Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: ivreg2, bw vs cluster--2nd attempt


From   "Mark Schaffer" <[email protected]>
To   [email protected]
Subject   Re: st: ivreg2, bw vs cluster--2nd attempt
Date   Thu, 15 Sep 2005 23:07:45 +0100 (BST)

CR,

> Sorry if this is a double-posting.  I got an error message
> after the first attempt and couldn't tell whether it went
> through.
>
> I have a question about different approaches to robust
> standard errors using ivreg2 with panel data.  My
> understanding is that ivreg2 with the bw() option produces
> Newey-West standard errors.  In the panel context, I believe
> that using the cluster option is analagous to applying
> Newey-West where we allow for all lags to be important.
> Hence, I would expect that using the cluster option is
> equivalent to speciying robust and bw(t), where t is the
> number of time periods in the data.  However, I see that the
> results are not the same.  Here is an example:
>
> use http://fmwww.bc.edu/ec-p/data/macro/abdata.dta
>
> tsset id year
> ivreg2 n w k ys, bw(9) robust small /* newey-west? */
> ivreg2 n w k ys, cluster(id)
>
> So either my understanding of newey-west and cluster is wrong
> or my attempt to implement them is wrong (or both!).

It's the former.  The kernel-based approach (Newey-West and its relatives)
is quite different from the cluster-robust approach.

In the cluster-robust approach, the within-group autocorrelation can be
completely arbitrary.  The asymptotics rely on the number of groups
(clusters) going off to infinity.  Hence this approach is well-defined
only for panel and panel-like data.

In the kernel-based approach, the asymptotics rely on the number of
periods going off to infinity.  Hence this approach is usually used for
pure time-series applications, since most panels these days are large-N
small-T panels.  However, you have to make some assumptions about the
asymptotic behaviour of the time series for the approach to be valid
(roughly speaking, how quickly the autocorrelations take to die out).

Hope this helps.

Cheers,
Mark

> Does
> anyone have insights into these issues?  Thanks.
> *
> *   For searches and help try:
> *   http://www.stata.com/support/faqs/res/findit.html
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>


Prof. Mark Schaffer
Director, CERT
Department of Economics
School of Management & Languages
Heriot-Watt University, Edinburgh EH14 4AS
tel +44-131-451-3494 / fax +44-131-451-3294
email: [email protected]
web: http://www.sml.hw.ac.uk/ecomes



__________________________________________________________________

DISCLAIMER:

This e-mail message is subject to http://www.hw.ac.uk/disclaim.htm
__________________________________________________________________

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index