 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: fully modified OLS and dynamic OLS
Bob:
Looks like we have similar suggestions and references in mind. I'm
curious how to use -reg3- in this instance: would you specify lagged
dependent variables as the instrument set? I'm trying to envision
how to fit a simple income-consumption cointegrating relationship
(e.g., the permanent income hypothesis, PIH), or maybe an application
to purchasing power parity (PPP), into a 3SLS framework -- if you
could provide more detail (or some example code), I suspect both
Grant and I would benefit.
You are right to note Saikkonen (1991) as the main reference for
dynamic OLS, but I suspect the technique of Stock and Watson (1993)
-- who propose what might be called "dynamic GLS" -- is more commonly
used. (Aside: is there a second (2003) edition of Maddala & Kim?)
Grant:
Bruce Hansen has Gauss and Matlab code to estimate cointegrating
relationships by FM-OLS that likely could be adapted to Stata/Mata:
see <http://www.ssc.wisc.edu/~bhansen/progs/jbes_92.html>. If you are
interested in Peter Pedroni's panel FM-OLS estimator, your best bet
still looks to be RATS: see <http://www.estima.com/procs/PANELFM.SRC>.
Hope this helps,
Mike
References (thereby denying Nick a datapoint suggesting economists
don't provide proper citations):
Hansen, Bruce, "Tests for Parameter Instability in Regressions with I
(1) Processes," Journal of Business and Economic Statistics, v.10 n.
3, 1992, pp. 321-335.
Maddala, G.S. and In-Moo Kim, "Unit Roots, Cointegration, and
Structural Change," (1st edition?) Cambridge University Press, 1998.
Pedroni, Peter, "Purchasing Power Parity Tests in Cointegrated
Panels," Review of Economics and Statistics, v.83 n.4, 2001, pp.
727-731.
Pedroni, Peter, "Fully-Modified OLS for Heterogeneous Cointegrated
Panels," Advances in Econometrics, v.15, 2000, pp. 93-130.
Saikkonen, Pentti, "Asymptotically Efficient Estimation of
Cointegration Regressions," Econometric Theory, v.7 n.1, 1991, pp. 1-21.
Stock, James and Mark Watson, "A Simple Estimator of Cointegrating
Vectors in Higher Order Integrated Systems," Econometrica, v.61 n.4,
1993, pp. 783-820.
On Mar 18, 2009, at 11:25 PM, Robert A Yaffee wrote:
The fully modified OLS was developed by Phillips and Hansen in 1990.
Their option corrects for endogeneity and serial correlation. The
Newey-West
option is the closest thing to that but it does not necessarily
correct
for endogeneity in a multivariate system. Perhaps the 3sls in the
reg3
command would give you a close approximation.
On Mar 18, 2009, at 11:36 PM, Robert A Yaffee wrote:
Grant,
As for the Dynamic OLS, if this is the method put forth by
Saikkonen
in 1991, this could be done with Stata easily.
As mentioned in GS Maddala and I.M. Kim's Unit Roots,
Cointegration,
and Structural Change 2003, Cambridge University Press, p. 163:
the model regresses a y(1,t) on a y(2,t) and a d.y(2,t-j)
would be formulated as a regression
of y(1,t) = B'y(2,t) + Sum (from j=-k1 to K2) b(j)d.y(2,t-j) + v
(t) ,
where k1 and k2 are selected to increase at an appropriate rate up
to T.
The method adds leads and lags to y(2,t) but not to y(1,t).
You could estimate this with 3sls.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/