This note is in reply to Scott Moss's inaugural address, available on the web
at <http://www.cpm.mmu.ac.uk/cpmrep56.html>. In the course of private
correspondence with Scott regarding this essay he suggested this list as a
venue for a public response. I apologize if you think this list is an
inappropriate venue, although it is at least tangentially relevant since the
essay's point is to justify social simulation modeling.
I do not wish to argue against such modeling: I have done such modeling
myself. My point here is to take issue with Moss's serious, sweeping charges
against economists in general. Moss claims that economists are
"intellectually dishonest" "bad scientists" who should be "ignored by serious
social scientists." Moss gives several examples to demonstrate his claims. I
will briefly address each.
The first is "the Humbug Production Function." This addresses the validity
of Robert Solow's method for separating changes in national output caused by
changes in the supply of factors of production from changes in technology.
The details are irrelevant to Moss's argument: Moss wishes to show that
Solow's technique has been shown to completely wrong, even disavowed by
Solow, and is yet used anyways, thus showing the "intellectual dishonesty" of
the economics profession. Moss is very much mistaken.
The central issue revolves around a critique of Solow published in 1974 by
Shaikh. Moss tells his readers that Shaikh's critique demolishes Solow's
technique, a result Solow himself "accepted in substance" and which has
"never been successfully refuted." Moss proceeds to provide citation counts
showing that Solow is still oft cited whereas Shaikh is largely ignored,
allegedly demonstrating economists ignore criticism which undermines accepted
methods. In reality, Solow provided an absolutely devastating reply to
Shaikh:
Solow, R. (1974) "Law of Production and Laws of Algebra: the Humbug
Production Function: Comment," Review of Economics and Statistics
56:1, p121.
It is common practice when providing such a comment to thank the author,
comment on the thoughtfulness of the piece, and so on. Solow provides no
such banalities and goes straight for the jugular, opening his one-page reply
with:
Mr. Shaikh's article is based on misconception, pure and simple.
He proceeds to very clearly explain why the "Humbug Production Function"
Shaikh and Moss refer to is "nonsense." He closes his reply with "The humbug
seems to be on the other foot." Solow comes as close as one can in academic
journal to simply asserting Shaikh doesn't have the foggiest idea what he's
talking about. Solow certainly, unambiguously, forcefully and entirely
rejects Shaikh's critique.
Readers interested in the details of the issue are invited to read the
articles. The methodology is not the issue, the issue is whether Moss has
provided evidence to bolster his bold claims about economists. He certainly
has not: Moss claims that Solow "accepted in substance" what Shaikh had to
say, that Shaikh's "critique was never successfully refuted," whereas that is
clearly not the case. It is difficult to interpret Moss's claim as anything
other than rather spectacular intellectual dishonesty itself, which is
signally ironic.
Moss's second example bolstering his case is the response to a 1968 article
by Roy Radner. Quoting Moss:
Consequently, general equilibrium cannot exist unless individuals
have unlimited computational capacities. This would seem to be an
important result since it states formally that the theoretical basis of
economists' views of markets requires buyers and sellers in all markets to
have unlimited computational capacities. Though by no means as influential
as the Solow paper discussed above, the Social Sciences Citation Index
records 68 citations of the Radner paper since 1981. None of those papers
address Radner's conclusion that unlimited computational capacity is a
necessary condition for equilibrium when spot trading takes place over
time.
We are apparently supposed to believe that Radner's result is being ignored
because of its serious consequences, thus again revealing that economists
are dishonest and unscientific. People do not have infinite computational
ability, therefore general equilibrium theory must be wrong, therefore we
must ignore Radner (Moss's readers first must ignore the fact that Radner's
article was published in a popular mainstream journal.)
In fact, economics, like all scientific disciplines, uses models which
simplify reality. Making rather implausible assumptions about computational
abilities is just one such assumption. In this case, relaxing that
assumption forms both a research agenda looking into its consequences and is
often an issue in various models. Moss's attack here is no more substantive
than someone leaping up at the back of a lecture hall and complaining that
people don't actually have infinite lives, that there are more than two
countries in the world, or that there does not exist a continuum of
consumers. Moss is also mistaken when he asserts that no one has ever
addressed Radner's point; this point was previously made in slightly
different form by Savage (1954) in an oft-cited paper and Radner's specific
point has been discussed by Mongin and Walliser (1988), Smith (1991) and
Lipman (1991). More broadly, many economists have investigated issues in
computation and their implications. Moss is simply wrong, and even if he
were correct that no one has "addressed" this issue, he still would not have
demonstrated that economists are all "intellectually dishonest" "bad
scientists."
Moss's third example is more than a little vague. He briefly explains what
rational expectations means, then provides what he claims is the methodology
used by RE modelers:
1. Write down a rational expectations model.
2. Determine the equilibrium configuration of that model.
3. Replace the rational expectations agent with multiple agents represented
by genetic algorithms.
4. Simulate the model devised in step 3.
5. If the simulation converges to the corresponding rational expectations
equilibrium, write up the results and send to a journal.
6. If the simulation does not converge to the corresponding rational
expectations equilibrium, revise the model and/or the genetic algorithm
and go to step 2.
Moss's evidence for this questionable methodology is:
Steps 5 and 6 of this procedure have been specified inductively on the
basis of questions asked by me at seminars and workshops where such papers
have been read.
This is, of course, less than compelling. First, it is simply not the case
that all RE models are solved via genetic algorithms, indeed, that is to the
best of my knowledge not even a common solution technique for this type of
model (linearizing the marginal conditions around a steady state is one
popular solution method). And even if Moss's claim were correct, what of it?
If someone tries a numerical technique to find an equilibrium point in a
model and that attempt fails to converge to the equilibrium, prompting the
modeler to try again or use a different technique, why on earth is that
evidence the modeler is acting dishonestly or unscientifically?
Moss proceeds to claim that the fact economists never make simplifying
assumptions to make the model "more suitable for analyzing the problem" is
also evidence of his charges. This is unremarkable: simplifying assumptions
are made to make a complex reality less so and thus amenable to analysis.
All models (even verbal "models") in every discipline are of this form; one
makes simplifying assumptions because one has to. _Of course_ such
simplifications make the model less like reality.
Finally, Moss claims that economists never employ anecdotal evidence, which
we are supposed to believe is a serious fault. Moss misconstrues his
colleague's comment that he "does not believe anecdotal evidence." Lots of
economic ideas and theories are have a genesis in anecdotal observations
about how the world works. The comment decries use of such evidence to test
and measure such theories. _That_ would be unscientific: anecdotal evidence
is, of course, subject to massive selection biases amongst myriad other
problems. One can only guess that if economists did regularly use anecdotal
evidence to test theories and measure relationships, Moss would have instead
chastised them as unscientific for doing so.
In short, Moss has not supported his claims. I would strongly suggest to
Professor Moss that he make his case for the superiority of his simulation
methods on scientific grounds: show how they outperform methods in the
received literature. The end-run approach of trying to show that all of
economics is not only wrong but dishonest, and therefore Moss's own methods
are better on general principles, leaves much to be desired scientifically
and otherwise.
Chris Auld (403)220-4098
Economics, University of Calgary <mailto:[log in to unmask]>
Calgary, Alberta, Canada <URL:http://jerry.ss.ucalgary.ca/>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|