HC2 HC3 standard errors with clusters

[FernandoRios]

Dear all
I have one question regarding robust standard errors with the command regress.
Does it make sense to combine HC2 or HC3 robust standard errors with clusters? I know it is possible to request them with regress using the old syntax

webuse dui, clear
gen clt=ceil(runiform()*10)
regress citations fines i.taxes i.csize i.college , hc3 cluster(cit)

But with the more up-to-date syntax, requesting HC3 clustered standard errors does not seem possible.

** Something like this does not work
regress citations fines i.taxes i.csize i.college , vce(cluster clt, hc3)

Thank you in advance
Fernando

[John Ham]

Dear Subscribers,

Does anyone have an answer to Fernando's question? I have the same question.

Thanks, John Ham

[FernandoRios]

Hi John
this was a question I asked when I was working on a particular command, and new far less about standard errors
short answer
in Stata 18 doing this is possible it’s one of the few features
with earlier versions you need to program it. And is not that easy.
there is a recent paper on how to obtain standard errors with clusters (don’t have the reference on hand right now) that tells you how to do it
hth

[Carlo Lazzaro]

Fernando and John (welcome to this forum, John):
while Fernando's second code gives back:

option hc3 not allowed
r(198);

I notice that, with or without -hc3-, the results of the first regression do not change:

. webuse dui, clear
(Fictional data on monthly drunk driving citations)

. 
. gen clt=ceil(runiform()*10)

. 
. regress citations fines i.taxes i.csize i.college , hc3 cluster(cit)

Linear regression                               Number of obs     =        500
                                                F(5, 50)          =      62.82
                                                Prob > F          =     0.0000
                                                R-squared         =     0.7183
                                                Root MSE          =     5.2293

                             (Std. err. adjusted for 51 clusters in citations)
------------------------------------------------------------------------------
             |               Robust
   citations | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
       fines |  -7.690437   .5098768   -15.08   0.000    -8.714555   -6.666319
             |
       taxes |
        Tax  |  -4.493918    .584169    -7.69   0.000    -5.667255    -3.32058
             |
       csize |
     Medium  |   5.492308   .5543663     9.91   0.000      4.37883    6.605785
      Large  |   11.23563   .7562745    14.86   0.000      9.71661    12.75465
             |
     college |
    College  |   5.828441   .6550647     8.90   0.000     4.512704    7.144177
       _cons |   94.21955     5.2879    17.82   0.000     83.59849    104.8406
------------------------------------------------------------------------------

. regress citations fines i.taxes i.csize i.college , cluster(cit)

Linear regression                               Number of obs     =        500
                                                F(5, 50)          =      62.82
                                                Prob > F          =     0.0000
                                                R-squared         =     0.7183
                                                Root MSE          =     5.2293

                             (Std. err. adjusted for 51 clusters in citations)
------------------------------------------------------------------------------
             |               Robust
   citations | Coefficient  std. err.      t    P>|t|     [95% conf. interval]
-------------+----------------------------------------------------------------
       fines |  -7.690437   .5098768   -15.08   0.000    -8.714555   -6.666319
             |
       taxes |
        Tax  |  -4.493918    .584169    -7.69   0.000    -5.667255    -3.32058
             |
       csize |
     Medium  |   5.492308   .5543663     9.91   0.000      4.37883    6.605785
      Large  |   11.23563   .7562745    14.86   0.000      9.71661    12.75465
             |
     college |
    College  |   5.828441   .6550647     8.90   0.000     4.512704    7.144177
       _cons |   94.21955     5.2879    17.82   0.000     83.59849    104.8406
------------------------------------------------------------------------------

[Joseph Coveney]

Does anyone have an answer to Fernando's question? I have the same question.

Have you seen this thread? In particular, take a look at Enrique Pinzon's post there, and see whether it helps.