Regression using regular standard errors

[Sophie Starck]

Hi,

I am running a regression and my validity tests tell me to use regular standard errors (that robust are not needed). I have previously only used robust ones using the code vce(robust). What do I type in if I want to use regular ones? Or do I not need to type in anything at all?
This is the command I want to write (but use regular se instead of robust):

qui xtreg Ålder ROA Kortfrist_sku Likviditet ln_storlek Ränta, vce(robust)
qui outreg using Regression.doc, nocons addrow(Slumpmässiga effekter,Nej) se bdec (3) starlevels(10 5 1) sigsymbols(*,**,***) starloc(1) summstat(r2 \ N) summtitle(R2 \ Observationer) summdec(2 2 0) title(Tabell X. Regression \ Beroende variabel: ROA) ctitle("", "" \ "Förklaringsvariabler" \ "" "(1)") blankrows nolegend varlabels replace

Kind regards,
Sophie Starck

[Andrew Musau]

my validity tests tell me to use regular standard errors

For linear models, the recommendation is to always use robust standard errors with cross-sectional data. The reason is that these are valid both under the assumption of homoskedasticity and under the assumption of heteroskedasticity. As you are using xtreg, this implies that you have panel data. -vce(robust)- is clustering on the panel variable. In wide panels \(N>T\), the standard errors are valid under the assumption of homoskedasticity and no serial correlation and under the assumption of heteroskedasticity and arbitrary forms of serial correlation. But to directly answer your question, specifying no -vce()- options will give you conventional standard errors.

[Sophie Starck]

Thank you for your response. I will now use robust standard errors instead.