st: Panel cointegration: comparison among xtpmg, xtwest and xtmg

[Paulo Regis <pauloregis.ar@googlemail.com>]

Dear Stata users,

I am working with a panel of 12 countries and 51 years to study a
panel cointegration relationship in the exchange rate. To my
knowledge, there are three Stata commands that may be of help: -xtpmg,
-xtwest and –xtmg. Since all of them perform the Mean-Group Estimator,
I have compared them to obtain similar outputs.

However, I am having some problems with the coefficients of the long
run relationship. Down, you have two examples using the data from
Pesaran, Shin and Smith (1999) (this is also the data that Blackburned
and Frank, the creators of xtpmg, use to explain how xtpmg works, and
can be downloaded from ideas repec at:
http://ideas.repec.org/c/boc/bocode/s456868.html
). The first example is the simplest, I could come up with. The second
example is a bit more complicated but it is more clear the results are
not easy to compare.
The output of –xtpmg and –xtwest are very similar: the coefficients
match. However, I am having problems with -xtmg. The coefficient on
the short run adjustment coefficients match but the coefficients in
the long run adjustment do not.