Thanks; this is useful.
________________________________________
From: Ariel Linden, DrPH [ariel.linden(a)gmail.com]
Sent: Wednesday, August 24, 2011 6:16 PM
To: 'Gary King'; Nel Dutt
Cc: cem(a)lists.gking.harvard.edu
Subject: RE: [cem] Using CEM with Panel Data in Stata
I would argue that it depends on whether the analysis is intended to be longitudinal or
cross-sectional. If a firm is matched to others in one year, but then matched to others in
the next year, they may not be following the same trajectory over time. This may be
irrelevant if the intent is to show cross-sectional results at each time interval.
However, if the intent is to show the trajectory over time, perhaps it makes sense to
match at the baseline (or first time interval) and then observe the trajectories over
time.
Ariel
From: owner-cem_at_lists_gking_harvard_edu(a)mail.hmdc.harvard.edu
[mailto:owner-cem_at_lists_gking_harvard_edu@mail.hmdc.harvard.edu] On Behalf Of Gary
King
Sent: Wednesday, August 24, 2011 6:09 PM
To: Nel Dutt
Cc: cem(a)lists.gking.harvard.edu
Subject: Re: [cem] Using CEM with Panel Data in Stata
there's nothing necessarily wrong with this strategy. the only question for you is
whether those matches are really sufficiently equivalent. matching different years means
that for sure there will be some differences (even if only the last digit of the year!),
but you're going to need to decide whether those make a difference in your analysis
and the effect of that on the outcome is important enough.
Gary
--
Gary King - Albert J. Weatherhead III University Professor - Director, IQSS - Harvard
University
GKing.Harvard.edu<http://gking.harvard.edu/> -
King@Harvard.edu<mailto:King@Harvard.edu> -
@kinggary<http://twitter.com/kinggary> - 617-500-7570 - Asst 495-9271 - Fax
812-8581
On Wed, Aug 24, 2011 at 5:44 PM, Nel Dutt
<nilanjana.dutt@duke.edu<mailto:nilanjana.dutt@duke.edu>> wrote:
Hi all,
I am using CEM on a panel dataset with the k2k option. I am currently matching
observations by year (and a few other covariates). I was wondering if there are any
econometric problems while running regressions (post matching) with a matched sample which
contains the following situation:
in year t observation_1 is matched with observation_2, but in year t+1 observation_1 is
matched with observation_3. That is, in different years the same firm is matched with
different firms.
Thanks,
Nel
-
cem Mailing List, served by Harvard-MIT Data Center
Send messages: cem(a)lists.gking.harvard.edu
[un]subscribe Options:
http://lists.gking.harvard.edu/?info=cem
More information on cem:
http://gking.harvard.edu/cem