[OPE-L:6204] (Fwd) Re: Re: econometrics

From: Andrew Brown (Andrew@lubs.leeds.ac.uk)
Date: Mon Nov 19 2001 - 11:00:27 EST

message from patrick meant for the list follows:

------- Forwarded message follows -------
Date sent:      	Mon, 19 Nov 2001 11:02:55 -0500
To:             	Andrew@lubs.leeds.ac.uk
From:           	"Patrick L. Mason" <pmason@garnet.acns.fsu.edu>
Subject:        	Re: [OPE-L:6200] Re:  econometrics

There will never be a perfect statistical model. Some variable will always 
be omitted. No empirical variable will ever match precisely its theoretical 
concept. Functional form questions will always be present. Perfect 
explanation of past events will never be sufficient to provide perfect 
prediction of future events. (Neural network models are great at 
explanation and lousy at prediction). For these reasons and many others we 
typically need a large number of studies using many data sets and 
approaches to  establish the nature and extent of expected correlations. 
Even when we have a well known statistical correlation, e.g., individual 
income is positively correlated with individual years of education and 
positively correlated with individual quality of education, it does end 
theoretical debate over what the correlation represents.

Descriptive statistics should be used.
Regressions should be used.
Calibration should be used.
Experimental economics should be used.
Simulation should be used.

Economics is doomed to remain an imprecise science. So, we should always 
use a wide variety of techniques to establish the nature and extent of 
quantitative relations between variables.

peace, patrick l mason

At 10:50 AM 11/19/01 +0000, you wrote:
>random comments on Allin's comments:
>1) Whether the only game or not, the trouble I find with typical
>econometric methods (typically based on multiple regression), is
>that the theoretical assumptions are *always* broken - to this day I
>am left wondering what the meaning of a regression coefficient and
>loads of test results is when the assumptions that would give
>meaning to the coefficients and tests do not hold. This is a very
>strange 'science' where no really seems to know what they are
>doing or why...and no one ever articulates the true mode of
>reasoning whereby they attach meaning to results.
>2) Does 'calibration' count as another game so suggesting there is
>another game in town after all?
>3) Why do you (Allin) reject Tony Lawson's characterisation of
>neoclassical economics as deductivist? As you know, Lawson's
>arguments tie econometrics closely to neoclassical economics, if
>they are accepted.
>On 18 Nov 2001, at 23:37, Allin Cottrell wrote:
> > On Sun, 18 Nov 2001, michael perelman wrote:
> >
> > > I tried to show one of the problems with Marxian econometrics.
> > >
> > > www.ucm.es/wwwboard/bas/messages/246.htm
> >
> > Eh?  This is Spanish-language get-rich-quick Internet spam.  What is
> > the relevance?
> >
> > Like Patrick, I think econometrics is not really optional.  In the
> > broad sense (that used by the founders of the Econometric Society in
> > the 1930s) econometrics is simply quantitative economic analysis.  In
> > the somewhat narrower sense that the term 'econometrics' has acquired
> > since WWII, it is quantitative analysis based on probability theory,
> > or more specifically the theory of statistical inference as elaborated
> > by Bayes, Neyman and Pearson, and R.A. Fisher (there are differences
> > between these three approaches but modern econometrics comprises them
> > all).  I don't say that these theories are beyond question, but in the
> > last half-century nobody has proposed any alternative paradigm for
> > testing/quantifying theory against non-experimental data.  They are,
> > collectively, the only game in town.
> >
> > I don't see any inherent connection between neoclassical economic
> > theory and econometrics.  They just "happen to go together" in that
> > neoclassical theory has been dominant in the period since econometrics
> > came to maturity; it has therefore provided most of the hypotheses
> > that econometricians have wanted to test.
> >
> > Jerry mentioned some "other approaches" to quantitative analysis,
> > including input-output analysis and "chaos theory".  These are of
> > interest but they are not in competition with econometrics; they are
> > both theoretical frameworks rather than methodologies for empirical
> > analysis as such.
> >
> > Allin Cottrell.
> >

------- End of forwarded message -------

This archive was generated by hypermail 2b30 : Sun Dec 02 2001 - 00:00:06 EST