st: musings on the vce architecture: ways to implement improvements to robust and clustered standard errors

[László Sándor <sandorl@gmail.com>]

Hi,
Guido Imbens and Michal Kolesár have a new paper out on a simple
procedure to improve the final sample properties or "robust VC"
estimators, incl. the case with clustering:
http://www.nber.org/papers/w18478

As Stata is the weapon of choice for most applied microeconomists
doing empirical work, where robust standards and clustering are almost
obligatory, I wonder how easily these suggestions could be added to
Stata. Is the -, vce()- (or -, robust-, or -, cluster()-) architecture
easy to modify? Or end-users could only code up duplicates of the
original commands with some tweaks to the VC estimation? In any case,
can we hope from fellow StataCorp readers that such implementations
can be coming in, say, Stata 13? (I know, I know, nobody will confirm
this or the contrary — consider this email a shout-out to the Imbens
papers and a feature request, not a real question.)

I am also asking because last year Imbens had another paper with
reasonable improvements to standard procedures about standard errors
for many applications, which again looks fit for tweaking something in
the vce/robust/cluster options, not writing improved -regress- etc.
commands: http://www.nber.org/papers/w17442

Thanks,

Laszlo

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/