# Giffen Effect as Good Control

[From Rick Marken (2002.11.14.1050)]

Bill Powers (2002.11.14.0647 MST)--

There must also be a preference
for the good with the higher price per unit value, so the person always
buys as much as possible of it, filling in with the other good when the
budget is tight to keep the total value at the reference level.

Yes. My quick experiments with the model suggest that the preference for the
higher price per value commodity (in terms of the amount consumed when there are
no budget constraints) must be at least equal to the preference for the lower
price per value commodity in order for the Giffen effect to occur.

You will notice that in your clever demo, the preference for meat lies in
the person doing the demo rather than in the conditions of the demo. If you
satisfy your caloric requirement with bread alone, which you can do in your
demo, there will be no Giffen effect.

Yes. Correct.

>So the Giffen effect will only be shown by a system that has learned the
>_skill_ of controlling both V and B by properly increasing consumption of the
>good with a lower price to value ratio when the price of this good increases.
>This is likely to be a _learned_ skill.

Perhaps you can see now that this is not true. If you have the required
needs and preferences, the Giffen Effect _will be seen_.

Actually, I think it still is true, even if the system has the required needs and
preferences. What you need besides the required needs and preferences are the
correct connections between control systems and outputs. For example, in my model
of the Giffen effect, consumption of meat (X in your equations) is proportional to
the sum of the outputs of both the calorie (V) and budget (B) control systems
(Ov+Ob). Consumption of bread (Y in your equations) is proportional to the
difference in output of the calorie (V) and budget (B) control systems (Ov-Ob).
When the systems are connected in this way, an increase in the price of bread (Y)
causes the output of the budget system, Ob, to go negative (saying "buy less")
with the result that there is a _decrease_ in meat (X) consumption and an

This is what I meant when I said that the system must be "wired" appropriately to
show the Giffen effect. If the system is not wired appropriately it will not show
the Giffen effect even if it has the required needs (for calories and staying
within budget, in this case) and preferences (bias for consuming meat relative to
bread). For example, you could have an "inappropriately" wired system whose
consumption of both meat and bread is determined by the sum of the outputs of the
calorie and budget control system. This system works fine (meets its caloric and
budgetary requirements) as long as it's budget always exceeds the cost of
commodities needed to control its caloric intake. But when this system's budget
goes below this level it's SOL. Depending on the details of the weightings of the
outputs of the calorie and budget control systems it will lose control of budget,
calories or both.

That's why I conclude that the Giffen effect is likely to be displayed only by
systems that have lived within the constraints of budget and expense that are
associated with observation of the Giffen effect. Such systems will have learned
(become "wired up") so that they will take the rather non-obvious action (buying
more of a product that is _increasing_ in price) necessary to keep both their
budget and their caloric requirements under control.

Best regards

Rick

···

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

[From Bill Powers (2002.11.14.1250 MST)]
Rick Marken (2002.11.14.1050)–
I’ll meet you halfway on this one. Yes, the system has to be wired so as
to reduce meat consumption and/or increase bread consumption when the
expenses tend to go over budget. That probably does requires some
learning.
However, I don’t think there is any other simple solution to the problem
of a conflict between budget and needs, so anyone who reorganizes is
likely to come across this solution eventually (what happens to those who
don’t?). The anaytical solution merely shows the conditions necessary for
both the budget goal and the value goal to be met simultaneously, given
the available choices, without discussing the “wiring” needed
to accomplish this end.
Attempting to generalize a bit, this effect turned up here when there
were two means to the same end, one of them being cheaper per unit value
than the other. Another condition was that the same means have other
effects which are also judged with respect to reference levels – that’s
what “preference” means. Apples and lemons can both satisfy
reference levels for calories or tartness, but lemons would be preferred
over apples if there were a reference level for vitamin C intake, or
freedom from scurvy. So apples would become the “giffen good”
relative to lemons; one would desire lemons to ward off symptoms of
scurvy, but below a certain income range, increasing the price of apples
would require purchasing fewer lemons and more apples in order to get
enough of the other value, which I called calories or tartness. So
does it seem that the Giffen phenomenon in general entails a curtailment
of whatever it is that is “preferred?” You can have all the
lemons you need to avoid scurvy at the same time you satisfy your need
for calories, provided you have enough income. If your income falls below
a critical level, and if there is another source of calories that
provides less vitamin C, you can meet your need for calories but only at
the risk of scurvy.
This isn’t the whole picture yet, but I think we can zero in on a general
way of describing the Giffen Interaction (how’s that for a term?).
Different goods satisfy different needs or desires, when obtained in
enough quantity. If income drops below the level needed to satisfy all
needs and desires, then only some of them can be fully satisfied, and the
Giffen effect will come into play when there are price
differences.
I’m sure that can be said better and more correctly, but it’s a start.
Bill W., I think this approach, if we can master it, will provide a very
sophisticated and useful alternative to the demand side of supply and
demand. Also, I wouldn’t be at all surprised if there were a counterpart
of this Giffen Interaction on the supply side. Start with two different
goods that are being made to generate producer income, the manmufacture
of one of which has other desirable side-effects for the producer. Now
manipulate the demand for the two products, and I’ll bet you could find a
situation where the producer would have to make more of the product for
which demand decreased. Maybe that’s not exactly the right way to
put it, but you get the drift.

Best,

Bill P.

···

[From Rick Marken (2002.11.14.1520)]

Bill Powers (2002.11.14.1250 MST)--

I'll meet you halfway on this one. Yes, the system has to be wired so as to
reduce meat consumption and/or increase bread consumption when the expenses tend
to go over budget. That probably does requires some learning.

However, I don't think there is any other simple solution to the problem of a
conflict between budget and needs, so anyone who reorganizes is likely to come
across this solution eventually (what happens to those who don't?). The
anaytical solution merely shows the conditions necessary for both the budget
goal and the value goal to be met simultaneously, given the available choices,
without discussing the "wiring" needed to accomplish this end.

I agree. I just think it's important to understand that the Giffen effect is not
_determined_ by circumstances. Conventional explanations of the Giffen effect (to
the extent that I can understand them, and that's not much) give the impression
that raising the price of certain goods ("Giffen goods") automatically causes
higher consumption. The violation of the expected relationship between price and
demand occurs because of some special property of "Giffen goods". At least that's
the way it _sounds_ to me.

I think it's important to understand that, from a control theory perspective, the
Giffen effect represents a learned approach to dealing with a conflict. And it's
not necessarily the case that everyone confronted with this conflict will learn
the "Giffen" approach to dealing with it. What happens to those who don't? They
become malnourished, they steal, they move, etc. All of these things have been
observed to happen just as often (perhaps more often) than Giffen consumption
(increased consumption of a lower price per value good as the price of that good
increases).

I think the lesson of the Giffen effect is that consumption is a means of
controlling perceptual variables within the constraints imposed by budget,
personal principles and intrinsic needs. In other words, consumption is not just
"a show put on for the benefit of an outside observer" (such as an economist).

Best regards

Rick

···

--
Richard S. Marken, Ph.D.
The RAND Corporation
PO Box 2138
1700 Main Street
Santa Monica, CA 90407-2138
Tel: 310-393-0411 x7971
Fax: 310-451-7018
E-mail: rmarken@rand.org

[From Bill Powers (2002.11.14.2051 MST)]

Rick Marken (2002.11.14.1520)--

>I agree. I just think it's important to understand that the Giffen effect
is not

_determined_ by circumstances. Conventional explanations of the Giffen
effect (to
the extent that I can understand them, and that's not much) give the
impression
that raising the price of certain goods ("Giffen goods") automatically causes
higher consumption. The violation of the expected relationship between
price and
demand occurs because of some special property of "Giffen goods". At least
that's
the way it _sounds_ to me.

As far as I can see, a "Giffen good" isn't any particular kind of good;
it's simply the cheaper alternative under the particular set of
circumstances (and goals, of course) that results in an increase of price
producing an increase in demand.

I'm with you, of course, in considering this a subset of a general study of
human nature.

>I think ... consumption is not just "a show put on for the benefit of an
outside observer" (such as an economist).

Let's not forget that there are many brands of economists, at least two of
whom are on our side. There may be more in the future if we are lucky (and
smart).

Best,

Bill P.

[From Rick Marken (2002.11.14.2030)]

Bill Powers (2002.11.14.2051 MST)--

Rick Marken (2002.11.14.1520)--

> I think ... consumption is not just "a show put on for the benefit of an
> outside observer" (such as an economist).

Let's not forget that there are many brands of economists, at least two of
whom are on our side. There may be more in the future if we are lucky (and
smart).

I think you have to recalibrate your radar again;-) I wasn't criticizing
economists in general or any economist in particular. Maybe I should have left
out the parenthesized part but I was talking about the kind of behavior
economists observe. But I could just as well have said "such as a psychologist"
or "such as me". It was said in that spirit.

view of consumption prevalent in economics. Consumption is done to control (well,
maximize, not quite the same thing) utility (U). So from this point of view
consumption is done for a purpose: the purpose of maximizing U, which is a
function of consumption. The problem, of course, is that this approach assumes
that all consumption is done to "control" one variable: U. The Giffen effect,
however, seems to require an explanation in terms of control of at least two
variables. Now that I think of, an article on the Giffen effect that I read
yesterday did, indeed, involve two utility variables, which they called U' and
U''. Maybe the complex equations I saw in that article boil down to the equations
you posted yesterday.

Best

Rick

···

--
Richard S. Marken
310 474-0313

[From Bill Powers (2002.11.16.0910 MST)]

Rick Marken (2002.11.14.2030)--

control >theory view of consumption prevalent in economics. Consumption is
done to >control (well, maximize, not quite the same thing) utility (U).

There's a very simple relationship between maximizing utility and
minimizing error, but it makes a huge difference which one you try to
perceive and control. Utility of a variable reaches a maximum at some value
of the variable, so as the variable passes through this value, the utility
rises to a peak and then falls again. Suppose that when the variable has a
value of 60, the utility is at 90% of maximum, and also suppose that
utility is the _only_ measure you have. Should you increase or decrease the
value of the variable? There's no way to tell: you must first make a small
arbitrary change in the variable and see whether the utility increases or
decreases, So every time you look at the measure of utility, you have a
50-50 chance of making the first correction the wrong way (as in tuning a
radio or manually focusing a camera).

On the other hand, if you know the maximum-utility value of the variable
beforehand, you can use that value as the reference level, and see
immediately whether the variable is too small or too large. The first move
is always in the right direction, and you achieve "maximum utility" without
having to bother with measuring utility. The whole concept of "maximizing
utility" is just the wrong one for explaining any kind of behavior.
"Correcting error" is far simpler to model, and to do..

Maximizing utility also has another problem: defining it. What is the
maximum utility achievable by drinking water? Obviously, this depends
greatly on how much water you've drunk recently. What's the maximum utility
of car-ownership? If owning one car provides a certain level of utility,
does owning two of them produce more utility, or less? There's no one right
answer. If you have a small budget, licensing and ensuring the second car
could be beyond your means so adding it is a net expense.

In the coming revolution in economics, I think one of the first concepts to
be discarded should be that of utility.

Best,

Bill P.

···

consumption is done for a purpose: the purpose of maximizing U, which is a
function of consumption. The problem, of course, is that this approach assumes
that all consumption is done to "control" one variable: U. The Giffen effect,
however, seems to require an explanation in terms of control of at least two
variables. Now that I think of, an article on the Giffen effect that I read
yesterday did, indeed, involve two utility variables, which they called U' and
U''. Maybe the complex equations I saw in that article boil down to the
equations
you posted yesterday.

Best

Rick
--
Richard S. Marken