Dear All,

The question that has been bothering me for some time is how to handle
real versus nominal prices in an econometric setting.

I suppose the question starts off with the conventional view that when
dealing with prices, one really needs to deflate the price series with a
CPI index in order to identify real price movements versus inflationary
effects.

The next stage is that one must determine what particular CPI index to
use. Most people just take the general CPI index and deflate their price
series, but is this the appropriate one to use? For instance, if you are
looking at a Manufacturing price series, e.g. the price of new cars, is
it appropriate to deflate it by a CPI that includes food prices and
mortgages in its basket of goods?

I would contend not, surely deflating a price series by some totally
arbitrary index would add-in distortion rather than eliminate it. The
appropriate index to me would be a Car price (or some higher level
manufacturing aggregate) index sub-component of the CPI.

This realisation - that a general CPI is not the appropriate deflator
and that a commodity specific one is - leads me onto the question which
made me to think very deeply about what I was really doing with my
econometrics:

Why do we deflate price series?

The answer I came up with was that we need to deflate price series to
counteract inflationary effects. This is a very simple answer, but I
suddenly saw the connection between real and nomial prices and
non-stationarity.

My question for the mailing list, therefore, is whether we can
legitimately use our tools for non-stationarity and cointegration in the
analysis of nominal price series to take into account inflationary
effects that would normally be catered for by using real prices.

My opinion is that by incorporating trend stationary (TSP) and
difference stationary (DSP) processes into the DGP for a particular
(nomial) price series explicitly takes into account the inflationary
effects impacting on that price series and that the reason why the
deflation of a price series by a CPI was originally developed was that
the tools for non-stationarity were not yet developed.

So my contention is that, for a price series, one should always run
their ADF tests with a time trend and that for a VEC model one should
always have intercepts and trends. The question is whether to have
restricted or unrestricted intercepts and trends. My gut feeling is that
they should be unrestricted.

Now, this last point is a bit controversial and thus I would like some
opinion from the mailing list on this.

The rationale for not having an explicit trend in the VEC is that there
is already a linear trend in the long-run cointegrating relations. My
point is that if we are interested in accounting for inflationary
effects in the short run deviations then we must have a time trend in
the short-run deviations as well as first differences - to account for
the TSP and DSP effects - not only in the ECM.

The rationale for not having an unrestricted intercept and trend (but
having a restricted intercept and trend) is that the nature of the trend
varies with the number of cointegrating relations when the intercepts
and trends are not restricted. This, apparently, is an unsatisfactory
property. My feeling is that this is actually a good property. The trend
term in the short-run component of the VEC changes as the number of
ECM's increase (with their linear trend component inside) indicating
that there is a "swapping" of DGP effects between the inflationary
effects and long-run trend effects as additional long run relations (the
ECMs) are added to capture more of the DGP.

So, to sum up:

(1) Can we use nominal prices and non-stationarity/cointegration as a
replacement for real prices?
(2) Can we explicitly incorporate intercepts, time trends and first
differences to capture the TSP and DSP effects of inflation or do we
only use intercepts and trends if we have a valid economic reason to do
so?
(3) Can we use intercepts and trends in VEC models to capture the
inflationary effects in the short-run component of the model?
(4) Can we use unrestricted intercepts and trends in the VEC or must
they be restricted?

Thanks for your time


Regards

Tim Purcell


--
****************************************************************

 Tim Purcell                 _
 School of Natural and       0]                         _
 Rural Systems Management   ||                         [0
 University of Queensland \/||\ --|-------  ------|-----||-\
 St. Lucia 4072             /\                        __||  \
 Australia                 /  \                      /    \
                          /_  |_                   _/      \_
 Tel: +61733652174
 Fax: +61733659016
 Email: [log in to unmask]

 ****************************************************************