Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: Re: Econometrics question
Kit Baum <firstname.lastname@example.org>
st: Re: Econometrics question
Tue, 30 Mar 2010 07:43:52 -0400
On Mar 30, 2010, at 2:33 AM, Kookootche wrote:
> My results look at the regular OLS estimates and then the country FE and the year FE estimates, to try to understand the differences of these effects in the cross-section and over time. In principle, I also think I should have country and year effects simultaneously... however in practice this may not be a good thing because some countries only issue 1 or 2 bonds per year or none at all, which would leave me with very little variation to explore. Nonetheless, when I include both country and year FE, my results look very similar to the results with only country FE.
> Also, note that in the results below I'm also controlling for "global" economic conditions (U.s. interest rates, 10-year swap spreads, volatility of EMBI spreads...) -- not individual episodes of crises, but these should pick up some of these simultaneous changes in spreads you suggest).
> These are the results you suggested I run:
The FE model for the full sample has a decent within R^2 (0.26). I presume L1ggdp is the gdp growth variable. For the full sample, it is negative and significant.
For the early (pre-98) sample, the within R^2 falls to 0.08, and although the point estimate of the gdp growth variable is still negative (albeit closer to zero), it cannot be distinguished from zero.
For the later (98+) sample, the model fits better overall, with a within R^2 of 0.33. The gdp growth variable is completely insignificant.
It seems to me that these results indicate that the partial effect of GDP growth on spread, taking all other factors into account as you have, is identified in the full sample, but the data do not have the power to precisely estimate it in the earlier years, and it seems quite irrelevant in the later years, cet.par.
Given that the number of countries entering the two halves of the analysis and the full analysis all differ, there is no simple algebra that relates a coefficient in the subsamples to the full-sample coefficient. This is, as you say, a highly unbalanced panel, and the algebraic constraints that might be present in a balanced panel simply do not apply here.
PS> Quite correct, in official xtreg, fe, you cannot cluster by anything but the panel unit variable or some variable that nests clusters (e.g., countries within regions).
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: