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

Re: st: AW: Matching Estimator

From   Austin Nichols <>
Subject   Re: st: AW: Matching Estimator
Date   Thu, 10 Dec 2009 12:33:54 -0500

Erasmo Giambona--
Whether you want ATE or ATT depends on the substantive question you
want the answer to.  Imagine you have all the data, including
(unobservable) counterfactual outcomes for each individual, and figure
out what calculation you would want to make.

Do you want to know:

What is the effect of the last year of education (on earnings) for
those who have 17 years of educ?  (ATT)

What would be the effect of the next year of education (on earnings)
for those who have 16 years of educ, if we made them all get 17 years
of educ? (ATC)

Or the average of those two effects, using the pop weights of those
with 16 and 17 years of educ? (ATE)

If you think people are rationally deciding to pursue more educ when
it gives a higher return, then ATT>ATC but that need not be the case
in reality.

On Thu, Dec 10, 2009 at 9:50 AM, Martin Weiss <> wrote:
> -----Ursprüngliche Nachricht-----
> Betreff: st: Matching Estimator
> Dear All,
> I am trying to use the Abadie/Imbens (2002) estimator that one can
> obtain with "nnmatch". Can somebody provide some intuition on the
> difference between Average Treatment Effect (ATE) and Average
> Treatment Effect on the Threated? In prticular, what does it mean that
> in the case of the ATE "all observations are matched to their nearest
> m neighbors of the opposite treatment group" (from help nnmatch)? Why
> would one want to match the firms in the control group?
> Any suggestions/thoughts on the issue would be appreciated.
> Erasmo
> References
> Abadie, A. and G. Imbens. 2002. Simple and Bias-Corrected Matching
> Estimators. Tech. rep., Department of Economics, UC Berkeley.

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index