The
Blinder-Oaxaca
Decomposition
Uses:
Decomposes
outcome
variables
into
explained/unexplained
variation.
For
example
decomposes
wages
into
explained
variation
(e.g.
level
of
schooling)
and
unexplained
variation
(e.g.
discrimination,
new
policies
or
structural
changes).
Example:
Male/Female
Wage
Differentials
W
=
wages;
=
intercept;
=
coefficient
of
years
of
schooling
(S
);
=
error
term;
M
=
men;
F
=
women.
WF = F + F SF + F
WM = M + M SM + M
M W
F
*We
want
to
understand:
W
We
want
to
look
at
wage
differentials
between
men
and
women,
so
we
construct
a
counterfactual
equation
(intercept
&
coefficient
replaced
with
those
from
mens
equation).
WF = M +
M SF
*We
want
to
understand:
+ F
M
W
F = W
M
W
F + W
F
W
F
W
Difference
due
to
education
Unexplained
difference
The
Blinder-Oaxaca
decomposition
equation
is
a
combination
of:
Characteristics
effects
(explained
variation):WM
F =
W
Visual
Examples:
M ( SM
M ( SM
F = (M
W
Coefficients
effects
(unexplained
variation):
WF
M
W
F =
W
SF ) + (M
F ) + (
F ) + (
Total
Difference
Characteristics
Effects
Discrimination
Effects
F ) SF
F )
S
F )SF
Stata
Applications:
As
of
2010
an
oaxaca
command
is
available
in
Stata
that
computes
the
Blinder-Oaxaca
decomposition.
This
should
come
standard
on
new
versions
of
Stata,
but
if
you
dont
have
it
you
can
download
it
from
SSC
(the
makers
of
Stata).
*decompose
&
decomp
are
older
versions
of
the
same
command.
Syntax:
Relevant
Sources
for
Further
Reading:
Blinder-Oaxaca
Decomposition
Technique."
International
Encyclopedia
of
the
Social
Sciences.
2008.
Encyclopedia.com.
<http://www.encyclopedia.com>
Explaining
Differences
between
Groups:
Oaxaca
Decomposition.
Health
Equity,
Ch.
12.
The
World
Bank.
<http://siteresources.worldbank.org/INTPAH/Resources/Publications/459843-
1195594469249/HealthEquityCh12.pdf>
A
Stata
implementation
of
the
Blinder-Oaxaca
decomposition.
The
Stata
Journal.
May
2008.
ETH
Zurich
Sociology.
<http://www.ssc.wisc.edu/~jmuniz/jann_oaxaca.pdf>
St:
New
version
of
-oaxaca-
available
from
SSC.
The
Stata
Listserve.
Jan
2010.
<http://www.stata.com/statalist/archive/2010-01/msg00042.html>