Agent-Based economic models

[From Bill Powers (2004.01.22.1133 MST)]

Looking around the net for computer models in economics, I came across the
work of Dr. Charlotte Bruun (Aalborg University, Denmark), and a paper
called "Agent-based Keynsian Economics: Simulating a Monetary Production
System Bottom-Up." The paper can be found at www.socsci.auc.dk/~cbruun.

I wrote to Ms. Bruun and confessed to having ambitions to develop an
economic model depite my ignorance in economics. She said

I am an economist, but I don't believe that we are a sacred species - and
I don't believe that you have to know all economic theory to contribute to
the subject. Contrariwise I do think that economics as a science is too
stuck in "the usual assumptions". There are, however, many interesting
things to be found in the literature - but not necessarily the part of the
literature that is entailed in the usual curriculum.

Her program is written in Turbo Pascal 7.0, but she said it won't run on
her new computers. I was happy to send her the patched turbo.tpl that gets
rid of the speed problem, so maybe she will be able to compile my stuff,
too, which I have sent her.

I explained the Test Bed idea and she immediately recognized it, saying "I
was glad when you mentioned economic accounting since my belief is that
accounting identities are vital for this 'system rationality'."

Coincidentally, I found a paper I wrote last year on the Test Bed project
that was titled "Toward an agent-controlled simulation of the US economy",
dated February 2003. I'll attach it here just for general interest.

So, fellow Test Bed Mates, we may get some help after all.

Best,

Bill P.

Toward an agent.doc (65.5 KB)

[From Rick Marken (2004.01.22.1530)]

Bill Powers (2004.01.22.1133 MST)--

Looking around the net for computer models in economics, I came across the
work of Dr. Charlotte Bruun (Aalborg University, Denmark), and a paper
called "Agent-based Keynsian Economics: Simulating a Monetary Production
System Bottom-Up." The paper can be found at www.socsci.auc.dk/~cbruun.

This looks very interesting. I'll try to read it tonight.

Coincidentally, I found a paper I wrote last year on the Test Bed project
that was titled "Toward an agent-controlled simulation of the US economy",
dated February 2003. I'll attach it here just for general interest.

Figure 1 in that paper reminds me of what I think was the main problem with
my H. economicus model: it didn't keep separate track of money (R in your
model) and goods and services (V in your model). Well, it did have a
separate cumulative variable for goods and services (Q in my model) and cash
reserves (PQ in my model) but they were not controlled separately.

I just looked and saw that I had implemented the spreadsheet version of the
model shown in Figure 1. I actually think this was a pretty good start. I
look at it as an aggregate model of the economy. The plant is the aggregate
producer. The "households" are the aggregate consumer. I think the next
step might be to add a bank and a government. The bank will just lends money
into the system and keep track of how much it is owed and the government
would tax the cash reserves of recipients of capital and wage income; that
tax income would then go into purchasing inventory from the plant for the
aggregate consumers. I think that, for now, we can capture income disparity
in terms of the per capita wealth of the capital and wage consumer.

Why don't you try to add a bank and a government to the model and I'll see
if it makes sense and try to implement it in the spreadsheet?

By the way, re H. economicus, I realized last night that I didn't put growth
in independent of leakage in as an assumption of the model. What I put into
the model was TCP's "intrinsic growth rate". But I put it in as a growing
reference for PQ rather than as a spontaneous increment in Q, as TCP did. I
could justify the idea of a growing reference for PQ as a reflection of the
growing population. The 13% yearly increase in my model is a bit high for
population increase but 13% was TCPs estimate of the nominal growth rate
without leakage. I thought that leakage, as a disturbance to PQ, would
produce a chronic lowering of growth rate. But, of course, it didn't
because money is automatically put into the economy to make up for the loss.
This is a result of TCPs auto inflation, which happens magically in TCP's
model but is the result, in my model, of an agent (the composite manager)
raising prices to make up for the loss of income due to inflation. It may be
a mystery where the money comes from that makes up for the leakage. I
presume it is just borrowed from bank. But this debit is kept track of in
the model. The books do balance in my model. The main flaw in the model was
not properly dealing with money and goods/services separately. But I was
following TCP's lead. I still think his basic circular flow approach is
right. I just think that we can do the modeling a bit better now.

Anyway, let's keep trying to develop the (aggregate) agent-based economy you
described in you paper last Feb (I can't believe it's been a year since I
worked on that).

Best

Rick

ยทยทยท

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400