How to solve auto-correlation using unbalanced panel with gaps

[Alexandra Maso]

Hi all,
I am dealing with unbalanced panel data with gaps, it is a survey to manufacturer companies between 1990 to 2012. I am using xtreg, vce (cluster varlist) fe in stata 16.1.

xtreg growth innov exp rec lnage lnpertot size profg i.naceclio i.year, vce (cluster id) fe

Where growth is the log (sales n/ sales n-1) and innov exp and rec are independent variables regarding innovation and market dynamism, all IV are dummies. The other variables are control variables.

My doubt is that I have found serial correlation using xtserial:

And when I run the Portmanteau test -xtistest-, it confirms autocorrelation and is not balanced for gaps?

I have two questions:

1. Do you have any suggestion to solve these serial correlation problems?
2. Can you provide me information regarding the meaning of this "gaps" problem?

Thank you in advance for your help.

PD. This is my first time in the Stata forum, if I am not posting correctly, I will appreciate your comments to improve for the next times.

Best Regards,

Alexandra.

[Carlo Lazzaro]

Alexandra:
welcome to this forum.
1) -vce(cluster panelid)- option deals with both heteroskedasticity and serial correlation. Hence, you're code still takes serial correlation into account;
2) as you can read in -xtset- entry, page 512, Stata .pdf manual, the -but with gaps- note simply means that, within some panels, one or more years are skipped (or, better phrased as per Stata .pdf manual,

...within some panels,some time values are missing.

.