# [OPE-L:6209] Re: Re: econometrics

From: Patrick L. Mason (pmason@garnet.acns.fsu.edu)
Date: Mon Nov 19 2001 - 17:09:50 EST

```Jerry:

Thanks to your history of thought posting I don't see the point of
disagreement.

Consider the snipet below.

> > Consider the simple regression coefficient b = cov(x,y)/var(x). Why is
> > this  confined to neoclassical economics?
>
>Continue.
>So far you don't have a specification of an econometric model. As one
>continues on with the specification of a model and the parameters and
>assumptions of the model, then marginalist concepts including factors
>of production and production functions (and usually linearity) are
>smuggled in.

The coefficient is the econometrics. Everything after that and before that
is theory.

Y = f(X)                 is a (general) theoretical proposition

Y = BX                  is a specific linearization of the theory. No
'metrics yet.

Note: linearity is not a marginalist concept. It is a convenient concept.
Sometimes appropriate, sometimes not. In any case, there are non-linear
techniques.

Further, it is economic theory that tells us where B = 0, B < 0, B > 0, B <
1, B > 1, etc.

Since X and Y are theoretical concepts we will never observe either as a
precisely measured variable in the actually existing economy. For example,
we may have a theory drawing on prices of production - which are never
observed anywhere. All we ever have are market prices. Even then, they
aren't measured precisely. So, what we get is

y = bx + e,

where y, x are the imprecise empirical measures of Y, X, respectively, and
e is an error.

It is theory, not econometrics that imposes structure on the error term.
The econometrics simple tells us how to obtain b, given the theoretical
assumptions regarding the error term and all of the assumptions we are
willing to accept regarding the relationship between x and X and y and Y.

For me, econometrics = statistics. So, when I calculate simple descriptive
statistics I'm doing low brow econometrics.

Let me accept your argument that marginalists created regression analysis,
i.e., "econometrics." That does not imply that the procedure cannot be
divorced from the theory.

See, for example, Semmler, W., Competition, Monopoly and Differential
Profit Rates, (New York: Columbia University Press) 1984.

As for my working paper, I've attached a copy. Among others, Marxists have
spent a great deal of time discussing the social construction of race (and
gender). Effectively, this means that racial identity is an endogenous
economic outcome.

So, taken by itself, y = bx + d*race + e, where "race" is a dichotomous
variable, is not a particularly enlightening (no pun intended) equation.

I use evolutionary evolutionary game theory to model racial identification
among Hispanic Americans. The model has multiple equilibria.

I then use standard regression analysis to determine which equilibrium is
likely to exist in the US for Mexican Americans, Cuban Americans, and
Puerto Ricans. This is a preliminary paper, but the results thus far are (I
think) quite interesting.

peace, patrick

At 03:43 PM 11/19/01 -0500, you wrote:
>Re Patrick's [6205]:
>
> > Let's create a list of everything appropriated by neoclassical economics.
> > 1. english
> > 2. mathematics
> > 3. statistics/econometrics
> > 4. formal education
> > 5. professional refereed journals
> > Clearly, we cannot reject something just because neoclassicals have
> > appropriated it.
>
>The history of thought question here is whether econometrics was
>'appropriated by' or 'created by' marginalists.  *If* one was to define
>econometrics simply as statistical analysis by economists, then there
>were certainly  pre- and non-marginalist  'econometric' studies.
>Yet, I think this identification of statistical analysis with econometrics
>is far too an all-inclusive understanding that does no recognize the
>specific historical origins of econometrics as a field of study _per
>se_.  I especially challenge your claim in [6202] that Keynesian and
>institutional economics 'gave rise' to the intensive use of econometrics.
>I think that the emergence of econometrics as a distinct field of study
>can be traced back to the founding of the Econometric Society in 1930
>which was a consequence of an earlier meeting of Ragnar Frisch of
>Oslo, Charles F. Roos of Cornell and Irving Fisher of Yale. A pioneer
>of 'statistical economics' was Henry L Moore.  So who appropriated
>what? Both Fisher and Moore met Walras and Pareto on trips to
>Europe and Moore had "innumerable trips to Europe -- seeking out
>Walras in 1903, Pareto in 1908, Bortkiekicz in 1912, and others,
>but maintaining correspondence only with Walras" ("Henry L. Moore
>and Statistical Economics" in George J. Stiglitz _Essays in the
>History of Economics_,  University of Chicago Press, 1965, p. 345).
>The subsequent history of econometrics will also show, I believe, that it is
>a field that was not only created by marginalists but has been dominated
>by them from its earliest inception to the present day.  An examination of
>the statistical works of non-marginalists like Kalecki and Kuczynski will
>show that they produced statistical, rather than specifically econometric,
>studies.  Not all number-crunching of economic data can be said to be
>'econometric', can it?  If that were the case, then I had a former job as
>an econometrician! (see below).
>
> > Consider the simple regression coefficient b = cov(x,y)/var(x). Why is
> > this  confined to neoclassical economics?
>
>Continue.
>So far you don't have a specification of an econometric model. As one
>continues on with the specification of a model and the parameters and
>assumptions of the model, then marginalist concepts including factors
>of production and production functions (and usually linearity) are
>smuggled in.
>
> > What's wrong with non-parametric estimation, which attempts to recover the
> > functional form from the data rather than imposing the functional form on
> > the data?
>
>Nothing _per se_. It depends on the issue/data that one is
>studying/observing/
>collecting.
>
> > Jerry, please know neoclassical theorists and libertarians often many of
> > the objections that you are making. Theorists of all persuasions are often
> > hostile to econometrics and any other quantitative analysis. Why? Because
> > theorists enjoy speculating without the constraints of relevance!!! :):):)
>
>To repeat: I don't have objections to quantitative analysis as such. Indeed,
>I had a job once (for the UAW Research Department in Detroit) where
>90% (or more) of my time was spent crunching numbers. btw, I even liked
>it!  I've also taught statistics in the past -- and enjoyed that as well.
>There's something quite enjoyable about occasionally descending from the
>peak  of 'high theory' :-)  and getting one's hands dirty with some very
>concrete
>data.
>
>Yet, when you refer to theorists who 'enjoy speculating without the
>constraints of relevance'  you *must* be referring to econometricians
>themselves rather than the critics of econometrics!
>
>I have an empirical test for you: ask your students to go to the university
>library and read any econometric article from any economics journal.
>Then ask your students how many of  them considered the article to be
>'relevant'!  Indeed, one could argue that econometric studies often are
>striking and glaring examples of the utter *irrelevance* of mainstream
>theory. What's worse -- they are not only irrelevant (by and large) but
>(generally) *boring* to the point of  causing tears to be shed (you could
>also ask your students whether they thought the econometric article was
>'boring' or stimulating. I have my intuitions about what their responses
>will be).
>
>I have no doubt, though, that your working paper combining evolutionary
>game theory with econometrics and Marxian theory will be anything but
>boring or irrelevant. Could you tell us more about it? (you might even
>consider posting it for comments from those on the list. I even promise
>to be in my 'listening'  [i.e. 'lurking] rather than 'talking' mode).
>
>In solidarity, Jerry

```

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