output
variable can thus be identified as the rate of spending money, which takes
place in the environment of the control system. This action affects the
environment so as to cause goods to appear in the consumer's inventory.

So far this has nothing to do with value.

You have already introduce the notion of "goods." So, your claim that
nothing so far has anything to do with value would appear to be
contradictory.

>The subject of value comes up

when we ask why this entity is acquiring goods.

No, you already introduced value before this point. You can not talk about
goods, not consistently at least, without getting into the value issue.

In PCT, the simplest

one-level answer is simply "Because this control system has a reference
level for possessing a certain inventory of goods." That completely
explains the actions and the fact that the actual inventory will approach

a

particular value and stay there (if possible). Of course it does not
explain why the reference level for these goods has been set to a
particular level.

Yes.

That question opens up the whole hierarchy of control

systems: the ends toward which this control process is only a means, and
the even higher levels for which even those higher ends are only means.

Right.

And the hierarchy itself, however many levels it may contain, is finite,

so

we eventually come to a highest level of reference signal, and no longer
can explain its setting by referring to still higher systems. In my scheme
this is where the reorganizating system comes in and connects to what I
call "intrinsic variables." That is also where the evolution of species
comes in.

There is much in what you say above that I am in agreement with. I've been
aware of this for quite some time. I wasn't as you say "searching" for such
a conception. I have been familiar with these conceptions for for a decade
or two and I continue to find them valuable.

My hypothesis (adopted from Ashby) is that there is a set of basic
variables which serve as markers for the life support systems.

I would agree with this.

There are

enough of them that if they are all kept close to particular values,
chances are that the physical organism is working properly in all

important

regards.

"Physical organism" seems redundant to me, and thus seems to me to be
introducing an ideological twist into the discussion.

More to the point, any deviations of these basic variables from

what we may as well call their "intrinsic reference levels" indicates that
something is going wrong that needs to be fixed.

Usually this is the case.

I have not tried to

propose every possible way in which "intrinsic errors" of this kind might
be corrected, but have focused on one basic and especially powerful method
that I call "e. coli reorganization." This process alters the organization
of the brain at random, and ceases to alter it only when all intrinsic
variables are once again at their respective intrinsic reference levels.
Those intrinsic reference levels, of course, are a product of the

evolution

of the species.

Right. And, I think in almost all most all of the above you are making a
useful point.

So I have proposed a mechanism by which the actual, objective

Here, in the use of "objective" it appears to me that you are introducing a
slant into the argument which is unnecessary. I am not an "objectivist."
You could omit "actual" and "objective." Would , or how would, doing so
change the meaning of the sentence?

needs of the

organism at a very basic level can end up establishing reference signals,
or relationships among reference signals, in the brain -- establishing the
values that the control systems in the brain seek to achieve via actions
on the environment.

>In economics there is nothing going on that _is_ physical. Now, what
>is going on is _connected_ with the physical and biological world.
>But, there is no physical metric involved. The commodities in economics
>may be counted and measured in physical units quarts of milk, of whisky
>or lubricating oil. However, the goodness or economic value of these
>different quarts can not be added up by adding their _thingness_.
>Using orthodoxy's conception of maximization what economists say they
>are doing is adding up how much utility is contained in each good.
>This isn't in my view an adequate approach to measuring economic value.

I agree, because it doesn't make the connection between the objective

needs

There is a sense in which I agree with you about the character and nature of
needs. I think your system provides the best explaination of this that is
availible.
I don't think, however, that it is useful to describe this
structure of behavior in terms of "objective needs."
Your talk about "objective needs" is in my view a unnecsessary injection of
a materialist ideology.

The term "objective" which you use here is a misnomer. The "needs" are not
in fact all needs for objects. Some of the needs are for relationships.
Positivism, it seems to me, puts barriers that are unneccessarily.

of the organism and the detailed reference conditions that the organism
pursues.
....
>A control theory approach to economic behavior makes possible an
>analysis of economic behavior which avoids the many paradoxes that
>pop-up when it is assumed that valuation, all valuation, can be
>measured in terms of some commensurable substantive unit.

Yes, I agree that searching for some common standard of value is futile.

This isn't quite what I said, or at least quite what I intended to
communicate.

Values do not inhere in the thing valued, but in the system that is able

to

specify values for different things.

I think this statement may contain a subtle mistake, but I don't think that
in the present discussion it is one that needs to be considered now.

There is no way to compare the value

of a symphony with the value of a chocolate soda. "Indifference curves"

are

not up to the job. And a common metric would deny the fact that reference
signals are always being adjusted by higher systems: values are not fixed.

Yes. I think that we are in complete agreement here.

To expand on this: There may be a basic contraction in the idea that we
construct a simple measure of value-- such as a price index -- and expect
that it will provide a measure of value such that incomes and commodities
from 1900 can be compared to incomes and commodities in 2000. I am not
saying that such a comparison is either impossible or meaningless but there
are formitable difficulties involved.

But let's not confuse that with the physical nature of
PCT.

Nor should you confuse economics with physics.

One of the reasons that I do not use the caption PCT is that I have been
aware of how you conceive of PCT in physical terms. There is, of course, a
physical aspect to control theory, but this physical aspect can be thought
of without bringing in a monistic, objectivist, philosophic world view.

A reference signal s (if the model is correct) a realhysical thing inside

a brain,

not an abstract concept.

I do not mean to be disrespectful in anyway. However what you are saying
here does not appear to me to be useful. This statement in my view is
precisely what a positivist philosophic stance requires. I simply don't see
any reason for you to place your work in a positivist context.

All concepts including "materialism" are concepts. And, the materialists
are among the purest idealist around.
Furthermore, all concepts are "absract."

The actions by which we make the environment

change to move our perceptions closer to their reference levels are not
abstract, but ordinary physical actions

Ordinary Perhaps. But, there is no need for me to think about this is
positivistic terms.

involving forces and masses and the

expenditure of energy.

I have no problem, rather I would insist that value relations involve these
physical conceptions, but value is neither not a substance.

They consist of such simple processes as getting out

a wallet and passing money across a counter,

Here is where I think your positivism gets you in trouble. "Passing money
across a counter" isn't simple. Thinking that it is simple when it is not,
creates a problem.

and reaching out to accept a

package to take home. PCT applies to all levels of organization, not just
the higher ones. It forms a bridge between thought and action and it gives
abstract-seeming ideas like wants and intentions (and values) actual
physical effects on the world.

I can agree in part to this. But, my agreement does not extend to the point
where the description takes on a positivistic interpretation.

I don't disagree with you that PCT provides a description of behavior in
causal terms or physical terms. What I am arguing is that the category
"value" is not a substance.
Economics is not physics.

>The difficulty I had with understanding what you meant by the
>"environment" had its source in my not being aware that you were
>thinking about the economic environment in physical terms.

To me, there is no difference in kind between the organization of control
systems in the brain and the physical entities in the world outside the
body, the environment; they all coexist in the only world there is.

If the only world that you recognize is a positivist one ( I am not sure at
this point if you think of yourself as a positivist but it appears likely )
and the only world is a materialistic one then it would appear that you deny
the reality of values. You seem to be a monist. I think I prefer to be a
pluralist in the sense that it appears useful to me to think in terms of a
world view in which there are different kinds of values. One of the things
that I find distasteful about orthodox econmics is that really only has one
good-- utility or whatever it is that maximization maximizes.

Inside

my models, there are "mental" processes like wanting and comparing and
perceiving (and valuing), but just outside them, the outputs of the model
enter the ordinary physical world where they cause things to happen, and
among those things are the happenings that are represented in our
perceptions, back inside the model. This keeps the model from floating off
into imaginary conceptual spaces like a hot-air balloon, georgeous but
insubstantial:

The accusation of "insubstantiality" doesn't frighten me. I don't think of
experience including values within the category of experience, in terms of a
monistic materialism. And, as far as I know all "conceptual spaces" are
immaginary.

it is anchored in the world of experience at both ends: in

the experiencer, the brain, and in the world outside the brain.

There is, however, no place in your world for values-- not at least I as
think of values. For you if it not "substantial" then it isn't valuable.
The difficulty arises when the notion of "substantiality" is itself
examined. I may have said it above, but it is worth repeating-- the most
dedicated idealists are materialists. And, the monistic materialists are
the most pure of heart in this respect.

What you say has an appealing plausiblity. I would agree that "experience"
is connected to a reality or realities. But, experience is not a substance.
What experience is has yet to be worked out with any adaquacy, but I think
control theory can be helpful. Experience has an unavoidably economic
aspect to it. The orthodox conception of the economy has some problems--
such as the persistent problem of the Giffen paradox. And, control theory
provides what think is a good explaination for the behavior that is
considered paradoxical in the orthodox context. One of the things that
attracted me to the paradox-- besides its being a fundamental issue in
economics-- is that the behavior of an economic agent was obviously tied to
the calorie.

So you needn't worry that I'm leading you down a garden path;

Said the spider to the fly. But, there isn't much danger of you leading me
into a materialist monism.

my remarks concerning numbers had merely

"Merely?" Come on, it was a hard sell for stale snake oil. Fess up to
ideological creedo as a monistic materialist.

to do with the physical process by which we

take possession of goods by handing money to the seller. The money and the
goods are aspects of the physical environment (even if the money is only a
mark in a ledger or a magnetized domain on a disk drive).

I acknowledge that when modeling it is neccesary to be explicit about what
it is that is being modelled and that I should be more careful in attaching
dimensions to the terms being modelled. However, "Economics is not
Physics." Positivism, objectivism, materialism have not provided an
adaquate basis for economic theory.

Our perceptions

of the goods, the reference signals that give them value, and the
comparison that leads to output action, are part of the brain's operation
-- which is part of the physical world, but a rather specialized part.

I have found this discussion most informative. Not that I learned anything
new about control theory in the course of it. But, I never would have
suspected that you take a monistic materialism so seriously. I don't think
there is much of any chance of your converting me to your old time religion.
And, I am quite sure that you are not interested in considering switching to
what I view as a more adequate set of preconceptions.

The result? For the time being value issues such as how we compute a price
index, how we evaluate the worth of a commodity, are in a state of flux, and
so is the conception of the economic environment. So, for the time being
their isn't anything that is any better than defining consumption as a real
number. We can talk about it, but we can't really define it, and we can't
measure it-- we pretend to, but really we don't have much of any idea.

Bill Williams