[From Bill Powers (2003.02.17.1834 MST)]
Seems like a pity to waste all this work. Anybody know any economists who
might find something interesting in a simulation?
Best,
Bill P.
[From Bill Powers (2003.02.180735 MST)]
Rick Marken (2003.02.17.2020)--
> Anybody know any economists who
> might find something interesting in a simulation?
I know quite a few (30% of the degrees at RAND are in Economics). So if you
(we?) get the simulation running and can show how it can account for some
macroeconomic data, I will be happy to run it by some economists here.
I think that to achieve credibility, we need an economist to join in the
effort of completing the model. I have been working out of a very
peripheral acquaintance with economics and a reliance on what seems to me
to be common sense. In my mind, Bill Williams was to serve as the economic
expert who knew what issues needed to be dealt with and what economists
would expect a model to do, as well as providing programming expertise. We
still need that kind of advice. I wish it were to come from Bill, but at
least for now I guess that's not to be.
Perhaps the person we need doesn't exist. If I were writing a Help Wanted
ad, I would try to explain that we need a expert economist who is up to
date on everything happening in that field, on economic history, and on the
major theories of economics -- but who has not signed on with any school of
thought and is willing to examine any proposal as a potential revision of
the field. This implies that this person is dissatisfied with economics as
it stands and perhaps even with his own work in that field, and is looking
for something that might, with development, be better.
Were you able to see the diagram I posted a few days ago? Actually, by
mistake, I sent it twice (uncorrected), the second time with the erratum
post saying that the minus sign between Yk and Kmul should have been an
equal sign. This is a macroeconomic model, but each part can potentially be
expanded into an array of subsystems with a distribution of properties, so
there is a relatively straightforward path, somewhere in the future, to a
micro model, or semi-micro (I don't think we're ever going to carry it to
the level of individual companies or consumers!). Before we think of that
expansion, we need to add the major entities that are missing from the
model, including those we don't know about but which an economist could see
are not there. Two big ones we know about are banks (to create money) and
the government (government could be part of a general not-for-profit
sector). We need to add growth of three kinds: growth of population, growth
of productivity (and concurrent demand for investment), and growth of goals
for consumption. These are interdependent to some extent.
These major sectors have properties that represent the average behavior of
a large group of people who do things in similar ways, in the way I have
managers controlling inventories by adjusting prices. Not all managers will
have similar inventory goals, or similar output gains, but the net effect
of all of them acting in parallel will be to establish some virtual
reference level for inventory and some average dependence of price on
inventory error. If we were to think of or be told of a class of managers
who do _not_ control inventory in this way, we would characterize what they
do do as required, and add another entity to the model.
Another major sector missing from the model is retail, or small business.
It could be that a retail organization will resemble a primary producer in
its control structure, but there may be differences significant enough to
warrant adding a retail entity. On the other hand, perhaps this "entity
creep" is simply the logical route toward turning the macro model into a
somewhat more detailed model that begins to have micro features.
What's missing at the macro level is, of course, competition. Some
competition may be introduced when we give certain members of the
capital-income group (owners or entrepreneurs) control over the plant, and
over their own share of capital distributions. If they try to make their
own share too large, the wage-share will not be able to support the workers
and in one way or another they will fail to keep the plant producing, and
of course their consumption will fall, and so forth.
The point we need to make in a Help Wanted ad is that the basic philosophy
of the model is approximately established, but its direction of development
will have to be decided step by step as we go. Some future directions
suggest themselves, but nobody knows what form the model will have taken by
the time it's complete enough to compare with the behavior of the real
system. So an economnist who participates in the development will have an
essential role to play.
Best,
Bill P.
[
From Bill Powers (2003.02.18.1408 MST)]
Rick Marken (2003.02.18.1240)--
>> I think that to achieve credibility, we need an economist to join in the
>> effort of completing the model.
I disagree. I think you (we?) need to build a model that accounts for some
basic
macroeconomic data with very few parameters.
That's what TCP tried to do, although he didn't get as far as an actual
model, and apparently (according to Bill W.) he misindentified one of the
pieces of "basic macroeconomic data," the investment rate. Were there other
misidentifications? Is it actually feasible to suppose that the composite
producer not only produces increasing amounts of goods, but the increasing
amount of money necessary to supply the needed buying power? What is the
effect of interest in such a money-creation scheme?
It may be possible to construct a "zero-th approximation" model as you
propose, in which all relationships are boiled down to a very small
handful. But how do we know where the threshold of oversimplication is? I
don't doubt that the simplified model will run, but showing that it is not
omitting anything important takes more knowledge than I, and perhaps even
you, have. Verifying that it contains no logical errors or simple typos is
also not a trivial problem.
Bill W. sent me his Veblen-Duessenberry (?) program, saying that it was
running and doing some interesting counterintuitive things, which he was
learning from. One of my last communications to Bill on that program,
yesterday, alerted him to two errors in the control-system portion which
have a drastic effect on the behavior of the system: one involved
substituting a variable bi, which I believe may be the integral of the
budget (??) error (many ???) for the variable be, which may be the budget
error itself. With the errors corrected one possible way, the program
simply stopped on a range error with a huge consumption figure. With the
error corrected in the other possible way, the effect of reducing certain
reference levels was totally different from the way the program behaved
when I got it, which of course means that Bill now has to unlearn and
forget, if he can, whatever he thought the model was telling him.
The problem with simulations is that they almost always run and produce
some behavior. What they don't tell you is whether to believe that
behavior. This is why you have to understand every detail of the model,
know what behavior you expect of it, and test to see that it occurs. In
Econ002 or so, one of the early versions, everything seemed to be running
correctly, but when I finally left it running and went off to get some tea,
when I came back it had about ten times as much money in all the cash
reserves as it started with. If I had believed what the model seemed to be
telling me, I would have said, in superstitious awe, "It's showing growth,
and I didn't even have to program it in!" I confess that such a thought
flitted through my mind before I faced the reality that the books weren't
balancing. It took quite a few hours to track down where the extra money
came from (one quantity was being added instead of subtracted, so it was a
simple sign error -- but the program was perfectly happy to run with the
wrong sign in it).
Econ004 is obviously quite a bit more complex, with more control systems
and more entities in it, than your circular-flow model. So it's not quite
as "macro" as your model is, which could be considered a disadvantage. On
the other hand, it is easily expandible to include still more details, and
thus to provide more potential ways of testing it against real data. I
don't know that either approach is inherently better, but they are
certainly different.
And my point is that I am keenly aware of needing some advice from a real
economist to carry my model further, because I'm not at all sure how the
details of the real system work. It would be silly to include something in
the model that never happens in real life. Perhaps you can get along
without such advice, but I can't (even if I end up arguing with half of the
advice I do get).
So if there are any economists at RAND who look halfway compatible, maybe
you could show them this correspondence and explain the diagram to them,
and see if there are any takers of the bait. I would appreciate it.
Best,
Bill P.
[From Bill Powers (2003.02.19.1235 MST)]
Rick Marken (2003.02.19.1130)]
I guess I don't understand why you're now so concerned about getting help from
economists on a control model of the economy when you were obviously perfectly
capable of developing a control model of behavior without soliciting help from
behavioral scientists.
I haven't read nearly as much economics as I did behavioral science; I
never took a course in economics, but I've had quite a few undergraduate
psych courses, and a year in psych graduate school. I just don't feel very
confident about offering something that covers the way the economy works
correctly without at least some arguments with economists. I can't look at
my own economic behavior and generalize from it; a great deal depends on
the actual rules of the game as it is played rather than on basic principles.
Well, I don't want to try to argue you into something you don't want to do,
so I'll just snoop around and see what sort of help is out there. Maybe
your friend will like the spreadsheet model.
Best,
Bill P.
[From Bill Powers (2003.02.20.0800 MST)]
Maybe my friend will like the diagram and description. Why not send me the
latest
version of the diagram and a write up and I'll sit down with him next week
and see
what he thinks.
Working on it. I have to go into town for a while, but will finish it up
this afternoon.
Best,
Bill P.
[From Bill Powers (2003.02.20.1913 MST)]
Rick Marken (2003.02.19.2045)--
>Maybe my friend will like the diagram and description. Why not send me
the latest
>version of the diagram and a write up and I'll sit down with him next
week and see
>what he thinks.
Here is a short pseudo-paper with the diagram embedded. It's less than 60K
long so I guess Word compresses graphics. I hope the diagram actually shows up.
Best,
Bill P.
Toward an agent.doc (65.5 KB)
[From Bill Powers (2003.02.21.0921 MST)]
Rick Marken (2003.02.20.2030) --
>In the meantime, I can't seem to find Xw in the wage earning household
record.
>Nevertheless, Xw defined by the assignment statements in procedure
>wconsumerlevel1. Does this matter?
It's there in my copy of econ004b.pas, but I'm attaching it as a text file.
Note that Cw has changed everywhere to wC (ditto for Hw,k to k,wH),
allowing Gw and Gk to be changed to Cw and Ck, for "consumption".
Consumption is the rate of buying.
Best,
Bill P.
ECON004B.txt (94 Bytes)
[From Bill Powers (2003.02.21.1914 MST)]
Rick Marken (2003.02.21.1520)--
>By the way, I gave your (excellent) write up to a very senior person here
who I like
>very much and is a very good modeler (OR type). He's not an economist but
I thought I'd
>just see what kind of reaction I get from him. When I told him about what
you were
>doing, however, he said exactly what I had said: you don't want an
economist, you want
>a banker, an accountant, or a CFO.
Why don't I want an economist? I thought economists were supposed to study
the economy, a much broader subject than the three specialties mentioned.
Is this a subtle way of saying economists don't really know anything? Are
bankers, accountants, or
CFOs interested in models? Wouldn't all three be needed?
I looked up "economic models" on the web, and came up with
http://www.unibw-hamburg.de/WWEB/math/uebe/modelle/titelseite.html
which is a list of macroeonomic models compiled at the University of Hamburg.
"Those who want to comment on, and/or augment our list, please contact us
at Institut f�r Statistik und quantitative �konomik, Universit�t der
Bundeswehr Hamburg, Holstenhofweg 85, D-22039 Hamburg, Germany (see above).
The paper printout editions of this list prior to this publication will
generally be discontinued after completion of this page."
Unfortunately the models are summarized with an absence of explanation, I'm
not even sure what kind of models they are (a very few are described as
"simulations", but I haven't seen anything helpful there).I will search
further on the Web.
Best,
Bill P.
]
[From Rick Marken (2003.02.17.2020)]
Bill Powers (2003.02.17.1834 MST)]
Seems like a pity to waste all this work. Anybody know any economists who
might find something interesting in a simulation?
I know quite a few (30% of the degrees at RAND are in Economics). So if you
(we?) get the simulation running and can show how it can account for some
macroeconomic data, I will be happy to run it by some economists here.
Best regards
Rick
···
---
Richard S. Marken
MindReadings.com
marken@mindreadings.com
310 474-0313
[From Rick Marken (2003.02.18.1240)]
Bill Powers (2003.02.180735 MST)]
I think that to achieve credibility, we need an economist to join in the
effort of completing the model.
I disagree. I think you (we?) need to build a model that accounts for some basic
macroeconomic data with very few parameters. Once you have the model set up so
that it is clear how it works and what predictions it makes then I'd be happy to
sit down with an economist and see what (s)he thinks. I bet I could find someone
who would be willing to get involved if I were able to make a convincing
demonstration of concept using the model and real macro data.
Best regards
Rick
···
--
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
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 Rick Marken (2003.02.19.1130)]
Bill Powers (2003.02.18.1408 MST)
Rick Marken (2003.02.18.1240)--
>> I think that to achieve credibility, we need an economist to join in the
>> effort of completing the model.>I disagree. I think you (we?) need to build a model that accounts for some
>basic macroeconomic data with very few parameters.That's what TCP tried to do, although he didn't get as far as an actual
model, and apparently (according to Bill W.) he misindentified one of the
pieces of "basic macroeconomic data," the investment rate.
I guess I don't understand why you're now so concerned about getting help from
economists on a control model of the economy when you were obviously perfectly
capable of developing a control model of behavior without soliciting help from
behavioral scientists. Why do you think economists know any more about the economy
than behavioral scientists know about human behavior? If you want to know how
components of the economy (like investment, banking, etc.) work I think the people
to ask about it are those involved with those components of the economy (brokers,
bankers, etc).
It may be possible to construct a "zero-th approximation" model as you
propose, in which all relationships are boiled down to a very small
handful. But how do we know where the threshold of oversimplication is?
Why do you think an economist would know? They're just cranking out complex
regression analyses. I think the kind of modeling you're doing will be as
unfamiliar to an economist as it is to a behavioral scientist.
I don't doubt that the simplified model will run, but showing that it is not
omitting anything important takes more knowledge than I, and perhaps even
you, have.
Right. That knowledge is most likely to be in the heads of the worker bees
(bankers, etc), not economists. I do think economists would be wonderful to have
as participants, but only after they understand what you're up to. That's why I
think it would be best to have something tangible for them to look at. They might
look at it and say "Oh, that's nothing but..." or "That's all wrong.." but then,
hopefully we can find out who has already done it and what's missing.
And my point is that I am keenly aware of needing some advice from a real
economist to carry my model further, because I'm not at all sure how the
details of the real system work. It would be silly to include something in
the model that never happens in real life. Perhaps you can get along
without such advice, but I can't (even if I end up arguing with half of the
advice I do get).
My preference would be to get advice about the mechanics of the model from the
worker bees: the people who know how various components of the system work. Once
the model is working (and accounting for data) I would then be willing to show it
to some of my economist friends here.
So if there are any economists at RAND who look halfway compatible, maybe
you could show them this correspondence and explain the diagram to them,
and see if there are any takers of the bait. I would appreciate it.
I would rather wait. I'd feel more conformable if I had a well annotated working
model to show them that did something cool, data prediction-wise. Oh, and my
closest economist friend is on a Mac. So when I can get an Excel version running
I'll show the model and diagram to him and see what he thinks.
Best regards
Rick
···
--
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
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 Rick Marken (2003.02.19.2045)]
Bill Powers (2003.02.19.1235 MST)
Well, I don't want to try to argue you into something you don't want to do,
so I'll just snoop around and see what sort of help is out there. Maybe
your friend will like the spreadsheet model.
Maybe my friend will like the diagram and description. Why not send me the latest
version of the diagram and a write up and I'll sit down with him next week and see
what he thinks.
Love
Rick
···
--
Richard S. Marken
MindReadings.com
marken@mindreadings.com
310 474-0313
[From Rick Marken (2003.02.20.2030)]
Bill Powers (2003.02.20.1913 MST)--
Rick Marken (2003.02.19.2045)--
>Maybe my friend will like the diagram and description. Why not send me
the latest
>version of the diagram and a write up and I'll sit down with him next
week and see
>what he thinks.Here is a short pseudo-paper with the diagram embedded. It's less than 60K
long so I guess Word compresses graphics. I hope the diagram actually shows up.
OK. I'll keep you posted.
In the meantime, I can't seem to find Xw in the wage earning household record.
Nevertheless, Xw defined by the assignment statements in procedure
wconsumerlevel1. Does this matter?
Best regards
Rick
···
--
Richard S. Marken
MindReadings.com
marken@mindreadings.com
310 474-0313
[From Rick Marken (2003.02.21.1520)]
Bill Powers (2003.02.21.0921 MST)]
Rick Marken (2003.02.20.2030) --
>In the meantime, I can't seem to find Xw in the wage earning household
record.
>Nevertheless, Xw defined by the assignment statements in procedure
>wconsumerlevel1. Does this matter?It's there in my copy of econ004b.pas, but I'm attaching it as a text file.
Note that Cw has changed everywhere to wC (ditto for Hw,k to k,wH),
allowing Gw and Gk to be changed to Cw and Ck, for "consumption".
Consumption is the rate of buying.
OK Thanks. I've printed this out and I'll try to make the changes this weekend. To my
amazement, however, I find that I have already managed to produce a working spreadsheet
version of Econ004. I'm just not sure it works correctly, so I'll have plenty of
debugging to do this weekend. Do I have the latest executable version? Part of the
debugging will have to involve making sure the spreadsheet does what the program does.
By the way, I gave your (excellent) write up to a very senior person here who I like
very much and is a very good modeler (OR type). He's not an economist but I thought I'd
just see what kind of reaction I get from him. When I told him about what you were
doing, however, he said exactly what I had said: you don't want an economist, you want
a banker, an accountant, or a CFO.
Best regards
Rick
···
---
Richard S. Marken, Ph.D.
Senior Behavioral Scientist
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