Fixed effects and dynamic ordinary least squares

[matteo abbruzzese]

Dear all,

I am currently working on a research topic related on the environmental disaggregate renewable energy effectsts.
I am estimating some regression models with the objective of evaluating the effectiveness of renewable energy diffusion on the ghg emission.


The models are:

Yit= B0 + B1 RES+ B2 FOSSIL+ uit


Where: Res= share of renewable energy production; Fossil= share of fossil energy production; E= ghg emission


I would like to know what is the differece beetween fixed effects and DOLS (dynamic ordinary least squares).
I have 21 years and 28 countries.


I would very much appreciate some thoughts on this problem.
Thanks in advance!

Matteo

[Carlo Lazzaro]

Matteo:
if you have (as it seems) a panel dataset and your dependent variable is continuous, you may want to consider:
1) -xtreg, fe- for fixed effect panel data regression (however, you should not rule out a random effect specification on an a priori basis). As your T dimension is slightly smaller than the N one, you should, in all likelihood, impose -cluster- (or -robust-) standard errors to take both heteroskedasticity and (even more important) autocorrelation into account;
2) -xtabond- for dynamic panel data regression.

[matteo abbruzzese]

Thanks Carlo Lazzaro,

Yes I have a panel dataset and all variables are continuos. On my topic the literature use DOLS a lot, but I Idid not understand the difference between OLS and DOLS.
I never used xtabond yet. So I know xtreg very well. What is the main differences beetwen xtreg and xtabond??

[Carlo Lazzaro]

Matteo:
-xtabond- allows lagged dependent variable; as you know, a lagged dependent variable in -xtreg- makes the regression biased.

[matteo abbruzzese]

Thanks Carlo Lazzaro,
as I told you, the analyzed literature often uses the DOLS estimator (Dynamic Ordinary Least Squares for Cointegrated Panel Data) of Kao and Chiang (2000). This estimator give the long-run elasticities and it seems very interesting for my topic. However, I did not find detailed information on DOLS in Stata; indeed the authors use Eviews. This method is based on the cointegration relationship between the variables analyzed in the models.
I found only the "xtdolshm" command on stata, but the output is different from the analyzed literature because in Stata the output give z-statistic and in Eviews t-stastic.

[Carlo Lazzaro]

Matteo:
slide 20 of https://www.ssc.wisc.edu/~bhansen/390/390Lecture22.pdf seems to cover (marginally) what you're interested in.
Sorry I cannot be more helpful.
An aside from the pedantic corner: whenever you mention a contribution, please share full reference (see the FAQ on why this is a good idea). Thanks.