[gov3009-l] Das Presentation

Shigeo Hirano shirano@fas.harvard.edu
Tue, 27 Nov 2001 11:37:17 -0500 (EST)

Hi Everyone,

In tomorrow's applied statistics seminar, Mitali Das, assistant professor
of economics at Columbia University will present her paper "Estimation of
Models with Endogenous Covariates, without Instrumental Variables".
Although the paper is not being distributed before the seminar you can
find the abstract below.

See you tomorrow!


"Estimation of Models with Endogenous Covariates, without Instrumental

Mitali Das
Assistant Professor
Department Economics, Columbia University


        This paper introduces restrictions from economics into the
estimation of models with endogenous covariates. Under weak conditions
such as Spence-Mirrlees single crossing, it is shows that slope
parameters of standard regressions models, e.g. simple linear, binary
choice, censored and truncated models, are identified. Unlike a very large
literature on estimation of such models, this paper does not depend on
instrumental variables setting them apart from a long line of
methodological and empirical work on structural estimation that began with
Koopmans (1949).
The small sample performance of the estimators is studied in a small scale
simulation study. It is also applied to a returns to education problem. In
this application, it is found that the proposed estimators that use no
instruments yield the same behavioral response as a linear IV estimator in
Card (1995) which uses an instrument known to be ``valid'' in the
empirical literature.

Shigeo Hirano                     Political Economy and Government
shirano@fas.harvard.edu                         Harvard University