st: RE: Propensity Score Matching

[Joe Canner <jcanner1@jhmi.edu>]

Tarik,

Not only is it acceptable to include variables in both the matching process and in later regression analyses, in my opinion it is almost obligatory.  Propensity score matching only matches on the probability of receiving the "treatment".  Accordingly, two subjects could have identical propensity scores but have a totally different set of baseline characteristics.  Therefore, it will still be important to account for these characteristics in your regression.  Propensity score matching gives you roughly comparable groups (at least with respect to the matching variables) but doesn't entirely eliminate the effect of the baseline characteristics on your outcome.

Regards,
Joe Canner 
Johns Hopkins University School of Medicine

-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Tarik Demp
Sent: Tuesday, March 11, 2014 2:04 PM
To: statalist@hsphsun2.harvard.edu
Subject: st: Propensity Score Matching

Dear statalist,

I have a basic question related to a propensity score matching. The model
is: I have start-up companies, some of them ask their bank for a loan, some
of them ask their family/friends. Now, I want to assess how this choice
affect the costs of the loans. Therefore, I run the following propensity
score matching:

psmatch2 bankloan age of company industry of startup etc., neighbor(10)
common outcome(loan costs)

bankloan equals 1 if they got a loan from the bank and 0 if they got a loan
from their family/friends. After having matched bank-financed startups
(treated) to family-/friends-financed start-ups (untreated), I can compute
how much more bank-financed start-ups pay compared to their peer group of
matched family-/friends-financed start-ups.

My question now is: If I would like to explain what increases the premium
bank-financed start-ups have to pay over their comparable non-bank-financed
startups can I use the variables I used in the probit estimation? So, could
I use the propensity score matching just as peer selection tool and re-use
the variables again as e.g. age of company might explain the premium
bank-financed start-ups have to pay compared to other start-ups?

Thank you for your help
*
*   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/

*
*   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/