Scott's response contains nothing relevant to the alleged point of this
discussion except:
PS: Chris keeps writing that I said that rational expectations
models are solved by using genetic algorithms. That is wrong.
I did not say that. I also did not say that no one has ever cited
the Radner result.
Scott is right that I misread him on the first point. I stand corrected
(although my more substantive point regarding convergence of numerical
methods stands). But I never claimed that he said that no one has ever cited
the Radner result, rather, I claimed that Scott is mistaken in his assertion
that no one has ever addressed Radner's result regarding computational
ability.
Scott responds to my simple argument regarding models with finite vs
infinite lives with a discussion of modeling behavior under uncertainty. I
can't quite figure out the relevance of that discussion, nor can I find where
Scott addresses the point that infinite lives is actually a relatively
innocuous assumption.
I choose not to dignify Scott's charges that I mislead students, that I
was mislead as a student myself but wasn't capable of seeing it, or that my
simple argument was "ingenuous" with responses. I would ask Scott to please
try to be polite and restrain himself from offensive comments.
Scott suggests that economics is devoid of empirical tests of
theoretical models:
What economists do claim is that the test of a model is the goodness of
its predictions. So what is the record here on the predictive accuracy
of any of the models under discussion? Indeed, what is the record on
the predictive accuracy of any economic models at all?
I'll make my own prediction here: No answer to these questions will be
forthcoming from economists.
I'll echo a theme of Leigh's and comment that most of Scott's accusations
would carry more weight if this were 1965 rather then 2000. Economics has
become a very empirical discipline. More than half of published studies are
empirical, and "model validation," which I take to mean testing to see
whether the insights models give about how the world works are consistent
with empirical evidence, is a very, very common undertaking. With respect
to the models under discussion, one can find an extensive survey of the
equally extensive empirical literature on growth modeling and accounting
following Solow's contributions in:
Fagerberg, J. (1994) "Technology and International Differences in Growth
Rates," Journal of Economic Literature 32:3 pp1147-1175.
With respect to the real business cycle literature (the research agenda
encompassing the rational expectations methods Scott refers to), most of
the papers on that subject could be reasonably deemed as exercises in
"model validation." See, for one obvious example, the _seminal_ paper
in this literature,
Kydland, F. and E. Prescott (1982) "Time to Build and Aggregate
Fluctuations," Econometrica 50:6 pp1345-1370,
with particular relevance to section 5, "Test of the theory." I suppose
it's also worth noting that a new empirical method, the generalized method
of moments, was developed specifically to test this sort of model, see
Hansen, Econometrica 1982.
Perhaps Scott will respond that this sort of procedure is not "model
validation" because it is not testing the realism of each underlying
assumption. But we don't need a test to know that, for instance, people
don't live forever, or that there is not a C1 measure of consumers, or that
the world isn't composed of two countries producing two goods with two
factors of production. Scott says he's not simply complaining that models
make unrealistic assumptions, so I don't see how he can assert that models
that make such assumptions are not "valid" on those grounds alone.
Chris Auld (403)220-4098
Economics, University of Calgary <mailto:[log in to unmask]>
Calgary, Alberta, Canada <URL:http://jerry.ss.ucalgary.ca/>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|