From Bill Williams 21 January 2004 8:30 AM CST]
[Martin Taylor 2004.01.20.2300]
>[From Bill Williams 20 January 2004 7:00 PM CST]
>To start with say the rate of income is 100 units.
...
>. The analysis here is, as has been agreed by all parties, a
>simultaneous
>aggregate analysis.I'm having a problem understanding your discussion. Aren't these two
statements--income having a "rate" and analysis being
simultaneous--incompatible?If in some way they aren't incompatible, how is it possible to get
any kind of dynamical analysis whatever out of a static base?Sorry to be so dumb.
First, Not to worry. And, I am quite sure "dumb" is not a problem.
And, second: There are other's lurking about on CSGnet who are far
more qualified than I am to expound aspects of this stuff. Bear
with me a bit here and permit me to explain my qualifications, and
limitations in regard to this question, and its history.
Economists, or at least some eminent economists have been aware since
the 19th century that the problem of time was one of the principle
difficulties that had and was being experience in developing an adequate
theoretical foundation for economic analysis. Alfred Marshall (1895)
If you look at the very extensive correspondence that was a background
to the publication of Keynes _General Theory_ it is obvious that
questions about equation and sequence occupied a large portion of the
time and effort of the circle around Keynes. And, despite all the
effort they didn't really get it quite right. In Veblen's lectures
delivered at Harvard (1899-1900) "Preconceptions of Economic Science"
The problems of time, causation and agency make up one of the core
threads of Veblen's argument, and especially his conclusion. Joan
Robinson's Clark lecture (1972) was in large part devoted to the
issue of time in combination with the question of whether economic
agents can be assumed to have perfect foresight or not. And, another
of her most important papers was devoted almost entirely to the
question of time. John R Common's (1932?) book _Institutional
Economics_ ends with a comment about the question of "futurity" -- by
which Common's meant the complex of issues revolving about social
structure, causation, and human agency. I would add to this concern
of the economists, John Dewey's (1938) similar reflections expressed
in his chapter on "Sequence and Causation" in his _Logic_.
Throughout the 2Oth century work was going on attempting to sort
out the issues involved in correctly handling economic processes,
value theory, and human behavior in time. I entered into these
efforts in the mid-196O's when I was a student of J. F. Foster
at the University of Denver. In alignment with the work mentioned
above I wrote a Master thesis in 1969 entitled "Equilibrium and
Equation and Marshall and Keynes." Then as a doctoral dissertation
I wrote on Mathematical aspects of Veblenian Economics.
While the discussion above doesn't contribute anything, not yet,
to Martin's question concerning a perception that there may be
an inherent conflict between a simultaneous and a time rate
analysis, I think it does provide an indication that the problem
of the role of time in analysis has for more than a century been
perceived to be a critical problem for theoretical economics.
Representative of this literature is a paper by William Fellner
(1944) "Period Analysis and Timeless Equilibrium" Quarterly
Journal of Economics Volume 58 Number 2 February (p. 315-22.).
I think Fellner's paper may come fairly close to Martin's
question in that they share some similar assumptions about
there being limitations regarding the representation of change
in a systematic analysis-- particularly in a system of
simultaneous equations. Prefacing what I have to say about
this issue with an elaborate introduction is intended to
indicate that the question has considerable history. Even
Adam Smith worried about the issues involved.
So, after a century of efforts to clarify, if not yet to
attempt to solve this cluster of problem in economics and
social theory, it would be reasonable to ask what there is
to show for these efforts. The current answer is not much--
even with respect to clarification of the issues.
However, there is grounds for hope that not only can the
issues be clarified, but they might possibly even be
solved. One of my reason for thinking this is the success
in electrical engineering of a program called "Spice code."
What Spice code does is to numerically simulate analog
electronic circuits. Since Powers often describes models
of behavior in terms of analog electronic circuits, the
fact that such circuits can, very successfully, be modeled
by numerical programs suggests that there is a two way
street here.
When I finished a doctorate, one of the first things I did
when I could find the money was purchase an oscilloscope so
I could learn enough electronics to do analog simulations.
At the time, and this was before cheap desktop computers,
and in a context in which I didn't have access to mainframes,
or even mini-computers, analog simulation appeared to me
to be the only way to learn about modeling as a process
in time. At that point my vantage point was roughly the
one that led Martin to pose his question about "time
rates" and simultaneous equations. (I may be
misrepresenting Martin here, but I'm not doing so
intentionally.)
In anycase, what I found when I looked into the issue
was that there really wasn't any problem. But, it took
some determined "looking into" the question to see this,
and some crucial help from a friend who was an electrical
engineer. In physics, and some fields of engineering
simultaneity, or simultaneous equations, and time rates--
even accelerated time rates get along just fine. The
methods required for obtaining solutions to such systems
of equations may pose a problem, but there are methods
available-- if the problem you have justifies the effort
involved. Plus you really do have to know what you are
doing to make use of the classical mathematical methods.
But, conceptually, the "in principle" stage of things
presents no real problem. The example of Newtonian
physics -- See David Berlinski's _Newton's Gift_, or
his _Tour of the Calculus_ which provide a very accessible
viewpoint upon this conceptual world. A world consisting
of systems of simultaneous equations of time rates-- has
formed the core of the classical Newtonian tradition in
mathematical physics.
Now I return to Martin's post, where he says,
If in some way they aren't incompatible, how is it possible to get
any kind of dynamical analysis whatever out of a static base?Sorry to be so dumb.
I don't know what could possibly be dumb about your question.
However, I don't understand something here. You are talking
about "dynamical analysis" and a "static base." Where I was
talking about an aggregate analysis and simultaneous equations.
(But, who knows what all I've said!) If you read "static" where
I wrote "simultaneous" then it was miscommunication-- or maybe
I mis-wrote. Anyway I don't see that the sort of Newtonian
system with dynamics changing time rates in a system of
simultaneous equations presents any problems-- at least as a
system of representation.
For me, I thought the several years I spent working directly
analog circuits with op-amps and feedback circuits at a simple
electronics bench was extremely valuable. It was a way of
becoming familiar with a phenomena -- feedback loops, or
control theory -- without much in the way of analysis or
verbal description.
But, then with cheap digital computers it became possible
to escape the classical world of mathematical analysis for
simulation. (See David Berlinski 2OOO _The Advent of the
Algorithm_ for a fuller description of this escape that
the newly available computation power permitted. )
This new algorithmic approach to phenomena has the great
advantage that it can calculate what sometimes, or often,
couldn't be easily analyzed. So, in a sense, or perhaps
literally it becomes possible to go from intractable,
difficult to represent systems, or systems that are
changing to results of reasonable accuracy by
computation -- which opens up enormous possibilities.
The classical mathematical methods which could be
applied to feedback amp design, Black, Bode,
Nyguest made it possible to do the engineering
required to create industrial products. But, these
methods were too demanding to be of much use to
psychologists, or even more so to economic and
social theorists. Herbert A. Simon 1952 "On
the Application of Servomechanism Theory in
the Study of Production Control." used the
classical methods in the analysis of a single
item of inventory regulated by a control loop.
To do so required 2O some pages of advanced
mathematics to generate almost trivial results. Simon
didn't expand upon this work, and neither did
anyone else for a number of years.
However, within the mainstream of the economics
profession, the issue of static's and dynamics
has been a major concern of theorists from the
end of the 19th century till today. Control
theory today is in very widespread use in
economics. The Bellman Equations form the core
of contemporary economic dynamics. However, the
Bellman equations are employed in order to have a
method to use to consider how an economy ( a
theoretical model that is) will react to
"disturbances." The problem, as I see it is,
that the theoretical model that the Bellman
equations animate is a static model based upon
atomistic, autistic, non-cultural economic
agents whose behavior consists exclusively of
maximizing. Along with maximizing, the
assumption of equilibrium, the role of capital,
and competition form the core contemporary
orthodoxy. This assembly, the neo-classical
theoretical core, is inherently a static model.
Most economists assume that this static
neo-classical analysis is the _only_ systematic
method that will _ever_ be available for
economic analysis.
It hasn't yet occurred to the economist
that it would be possible to replace the
principle of maximization with control theory
principles. There are I think some readily
identifiable reasons for this. The first
may be the scale of the problems involved
in shifting the core of economic theory
from a system based upon maximization to a
system based upon control theory or any
other set of principles. The second is
the near absence of resources to carry on
a project which, almost no one believes
is possible, and not many power brokers,
or gate keepers are favorably inclined.
Despite the existence of a situation that is
institutionally and officially unhelpful,
I think I can see the development of a
situation in which control theory will be
applied in a fundamental reconstruction of
economic theory. I see increasing numbers
of economists who are despite being heterodox
also mathematically proficient, some of these
are familiar with programming, and others are
familiar with control theory-- often through
previous industrial employment.
In recent posts I've suggested that an analysis
of economic transactions has to take into
account some special features of the economic
processes which are not immediately obvious.
A transaction is quite unlike the processes
to which control theory ordinarily applied. This
can be expected to create unexpected difficulties.
Until the required skills-- a combination of
mathematical and control theory proficiencies,
awareness of psychological or agency theory,
and economic theory-- are more widely
distributed than now the solution to many
problems may depend upon a matter of chance
distribution of skills.
The way that efforts to apply control theory
to problems in economics can go wrong, are
illustrated by Rick Marken's misadventures
over the last several years. Viewed from the
side of engineering control theory, (Bill
Powers) Marken's effort amounted to a "giant
leap-- in the wrong direction." View from the
side of theoretical ( heterodox ) economics
(Bill Williams) Marken's efforts were so
poorly informed that they repeated,
unknowingly many of the classic mistakes that
had already long since been experience in
economics-- the adoption of a out-dated theme
from the history of business cycle theory,
T. D. Powers' attempt to develop a theory of
depression and under-employment the Leakages
thesis. Marken's problems were exacerbated
by a paranoiac view of the economics profession,
and an extreme disinclination to listen to
criticism.
There obviously more that could be said, but this
May be a stopping point.
Martin, if I've mis-interpreted your question, I'd
Welcome a correction.
Or, if people more qualified than I am to answer
Martin's question feel so inclined, I think they
Should feel free to expand, correct or criticize.
I'm attempting to communicate a point of view about
an issue that is I am convinced crucial to a
theoretical understanding of control theory and
Its application-- especially to the case of economics.
There aren't at present any usable guides for applying
control theory to economic issues-- at present no one
commands as a part of their own knowledge the skills
and knowledge to apply control theory to economic issues
with entirely confident results. Rick is entirely
confident, but he lacks either the required skills or
knowledge. Bill Powers admits, some of the time, to an
unfamiliarity with economic questions. I think there
may be, instead a problem of his knowing too much that
isn't so. And, I will admit to a minimal level of
proficiency in programming, a very questionable
understanding of control theory, and other deficiencies
that those who find the topic more interested than I do,
can expand upon.
So, the issue remains, what will the "cranks" the
"crack-pots" and in Lionell Robbins phrase-- "slightly
disturbed spirits" manage, despite the difficulties,
to do?
Bill Williams