<>
Daniel said
I have a general question about how to handle overlapping observations
in a fixed (or random) effects panel regression. Specifically, I have
quarterly observations of investment, cash flow, and a number of other
variables. The model specifies a yearly relationship where investment
in a specific product line over the 12 months is dependent on the cash
flow over the same period and the asset base at the beginning of the
12 month period.
If I simply run the regression with all of the observations, I must be
double counting information. It seems like I only have 1 real
independent observation for every 4 quarterly observations.
If the model says that investment over the last four quarters depends
on current cashflow and the asset base four quarters ago, that
relationship should hold every quarter; there's nothing sacred about
the last quarter of the calendar year. So something like
webuse grunfeld,clear
// tsmktim from ssc
tsmktim yq, start(1990q1) i(company)
g inv12 = invest + L.invest + L2.invest + L3.invest
g kstock12 = L4.kstock
g cashflow = D.mvalue
xtreg inv12 cashflow kstock12, vce(cluster company) fe
makes full use of all available observations in the panel.
KIt
Kit Baum | Boston College Economics & DIW Berlin | http://ideas.repec.org/e/pba1.html
An Introduction to Stata Programming
| http://www.stata-press.com/books/isp.html
An Introduction to Modern Econometrics Using Stata | http://www.stata-press.com/books/imeus.html
*
* 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/
