re: st: Clustering Standard Errors versus Dummies

[Christopher Baum <kit.baum@bc.edu>]

<>	
Natasha asks:

Can anyone please explain me the difference between clustering the
standard errors and including dummies.

So for example if am estimating an augmented production function on an
unbalanced panel dataset where I observe firms over time and I include
firm specific fixed effects. Can anyone please explain me the need
then to cluster the standard errors at the firm level?

Different assumptions are involved with dummies vs. clustering. 

Including dummies (firm-specific fixed effects) deals with unobserved heterogeneity at the firm level that if ignored
would render your POINT estimates inconsistent. Assuming i.i.d. errors (as xtreg, fe does, as it is just OLS) 
may lead to significant bias in the INTERVAL estimates (that is, standard errors) if there are significant within-panel
error correlations (that is, firm i's error is correlated over time with other errors of firm i). 

I am working, with coauthors Mark Schaffer and Austin Nichols, on a presentation on clustering for the July BOS'10 Stata Conference. It will be available on the web after the conference.

Kit



Kit Baum   |   Boston College Economics & DIW Berlin   |   http://ideas.repec.org/e/pba1.html
                              An Introduction to Stata Programming  |   http://www.stata-press.com/books/isp.html
   An Introduction to Modern Econometrics Using Stata  |   http://www.stata-press.com/books/imeus.html


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