re: st: Newey-West HAC estimators

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

<>
Elizabeth said

I'm trying to figure out the optimal lag selection for the Newey and West (1987, 1994) heteroskedasticity and autocorrelation consistent (HAC) covariance matrix estimator constructed using the Bartlett kernel.  

(1) Is there a rule of thumb for lag selection based on a panel sample size of around 14,200 firm-year observations?  
(2) What other criteria should I follow to select the appropriate lag lengths?


What matters is not the number of firm-years but the number of years (or max # of years) available. Try running Baum-Schaffer-Stillman ivreg2 (SSC) on one time-series from your panel, e.g.

. webuse wpi1

. ivreg2 wpi t, robust bw(auto)

and see what automatic BW selection produces (in this case 20). The help file for ivreg2 and B-S-S SJ papers (freely available) discuss the rules of thumb as well as the algorithm used for auto bw selection.

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/