A basic problem with "overshooting" is that equations that do overshoot are
victim to the "never the same" paradigm that econometric parameters can never
capture. Soooo, we have things like "Chow Tests" that suggest some profound
"structural change" has taken place. Ex post, dates like 1873, 1884, 1893,
1907, 1929, etc. become "secular/cyclical" points in time, that with a long
enoug time series can be incorporated into a model.
A "good" econometrician would take up Doug's challenge: pre-Bretton Woods
was characterized by an exchange rate process that was qualitatively different
than post-1944. Of course the "encompassing principle" would have you throw
in all sorts of variables and then leave 'em out during one or the other
period.
Jason
"Here's the result, now get me the econometrics to back it up..."
Alfred E. Newman
