alvarez, m. & brehm, j. are americans ambivalent towards racial policies
TRANSCRIPT
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
1/31
Midwest Political Science Association is collaborating with JSTOR to digitize, preserve and extend access to American Journal
of Political Science.
http://www.jstor.org
Are Americans Ambivalent Towards Racial Policies?Author(s): R. Michael Alvarez and John BrehmSource: American Journal of Political Science, Vol. 41, No. 2 (Apr., 1997), pp. 345-374Published by: Midwest Political Science Association
Stable URL: http://www.jstor.org/stable/2111768Accessed: 17-08-2014 02:45 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available athttp://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 contentin 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 [email protected].
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/http://www.jstor.org/action/showPublisher?publisherCode=mpsahttp://www.jstor.org/stable/2111768http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/stable/2111768http://www.jstor.org/action/showPublisher?publisherCode=mpsahttp://www.jstor.org/
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
2/31
AreAmericans
mbivalentowards
Racial
Policies?*
R. Michael Alvarez, California nstitute f Technology
John
Brehm,
Duke
University
Theory:
he
variability
n
ndividualespondent'sonsiderationsver
acial olicy
may
e
due
to
uncertaintyr
to ambivalence. mbivalences distinctrom
ncer-
tainty
n
that t stems
rom
ncommensurable
hoices,
ndcannot
e
altered ith
additionalnformation.
Methods:
sing heteroskedastic
robit echnique,
e
consider
ix
separate
ore
beliefs otentiallyelevantowardsacial olicy hoice modem acism,ntiblack
stereotyping,uthoritarianism,
ndividualism,
nd
nti-Semitism)
or our
ifferent
policy
hoices.We evaluate wo
eparate odels
or he
ource
f
ndividualari-
ance:conflictingalues nddirect ffectsf values.
Results: ur nalysisndicateshatmodem acism
rumps
ival
xplanatory
ari-
ables
n
explanations
f
racial
olicy hoice,
nd hat
ariability
n
attitudes
oward
racial olicys due to
uncertainty,
ndnot o ambivalence.
Few contemporary
olicydebates re
as
conflictuals the
debatesover
policies ntended o redress acial nequality. litesconflict ver he pecific
terms fpolicy,over the
appropriateness
f
particular oals,
and over the
symbols
nd
rhetoric hatdefine he issues.
For
some
individuals,
acial
policy
wears all the
signs
of a
policy
debate that
yields
nternal
onflict,
or
ambivalence
ver
policy
choices. Our
specific
esearch
uestion
enters
on the sources of individual
mbivalence,
manifested s
greater
mplicit
variation
n
the rangeof considerations
n
response
o
survey uestions.
If
one thinks f a
survey esponse s a sample from n
individual's
range fconsiderationsverpolicy hoice Zaller 1992), then here re two
more
mportant
ttributesf the
range
of
considerations:ts central en-
dency which
we
might ssumeto be the
xpected urvey
esponse)
s well
as its
variance
the
diversity
f
potential
nswers
respondent
might up-
ply). Typicallyresearchers ave been concernedwith
modeling
nly
the
*Anearlier ersion fthis aperwaspresentedtthe 3rdAnnualMeeting f heMidwest
Political cienceAssociation,pril -8,
1995, hicago,L. We appreciatehe ommentsf
Morgan ousser, ynn anders, aulSniderman,aura toker,nd heDuke-UNC olitical
Psychology roup.We thank aul
Sniderman,hilip . Tetlock, ndThomas iazza for
use
of
their 991Race
and
Politics urvey. lvarez hanksheJohnM. OlinFoundation
for
upport
f his research.
Editor's ote: he paper eceived heJohnprague ward romheMidwest olitical ci-
enceAssociation
n
1996.
American ournal fPolitical cience,Vol. 41, No. 2, April
997,Pp. 345-374
?
1997bytheBoard fRegentsf theUniversityfWisconsin
ystem
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
3/31
346
R.
Michael
Alvarez nd John
rehm
firstomponent
f the
urvey esponse-choices
bout
policies.Under-
standing
he actors
hich
etermine
ndividualesponses
opolicy hoice
questionss mportant.ut t lsomighte mportantoexaminehe econd
component
f the
survey esponse-the
variation
n individual
olicy
choices,
r the
range
f
considerations
hey ringo bear
n their
urvey
response.
arger ariation
n
responses
rom
n individual ight e
due
to
unreflective
esponse
anufactured
tthe
oor,
r t
might
e
symptom-
atic
of
nternalized
onflictver he
pecificolicy
omain.
The
firstorm fresponse ariability
s uncertainty,istinguished
rom
other ormsfresponse
ariability
n
that
dditionalnformationelps
he
respondento harpenisorher pinions.ncertaintybout ublic olicies
is a
functionf a
lack of
nformation,
ither temming
rom he
respon-
dent's wnpersonal
nformation
osts rfrom eficiencies
n
the ransmis-
sion of
informationrom lites o citizensAlvarez
1997;
Alvarez nd
Franklin
994;
Bartels 986). Uncertaintyan
always
be
resolved,
ow-
ever, y
additional
nformation.
The second
orm f
response
ariability
s
ambivalence,
herein
ddi-
tional
nformationnlyheightens
he
nternalized
onflict
Liberman
nd
Chaiken991;Tetlock 986),makinghe olicyhoicesmore ifficultnd
responses
more ariable. mbivalence
rises
n a choice ituation hen
the
hoice
recludes
ompromise,
uch hat
pting
or ne
alternativere-
cludes
he ther.
iterally,
ambivalence means
the imultaneousxis-
tence fconflicting
motions
Webster's
New
UniversalUnabridged
ic-
tionary,
nd
ed.,
S.V. ambivalence ),
.e.,
that policychoice
is
simultaneouslyood
ndbad.The roleof
additionalnformationoes
not
reduce
henumberf factors elevant
n an ambivalenthoice ituation;
instead,dditionalnformationakes he imultaneousactorsmore a-
lient.f twoof
a respondent'sorebeliefs onflictver policy hoice,
no amount f
additionalnformationan narrowhe esponse
ariance.
The different
rigins
f
response
nstability
re
consequential-the
doorsteppinion
uggests
hat
espondents
re
ll-equipped
o
participate
in
the
ebate,
hile
he
nternally
onflictualpinion
means
hat he
ebate
itself
s difficult.ariation
ueto ll-formedr
uncertain
esponses
ndicts
the
respondent
r the
political
rocess,
whilevariation
ue
to
difficult
choices
ndambivalencedentifies
he
very
marrow
f
politics.
Aswehave rguedlsewhereAlvarezndBrehm 995), source f
conflict
ver
olicy
hoices anbe conflictetween
orevalues. f
multiple
core
values
ertain
o
a
policy
hoice, et heymplicate
ifferentnswers
in
the
debate,
hen
espondents
ho
heedthose alues
will
be conflicted
in their nswers.
ifficult
olicy
hoices
ypically
nvolve uch onflicts.
Abortion
olicy
s
difficultor
he espondent
ho
prizes
woman's
epro-
ductive
hoice,
ut lso believes hat
umanife
begins
well
before
irth.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
4/31
ARE AMERICANSAMBIVALENTTOWARDS
RACIAL POLICIES? 347
Environmental
olicy hoices re tough or he
respondent ho values
clean
ir
ndwater, et egretsmposingny osts
n ndustryhatmight
leadtowardshe oss of obs. The decision o go to war gainst foreign
tyrants conflictual
or hosewho ue ommittingmericanroopsobattle.
Do conflicting
orevalues ndercutacial
olicy? ccordingo everal
prominentcholarlyreatments,he nswers yes.
Rokeach 1973) denti-
fies significantraction
frespondentsorn etweenwo
erminalalues,
freedomnd
quality,ver iverse olicy hoices,ncludingace.Katz nd
Hass 1988)
attributehites' onflictver acialpolicy o two ompeting
value rientations,Protestantthic nd
humanitarian-egalitarianrienta-
tion.
Do these
ompeting
orevalues
mplicate
nternalized
onflict?gain,
the
nswer
n the
cholarshipppears
o be
yes.McConahay1986) argues
that
he ifference
etweenhe
espondent's
eactivity
o
old-fashioned
racism cales racial uperiority)ndthose
measuringmodem racism
denotes zoneof mbivalencebout ace.Katz
nd
Hass
1988)
and
Katz,
Wackenhut,
ndHass
1986) argue
hatwhite
ubjects
xperiencesycho-
logical ensionnddiscomfortn
response
o he
resencef ues ctivating
conflictedacial ttitudes.aertnerndDovidio 1986)demonstratere-
sidual ore
f versive
acism,
ven
mong ubjects
rofessingeneralized
support
or acial
policy.
To even hemost
ursory
tudentf he
cholarship
n racial
ttitudes,
it s bluntlybvioushat
esearchersollide ver he asicforces hat rive
racial ttitudes.ne
groupKinder 986;
Kinder ndSears
1981;
McCona-
hay 1986)contendshat
symbolic acism, combination
f
antiblack
affect ith raditionalmerican
alues,
riveswhite esistanceo racial
policy. y
this
rgument,
hiteswho
oppose
uch
policies
s
affirmativeaction rbusing n thegroundshat lacks regetting ore han hey
deserve,
remotivated
y
a form fracism hat as
replaced
vert
xpres-
sions
f
racial
uperiority.
niderman
ndhis
colleaguesSniderman
nd
Hagen 1985;
Snidermannd Piazza
1993;
Snidermannd
Tetlock 986)
argue
hat
ymbolic
acism
ails
because
t
confoundshe
policy
hoice
with he
ttitude,
hile
t
the
ame ime
gnoring
he
ontinuingresence
of
simple
ntiblackffects a source fwhite
pposition
o racial
olicy.
We take he evel
f
onflictbout he ources
nd
meaning
fwhite
pposi-
tion o racialprogramss onepieceofevidencehatndividualttitudes
may
be in
conflict.
Both
roups
f
partisans
n the
ebate avedirect vidence bout
ndi-
vidual mbivalenceoward ace.When
nealters he
uestion
n
relatively
minor
ays,
necanelicit
ntirely
ifferent
esponses,ndeed,
vendiffer-
ent
patterns
f
responses. mong
ther
ersistent
ifferences,
inder nd
Sanders
1990)
demonstrate
hroughsplit-formurveyesign
hat
raming
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
5/31
348
R.
Michael
Alvarez nd John
rehm
thedebate s
a
question
f
reverse
iscriminationaused
egalitarian
respondentso see affirmative
ction s
undesirable,
hereas frame f
unfair dvantages ausedegalitarianso supportffirmativection.
SnidermanndPiazza
1993,
178)
conclude
n
their
ook:
Of
the
myriad
indings
e
have
reported
n
contemporary
mericana-
cial
attitudes,
he netowhichwe
ourselves
ttach
he
most
mportance
s the
pliability
f
the
policy
ositions
f
substantial
umbersf
whites n
specific
issues frace. t
has
ong een ssumed hat
hitesre
dug
non
racial
ssues.
In
fact,argenumbersfwhites anbe dislodgedromhepositionsheyhave aken nmanyssues f
racebycallingheir
ttentiono
countervailing
considerations.
The deepest
rony
f
the
dispute
s
that oth
roupsf
partisanscknowl-
edge
hat he
raming
f he
ebate
eterminesow
espondents
ill
nswer
questions
bout acial
olicy,
nd hat
oth
roups
f
partisanselieve
hat
the
ability
f
response
choes
he
broader
olicy
ebate
mong
lites.
The differencesetween ncertaintynd ambivalenceverpolicy
choice
oint
otwo riteriao
distinguish
he
oncepts.
ncertaintyeduces
with dditional
nformation,
hile
mbivalenceoes
not;
hence,
f one
s
ableto
demonstratehat etter
nformed
espondents
ave ess
variancen
their
esponses,
t
ndicates
ncertainty.
mbivalencetems rom
choice
between
ncommensurables,
nd
thedominant
iterature
oints o the
ole
of
competingalues;
ence,
f
one s
ableto
demonstratehat s
competing
values
ecome
ncreasinglymportant
o
an ndividual
hat
ndividual's
e-
sponse etbecomesmorevariable,hen tindicatesmbivalence.hesecriteriastablish
ell-definedonditionso
adjudicate etween
mbiva-
lence nd
uncertainty.
Our
point
s not
to
settle,
r
hardly
o
enter,
he
rgument
ver
he
existence
f
symbolic
acism
s a distinct
ource f
racial
ttitudes.ur
specific
nterests to
assesshow evels
f
support
or
hese alues
ccount
for oth he
olicy
hoices ndthe
variability
f
those hoices.
ur
paper
has
three arts.We
first
evelop
ix
scales
for
ertinent
aluesfromhe
1991 Race
and Politics
urvey:ymbolic
acism,
ntiblack
tereotyping,
authoritarianism,nti-Semitism,galitarianism,nd individualism. e
then-turn
owards
n
evaluation
f the ffect
f these
alues
upon
policy
choices n
this
urvey.
sing heteroskedastic
robit
model,
we look at
both
he
probability
f
support
or
he
policy ndthe
mplicit
ariancen
that
robability.
inally,
e
reflect
pon he
elationshipetween
nternal-
ized
conflictt
the evel
f
ndividual
urvey
esponseo the
ebate
mong
policymakers.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
6/31
ARE AMERICANSAMBIVALENTTOWARDS
RACIAL
POLICIES? 349
1. Measurement f Racial
Values
We constructur orebeliefs cales
n the asisofthe1991Race and
Politics urvey, data et ollected ythe urvey esearch enter fthe
University
f
California,erkeley.'
he
unique spect
f theRace and
Politics
urveywas
the
profusion
f
split
uestionnaires,andomlys-
signed o respondents,adepossible ya
Computer idedTelephonen-
terviewpproach.
achof he our
ependent
ariables e consider
ltered
importantomponentsfthe uestion.orpurposesf scaleconstruction,
however, e were tymiedythe plit
amples, ndemployednly hose
questions
hichwere
sked f all
respondents.
We useconfirmatoryactornalysisodevelopix cales, achmeasur-
ingoneof the oncepts hich ssume aryingegrees f prominence
n
scholarlyxplanations
or
variation
n
racial ttitudes.he overall it or
our
confirmatory
actormodel s
adequate
with Goodness
f Fit
ndex
of 0.86.2The firstcale s a measure f
modem acism.
According
o
the
various
uthors
e.g.,
Kinder
986;
Kinder
nd
Sanders
990;
Kinder
bTheata re available hroughhe urvey esearch enter. urrecode SAS), con-
firmatoryactorSAS), andheteroskedasticrobitilesSHAZAM)will e available hrough
the nter-universityonsortiumor olitical
nd Social Research eplicationrchive. he
survey as a telephonenterviewased nrandom-digitialing sing
stratified
wo-phase
sample electionrocedure.hefirsthase f he
rocedureampled
rom
nownrea odes
andprefixes,ppending four-digitandom umber
o generate completeen-digitele-
phone umber.he second hasedrew isproportionately
romample trataontainingt
least ne known esidentialumber,lthoughrawing
lso fromtrata here here as no
known esidential
umber.Known
residential
umbers ere rawn rom
tape
reated
by DonnellyMarketingervices.
ee
Casady
ndLepkowski1991) for urtheretails f
the ampling ethodology.he target opulation
onsistedf all English-speakingdults
over18 years ld,residingnhouseholds ith elephones,ithinhe 8 contiguoustates.
The total
ample
ize was
2,223,
with
response
ate f 65.3% a response ate
n
excess
of
typical
ates or
elephoneurveys).
2We onstructhe cales sing confirmatoryactor odelwith he AS routineROC
CALIS, multiplying
he ndicator
ariables y
thefactoroadings
n
order o obtain he
scales.We free he nter-indicatorovariances
8,
in the tandard
ISREL notation)nly
for airs f ndicators ithin factorfixingnter-indicator
ovarianceso zero
lsewhere).
We free he ovariances
mong
he actors
k).
One
factor
oadingX)
n
eachfactors fixed
at1.0to set he cale.Each scale s coerced o0-1 bounds,
here 1
denotes
hemaximum
observedevel or he cale e.g.,
1
for
uthoritarianism
enotes
respondent
ho
rovidesthemost uthoritariannswersf he
ample).
We reversehreef he calesmodem acism,
authoritarianism,
nd
galitarianism)
o
reflectur
reference
or
nterpretation
f he cales.
The reader hould ote
hat hemaximumndminimumevels or ach caledo
not
eflect
absolutes,ut nly
hemaximumrminimumbserved or he ample. hereliabilitiesas
measured ith Cronbach
lpha)
re 58 formodemacism,63
for ntiblack
tereotyping,
.65
for
nti-Semitism,
82 for
uthoritarianism,
40 for
ndividualism,
nd 52 for
galitarian-
ism.The Cronbach lphaunderestimateseliability
or
on-generic
cales
Bollen 1989,
217). Nonetheless,he eliabilityor he ndividualismtem
n
particulars quite ow.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
7/31
350
R.
MichaelAlvareznd John rehm
and
Sears
1981;
McConahay
986)
symbolicrmodem
acism
enotes
conjunction
f
antiblack
ffect ith
raditionalmerican
alues, aking
formnthe ense hat lacks rereceiving orettentionromovernment
or other
dvantages
han
hey eserve.3
We
located hree
uestions hich ap nto
he dea of
modem acism.
One of
these
uestions
s a
simple
hree
oint
ikert cale
asking
espon-
dents
o
rate he
mount
f
attention
hat overnmentas
been
paying
o
minorities.he second
ndthird
ndicatorssk how
angry he
espondent
feels bout
iving
blacks nd
otherminorities
pecial
dvantages
n
obs
and
schools,
ndthe ther
bout
spokesmen or
minoritiesho re
al-
ways omplaininghat lacks rebeing iscriminatedgainst. 4
This
scale
probably
ill
not
atisfy
ither
roup fpartisans
n
the
debate
ver
herelevance f
modern acism.
One of Snidermannd
Tetlock's
hief
bjectionso
themodem acism
oncept
s that he cales
often reat
olicy
hoices s
independent
ariables
n
the
amemodels hat
purport
o
explain olicy hoices s dependent
ariables.
...
[I]t
s
gratuitous
o
equate pposition
o
affirmativectionwith acial
prejudice-gratuitousecausetwould therwiseepossible oexaminehe
actual elationetween
he
wo,
nd hus stablishs a
matterf
fact,
ndnot
of
definition,
ow
and to
what
egree
he wo
re
connected.
uite imply,
defining
pposition
o affirmative
ction s racism
recludes
alsificationf
the
rediction
hat he wo
re ndeed
elated,
t
the ost f
making
he
elation
between
hem
tautologySniderman
ndTetlock
986,
135).
In
order o be
sensitiveothese
bjections, e
replicatell of our
models
excludinghemodem acism
cale,
nd
furthereplication
here e
only
include he ast tem complaining ). ut ndefense fthemeasure, e
note hat he
pecificolicy
eferents
n
themeasures
re rather
blique.
3McConahay1986)
prefershe erm
modern acism,
ith he
xplicitdeathat he
underlyingacial ttitudes
re
possible nlynthe
post ivil
rightsmovement
eriod, nd
because
old-fashioned
acismmightlsobe
symbolici.e.,
not
roundednrealisticroup
conflict).
inder
1986)prefers
he ermsymbolic
acism,ince
he traditional meri-
can values nvoked
ythe
dea arehardly
modern. Neither
abel s perfect.
e optfor
modern acismn he ontextf he pecificuestionsince llreferobliquely)opolicies
only
n
practiceince he
1960s.
4The
how
ngry ariablesre
caled rom
(doesn't other)o
10
extremelyngry),
while hefirst
uestion
s
scaled
oppositelyrom
(too much
ttention)o 5
(notpaying
enough
ttention).
onsidering
he
angef
he
ndicators,
he
ndicatorsoad
roughlyqually
on
the nderlying
cale,
with
stimatedactor
oadings
f -.27
(anger bout
iving
lacks
andother
minorities
pecial
dvantages)
nd-.34
(anger bout
pokesmen
or
minorities
who
are
always
omplaining);
he
ttentiono
minorityroblemsndicatoras
been
on-
strained
o a
coefficientf 1.0.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
8/31
ARE AMERICANSAMBIVALENTTOWARDSRACIAL POLICIES? 351
The attention ariable sks therespondent
o evaluate he mount f
attention rom overnmento theproblems
fminorities,ardlyn the
same evel fspecificitys the ependentariables e discuss ater nthis
paper. he atterwoquestionssk about diffuse
anger provoked y
a policy special dvantages)r complaining
pokesmen.
The weaknessn these ndicatorss seen
from he dvocates f the
symbolic
acism
oncept
ould e the bsence f pecificeferentsotradi-
tionalAmericanalues.
f
symbolicacisms the onjunctionf antiblack
affect
nd
traditionalmerican alues, hen hese uestionsredecidedly
weakas far s theirmphasisn unearned dvantages. he anguage
here mphasizestoo much ttention r special dvantages hich r-
bits omewhereear hemeaningf unearneddvantages.
In
place
of themodern acism
cale,
Snidermannd Piazza (1993)
arguefor eparate
onsiderationf alternativealues: authoritarianism,
anti-Semitism,ndividualism,
nd ntiblacktereotyping.
n
a series
f
m-
pressive ivariatenalyses,niderman
ndPiazzademonstrate
hat
uthori-
tarianisms more tronglyorrelated ith pposition
o racial
olicy
han
measures f ndividualism
which hey
ake o be
the
oreof
themodern
racism rgument).
The uthoritarianismcaledraws pon
ix ndicatorshichorrespond
to three
spects
fthe lassic
F
scale ofauthoritarianismAdorno
t al.
1950,228-41)-conventionalism,uthoritarianubmission,nd authori-
tarian
ggression-albeit
o
varying egrees
f coherence.Convention-
alism referredo
rigid
dherenceo
conventional,
iddle-class
alues
and s
captured erewith Following od's
will
and Improving
tan-
dards f
politeness
n
everyday
ehavior.
Authoritarian
ubmission
meant
submissive,
ncriticalttitude
oward
dealizedmoral uthorities
of the n-group nd s measured erewith Preservinghetraditional
ideas
of
right
nd
wrong.
Authoritarian
ggression
eferredo a ten-
dency
o be on the ookout
or, nd
to condemn,eject, ndpunish eople
who
violate
onventional
alues and
s
measuredy Respect
or
uthor-
ity,
Strengthening
aw and
order,
and
Respect
for American
power.
All
six ofthe
oadings
or he ndicatorsre
strong,
here
pre-
5Each f he ndicatorss a responsen a 10 point cale oa questionf he orm How
aboutX? Ona scale from to 10,how mportants that o you? where denotes one
of he east mportanthings nd
10
denotes one of he bsolutely ostmportanthings.
(This cale s not eversed.)heconfirmatoryactornalysis stimatesoadings f 1.00for
Preservinghe raditionaldeasofrightndwrong, 91 for Respect or uthority,74
for Following od's will, 71 for Improvingtandardsfpolitenessneverydayehav-
ior, 76 for Strengtheningaw andorder, nd 78 for Respect orAmericanower.
This cale hus mits t east ixaspects f he scale: nti-intraception,uperstition,ower
and
oughness,
estructivenessnd
ynicism,rojectivity,
nd ex.One
might
onsider
he
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
9/31
352
R. Michael
Alvarez nd John rehm
servinghe
raditionaldeas
of
rightndwrong
nd respect or uthor-
ity arethe
trongestair.
Our caleforndividualismsmeasuredy hreetems: hemportance
of elf-reliancend
mphasizing
ndividual
chievement,nd Likertcale
assessingngertgovernment
nterference.6his
s
admittedlyweak cale
(with
he
owest
eliability),
nd
anyeffect
hat t
has
in
our
subsequent
analysiss
likely
obe attenuatedueto
the
nferior
evelofmeasurement.
Thenext
cale hatweemploynour
nalysismeasures
nti-Semitism.
SnidermanndPiazza
1993)
find
modest
approximately3) correlations
between
ssessmentsfnegative
ttributes
f
blacks nd
agreement
ith
elementsf nanti-Semiticcale. nseveralnalysesntheirook, nider-
man nd
Piazza utilize nti-Semitism
s a measure f
prejudice
ecause
of
[their] elief hat heheart f
prejudice
s
captured y
the
notion f
ethnocentrism1993,107).We use
five ikertcale
questionsfour oint
scales
which
o not
nclude
he
hedging esponse
f
neither
gree
nor
disagree ).
At face
value,
ne of the
five
uestions
oes not eem o be
anti-Semitic:
Most
Jews re ambitiousndwork ard o
succeed.
We
have woresponses.
he reader hould
ecall
hat he
urpose
fthe
cale
is tomeasure religioustereotypenorderodenote rejudicend thno-
centrism.he
significantart
of
the
apparently
nomalous
uestion
s
Most
Jews
n
ight
f constructionf a
stereotype.
ur econd nswer
is data-driven:henwe
delete
his
uestion
rom he
cale,
he
emaining
loadings lunge
y nearly
5%.]
Fourth,
e
constructscalefor
egalitarianism.
he
nominalonflict
between
galitarianismnd ndividualism
ppears
rominently
nthe
itera-
ture n
attitudesoward
acial
olicy.
atzandHass
1988)
find
n
experi-
mental ork hat rimingubjectso consider nevalueorthe therig-
nificantlyncreasedcores n
correspondingttitudes,utnot n
attitudes
corresponding
o
the
nonprimed
alue. n
other ords,he wo cales
per-
respect orAmerican
ower uestion o
tap nto he power nd
toughness spect f
theF
scale.
6Two f
he uestions
reof similar ormothose
sedfor he uthoritarianism
cale
(infact, art
f he ame atteryf
uestions):Self-reliance
nd emphasizingndividual
achievement. Self-reliance s fixed t
1,
nd
emphasizing
ndividual
chievement
nd
excellencentheob scalesat only 30.Likewise, (How about)Governmentfficials
interfering
nd tryingo tellus whatwe
can and can't
do
with urown ives scales at
only 15.
7The onfirmatory
actoroadings or
heAnti-Semitismcale re 1.00for
MostJews
are
ambitiousndwork ard o
succeed, 85 for MostJews remore
willinghan ther
people o use
shady ractices
o get head
n
ife, 90
for Most
Jews
elieve
hat
hey
are
betterhan ther
eople, 76
for
MostJews n
general
re
nclined
o be more
oyal
to
srael han o
America,
nd 86
for
Most
Jews on't
are
what
appens
o
people
who
aren't
ewish.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
10/31
ARE
AMERICANSAMBIVALENT
TOWARDS
RACIAL
POLICIES?
353
ate
ndependently,
ndeachhas the
potential
o
significantly
ffect
refer-
encesfor acial
policy. ipset
nd
Schneider1978,43),
in
a review f a
range f surveyata, ee thedynamicfattitudesowards acial olicy s
between wo alues hat re tthe ore f
theAmericanreed-individu-
alism nd
egalitarianism.
As
with
ur
scale for
ndividualism,
he scale for
galitarianism
s
somewhat eak.
While hefirst ariables fixed t
1
(moremoney
eing
spent
o reduce
nemployment),
he
next
ariable
oads
at
only 33. The
non-Likertcale
questionoadsat
half
hat
-.13). Again,we expect hat
the
ffectsfthe
galitarianism
cale n the
stimationillbe attenuated.
Last we develop measurefantiblacktereotyping.hescale per-
formsdequately.llofthenegative
ttributesaggressive,azy,
oastful,
irresponsible,
omplaining)avepositive,
arge, nd tatisticallyignificant
coefficients.8
We takenote
f
the orrelations
mong
hese ix
generated
cales
see
Appendix for
he ull
orrelationatrix). one
of
he cales re
trongly
correlated.he
highestairs
f
correlations
nclude
ntiblack
tereotyping
and nti-Semitism
.39),
nd
uthoritarianism
nd
ndividualism
.37).
Most
of the emainingairs f ndicatorsre correlatedt 25or ess. The ack
of
trong
ntercorrelationsllows
s
to
be
more onfident
hat ur ndicators
are
ndeed apping
nto
eparate imensions
f possible
alues
underlying
racial eliefs.
These
cales
llow
us to evaluate he
elative eight
fthe
ompeting
explanationsor
pposition
o
racialpolicy,
n
a mode hat
s
similar o
any
other
nalysis
f
policy
hoice.But we
argue
hat
nderstanding
he
choice s
only
neof he
nterestingspects
f he
roblem,
nd
hat nder-
standing
he
variability
n
that
hoice,
or
ach respondent,
lso matters.
We employ heteroskedasticrobitechnique,imultaneouslyodeling
the
probability
hat
respondent ight hoose particularacial
policy
alternativend
he
mountfvariance hichs
associated ith hat
olicy
choice
or he
espondent.
he next ection ummarizesur
pplication
f
this
method.
8Each fthe ndicatorsor he ntiblack
tereotypingcale
s
a responseo
a
question
of
he
orm How about
X? On
a scale
from
to
10,how
welldo
you
hink
hat
t
describes
most lacks? where he caleranges rom ( very naccurate )o 10 very ccurate ).
Since
this
cale
s
ultimatelyeversed,
ach of the
factoroadings
epresents
score n a
pro-blackirection.he
factoroadings
or
he ntiblack
tereotyping
cale we
construct
are:
Aggressive
r violent
1.00),Lazy .87), Boastful.63), rresponsible.77),andCom-
plaining.68).
We
have
eplicatedur nalysis ith second calefor ntiblacktereotyping,
where
hemeasuresnclude oth ositive nd negative
tereotypes.his alternativecale
has
some erious efects
n
that hepositive nd negativettributesppear o oad better
on
separate
cales han n
a
single
cale.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
11/31
354
R. MichaelAlvarez
nd John rehm
2. Variability
f Racial Policy Choices
Themechanicsf he eteroskedasticrobitechniquere nAppendix
B. Briefly,
he deabehind
he echniques to specifynd stimateimulta-
neouslywo elatedquations,
nefor he
robabilityf choice, he
ther
for hevariance f the
rroresidual.We refero
the
firstquation
s the
Choice
Model,
and to the econd quation
s the Variance
Model.
Note hat
hevariancemodeled y
this
pproach
s the ndividualespon-
dent's
ariance
n
choice, ot
he ariancecross he
ample. ur pproach
employs
nferentialtatistics
n order
o
model nd
estimate
his ariance
in
choice.
Theheteroskedasticrobitmodel an be applied o virtuallyny ype
of discrete olicy
hoice.For
example,
n our
previous
aper
wefound
substantialupport
or
hehypothesis
hat
many
mericansre mbivalent,
and
not
ncertain,
bout bortion
olicy
hoices,
nd hat his
mbivalence
is determinedythe onflict
etweenwo ore
values,
he alue fhuman
life ndrespect
orwomen's ightsAlvarez
nd
Brehm
995).
Here we
are nterested
n
examining
he ame
hypothesis,
xcept
n
the ontextfracial ttitudes.
s we
discussed arlier,
heres substantial
discussionnthe iteraturebout he ole hatmbivalencelays nracial
politics,
ven or ndividual
hoice.
What
as
been bsents a directxami-
nation
f
the
formf the esponse
ariabilityi.e.,
uncertaintyrambiva-
lence).Ambivalence
s the orm fresponse
ariabilityntails
wo riteria,
thatndividuals ho
re
better
nformedxhibitreateresponse
ariabil-
ity, ecause
hose ndividualsisplay
onflictetween orebeliefs
rval-
ues.
Using
he 991Raceand
Politics
urvey
nd urheteroskedasticrobit
technique,
e can examine olicy
hoice ariability,nd valuate
whether
it s ambivalencer uncertaintyhat haracterizesariable pinionsbout
racial
olicies.
2.1
Choices
AboutRacial
Policy
The
1991Race
and
Politics urvey
atahas
xtensiveata bout iffer-
ent
acial olicy
lternatives.
n
this
aper,
we focus ur ttention
n
only
four ifferent
ypes
fracial
olicy
hoices:
ederalet-aside
rograms
or
minorities,references
or
ualified
lacks
n
university
dmissions,pen
housingaws, nd
whetheroraise
axes o ncreaseducationalpportuni-
tiesforminorities.e useonly hese our ince heynvolvedimple i-
chotomous
yes-no)policy
hoices.9
From he
1991
survey ata,
werecoded our ependent
ariables or
9Many
f the emainingacialpolicy hoices
n these
ata nvolved rderediscrete
choices.We
reserve
he
nalysis
fthose uestionsor
uture ork, ince
he
conometrics
of a heteroskedastic
rdered robitmodel
s
beyond he
copeof this resent
esearch.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
12/31
ARE
AMERICANSAMBIVALENT
TOWARDS RACIAL POLICIES? 355
each ype fracial
olicy hoice. hefour inary ariables
re achcoded
so that indicatesn affirmativeesponse
therespondent
avors ederal
set-asidesorminorities,hey elieve hat here ught obe preferencesor
qualified lacks n
universitydmissions,
here ught o be openhousing
laws,and that axesought o be raised
or ducational
pportunitiesor
minorities)
nd0 indicates
negative
esponse. iven his
oding, achof
the our eteroskedasticrobitmodels reestimatinghe
robabilityf an
affirmative
esponse,
nd he oefficients
e estimatehould e nterpreted
with hat
n mind.
Theres one
omplicationebrieflyentioned
n
the
reviousection,
however. ne of the nterestingspects f the1991data nvolved heuse
of
randomizeduestion
xperimentshich oseddifferent
arietiesf ach
question
odifferentespondentsSniderman
ndPiazza1993).That iffer-
ent
uestion
ordings
ere
mployed
n
this urvey,hough,reatly
om-
plicates he nalysis
fthese ata.To dealwith he
heterogeneityhe ues-
tion
wording
xperimentsmpose
n each
dependentariable,
e include
in the hoicefunction
f each heteroskedastic
robitmodeldummy ari-
ables
whichmeasure hich
uestion
ording
ach
respondent
eceived.10
The firstependentariable s a three-versionuestionbout pen
housing:
1. (Neutral,
Dummy
1
=
0, Dummy2
=
0, 57% yes)
Suppose
there
were
community-wide
oteon a
general
ousing
ssue and that
there ere wo
possible
aws to vote n.
One aw
says
hat ome-
owners an decide or hemselves ho o
sell
their
ouses
o,
ven
if
they refer
OT to
sell
toblacks.
he
other
aw says hat ome-
owners annot
efuse
o sell to someone
ecauseof their ace or
color.Which aw wouldyouvotefor?
2.
(Property
ights,Dummy
1
=
1, Dummy
2
=
0,
35% yes)
Some
people
elieve hat omeowners
hould e free o decide or hem-
selveswhoto sell their ouse
o,
even f
they refer
ot o sell t
to blacks.For
example,
ome
peoplemight ay
t isn'tthat
hey
don'twant o sellto
blacks;
t's
ust
hat
hey
on'twant
obe told
what o do with heir wn
property.
n other
words, hey
eel hat
because t's their
roperty,
hey
hould ave
he
ight
o sellto
any-
onetheywant o.
How do
you
feel bout his? o you
think omeowners
hould e
able to decide
for
hemselves
hoto sell their
ouses
o,
even f
0Althoughhere
weremany ariations
n question ormat,e found ew tatistically
significantnteraction
ffects.
hat s,
the ffectf he ifferentuestion ordingsdenoted
here y dummies)
nly hiftedhebaseprobability,nd did
not
nteract
ith
ny
of the
scales o an appreciableegree.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
13/31
356
R. MichaelAlvarez
nd
John rehm
they refer
ot o
ell
o
blacks,
rdo you hink
omeowners
hould
not e allowed
orefuseo
sellto
someone
ecause
ftheirace
or
color?
3.
(Role of
government,
ummy
=
0,
Dammy
2
= 1,
38%
yes)
Some people
believe hat
he
government
hould
make n active
efforto
seethat lacks
an
iveanywhere
hey hoose,
ncluding
white eighborhoods.
thers elieve
hat his
s notthe
govern-
ment's usiness
nd t
should tay
ut of this.
How do you
feel?
(Is this
n area
thegovernment
hould tay
ut
of or
should he
government
ake n
active
fforto
see
that lacks an
ive
any-
wherehey an affordo-includingwhite eighborhoods?)
Both heproperty
ightsnd
role
of government
ramesntroduce
rationale
or
pposing
ights
fequal
access
tohousing
or
blacks.
The
property
ights
rame
nvokes ne
ofthe ery
traditional
alues
embed-
ded
n
themodem
acismoncept,
nd o advocates
f
this
oncept
might
expect
he
ign
n
the oefficient
obe
negative,
nd
izable
it
s).
We
do
not
ee
the amerelationship
o the ole
of government
rame,
lthough
t
introducesovernment
ntrusion,
nd
a potential
eactance
ffect.
he
dummyerm or he oleofgovernments alsonegativendsizable.The
coefficient
n
the ole
of
government
ummy
slarger
han he quivalent
for
roperty
ights,lthough
he
wo re statistically
ndistinguishable).
The second uestion
manipulated
four-versionplit
n set-asides,
specifically,
ederal ontracts
or lack
ontractors.
1.
(Dummy
=
0,
Dummy
=
0,38%
yes)
Many eople
elieve
..
2. (Dummy
=
1,Dummy
=
0,
35%
yes)
Many
lacks elieve
..
3.
(Dummy
=
1,Dummy
=
0,
39%
yes)Manypeople re de-
manding
. .
4. (Dummy
=
1,
Dummy
=
1,
38%
yes)
Many
blacks
re de-
manding
..
...
that here
hould e a
law to
make ure hat certain
umber
of
federalontracts
o
toblack ontractors.
hat
o
you
hinkbout
such
law-is it a
good
dea
or a bad idea?
The
people/blacks
ramencourages
acistso
reject
acial olicy
ased
on the equester.he believe/demandingramentroducesreactance
element. e
wouldhave
xpected
hat he igns
n Dummy ,
Dummy ,
and
he nteraction
f he wo
would
ll
be
negative.
nly
he oefficient
or
the
people/blacks
rame
s
negative,
nd t
s not
tatistically
ignificant.
The third
uestion
s a second
our-wayplit
bout
aising
axesfor
education,
lthough
e
opted
o
fold
ogetheresponses
n oneof
he
plits
(not
o use the
dummy).
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
14/31
ARE
AMERICANSAMBIVALENTTOWARDS RACIAL POLICIES?
357
1.
(Dummy
=
0, Dummy
=
0,
58% yes)Some eoplehave aid
that
axes
need
o be raised o take areofpressing ational
eeds.
Howdoyoufeel-would youbewilling ohaveyour axes aised
a little
n order o mprove
ducationnpublic chools?
2.
(Dummy 1, Dummy
=
0,
62% yes)Both he
residentnd
Congress,ncluding
oth emocrats
ndRepublicans,
ave
ecided
that
axesneed o be raised o take
areofpressing ational
eeds.
How
do youfeel-would youbewilling ohaveyour axes
aised
a little
n
order
o
mproveducation
n
public chools?
3. (Dummy
=
0, Dummy
=
1,
51
%
yes) Some eople
have aid
that axesneed obe raised otake areofpressing ational eeds.
How do
youfeel-would yoube
willing
o have
your
axes aised
a
little
n
order
o
mprove
ducational
pportunitiesfor
inorities?
4.
(Dummy 1, Dummy
=
1, 64%
yes)
Both hePresidentnd
Congress,ncluding
oth emocratsnd
Republicans,
ave
ecided
that axes
need
obe raised o take areof
pressing
ational
eeds.
How
do
youfeel-would you
be
willing
o
have
your
axes
aised
a little
norder o
mproveducational
pportunitiesfor
inorities?
Thefirstariation
Somepeople/Both
he
residentndCongress) anipu-
lates he tatus f
the dvocate. ne might
xpect
hat he
higher
tatus
f
thePresidentnd
Congress ould nduce
igher
evels f
support
or
ol-
icy.Sincewesaw no direct elevance f
the tatus f the dvocate
o the
racial aturef he
uestion,
e
neglectedo
nclude dummy.
he econd
question
ould
ncourageacistsoreject ducational
pportunities
ormi-
norities,ince he nly ignificant
ariation
n
the uestionpecifies
mi-
norities. hecoefficientorDummy wasnegative,ndsizable.The finaluestionhatwe explorenthis aper resentedwo
versions
of
questionselated
o
preferenceor ualified lacks
n
university
dmis-
sion.
1.
(Preference,ummy
=
0,31
%
yes)
Some
people ay hat
ecause
of
past
discrimination,ualified
lacks
hould
e
given reference
in
universitydmissions. thers
ay that
his
s wrong ecause t
discriminates
gainst
whites. ow
do
you
feel-are
you
n
favor
oforopposed ogiving ualifiedlacks referencenadmissiono
colleges
nduniversities?
2. (Extra
ffort,ummy
=
1,
64%
yes)
Some
people ay
that e-
causeof
past iscrimination,
n extra fforthould
e
made
omake
sure hat
ualified
lacks re
considered
or
niversity
dmission.
Others
ay
that his xtra
ffort
s
wrong
ecause t discriminates
against
whites. ow do
you
feel-are
you
n
favor f or
opposed
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
15/31
358 R. Michael
Alvarez nd John
rehm
to makingn extra
ffort
o make
ure
ualified
lacks
reconsid-
ered or dmissiono colleges
nduniversities?
The preference rame eflects
strongerolicy
han imple extra f-
fort. ence,
we
expected
he
ign
n Dummy to
be positive,nd twas
both ositive
ndstatisticallyignificant.
2.2 Specification
f the Choice Function
Other
han
hedummy ariables
n
the hoice unction,ur pecifica-
tion f this omponent
f themodelwas
driven
argely y theexisting
literature.negroup rgues hatmodem acism tructuresttitudesowards
racialpolicy.
Anotheruggests
hat he ffect fother rejudices
hould
be
included,
ncludingntiblack
ttitudes,eneralizedut-groupntipathy
(in
theform fanti-Semitism),nd authoritarianism.
third roup
on-
tends hat
he onflictf two orevalues-egalitarianism
nd
ndividual-
ism-determineacial ttitudes.
We allow
for
ix
different
cales,
s
discussed
n
the
revious
ection:
modemracism, ndividualism,
ntiblack tereotyping,
uthoritarianism,
anti-Semitism,nd galitarianism.ach s a scaleconstructednthe asis
of
confirmatory
actor
nalysis
fvarious
urvey
tems,
nd achhas
been
recoded o range rom to 1,
where indicates
he
maximumbserved
levelof the cale.
Ourprior eliefs bout
he
igns
n each of
the cales
are that
hey
hould
ll be
negative,
ith
he
xception
f
egalitarianism.
Higherevels
fmodem acism,
ndividualism,ntiblacktereotyping,
u-
thoritarianism,
nd anti-Semitism,
nd lower evels
of
egalitarianism,
should ll be associated
ith esser
upport
or
acial
olicy.
We addition-
ally
nclude he
espondent's
elf-placementn a liberal-conservative
cale
as a controlnthe hoicemodel.We imit he stimatesononblackespon-
dents.
2.3
Specification
f the Variance
Function
The
pecification
f
he ariance
unction
eflects
ur
desire o
test wo
competing
xplanations
or
ifferencescrossAmericans
n
thevariation
of
heiracial
olicy
eliefs.
s
we
argued bove,
ariation
n
policy
eliefs
may
eflect
undamental
ncertainty
bout he
olicy
hoice
nder iscus-
sion. nother ords, eoplemay implyack nformationbout he olicy
choices ndwhat
heymightmply,
ndthat
ncertainty
illbe reflected
in
thevariance
unctionf theheteroskedasticrobit
model Alvarez
nd
Franklin
994;
Franklin
991).
To
control
or he ffectfuncertainty,e nclude
n
the pecification
ofthe
variance
unctionvariable hichmeasures hatwe call chronic
information.his s a simple oliticalnformation
easure,
ased
on
the
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
16/31
AREAMERICANS
AMBIVALENTTOWARDS RACIAL POLICIES?
359
earliermeasures dvocated
y Zaller 1992). We use an additive cale
whichmeasures hetherhe espondentorrectlynew henumberfSu-
preme ourtmembersnd hemaximumumberfpresidentialerms.his
variable
s
coded o range rom to 1,where indicatesorrectnswers
to
both
actual
oliticalnformationuestions.We expect he stimated
coefficient
obe
negative, hichmplies
hat
with
ncreased
oliticalnfor-
mation he mount f
variance
n
policy
hoices iminishes.
The 1991 urvey ata lso
ncludes hree actualtemswhich re
race-
related:
he
ercentage
f
poor
who re
black,
he
ercent
rrested ho re
black,
nd the
percentagefblackmaleswhoare unemployed. e used
responsesothese hreeuestionsodevelop domain-specificacial nfor-
mationmeasure. lthough
hemechanics
f
the
difference
re
farfrom
clear,
hronicnformation
easures
egularlyutperformomain-specific
information
easures
Zaller
1992).
The rival xplanationor
variability
n
attitudesoward acialpolicy
is
ambivalencenducedy
core
beliefs
nderlyingacial ttitudes.
e set
twocriteria
n
order o
identify
mbivalence.
he firsts that dditional
informationhould ot educe,ndmay
n
fact eighten,
he
esponse
ari-
ability.hesecond riterias that esponse ariabilityhouldncreases
corebeliefs
nd
values onflict.rior
esearch
nstructs
s to
attend
o one
particularource
f
conflict,etween
galitarianism
nd ndividualism.o
the
xtenthatacial olicies chieve
galitarianismyrejectingndividual-
ism,
we
should
xpect
o ee
greater
esponse ariabilitymong
ndividuals
whoprize
oth
galitarianism
nd
ndividualism.
To test or he
ore
belief
onflict, e include he bsolute alues f
the
differences
f
evelsof
egalitarianism
nd
ndividualism.11
o
get
n
intuitionor
what his
perationalizationeans, ecall hat ach of these
core alue cales scoded o thatheminimumcore s0 and hemaximum
is
1.
Thus,
when
we use the
bsolute alueof thedifferences,hen
re-
spondent's
evel
of
egalitarianism
nd
ndividualismre
n
conflict
i.e.,
both re
highly rized
alues),
hen
we get
measure f
zero.
When he
respondent'sevelofegalitarianism
iffersrom he evel f ndividualism
then he
alues
renot
n
conflict,
nd
we
get positive
alue or he onflict
term.
n
this
articularase,
we
expect hat
f
the onflictf he wovalues
structures
esponse ariability,
hen his
an
only
ccur
when oth
alues
1It
s conceivablehat onflictmong he ther calesmight
lso lead to greatere-
sponse ariability.n
earlier ersionf his aper,nfact,ncludedll
possible ombinations
of scales
n a
similar est.None
of the esults o
follow
ary
ignificantly
ith
he esults
ofthe ully aturatedest.We preferhe implerest fconflict
etweengalitarianismnd
individualisms it s onebasedupon he tandingiterature. e can think
f no
similarly
motivatedeason oexpect onflictmong ny ther air f values.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
17/31
360 R. MichaelAlvarez nd John rehm
are
n
agreement.ence,
f
mbivalences the ppropriate
haracterization,
weexpect hat he oefficientnthismeasureobe negative
nd ignificant.
A secondwaynwhich hese calesmightnfluenceariabilitynracial
attitudess via a
kind
f measurementffect. nidermanndcolleagues
(SnidermanndHagen1985;SnidermanndPiazza 1993;Snidermannd
Tetlock 986)have rgued hat esearchershould ot
hold hat pposition
toracial olicy s an ndicationfracism,ince pposition
oracial olicy
might e groundedn nonracialbjections. his s tantamounto holding
that acists rerelativelyixedn their ppositionoracial olicy; ut hat
nonracists ight e quite ariablenattitudesbout acial olicy.We esti-
mate second etofheteroskedasticrobitmodelswhich nclude he wo
primary
acial ore
values,modem acism
nd antiblacktereotyping,
n
thevariance unction.urexpectations that eople
whoaremore acist
on these
cales
will
have ower ariance,
ence hat he
ign
f
thedirect
effectsfmodem
acism
nd antiblack
tereotyping
hould
e
negative. 2
2.4 Results of the HeteroskedasticProbit Models
We
present
urresults
rganized y
the
pecification
f
thevariance
function:he conflict pecifications in Table1, nd he linear peci-
fications
in
Table
2.
Each tablegivesestimates
ormodel oefficients,
standardrrors,
nd
X2
tests oth or he
general
it
f
themodel nd for
the
presence f heteroskedasticity.
Estimationf the hoicefunctionields esults
hich hed ight n
the urrent
ebates bout
which ore
beliefs
tructurehoices bout acial
policies.
Across
ll
ofthemodelswe
estimated,
nly
ne ofthe
ix
scales
significantlyffectedupportor acial olicy:modem acism. espite
he
weaknessesf he articular easure,nevery pecificationf he ariance
model,
or ach of thefour
ependentariables,
hemodem acism cale
is far nd way he argestoefficientn themodel.
n every pecification,
themodem
acism
ariables
statisticallyignificant
tp
<
.05 or better.
Furthermore,
he oefficientn modem
acism
s one of only
wo
of the
coefficientsn the ix varia-bleshat s always orrectlyigned negative).
We find his
o
be
somewhat
ronic,
hat he
modem acism
measure
er-
forms
est, espiteignificant
ffortsomeasure ival
ypotheses.galitari-
anism
s
also
a
significantredictor
f
racial
olicy hoice,
uch hat
hose
12We lso nclude deologyn thevariance unction,n order o control or similar
effectuchthat iberalsmaybe more ariable n their pinions bout acialpolicy han
conservatives.s the eaderwillnote elow, his ffect id notmaterialize. e explored
anotherariationfthismodelwherein e scaled deological elf-placemento reflectx-
tremismi.e., xtremeiberals nd xtremeonservativescore tthemaximum, oderates
at theminimum),ndthere as againno effectf deology n variance.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
18/31
Table
1.
Heteroskedastic
robitEstimates,
991
Race and Politics
Survey:Conflict
pecificationf
Values
Variable
Housing Set-Asides
Taxes University
Choice Model
Constant
1.235* 0.350
0.0-19
0.246
(0.337) (0.250) (0.713)
(0.258)
Dummy
-0.536* -0.138
-0.003 1.022*
(0.153) (0.100) (0.707)
(0.160)
Dummy
-0.591* 0.063
.019
(0.167) (0.089)
(.073)
Dummy X Dummy -0.036 -.003
(0.126) (.707)
Modem Racism
-1.006* -1.292* -0.866* -1.841*
(0.297) (0.284) (0.374)
(0.316)
Individualism
0.129
-0.741*
0.309 -0.608*
(0.271) (0.298)
(0.203) (0.277)
Antiblack
tereotypes -0.229 0.088 0.002 0.271
(0.240) (0.207)
(0.164) (0.220)
Authoritarianism -0.450
0.148 0.053 0.154
(0.258) (0.193) (0.156)
(0.230)
Anti-Semitism -0.428* 0.626* -0.257 0.416*
(0.232) (0.234)
(0.177)
(0.201)
Egalitarianism
0.362 0.566* 0.402
0.617*
(0.255) (0.281)
(0.218)
(0.214)
Ideology -0.040
-0.059
-0.006
-0.058
(0.025) (0.028)
(0.014) (0.023)
Variance Model
Domain
Specific nformation
0.490 -0.102 -0.165
0.199
(0.338) (0.227)
(0.445) (0.232)
Chronic
nformation
-0.715* -0.758*
-1.046*
-0.217*
(0.295) (0.215) (0.414) (0.201)
lEG
-
INI
0.012 -0.075
-0.185 0.234
(0.404) (0.403)
(0.565)
(0.320)
Ideology
0.069 0.056
-0.029 0.009
(0.044) (0.046)
(0.060) (0.031)
N
1363
1396 1359 1494
X2
182* 207* 101*
399*
Het.Test
10.5* 13.8*
9.8 2.4
Dummyl
nd
Dummy2
re
question-specificariations
n
wording,
ee themain ext
for etails. p < .05.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
19/31
Table 2. HeteroskedasticrobitEstimates, 991 Race and Politics
Survey: inear Specificationf Values
Variable Housing Set-Asides Taxes University
Choice Model
Constant 1.450*
0.313
0.340
0.665
(0.490) (0.292) (0.215)
(0.424)
Dummy -0.608* -0.185 0.142 1.826*
(0.201) (0.112) (0.085)
(0.451)
Dummy -0.663* 0.059 -0.266
(0.225) (0.095) (0.106)
Dummy
X
Dummy -0.038 -0.126
(0.143) (0.103)
Modem Racism
-1.135* -1.674* -0.905* -3.016*
(0.394) (0.533) (0.293)
(0.780)
Individualism
0.096 -0.837*
0.105 - 1.007*
(0.282) (0.347) (0.175)
(0.481)
Antiblack
tereotypes
-0.243 -0.378 -0.061 0.259
(0.250) (0.305) (0.153)
(0.389)
Authoritarianism -0.538 0.071 0.321 -0.050
(0.299) (0.225) (0.213) (0.405)
Anti-Semitism -0.488
0.728* -0.247
0.194
(0.265) (0.287) (0.169) (0.292)
Egalitarianism 0.407 1.007* 0.475*
0.943*
(0.255) (0.364) (0.216)
(0.327)
Ideology -0.043 0.000 -0.031 -0.062
(0.027) (0.000) (0.017)
(0.039)
Variance Model
Domain
Specific
nformation
0.436 -0.120 -0.040 0.390
(0.339) (0.211) (0.259)
(0.235)
Chronic nformation
-0.679* -0.769* -0.724* -0.203
(0.302) (0.209) (0.274) (0.199)
ModemRacism -0.171 -0.320 -1.273* 0.515
(0.330) (0.334) (0.368) (0.264)
Antiblack
tereotypes
0.449 1.268*
0.719
0.459
(0.429) (0.424) (0.412)
(0.328)
Ideology
0.065
0.000 0.105
0.009
(0.045) (0.000) (0.048)
(0.032)
N
1363
1396
1359 1494
X2
182* 207* 101* 399*
Het. Test 12.5* 39.6*
19.1*
9.4
Dummy and Dummy2 requestion-specificariationsnwording,ee themain ext
for e-tails.
p
<
.05.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
20/31
ARE
AMERICANSAMBIVALENTTOWARDS
RACIAL
POLICIES? 363
respondentsho re
more
galitarianremoreikely o
supporthe acial
policy.
he
coefficientsor
galitarianismn the hoice
models realmost
always ignificant,lthoughlsosubstantiallymallerhan he orrespond-
ing
coefficientsor
modem
acism.13
Turningo he emainingour
ariables, e
found hatntiblacktereo-
typing as not
consistentredictor
f oppositionoracial
olicy.Under
the
onflictpecification,he
oefficientn
antiblacktereotypingas of
the
orrect
ignonly
nce.Under
he
inear
pecification,he oefficient
on
antiblack
tereotypingas negative
n
three fthe
our
ases,
butnever
statisticallyignificant.urthermore,e
have
valuatedhe ossibilityhat
the ollinearityf antiblacktereotypingnd modem acism ndermined
the oefficientorntiblack
tereotyping.
henwe
excludemodem acism
from he
choicemodel, ntiblack
tereotyping
oes not
ppreciably
m-
prove
s an
explanatoryariable.
None
of
the
remaining
hree cales
individualism,
uthoritarianism,
and
nti-Semitism)
chieves
ny
degree
f
consistency
cross
he our e-
pendent
ariables. he coefficientsn
the
hree
ariables re oftenncor-
rectly
igned, lthough
arelytatistically
istinguishablerom
ero.14
3We
eplicatedhe nalysis
n
two
ways. irst, e ran he
nalysis ithout
he
modem
racismmeasures,
nd here ere few ubstantive
hanges.
uthoritarianismnd ntiblack
stereotyping
ecame egative,tatisticallyignificant
ariablesor ach f
he our ependent
variables nder ach
f he hree ariance
odelpecifications.o the xtent
hat hemodem
racismmeasure works
because t ncludes olicy
hoices
s
evidence f
ttitudesoward
race i.e., hemodem
acismmeasure
s endogenous),
hen nemight referhe pecification
without odem acism. he
substantiveesults ith
espect o thevariancemodelwere
unchanged.
econd,we replicatedhe nalysis sing
nly one-item
easure f modem
racism, heremodem acismsmeasured y the mount f anger bout pokesmenor
minoritiesho re lways
omplaining. ith he
ne-item easure fmodem acism nly
three hanges
ccur o the esults.he estimatedffect
fmodem acism
cross ll models
is diminishedy
roughlyalf but emainsuite
tatisticallyignificant)
hile he stimated
effectsf egalitarianismnd
deology
n
the hoicefunction
ncrease.
ubstantively,his
means hat he ffectsf
modem acism
nd galitarianismre lmostf he
amemagnitude
on
the our acial olicy hoiceswe
examine, henwe use this
measure
f
modem acism.
Little hangesn
our stimates
f
thevarianceunction,ith nly he
stimatedffectf
chronicnformation
n
thehousingmodels
in
both
pecificationsf
the
variance
unction)
diminishinglightly,lthoughhe tandardrror or his oefficientncreasesnough o
renderhe arameter
tatisticallynsignificant.ur
ubstantiveonclusionsrenot nfluenced
by
this
measure f modem acism ither.
140f heracebelief
cales we
construct,
our
modem
acism,
ntiblack
tereotypes,
authoritarianismnd
anti-Semitism)
ere
relativelytrong
measures
n
our
confirmatory
factor
nalysis.
wo
egalitarianism
nd
ndividualism)
ere
dmittedly
eaker
measures,
however,
hich
may uggest
he
resence
f
measurementrror
n
these
wo
cales.
We
do
not elieve hat hismeasurementrror
nfluencesur
esults
n
n appreciable anner.
irst,
egalitarianismerformselatively
ell
n
he
models
resented
n
Table
1
and .
As
discussed
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
21/31
364 R. MichaelAlvarez nd
John rehm
Our primarynterests in thevariance
model.We turn irst o the
heteroskedasticity
ikelihoodatio
est, rinted
t the
ottom
f
each
table.
With degrees ffreedom,he ritical alue or heheteroskedasticityike-
lihood atio est n the onflictpecification
s 9.5 (atp
<
.05). The first
three
models xceed he utoff,mplyinghat esponsesreheterogenous.
The astmodel universitydmissions)oes
not xceed he ritical alue,
andwe
cannot eject
he
ssumption
fhomogeneousesponse. he inear
specificationas
five
egrees
f
freedom,
ndthe elevantriticalalue s
11.07 atp
<
.05). Again,
we can
reject
he ssumptionf homogeneity
inthefirsthreemodels, utnot hefourth.
Next, onsiderhe stimatedarametersf he ariancemodels. etter
(chronically)
nformed
espondents
re
ess
ambivalentbout acial olicy
than ess
nformed
espondents.
he
effect
f chronicnformation
n
the
variancemodel
s
alwaysnegative,
nd statisticallyignificanttp
<
.05
or
better
or
very
ariationfthe
housing
nd set-asides
uestions.The
coefficient
n nformations
negative
or he atterwo
uestions,lthough
only lightlyarger
han he
tandard
rror.)
he
effectf
domain pecific
information,owever,s
never
istiiiguishable
rom
ero
o a
statistically
significantegree.
Thedistinctionetween ambivalence
nd uncertaintys that d-
ditionalnformation
educes
ncertainty,
utdoes notnecessarilyeduce
ambivalence.f
the
respondent
s
in
a
state
f nternalized
onflict,
hen
additional
nformationnlyheightens
he tate f conflict. ith
ttitudes
toward
bortionolicy,
we demonstratedhat etter
nformed
espondents
weremore mbivalentnder
hard
policy
hoices
where
ore
values
were
n
conflict),nd
ess
ambivalent
nder
easy policy
hoices
where
onecorevaluedominates).
The consistent
egativeign
n
chronic
nformationmplies hat ari-
ance
in
attitudesoward acial
policy
s best
characterized
y
uncer-
tainty,
ndnot
y
ambivalence.
oes
this indinguggesthat
he e-
bate
on
racial
olicymightctually
ecome
onsensual,.e.,
that
ariance
in attitudes
ight
e reduced?We hold on further
nterpretation
f
the
effectsf chronicnformation
ntil he
onclusion.
in the ext,t s always orrectlyigned nd tatisticallyignificant,lthough ith weaker
effecthanmodem acism. econd, ndividualisms significantnfourf ightmodelsand
is correctlyignedn three);n the
ther
ourmodels, he stimatedffectf ndividualism
on racial olicy hoice annot e distinguishedrom ero.Also, hese rehighly onlinear
models,o
if
measurementrror aspresent
n
onevariable
hat
ught obias he emaining
coefficientsn unpredictableanners. otice hatwhenwe use egalitarianismnd ndividu-
alism o specify alue conflictn Table
1
that he ddition f these erms, hilepoorly
measured,o notchange he stimatedesults or he hoicefunctionn an appreciable
mannercompare o Table2). This tronglyuggests
hat
measurement
rrorn these
wo
scales s
not
nfluencing
ur stimatedesults.
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
22/31
ARE AMERICANS
AMBIVALENTTOWARDS
RACIAL POLICIES? 365
Thefirstf our
pecificationsfthevariance
model Table 1) posited
thatndividualariancen
attitudesoward acial
olicywas duetoconflict
betweenorevalues fegalitarianismnd ndividualism.or nstance, e
expected hat respondent
hohad high, oughlyqual evelsof esteem
for
oth
galitarianismnd ndividualismould
avegreatermplicitari-
ance n his
or
her
policy hoices. he
conflict
model ncorporateshe
absolute
alue f he ifferencentwo cales.Our
xpectationas that he
signs n
the oefficientsor
hese conflict erms ouldbe negative.
Instead, nly wo fthe
our ermsrenegative,
ndnone restatisti-
cally ignificant.
We obtain similar attern
rom
fully
aturatedetof
conflictermsmong he ixscales.)Overall, newouldhave o conclude
that
onflictetweenorevalues oesnot
ccount or ariancen
attitudes,
with
onflict easured
n
thisway.
We
set
two riteriao
distinguish
mbivalencerom
ncertainty:
hat
the
ffect f nformations to increase
mbivalence,
ecausedissonance
in corebeliefs hould e irresolvable. ith
espect o
the
fourmeasures
of
ttitudesowardacial
olicy
nthis
ssay,wefound
xactly
he
pposite:
informationeducesndividual
ariance,
nd here as no evidence f
on-
flictn corevalues s a source fvariance.
An
alternative
onceptionf
he
ariancenattitudes
oward
acial
ol-
icy s that he fit
ofattitudesoward acial
olicy
s a
function
f
racial
attitudesaries
y
he evel f
racial ttitudes.
n
other ords,acistsmight
have ittle
ariancen
opposition
o
racialpolicy, utnonracists ight e
quitevariable
n support
or
acialpolicy, ince
ppositiono that olicy
mighte
grounded
n
nonracialoundations.
n
order
o evaluate
his
on-
ception,
e
nclude irect
ffects
f
modernacism nd ntiblack
tereotyp-
ing the
wo
trongestredictorsf
oppositionoracial olicy) n the ari-ancemodelTable2). Ifthis lternativeonceptionolds,wewould xpect
that he
igns n
the
varianceoefficientsormodern
acism nd ntiblack
stereotyping
hould
enegative,.e., hat igher
evels feach ead to ess
variance.
Overall,
his
rgument
or he
ource
f
ndividual
ariation
ails.
Of
the
ight
oefficientsn
the
irect
ffects,
hree re
negative,
nd
only
ne
of
these o a
statisticallyignificant
egree.
oneof themeasures f anti-
black
tereotyping
re
negative,
nd ne s
statisticallyignificant.
ut here
is some ndicationhathe rgumentightoldformodernacism. hree
of
the our
erms
n
modernacism re
negative,
ne
to
a
statisticallyig-
nificant
egree.
he
fourth odel s
also
not
heterogeneous,y
the
ikeli-
hoodratio est.
An
implication
f
this ranklyrovisionalesults that t
is
touchy
o
nfer rom
pposition
r
upport
or acial
olicy
ack oracial
attitudes. ith t
east
wo f he our
olicy
hoices
set-asides
nd
axes),
thosewhowere owest n themodern
acism
cale
were f
widest ariance
This content downloaded from 200.130.19.155 on Sun, 17 Aug 2014 02:45:15 UTCAll use subject to JSTOR Terms and Conditions
http://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsphttp://www.jstor.org/page/info/about/policies/terms.jsp
-
8/18/2019 ALVAREZ, M. & BREHM, J. Are Americans Ambivalent Towards Racial Policies
23/31
366
R.
Michael
Alvarez and JohnBrehm
in
their
hoices,meaning
hat
hose
who
were, y
this
measure,
ot acist
mighttill e in oppositiono policy.
2.5
Magnitude fEstimated
ffects
The estimatedffectsf racial
values nd differentpecifications
f
heterogeneousesponses
llowonly or ough ests f
the
hypotheses
e
havepresentedarlier. ust s withmore ommon inary hoice
models,
the eteroskedastic
robit arameters
re
difficult
o
nterpret
n their wn.
Here
we
turn o another ayof presenting
he
stimated
ffectsf racial
values nd
deology ponpolicy
hoice;by examininghe first
iffer-
ences for ach ofthe mportantight-handide variables.
The first ifference ethodologys relativelyimple King
1989).
Herewe set
achright-hand
idevariable o the ample
mean
value;
with
these
alues nd
he stimatedoefficients,e candeterminehe robabil-
ity
hat
hypotheticalespondent
ould
ive particular
nswero
survey
question. hen, or particularndependent
ariable
say
modern acism
inthehousingmodel),we ncrease
he alue f hat ariable y wo ample
standard
eviations,
nd hen
alculate
he
robability
h