Re: Reply to B. Mitchell, P. Davidson

[ACSLKS@xxxxxxxxxxxxxxxx id <01HN1HR59RRK8Y71DC@xxxxxxxxxxxxxxxx>; Tue, 14 Feb 1995 13:55:39 -0400 (EDT)]

> FROM:  Paul Davidson
> "      Holly Chair of Excellence in Political Economy
> "      University of Tennessee, Knoxville, Tn. 37996-0550
> Cointergation of first differences implies structural relationships among
> rates of growth; with second differences it implies relationships between
> rates of change of rates of change, i.e., acceleration of economic varibles.
> Why can't there be cointegration among accelerating economic variables?
> Or even cointegration among rates of change in acceleration of variables (i.e.,
> 3rd differences)? etc. Etc. etc.
>
> Have a good day! Paul
> email address: PB108928@xxxxxxxxxxxxxx
> fax# (615)974-1686
> phone# (615)974-4221

I was responding to your message regarding B. Mitchell's random generation of
series, differencing, and testing for CI.  If there is a theoretical
relationship between/among rates of change, then the rates are used to test for
CI.  If there is a theoretical relationship between/among levels of variables,
then the levels are used to test for CI.  One does not simply difference on a
whim.  Consider everybody's favorite:  the Phillips curve.  Assuming that wage
growth is I(1) and unemployment is I(0), they would not be cointegrated and the
appropriate model would the first difference of wage growth (a second
difference) and unemployment--without an error correction.  What does it then
mean to difference these two variables and test for cointegration on the
differences??  Even if it was found that the first differences were
cointegrated, so what?--it does not fit the ECM specification and one is
involved in a fishing expedition.

By the way, Lynn Turgeon's E-Mail address is:

           ECOELT@xxxxxxxxxxx


Lonnie K. Stevans