Not nested Panel data

[Aya Elewa]

I am using an unbalanced firm-level panel data. I am regressing the market share of the firms on a set of independent variables that account for the firm characteristics.
I am running a fixed-effect model. My panel variable is the firm id.
I would like to cluster my error by sector to get rid of any serial correlation between firms in the same sector.

I am using this command
xtreg lshare ltariff ltfp lwages lage trade, fe vce (cluster sector)

However, I got this error "panels are not nested within clusters".

What should I do to correct for this?

Thanks in advance.

[Carlo Lazzaro]

Aya:
how did you -xtset- your data?

[Aya Elewa]

I have done the following:

xtset firm year



Thanks,
Aya

[Carlo Lazzaro]

Aya:
you should cluster on -panel.id- (which is not -sector-, but -firm- in your case).
There's no fix about that, but plugging -i.sector- among the set of your predictors:

xtreg lshare ltariff ltfp lwages lage trade i.sector, fe

As an aside (that reminds me of many years ago, when I was student in business administration, in a pre-laptop era!), -sector- should probably be -industry-.

[Aya Elewa]

Thank you a lot for you reply.

But I was having in mind to cluster by sector (industry in your words) as firms in the same industry have common characteristics that could be a source of correlation in the error term.

Using -mixed- in that case wouldn't be useful.

Thanks again!

[Carlo Lazzaro]

Aya:
the following Stata thread can be useful: https://www.statalist.org/forums/for...try-year-level