xtscc: update available from SSC

[Daniel Hoechle]

Thanks to Kit Baum, an updated version of xtscc is now available from SSC.

The new version of the xtscc program

• estimates pooled OLS, fixed effects (FE), and FGLS random effects (RE) regression models with Driscoll and Kraay (1998, Review of Economics and Statistics 80, 549-560) standard errors. Driscoll-Kraay standard errors are robust to heteroskedasticity, autocorrelation and general forms of cross-sectional dependence.
• handles factor variables as right hand variables (thanks to Sergio Correia for helping me with this)
• allows for analytic weights (i.e. xtscc can estimate "aweighted" pooled OLS, FE, and FGLS RE models)

Best,
Daniel

[Sinem Ates]

Dear All,

The help file for xtscc command in Stata says that " xtscc produces Driscoll and Kraay (1998) standard errors for coefficients estimated by pooled OLS/WLS or fixed-effects (within) regression". So it means that xtscc command is not for random-effects model. However when I run xtscc command with random effects model, it works and gives result.

I sent an e-mail to Daniel Hoechle (the author of xtscc), however could not get a reply. I found the post of Daniel here, in this post he says that xtscc can be used with random effects model, am I right? If so, why the help file says the opposite? Could you please inform me if I can rely on the results which Stata gives with the command xtcss,re?

[Sinem Ates]

Dear All,

I want to share Mr.Baum's mail which he sent me upon my question if I can use ttscc command with random effects:

The program listing itself (ssc type xtscc.ado) states that it does provide re estimates:


*
* (1) The dataset has to be tsset.
* (2) The procedure uses functions from Ben Jann's -moremata- package.
* (3) Version 1.2 of the program corrects an error in the computation of the df used for
* computing statistical inference.
* (4) Version 1.3 of the program adds option -noconstant- for estimating
* OLS regressions without intercept and for and option -ase- for estimating
* Driscoll-Kraay SE without small sample adjustment.
* (5) Version 1.4 of the program allows -xtscc, fe- to also estimate
* fixed effects regressions with analytical weights. Thereby, the
* within transform works as in Stata's official -areg- command.
* (6) Version 1.4 of the program allows -xtscc- to estimate random effects models. <------
* Thereby, the coefficient estimates match those from official Stata's
* -xtreg, re- command.
* (7) Thanks to Sergio Correia, version 1.4 of the program now also handles
* factor variables as explanatory variables.
*

[Sinem Ates]

Dear All,

Daniel's mail is as follows:


I recently completed an update to my -xtscc- program which also allows to estimate random effects models. So, yes, you can rely on these results. If you estimate a random effects model with -xtscc-, then the coefficient estimates (but not the standard errors) should coincide with those from estimating the same model with -xtreg, re-.

Best,
Daniel

[Victoria Consolvo]

Hi all,
I am wondering if there is a way that I can use xtscc but not for a pooled OLS regression. I'd like to run a LSDV estimation with fixed and time effects dummies but with the DK errors. Is there an option not to pool?
Thanks for any insight.