Bill Williams (16 November 2001 23:00 CST) to Bill Powers:
For more than a decade Bill Powers has made repeated requests that I join
him in colaborative effort to develop a general model of the market process.
I have declined, because as I have repeated told him, I view Keynes 1936
_General Theory_ as at least a proximately adaquate theoretical treatment
of the macro economy. Powers, however, without bothering to read Keynes
has strong opinions as to what Keynes acomplished-- nothing. This opinion
is derived from Bill Powers dad, an accomplished chemist who didn't find
it neccesary to read Keynes either. I will repeat once again, it is usually
prudent to familiarize oneself with the existing literature before spouting off
about it. ANd, the rewards for rediscovering the wheel are minimual. My
suggestion to Bill would be to find someone more suitable for a fools errand.
I'm happy to rush in and help with this errand. So that make two so far.
I think we need some criterion for evaluating the success of any model
we build. Since we are building a model of the macro economy I suggest
that the criterion of success be precise prediction of the behavior of
time varying macro economic variables. T.C. Powers did what is basically
a qualitative analysis of the predictions of the leakage model. I think
we should try to do better. I envision a model that predicts, say, time
variations in GNP as well as a simple control model predicts variations
in handle movements in a tracking task. Assuming that the macro economic
model is a control model, we should be able to predict the behavior of
the output variables -- the variables that protect controlled variables
from disturbance -- from the behavior of the disturbance variables. So,
for example, if leakage actually is a disturbance that is compensated
for by variations in money supply and production, then if we can measure
leakage correctly (from the model's perspective) then we should be able
to predict the output variables (money supply and production) almost
exactly. If we can produce a model that fits the data nearly perfectly
(RMS error of prediction on the order of 2% or less, say) I think people
would pay attention to the model even if we are not real economists.
Best regards
Rick
···
--
Richard S. Marken
MindReadings.com
marken@mindreadings.com
310 474-0313
Rick Marken (2001.11.17.1120) --
>Since we are building a model of the macro economy I suggest
>that the criterion of success be precise prediction of the behavior of
>time varying macro economic variables.
The motivation for modelling is the problem that we are limited by a kind of indeterminacy principle, where we can observe the behavior of a system as a black box, or we can open the box, dissect the system, and look at its internal structures, but a dissected system can't behave. The purpose of modelling is to understand the internal organization of that which is modelled. Consequently, another important criterion is that the organization of the model should correspond as far as we can determine to the organization of the system being modeled.
As we discussed not too long ago, a proposed model of aggregate entities can produce aggregate "behavior" that correlates well with statistical data while having no correlation with the control systems and their relations whose behaviors are statistically aggregated.
Bruce Nevin
···
At 11:20 AM 11/17/2001 -0800, Richard Marken wrote:
As we discussed not too long ago, a proposed model of aggregate entities
can produce aggregate "behavior" that correlates well with statistical data
while having no correlation with the control systems and their relations
whose behaviors are statistically aggregated.
Of course. So?
Models of the behavior of aggregates (like Ohm's law) can be quite
useful in themselves. And in this case, since the aggregate economic
model will be built from an aggregation of individual control systems,
we will know for sure that the behavior of the aggregate model emerges
from the behavioral organization of the individual entities from which
it is constructed. So, while (like Ohm's law) the behavior of the
aggregate economic model may differ from the behavior of the individual
entities that make it up, we will know (unlike Ohm's law, when it was
formulated) how the behavior of the aggregate is related to the behavior
its individual components.
Best regards
Rick
···
---
Richard S. Marken
MindReadings.com
marken@mindreadings.com
310 474-0313
Rick Marken (2001.11.17.1500)--
At 02:56 PM 11/17/2001 -0800
You suggest an analogy between aggregate econometric data and Ohm's law. Electrical current and voltage are measured more easily and reliably, and by less theory-laden means, than economic data. Using such data may lead to conclusions that were presupposed in (and by) the gathering of the data. I don't know that this is so, only something to be wary about. If a model built up from the interactions of autonomous control systems produces like econometric data -- and also data hitherto not observed because not looked for -- that will be convincing.
I have every confidence that you will persist as long as you find it rewarding to do so. I'm only suggesting that alternative approaches have value.
Since we are building a model of the macro economy I suggest
that the criterion of success be precise prediction of the behavior of
time varying macro economic variables. ...
If we can produce a model that fits the data nearly perfectly
(RMS error of prediction on the order of 2% or less, say) I think people
would pay attention to the model even if we are not real economists.
Yes, I should think so, but I'd settle for considerably less. The economy
may have a lot more variability in it than a tracking task has, and
essential data may prove hard to find. I think that Bill Williams' approach
of taking on the major unsolved problems of economics would also get us
somewhere. Of course to do that we would have to know what those problems
are, and agree that they need to be solved (a judgement that I guess every
person has to make privately). Some study is called for.
I have every confidence that you will persist as long as you find it
rewarding to do so. I'm only suggesting that alternative approaches
have value.
What alternative approaches do you have in mind?
I've presented a control theory approach twice at CSG meetings and
received nothing but resistance (except from Bill Powers, who offered
many constructive suggestions). Why is there so much resistance to a
control theory approach to economics? I have no doubt that there are
many problems with the details of my approach (and am hoping to get help
on these from economic experts). But I thought the overall picture of
the economy that I presented, as a collection of produce/consumers
(i.e.: input controllers), was a step in the right direction. Don't you
think so?
Best regards
Rick
···
--
Richard S. Marken
MindReadings.com
marken@mindreadings.com
310 474-0313
At 03:51 PM 11/17/2001 -0800, Richard Marken wrote:
I can't say about the resistance that you perceive. Except for the dubious status of Leakage, and the questions of data and structural correspondence that I just mentioned, I agree that it is a useful and constructive step, as far as it goes.
The alternative is to reconstruct economics from the bottom up, beginning with transactions among autonomous control systems, rather than from the top down, on the assumption that economists have measured the right things in appropriate ways and have not lost important distinctions in their aggregations.
The article by John Henry that Bill Williams cited a while back seems helpful.
What alternative approaches [to developing a model economy] do you have in mind?
Bruce Nevin (2001.11.17 16:23 PST)
The alternative is to reconstruct economics from the bottom up, beginning
with transactions among autonomous control systems, rather than from the
top down, on the assumption that economists have measured the right things
in appropriate ways and have not lost important distinctions in their
aggregations.
Could you be a little more specific about what you mean here. What
interactions do you think should be included in the model (with
280,000,000 people in an economy, each controlling 1000s of perceptions,
that's a lot of possible interactions)? How can we know what aggregate
variables should be measured by looking at individual interactions? Can
you give an example of the kind of "important distinctions" you think
might have been lost in standard economic aggregations? Perhaps you
could contrast your proposed approach with Bill Powers' (2001.11.16.0811
MST) reformulation of Bill Williams' economic analysis of the
relationship between income and investment that was appended to Bill
Williams (16 November 2001 23:00 CST) post.
Best regards
Rick
···
--
Richard S. Marken
MindReadings.com
marken@mindreadings.com
310 474-0313