没有合适的资源?快使用搜索试试~ 我知道了~
Unconditional Quantile Regressions 2009 Econometrica
需积分: 31 1 下载量 87 浏览量
2015-02-02
01:29:19
上传
评论
收藏 1.47MB PDF 举报
温馨提示
试读
22页
We propose a new regression method to evaluate the impact of changes in the distri- bution of the explanatory variables on quantiles of the unconditional (marginal) distrib- ution of an outcome variable. The proposed method consists of running a regression of the (recentered) influence function (RIF) of the unconditional quantile on the explana- tory variables. The influence function, a widely used tool in robust estimation, is easily computed for quantiles, as well as for other distributional statistics. Our approach, thus, can be readily generalized to other distributional statistics
资源推荐
资源详情
资源评论
Unconditional Quantile Regressions
Author(s): Sergio Firpo, Nicole M. Fortin and Thomas Lemieux
Source:
Econometrica,
Vol. 77, No. 3 (May, 2009), pp. 953-973
Published by: The Econometric Society
Stable URL: http://www.jstor.org/stable/40263848 .
Accessed: 31/10/2014 08:39
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
.
The Econometric Society is collaborating with JSTOR to digitize, preserve and extend access to Econometrica.
http://www.jstor.org
This content downloaded from 130.126.162.126 on Fri, 31 Oct 2014 08:39:16 AM
All use subject to JSTOR Terms and Conditions
Econometnca,
Vol.
77,
No. 3
(May,
2009),
953-973
UNCONDITIONAL
QUANTILE
REGRESSIONS
By
Sergio
Firpo,
Nicole M.
Fortin,
and
Thomas Lemieux1
We
propose
a
new
regression
method to evaluate the
impact
of
changes
in the distri-
bution of the
explanatory
variables on
quantiles
of
the unconditional
(marginal)
distrib-
ution of
an
outcome variable.
The
proposed
method
consists of
running
a
regression
of
the
(recentered)
influence function
(RIF)
of
the
unconditional
quantile
on
the
explana-
tory
variables. The
influence
function,
a
widely
used tool in
robust
estimation,
is
easily
computed
for
quantiles,
as
well as
for other
distributional statistics. Our
approach,
thus,
can
be
readily generalized
to
other
distributional
statistics.
Keywords: Influence
functions,
unconditional
quantile,
RIF
regressions,
quantile
regressions.
1.
INTRODUCTION
In
this
paper,
we
propose
a
new
computationally
simple regression
method
to
estimate
the
impact
of
changing
the distribution
of
explanatory
variables,
X,
on the
marginal
quantiles
of
the
outcome
variable, Y,
or
other
functional
of the
marginal
distribution
of
Y. The method
consists of
running
a
regres-
sion
of a
transformation
-
the
(recentered)
influence function
defined below
-
of the
outcome
variable
on the
explanatory
variables. To
distinguish
our
ap-
proach
from
commonly
used
conditional
quantile regressions
(Koenker
and
Bassett
(1978),
Koenker
(2005)),
we call our
regression
method an
uncondi-
tional
quantile regression.2
Empirical
researchers are
often interested
in
changes
in
the
quantiles,
de-
noted
qT,
of the
marginal
(unconditional)
distribution,
FY(y).
For
example,
we
may
want
to estimate the direct effect
dqT(p)/dp
of
increasing
the
proportion
of unionized
workers,
p
=
Pr[X
=
1],
on
the Tth
quantile
of the
distribu-
tion
of
wages,
where
X
=
1 if
the
workers
is
unionized
and X
=
0
otherwise.
In the
case
of
the
mean
/a,
the
coefficient
/3
of
a
standard
regression
of Y
on
X
is a
measure
of
the
impact
of
increasing
the
proportion
of
unionized
*We thank the co-editor and three
referees for
helpful suggestions.
We
are also indebted
to Joe
Altonji,
Richard
Blundell,
David
Card,
Vinicius
Carrasco,
Marcelo
Fernandes,
Chuan
Goh,
Jinyong
Hahn,
Joel
Horowitz,
Guido
Imbens,
Shakeeb
Khan,
Roger
Koenker,
Thierry
Magnac,
Ulrich
Millier,
Geert
Ridder,
Jean-Marc
Robin,
Hal
White,
and
seminar
participants
at
CESG2005, UCL,
CAEN-UFC, UFMG,
Econometrics
in
Rio
2006, PUC-Rio,
IPEA-RJ,
SBE
Meetings
2006,
Tilburg University, Tinbergen
Institute,
KU
Leuven,
ESTE-2007,
Harvard-MIT
Econometrics
Seminar, Yale, Princeton, Vanderbilt,
and
Boston
University
for useful
comments
on earlier versions of
the
manuscript.
Fortin
and Lemieux thank
SSHRC for
financial
support.
Firpo
thanks
CNPq
for financial
support.
Usual disclaimers
apply.
2The "unconditional
quantiles"
are
the
quantiles
of the
marginal
distribution of
the
outcome
variable Y.
Using "marginal"
instead of
"unconditional" would
be
confusing,
however,
since
we
also
use
the word
"marginal"
to
refer to the
impact
of
small
changes
in
covariates
(marginal
effects).
© 2009 The
Econometric
Society
DOI:
10.3982/ECTA6822
This content downloaded from 130.126.162.126 on Fri, 31 Oct 2014 08:39:16 AM
All use subject to JSTOR Terms and Conditions
954
S.
FIRPO,
N.
M.
FORTIN,
AND T.
LEMIEUX
workers on
the mean
wage, dji(p)/dp.
As is well
known,
the same
coeffi-
cient
/3
can
also be
interpreted
as an
impact
on the conditional mean.3 Un-
fortunately,
the
coefficient
/3T
from a
single
conditional
quantile
regression,
j8T
=
Fy\r\X
=
1)
-
Fy\r\X
=
0),
is
generally
different
from
dqT(p)/dp
=
(Pr[Y
>
qT\X
=
1]-
Pr[Y
>
qT\X
=
0])/fY(qT),
the
effect of
increasing
the
proportion
of unionized workers on the rth
quantile
of
the unconditional dis-
tribution of Y.4
A
new
approach
is therefore needed to
provide practitioners
with
an
easy way
to
compute
dqT(p)/dp,
especially
when
X
is
not univariate
and
binary
as in
the above
example.
Our
approach
builds
upon
the
concept
of the
influence function
(IF),
a
widely
used tool
in
the robust estimation
of
statistical
or econometric
mod-
els. As its name
suggests,
the influence function
IF(
Y;
v, FY)
of a distributional
statistic
v(FY )
represents
the influence of
an individual observation on
that dis-
tributional statistic.
Adding
back the statistic
v(FY)
to the influence
function
yields
what we call the recentered
influence function
(RIF).
One convenient
fea-
ture of the
RIF
is that its
expectation
is
equal
to
v(FY).5
Because
influence
functions can be
computed
for most distributional
statistics,
our
method
easily
extends to other choices of v
beyond
quantiles,
such
as the
variance,
the Gini
coefficient,
and
other
commonly
used
inequality
measures.6
For
the Tth
quantile,
the influence function
IF(Y;
qT,FY)
is
known to be
equal
to
(r
-
1{
Y
<
qT})/fy{qr)'
As a
result, RIF(
Y;
qT,
FY)
is
simply
equal
to
qT
+
IF(
Y;
qT,
FY).
We call the conditional
expectation
of the
RIF(
Y;
v, FY)
modeled as a function of the
explanatory
variables,
£[RIF(Y;
v,
FY)\X]
=
mv(X),
the
RIF
regression
model.1
In
the case
of
quantiles,
£[RIF(Y;
qT,
FY)\X]
=
mT(X)
can be viewed as an unconditional
quantile
regression.
We
show that the
average
derivative of the unconditional
quantile
regression,
E[mfT(X)],
corresponds
to the
marginal
effect on the
unconditional
quantile
of a
small location shift
in
the distribution
of
covariates,
holding
everything
else
constant.
Our
proposed
approach
can be
easily
implemented
as an
ordinary
least
squares (OLS) regression.
In
the case of
quantiles,
the
dependent
variable
in
the
regression
is
RIF(Y;
qr,FY)
=
qT
+
(r
-
1{Y
<
ÇtD/MÇt).
It is
easily
3The
conditional
mean
interpretation
is the
wage change
that a worker
would
expect
when
her union
status
changes
from
non-unionized
to
unionized,
or
/3
=
E(Y\X
=
1)
-
E(Y\X
=
0).
Since the unconditional mean is
fi(p)
=
pE(Y\X
=
1)
+
(1
-
p)E{Y\X
=
0),
it
follows that
drtpydp
=
E{Y\X
=
1)
-
E(Y\X
=
0)
=
p.
4The
expression
for
dqT(p)/dp
is obtained
by implicit
differentiation
applied
to
FY(qT)
=
P
•
(Pi[Y
<qT\X
=
l]- Pr[7
<
qT\X
=
0])
+
Pr[Y
<
qT\X
=
0].
3
Such
property
is
important
in
some
situations,
although
for
the
marginal
effects in
which we
are interested in this
paper
the
recentering
is
not
fundamental.
In
Firpo,
Fortin,
and
Lemieux
(2007b),
the
recentering
is useful because it allows
us to
identify
the
intercept
and
perform
Oaxaca-type decompositions
at various
quantiles.
6See
Firpo,
Fortin,
and Lemieux
(2007b)
for such
regressions
on the variance
and Gini.
7In the case of the
mean,
since the
RIF is
simply
the outcome
variable
Y,
a
regression
of
RIF(
Y;
fi)
on X is the same as an OLS
regression
of
Y
on
X.
This content downloaded from 130.126.162.126 on Fri, 31 Oct 2014 08:39:16 AM
All use subject to JSTOR Terms and Conditions
UNCONDITIONAL
QUANTILE
REGRESSIONS
955
computed
by
estimating
the
sample quantile
qT,
estimating
the
density
fY(qT)
at that
point
qT using
kernel
(or
other)
methods,
and
forming
a
dummy
variable
MY
<
qT], indicating
whether the value of
the outcome variable is below
qT.
Then we can
simply
run
an
OLS
regression
of this new
dependent
variable on
the
covariates,
although
we
suggest
more
sophisticated
estimation methods
in
Section
3.
We view our
approach
as an
important complement
to the literature
con-
cerned with the
estimation of
quantile
functions.
However,
unlike Imbens
and
Newey
(2009),
Chesher
(2003),
and
Florens,
Heckman,
Meghir,
and
Vyt-
lacil
(2008),
who considered
the identification of structural
functions defined
from conditional
quantile
restrictions
in
the
presence
of
endogenous
regres-
sors,
our
approach
is
concerned
solely
with
parameters
that
capture changes
in
unconditional
quantiles
in
the
presence
of
exogenous regressors.
The structure
of the
paper
is
as follows.
In
the next
section,
we
define the
key object
of
interest,
the "unconditional
quantile partial
effect"
(UQPE)
and
show
how RIF
regressions
for the
quantile
can
be
used to estimate the
UQPE.
We
also
link
this
parameter
to
the
structural
parameters
of a
general
model and
the conditional
quantile
partial
effects
(CQPE).
The estimation issues
are
ad-
dressed
in
Section
3. Section
4
presents
an
empirical application
of
our method
that
illustrates well the difference between our method and conditional
quan-
tiles
regressions.
We conclude
in
Section 5.
2. UNCONDITIONAL
PARTIAL
EFFECTS
2.1.
General
Concepts
We
assume that
Y
is observed
in
the
presence
of covariates
X,
so
that
Y
and X have
a
joint
distribution,
FY,x(;
•)
'•
R
x
X
-►
[0,
1],
and
X
c
R*
is
the
support
of X.
By
analogy
with
a
standard
regression
coefficient,
our
object
of
interest
is the effect of a small increase
in
the location of
the distribution of the
explanatory
variable
X
on the rth
quantile
of the
unconditional distribution
of
Y
. We
represent
this small location shift
in
the distribution
of
X in
terms of
the counterfactual
distribution
Gx(x).
By
definition,
the unconditional
(mar-
ginal)
distribution function of
Y
can be written
as
(1)
FY(y)
=
J
FYlx(y\X
=
x)
•
dFx(x).
Under the
assumption
that the conditional
distribution
FY\x(-)
is unaffected
by
this small
manipulation
of
the distribution
of
X,
a
counterfactual
distribution
This content downloaded from 130.126.162.126 on Fri, 31 Oct 2014 08:39:16 AM
All use subject to JSTOR Terms and Conditions
956
S.
FIRPO,
N. M.
FORTIN,
AND T.
LEMIEUX
of
Y,
G*Y,
can be
obtained
by
replacing
Fx(x)
with
Gx(x)8:
(2)
GY
(y)
=
J
FYlx(y\X
=
x)
•
dGx(x).
Our
regression
method builds on
some
elementary properties
of
the
influ-
ence
function,
a
measure introduced
by
Hampel (1968,
1974)
to
study
the in-
finitesimal
behavior of
real-valued
functional
v(FY),
where
v:Fv-+R,
and
where
Tv
is
a class of
distribution functions
such that
FY
e
Tv
if
\v(F)\
<
+oo.
Let
GY
be
another distribution in the
same class. Let
FY,t.GY
G
^
represent
the
mixing
distribution,
which is t
away
from
FY
in
the direction of the
proba-
bility
distribution
GY
:
FY,t.Gy
=
(1
-
1)
•
FY
+
1
•
GY
=
t
•
(GY
-
FY)
+
FY,
where
0
<
t
<
1.
The directional
derivative of v
in
the
direction of the distribution
GY
can be
written
as
(3)
Um
S?
HFy,,gy)-HFy)
=
dv{FYtt.GY)
Um
S?
/
=
dt
t=Q
=
JlF(y;p,FY).d(GY-FY)(y),
where
IF(};;
v, FY)
=
dv(FY,t.Ay)/dt\t=0,
with
Ay
denoting
the
probability
mea-
sure that
puts
mass
1 at
the value
y.
The von
Mises
(1947)
linear
approximation
of
the functional
v(FYjt.GY)
is
v(FY,t.GY)
=
v(FY)
+
t.j
IF(y;
v, FY)
•
d(GY
-
FY)(y)
+
r(t;v;GY,FY),
where
r(t\
v\
GY, FY)
is a
remainder term. We
define
the
recentered
influence
function
(RIF)
more
formally
as the
leading
terms of the above
expansion
for
the
particular
case where
GY
=
Ay
and
t
=
1. Since
/
IF(y;
v,
FY
)
•
dFY(y)
=
0
by
definition,
it follows that
RIFO;
v, FY)
=
v(FY)
+
j
IF(s; v, FY)
•
dAy{s)
=
KFy)
+
IF(y;i/,Fy).
Finally,
note that the last
equality
in
equation (3)
also holds
for
RIF(>>;
v,FY).
In
the
presence
of
covariates
X,
we can use the law of iterated
expecta-
tions
to
express
v(FY )
in
terms of the
conditional
expectation
of
RIF(_y;
v, FY)
8
Instead of
assuming
a constant conditional
distribution
FY\x(-\-),
we could allow the condi-
tional
distributions to
vary
as
long
as
they converge
as the
marginal
distributions
of X
converge
to one another.
This content downloaded from 130.126.162.126 on Fri, 31 Oct 2014 08:39:16 AM
All use subject to JSTOR Terms and Conditions
剩余21页未读,继续阅读
资源评论
sinat_25766775
- 粉丝: 0
- 资源: 1
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功