Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Match treated to not-treated firms


From   Austin Nichols <[email protected]>
To   "[email protected]" <[email protected]>
Subject   Re: st: Match treated to not-treated firms
Date   Thu, 20 Mar 2014 11:56:32 -0400

Laura-Sophie Henning <[email protected]>:
Per the Statalist FAQ, you are obligated to provide the provenance of
user-written commands, but none of them are required for what you
want, or what you say you want rather.  You say you want to sort your
data 300 times and pick the top 5 matches within group each time.  Why
you are choosing this method of picking "matched" firms is less clear
to me.  The commands you cite are supported by articles describing
their asymptotic properties, at least, and I'm not sure if you are
trying to construct counterfactual outcomes or run a downstream
regression (for which you would want SE that do not pretend that each
matched firm is drawn via a simple random sample of all
counterfactuals).

Try something like this, and read
http://www.stata-journal.com/article.html?article=pr0004 and
http://www.stata-journal.com/article.html?article=pr0005 for details.

forv j=1/5 {
 g match`j'=.
 }
forv i=1/300 {
 g d`i'=!(id==`i')
 egen m=mean(btmratio/(id==`i')), by(yr sic size)
 g distance=(btmratio-m)^2
 sort yr sic size d`i' t distance
 forv j=1/5 {
loc n=`j'+1
by yr sic size:replace match`j'=id[`n'] if id==`i'&t[`n']==0
  }
 drop not`i' m distance
 }



On Thu, Mar 20, 2014 at 8:12 AM, Laura-Sophie Henning
<[email protected]> wrote:
> Dear all,
> I want to identify matches for a set of treated firms and am not sure which matching command allows best to incorporate my matching criteria.
> My sample includes 300 treated and 25,000 not-treated firms. The binary variable "treated" (either 0 or 1) indicates treatment. The sample spans the years 2000-2011, across all 2-digit SIC codes.
> For each treated firm, I would like to find 5 non-treated matches (with replacement). I would want non-treated matches to be in the same year, 2-digit SIC industry, and size quintile (total assets), and to have the smallest possible absolute difference in terms of book-to-market ratio.
> I had a look at cem, nmatch and mahapick but did not find a way to account for all these criteria. Can somebody help? Thanks in advance!
> Kind regards,
> Laura Sophie Henning

*
*   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/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index