Regression: sic ommitted becuase of collinearity

[Irfan Ali]

Hello all,

I have a ten year data for 26 companies. I am running a regression after controlling for industry and fixed effect with the below command but getting the collinearity issue.

[xtreg CASH WOMENRATIO NWC TANGIBILITY LEVERAGE EFFECTIVE_TAX CASHFLOW SIZE CAPX i.year i.sic, fe robust cluster(gvkey)]

Result:

. xtreg CASH WOMENRATIO NWC TANGIBILITY LEVERAGE	EFFECTIVE_TAX	CASHFLOW	SIZE	CAPX	i.year	i.sic,	fe
note: 2810.sic omitted because of collinearity
note: 2834.sic omitted because of collinearity
note: 2840.sic omitted because of collinearity
note: 2844.sic omitted because of collinearity
note: 3021.sic omitted because of collinearity
note: 3241.sic omitted because of collinearity
note: 3300.sic omitted because of collinearity
note: 3674.sic omitted because of collinearity
note: 3711.sic omitted because of collinearity
note: 3714.sic omitted because of collinearity
note: 4210.sic omitted because of collinearity
note: 4512.sic omitted because of collinearity
note: 4813.sic omitted because of collinearity
note: 4841.sic omitted because of collinearity
note: 4911.sic omitted because of collinearity
note: 7372.sic omitted because of collinearity
note: 8090.sic omitted because of collinearity
note: 9997.sic omitted because of collinearity

After taking the first industry code (which is 2800) , rest of the industry code are ommittd citing collinearity?

Should i post more results or data to get the possible support here?

Thanks in advance.

Regards!

[Carlo Lazzaro]

Irfan:
as expected, -i.sic- is collinear with -fe-.
Hence, Stata omits it.
That said, please note that the best way to share what you typed and what Stata gave you back is using CODE delimiters (see the FAQ on this and other posting-related topics). Thanks.

[Irfan Ali]

Thanks once agin Carlo for jumping in quickly to save me.

I couldn't find the CODE delimiters that's why posted thi way. I will look again next time I am posting.

I am bit confused let's say if I ommit fe from my code sine sic is controlling for indusstry, will rest of the code suffice the purpose for year and industry fixed effects?

Thanks.

[Carlo Lazzaro]

Irfan:
- the CODE delimiters are invoked by clicking on th #shaped toggle of the advanced editor (A);
- if you go -fe-, all time-invariant parameters will be wiped out (and this is the case of -sic- code);
- you do not say whether the -fe- specification was actually supported by the outcome of the user-written command -xtoverid- (that replaces -hausman- whenever we invoke non-default standard errors);
- the industry -fe- is already included as a -panelid- in -xtset-.

[Irfan Ali]

Thanks again.

Here is how I set the tsset :

xtset gvkey year, yearly

does gvkey covers for industry fe since gvkey is used here as firm identifier?

[Carlo Lazzaro]

Irfan:
I would first decide whethet the analysis should focus on firms (as it seems) or industries.
-gvkey- covers for firm fixed effect: if it wasn't, you would have received a warning message from Stata, saying "repeated time values".
- -xtreg,fe- omits -sic- because it is a time-invariant predictor.

[Irfan Ali]

Thanks Carlo.

I want to have a regression analysis of my panel data of 26 firms and i just want to run the regression with both year and industry fixed effects. So the analysis should focus on firms, but i need to set fe for year and industry.

Now, i ran this command and it worked without any collinearity, but i am confused whether i have controlled for industry fe.

[xtreg CASH WOMENRATIO NWC TANGIBILITY LEVERAGE EFFECTIVE_TAX CASHFLOW SIZE CAPX i.year,fe robust cluster(firm)]


Regards.

P.s. i am using COMPUSTAT data

[Carlo Lazzaro]

Irfan:
no, you did not adjust (actually, you can only adjust for the other predictors, but not control for them) for industry.
And, since you use -fe- specification, you cannot adjust for -sic- as -sic-, being a time-invariant predictor, is wiped out by the -fe-machinery.

[Irfan Ali]

Apprecite your support.

Carlo, how can i then control for industry & year fe simultaneously?

Thanks.

[Carlo Lazzaro]

Irfan:
you may want to take a look at Sergio Correia's user-written programme -reghdfe- (type -search reghdfe- to spot it; then follow the instructions to install it).

[Phil Bromiley]

To elaborate on Carlo's comments, in Compustat, firms seldom change industries. This means fixed effects for firm take care of industry (since industry doesn't vary within firms).

xtreg CASH WOMENRATIO NWC TANGIBILITY LEVERAGE EFFECTIVE_TAX CASHFLOW SIZE CAPX i.year,fe robust cluster(firm)

Is a legitimate way to control for year and firm fe's which also takes care of industry. reghdfe would also do this.