文档介绍：
Whic h Momen ts to Matc h?
A. Ronald Gallan t
e T auc hen
Departmen t of Economics
Departmen t of Economics
Univ ersit y of North Carolina
Duk e Univ ersit y
Chap el Hill NC 27599-3305 USA
Durham NC 27708-0097 USA
Phone: 1-919-966-5338
Phone: 1-919-660-1812
Septem b er 1992
Last Revised Septem b er 1995
Supp orted b y the National Science F oundation. W e thank Laura Baldwin, Ra vi Bansal, John Coleman,
and An thon y Smith for helpful discussions and t w o referees for v ery men ts.
1
Running head:
Whic h Momen ts to Matc h?
Corresp onding author:
A. Ronald Gallan t Phone: 1-919-966- 533 8 or 1-919-966-2 38 3
Departmen t of Economics F AX: 1-919-966- 498 6
Univ ersit y of North Carolina e-mail: ron gallan ******@
CB 3305, 6F Gardner Hall ftp: , ,
Chap el Hill NC 27599-330 5 USA user:anon ymous, directory: pub/arg
2
Abstract
W e describ e an in tuitiv e, simple, and systematic approac h to generating momen t conditions
for GMM estimation of the parameters of a structural mo del. The idea is to use the score
of a densit y that has an analytic expression to de ne the GMM criterion. The auxiliary
mo del that generates the score should closely appro ximate the distribution of the observ ed
data but is not required to nest it. If the auxiliary mo del nests the structural mo del then
the estimator is as ecien t as maxim um lik eliho o d. The estimator is adv an tageous when
exp ectations under a structural mo del can b puted b y sim ulation, b y quadrature, or b y
analytic expressions but the lik eliho o d cannot b puted easily .
3
1 In tro duction
W e presen t a systematic approac h to generating momen t conditions for the generalized
metho d of momen ts GMM estimator [27] of the parameters of a structural mo del. The
approac h is an alternativ e to mon practice of selecting a few lo w order momen ts on
an ad ho c basis and then pro ceeding with GMM. The idea is simple: Use