[From Bill Powers (2011.07.18.1625 MDT)]
Adam Matic 2011.07.18 2223 CET]
Rick Maren (2011.07.18.1120)
It’s unlikely that you would have seen it in your business. The
Giffen
effect is expected only when people have a very limited budget; ie.
they are poor.AM:
In what market conditions? It seems relevant that buyers have only one
choice of rice and meat, and only one source of income. What if there are
competing producers with different prices?
The requirements for the Giffen Effect to occur are not complicated, and
as you indicate they can appear under many conditions.
The first condition is that the person is operating in such a way that
expenditures are close to the amount of income. You could split this into
budgets for this and budgets for that and look for Giffen Effects in each
budget area. There can also be Giffen effects arising from the existence
of two or more budget areas with a limit on total budget.
The second condition is that each good can satisfy more than one
requirement; a specific make of car, for example, can supply desired
transportation, and also can affect one’s prestige in the community one
wishes to impress, one’s safety in traffic, expenditures on gasoline, the
size of garage one needs to house it, and the cost of insurance,
registration, and repairs.
The third condition is that the importance given to goods is not
systematically related to the prices of the goods.
If the more important good is also the most expensive, and this
relationship holds from most to least important, the traditional law of
supply and demand will hold, and so will the interchangeability of goods.
That’s my instinctive guess; I haven’t worked this out in
detail.
However, if the most important good costs less than a more expensive good
that will satisfy the same requirement, but doesn’t supply something
desirable that the more expensive good does supply, the law of supply and
demand will work backward for the most important good. A rise in price of
the lower-priced good will cause a shift to buying more of it, and less
of more expensive goods,
Martin Lewitt (2011 July 18 1509 MDT) notes that the Giffen Effect also
works in the other direction: “Poor people using all their funds for
a unpalatable but affordable life sustaining food. Then that food drops
in price, perhaps due to a good harvest. They can now afford to forgo
some of that lower priced food and purchase a more palatable but more
expensive item instead.”
Nicely worked out.
This exercise is a first step toward a full analysis of the supply-demand
relationship. There are actually many goods, each of which can affect
more than one controlled variable, and each control system will act by
buying various amounts of all goods that affect its controlled variable.
To minimize total error in the collection of control systems, a
simultaneous equation must be solved, the solution specifying how much of
each good must be purchased at its current price. There will be no simple
relationship between price and amount purchased for any one good, or
between price changes and changes in the amount purchased – expecially
if we set up a model with many actors in it having a variety of
preferences.
Bill Williams’ model, having only one actor, two goods, and two
controlled variables, shows the possibility of the Giffen Effect. But the
Giffen Effect is only a very simple example of a much more complex
situation.
Fortunately, a control system model with many controllers, many inputs to
the controlled variable, and many outputs each affecting some or all of
the controlled variables, can solve these simultaneously equations
automatically.
I have resurrected an old Delphi program that I demoed in Los Angeles –
I think in 2002. This program can handle up to 500 control systems
each controlling a different controlled variable. The controlled
variables are perceptions generated by sensing different weighted sums of
500 environmental variables, and the output of each control system
affects, via another set of weights, the same 500 environmental
variables. The environmental variables can be the goods, and the
perceptions can be the controlled variable affected to varying degrees
(shown by the input weights) by each good. The output weights are
adjusted for stable control, which is done simply by making each control
system have an integrating output function, and making the matrix of
output weights be the transpose of the input weight matrix (demonstrating
that was the point of this demo). We could add a random array of prices
for the 500 goods, and another array of “importances” (gains)
for the control systems.
Note that this model does NOT assume that the controlled variables are
the quantities of the goods (or services). They are effects of obtaining
those goods and services, the effects being the real reasons for buying
anything. To be sure, one has a goal of obtaining food, and in order to
obtain it one trades money or work for it. But that’s not the end; the
reason for obtaining food is to fend off or reduce hunger or give it to
someone or enjoy the taste or feel proud of one’s status or any or all of
a number of other consequences of obtaining the food.
There are many ways to branch out from this model, but I think it will
provide a starting point for systematic development. And it will show
clearly that there is no simple law of supply and demand, and no simple
rule for deciding how much of one good provides as much benefit as one
unit of another good.
Best,
Bill P.
···