Spirals of Trust? The Effect of Descriptive Representation on the Relationship between
Citizens and Their Government
Author(s): Claudine Gay
Source: American Journal of Political Science , Oct., 2002, Vol. 46, No. 4 (Oct., 2002), pp.
717-732
Published by: Midwest Political Science Association
Stable URL: https://www.jstor.org/stable/3088429
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.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Midwest Political Science Association is collaborating with JSTOR to digitize, preserve and
extend access to American Journal of Political Science
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
Spirals of Trust? The Effect of Descriptive
Representation on the Relationship
Between Citizens and Their Government
Claudine Gay Stanford University
Research on black representation in
Congress emphasizes the material
merits of black congressional representation. Although scholarly
gains associated with black office
consensus remains elusive, empirical research has shed light on the
holding over the intangible goods
For more than a decade, social scientists have debated the substantive
links among race, legislative behavior, and policy outcomes favorable to mi-
associated with citizens' ability to
nority communities. We know comparatively little, however, about how
identify racially with their legislators.
constituents, both black and white, value black representation. What, if any,
This article considers the effect of
significance do constituents attribute to the race of their representatives
descriptive representation on the
and to the growing racial diversity of a legislative body traditionally domi-
relationships among citizens, legisla?
nated by whites? Prior research on minority political leadership at the local
tors, and the Congress. With data
level suggests that descriptive representation can favorably affect attitudes
from the 1980-1998 ANES, I show
towards public officials and institutions, with broad implications for the
that whites and blacks differ in the
political dynamics within American cities. As early as 1968, the Kerner
value they place on descriptive repre?
Commission identified the lack of black representation in city government
sentation. White constituents more
as a force exacerbating the political alienation and distrust that contributed
favorably assess and are more likely
to the urban unrest of the 1960s (National Advisory Commission on Civil
to contact representatives with whom Disorders 1968).
they racially identify. This tendency is
Here I address whether black representation in Congress affects citi-
partially explained by racial differ-
zens' political orientations. In particular, I ask: Does a constituent's ability
ences in legislators' ideological pro-
to identify racially with her member of Congress (MC) affect her percep-
files, but also reflects extrapolicy and tions of that legislator and of Congress as an institution? To what extent
explicit racial concems. Black con?
does a constituent's response derive from nonracial considerations such as
stituents place less significance on
shared interests and policy priorities? Drawing on 18 years of survey data
descriptive representation, although
from the American National Election Study (ANES), I show that white and
they are more likely to contact black
black constituents differ in the value they place on descriptive representa?
representatives. Although the relation? tion, in general, and black representation, in particular. White constituents
ships between legislators and their
more favorably assess and are more likely to contact representatives with
constituents are influenced by race,
whom they racially identify. This preference for white legislators is partially
perceptions of Congress as an institution are not affected by constituents'
ability to identify racially with their
representatives.
Claudine Gay is Assistant Professor of Political Science, Stanford University, Encina Hall
(West), Room 466, Stanford, CA 94305-6044 (cgay@stanford.edu).
The author thanks Chris Afendulis, Jack Citrin, Simon Jackman, Jennifer Lawless, Kira
Sanbonmatsu, Steve Voss, participants in the American Empirical Seminar at SIQSS and
the anonymous reviewers for comments and assistance on earlier drafts. This research
also benefited greatly from the comments of Fred Harris, Cathy Cohen, and the partici?
pants in the 2001 conference "Theorizing Black Communities: New Frontiers in the
Study of African American Politics" at the University of Rochester.
American Journal of Political Science, Vol. 46, No. 4, October 2002, Pp. 717-733
?2002 by the Midwest Political Science Association ISSN 0092-5853
717
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
718
CLAUDINE
explained
profiles,
gree,
GAY
by
racial
tributes play a critical
role in securing the trust and elec-diff
but
also
reflec
toral support of their
constituents. Fenno reports
on the
explicitly
racial
lengths to which members go to stress
their identification
c
constituents
place
less
with constituents
(1978, chapter 3). Recognizing
that
tion
per
black
or
black
though
tives.
tives
se.
More
narro
trust is built,
in part, on public gestures that signal
"I am
constituents
feel
one of you," a representative will adapt her language (e.g.,
legislators
"We in California") and even her attire (e.g., khakis at awho
they
are
Silicon Valley campaign
stop) to establish amore
link between
While
the
relatio
herself and her
constituents. The objective:
to win the
and
their
consti
general
accolade "she's a good woman,"
what Fenno de-
fines as the essence
of trust.
perceptions
of
ability
to
closing,
I
empirical
for
Congres
identify
Bianco (1994) attributes the premium constituents
raci
discuss
the
place
on identification to their desire for representatives
who share their values and interests. Personal attributes
research
on
the
more
i
m
need
to
widen
are assumed to signal policy
concerns; constituents,
bethan
study
o
havingthe
as policy-minded actors,
reason from "I am one
of you" to, as Fenno puts it, "I think the way you do and I
care about the same things you do" (1978,58). This belief
Race as a Basis for Political Trust
about common interests may determine the level of trust
established in the member-constituent relationship.
To secure favorable policy outcomes, constituents
Empirical research on black representation in Congress may give special weight to racial identification. Race as a
defines the political significance of race primarily in cue is a low-cost alternative to more detailed informa?
terms of legislators' behavior, focusing on policy respon? tion about a legislator's stands and, importantly, is virtu-
siveness and constituency service (Lublin 1997; Canon ally impossible to manipulate. Just as constituents might
1999). This research overlooks the effect of descriptive be expected to ignore attributes (e.g., hair color) that do
representation on constituents5 perceptions of their gov? not impart useful information about a representative's
ernment. Scholars have not ignored altogether the poten? policy concerns, they also may discount actions that are
tial symbolic value of descriptive representation but they the products of strategic political calculation (Bianco
have presented these "intangible goods55 as matters of in? 1994, Chapter 3; Fenno 1978, Chapter 3). A constituent's
dividual psychology without clear political implications.
ability to identify racially with her member of Congress
Yet, as Mansbridge argues, descriptive representation can may assume special significance because race is a "signal
"forge bonds of trust55 between legislator and constituent, that speaks louder than words, or at least with more
enhancing the "feeling of inclusion,55 which, in turn, meaning" (Bianco 1994,153). In research on black maymakes "the polity democratically more legitimate in ors, scholars have linked the response to black office
one5s eyes55 (1999, 641 and 651). In a representative de-
holding in part to assumptions about likely respon-
mocracy where perceptions of legitimacy are critical to
siveness, to the tendency to define black elected officials
systemic stability, the "spiral of trust55 set in motion by as the "specialized representatives of black people"
descriptive representation may have real implications for (Conyers and Wallace 1976,115).
political life (Williams 1998,172).
Studies of minority office holding at the local level
Constituents may also place a premium on racial
identification because of cultural forces or other extra-
underscore the potential significance of a citizen's ability policy goals. Favorable assessment of elected officials and
to identify racially with prominent political actors. Black governmental institutions may turn on factors such as
racial group consciousness or racial prejudice, which
among African Americans, contributing to more favor? have been shown to influence political attitudes among
able impressions of city government as a whole; and, con- whites and blacks (Dawson 1994; Kinder and Sanders
control of the mayor's office can enhance political trust
versely, the election ofa black mayor may precipitate a de- 1996; Reeves 1997). Additionally, constituents may value
cline in trust among whites, although the empirical
findings remain inconclusive (Abney and Hutcheson
1981; Howell and Fagan 1988; Bobo and Gilliam 1990;
Gilliam 1996).
Members of Congress behave as if their personal at-
more than policy responsiveness but also "accessibility"
and the "assurance that two-way communication is pos-
sible" (Fenno 1978, 239-40). Descriptive representa?
tion?perhaps by circumventing the social and psychological barriers that impair interracial communication,
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
THE EFFECT OF DESCRIPTIVE REPRESENTATION J19
dent's representative.
For r
or perhaps as a result of legislators5
personal outreach
white"natural55
Democrats,
the med
groups perceived to be their
constituen
black
Democrats,
six to
years.
may address such concerns.
Thus,
the ability
identi
For each
of
Con
racially with an elected official
may member
not trump
the
de
Americans
Democratic
for favorable policy outcomes,
but itfor
may
contribute A
sure
of ideological
levels of trust that exceed
expectations
aboutposition.
the tang
justed
using the
methodolo
policy gains associated with
descriptive
representation
Levitt, and
Snyder (1999)
In sum, the shared experience
"imperfectly
cap
across time. can
The form
median
tured55 by descriptive representation
thein
b
for respondents
represente
for greater trust in public
officials and
instituti
black Democrats,
(Mansbridge 1999, 641). 68.9;
This suggests
that the 90.4.
polit
Scores of in
items
on the
ANE
significance of black representation
Congress
may
entation
toward public
tend beyond the process of
representation
itself. offici
Assum
tions about the likelihoodconsider
of common
as we
the interests,
member-const
constituents
feel about concerns,
the elected officials who serve
extrapolicy and explicitly
racial
should l
them and the quality of the with
representation
they provide?
constituents to favor representatives
whom
they
Of particular
interest is the
degree to which
members
cially identify. In addition
to feeling
better
represente
in securing
the trust of the people they
reprethese legislators, I would succeed
expect
descriptively
represen
sent?an achievement
with implications for
subsequent
constituents, more convinced
of Congress5s
legitimacy,
electoral
support and, as Bianco (1994) argues, for the
more favorably assess the
institution.
amount of discretion legislators can exercise in their vot?
ing behavior. Although Bianco (1994) focuses on discrete
trust decisions involving high-salience policy proposals,
Data, Measures,
and
arguing that
trust is "conditional and contextual," he
Methodological
concedes Issues
that some legislators may be trusted more often
than others because they "hold a mix of actions or at?
tributes that
common
interestfrom
across a wide range
To test these propositions,
I signals
use
data
a pool
of policy
areas" (1994,152).2
This more generalized
sensebienn
sample of the ANES. The
sample
includes
ten
of trust is of interest
here.
election surveys conducted
between
1980 and 1
Four survey items not
assess levels
of trust:
(1) If you
Pooling multiple surveys increases
only
the
numb
a problem
that [your
current MC]of
could responde
do someof black respondents buthad
also
the
number
thing about, do you think s/he would
be very helpful,
represented by African-American
House
memb
somewhat
helpful, not very helpful,
or does it depend?
Nineteen hundred (1900)
respondents
self-identifie
(2) Do youinformation
happen to remember anything
special that
black; 13,028 as white. Using
available
in
[your
current MC] has done for
his/her785
district or
for the
nual political almanacs, I
identified
the
respond
of his/her district while
s/he has
been in Con?
(483 blacks, 302 whites)people
represented
by
black
mem
Using theinterview
(Feeling) thermometer,
how would
of Congress at the timegress?
of(3) the
(Novembe
you58
rate of
[your these
current MC]?respondents
(4) In general, do you ap- lived
the election year). All but
prove or disapprove
of the way [your current MC] has
majority-minority congressional
districts.
been handling his/her whose
job?
For each member of Congress
constituent
Additionally, respondents
were asked,
you (or
responded in the ANES, I assembled
data
on"Have
party
aff
your family
living here) ever contacted [your
ation and tenure. A totalanyone
of in
8941
respondents,
includi
all ofthe respondents with black representatives, were
lrThe
racial difference
in the median inflation-adjusted
ADA score
resented by Democratic
House
members.
Because
I
varies by region and is greatest in the South. The median ADA
interested primarily in the effect of a representative5s
score for respondents represented by Southern white Democrats is
on constituents5 attitudes,
analyze
only
the
respond
44.2; byISouthern
black Democrats,
92.6. Outside
the South,
the
medianTwo
scores for black
white and black
Democrats are 78.8 and 90.4, Gar
represented by Democrats.
Republicans,
Franks of Connecticut respectively.
(1991-1997) and J.C. Watts
2As Bianco
(1994) defines it,
trustthe
exists when
a constituent evaluOklahoma (1995-present)
served
in
House
betw
ates, or is prepared to evaluate, a member of Congress favorably
1980 and 1998; none ofthe ANES respondents are draw
regardless of how that member votes on a given piece of legisla?
from either of these twotion.
districts.
Tenure
is
measured
In the absence of trust,
a member receives
favorable
evalua-
tions only ifyears
she acts as her constituents
think best.
the number of consecutive
served
by the resp
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
J20
CLAUDINE
current
allows
GAY
MC]
any
these demographic characteristicsor
and political predispo-
me
to
test
sitions vary systematically
with the race
of the represen-
M
scriptive
tative andrepresent
are likely to affect citizens' political attitudes,
cation
between
they are incorporated in the subsequent analysis as con?legi
trol variables.
Secondly,
I
conside
institution.
Does
Finally, the analysis of
public opinion often is hamwith her MC contribute to a more favorable assessment
pered by the pervasive problem of incomplete data due to
of the institution? Such public approval is critical to the
item nonresponse in surveys. This research is no excep-
perceived legitimacy ofthe institution, its legislative pro- tion; although per-item rates of missingness are low, only
cess, and its policy outputs. Research on black mayors 50 percent of white and 48 percent of black ANES respon?
suggests that descriptive representation and trust in dents provided answers on all of the survey items of inter?
governmental institutions are closely linked. However, est. Political scientists typically have relied on listwise descholars have also shown that public attitudes toward letion when coping with the problem of missing data on
Congress are only weakly correlated with support for both explanatory and dependent variables. However, statcongressional incumbents (Cain, Ferejohn, and Fiorina isticians and methodologists have demonstrated repeat1987; Hibbing and Theiss-Moore 1995; Durr, Gilmour, edly that multiple imputation outperforms listwise deleand Wolbrecht 1997). As Durr, Gilmour, and Wolbrecht tion as a solution to the problem of item nonresponse
observe, "members of Congress manage to remain popu-
(Schafer and Olsen 1998; King et al. 2001). Multiple im?
lar and win reelection despite the low opinion the public
putation, which assumes that information in the observed
has of Congress as a whole55 (1997, 177). Cain, Ferejohn,
data provides indirect evidence about the likely values of
and Fiorina speculate that this disparity in judgments the unobserved data, can correct for the ineffrciency and
may stem from the disparity in popular expectations bias that result from listwise deletion. The method inabout legislators and the legislature: "The institution is volves the imputation of m possible values for each miss?
held up against programmatic standards more so than its
ing item in the data matrix and the creation of m com-
members, who instead are judged according to their per-
pleted datasets, each of which can then be analyzed using
sonal characteristics, accessibility, and solicitude for the
concerns of constituents55 (1987,202).
standard complete-data statistical methods. The statistical
results are combined across datasets using formulas that
One ANES item measures opinion in this area: Do formally incorporate missing-data uncertainty.
you approve of the way the U.S. Congress is handling
After verifying that my data met the conditions un?
its job?
der which multiple imputation would be preferable to
The ANES includes data on respondent's demolistwise deletion?namely, it is possible to predict that a
graphic characteristics and political predispositions, incell in the data matrix is missing?I used King et al.'s
cluding gender, age, educational attainment, income,
(2001) EMis algorithm (and accompanying software,
southern residence, party affiliation, and ideology. Prior
AMELIA) to implement multiple imputation.3 Each imresearch has documented the influence of such factors on
congressional approval and electoral support for Demo?
cratic candidates. Furthermore, an analysis of the ANES
revealed that the respondents represented by black
Democrats differ systematically from those represented
by white Democrats on some of these factors. The black
respondents with black MCs are, on average, older (45
versus 42 years), less likely to live in the South (36 per?
cent versus 67 percent), and more likely to self-identify as
Democrats (89 percent versus 82 percent) and liberals
(41 percent versus 33 percent) than are respondents rep?
resented by white Democrats. The white respondents
with black MCs are younger (44 versus 45 years) and
more likely to self-identify as Democrats (58 percent ver?
sus 51 percent) and liberals (33 percent versus 27 per?
cent). Mean educational attainment and income are
slightly higher among the black and white respondents
represented by black members of Congress. Because
3King et al. (2001) identify four conditions, a/Z of which must hold,
in order for listwise deletion to be preferable to multiple imputation: (1) The analysis model is conditional on X (i.e., explanatory
variables containing missing data), and the functional form is
known to be correctly specified; (2) There is nonignorable (NI)
missingness (i.e., the probability that a cell is missing depends on
the unobserved value of the missing response) in X, and there are
no other variables available that could be used in the imputation
model to predict X; (3) Missingness in Xis not a function of Y(i.e.,
the dependent variable containing missing data) and unobserved
omitted variables that affect Ydo not exist; (4) The number of observations left after listwise deletion should be so large that the efficiency loss from listwise deletion does not counterbalance the biases induced by the other conditions.
Although my analysis model is conditional on X, my data did
not meet the other conditions necessary in order to justify the use
of listwise deletion. In a series of regressions, I tested to see
whether missingness in each of the survey items of interest could
be predicted. Among the patterns I observed: Missingness on ideology, the explanatory variable with the highest rate of missing
data (16.85 percent), could be predicted by partisan identification,
gender, age, education, income, and level of political knowledge.
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
a
c
THE EFFECT OF DESCRIPTIVE REPRESENTATION J21
putation model uses a broad
array
of variables
(including
representation
consistently
is associated with higher
apall of the variables to be used
in the
subsequent
statistical
proval ratings,
greater familiarity
with a legislator's record
analysis) to predict the pattern
of missingness
of service to the district,
and more confidence in the qual-and to
generate five complete datasets.4
Each service.
dataset
was
ity of a legislator's constituency
For example, then
less
analyzed using the appropriate
statistical
model.
The re?
than one-third
of self-identified black
Democrats expect
sults listed in the proceeding
tables
the
re?
that a white
Democraticare
legislator
would becombined
"very helpful"
sults across datasets, calculated
using
the
formulas
in the resolution
ofa problem,
compared
to 40 percent of devel-
oped in Rubin (1987). The
goodness
of
fitby measures
the African
American constituents
represented
black
included in the tables are computed
averaging
across
Democrats. For whites by
and African
Americans, descriptive
the datasets.
representation increases the number of constituents who
can recall any particulars about their member's record
from one-in-six to more than one-in-four. Furthermore,
whites and African Americans are 48 to 79 percent more
Descriptive Representation
andwhothe
likely, respectively, to contact legislators
share their
racial
group
identification.
Member-Constituent Relationship
Thus, a simple crosstabulation suggests that descrip?
tive representation
favorably affects
the member-conOn average, constituents more
favorably
assess
represen?
stituent
relationship.
Table
1
also
reveals
the effect
tatives with whom they are able to identify that
racially.
Table
may beof
more pronounced
among whites than
among
1 reports mean evaluations
Democratic
legislators,
blacks; with
the exception of the "recall"
contacting
grouped by the race and party
affiliation
ofandrespondents.
measures, theadifferences
between
the evaluations of on the
While partisanship clearly exerts
strong
influence
white and black legislators
are larger for
white constitu?
member-constituent relationship,
with
self-identified
ents than they are for African
Americans.
In fact, AfricanDemocrats feeling better represented
by
Democratic
leg?
Americans
more
favorably
assess
white
Democratic
legis?
islators than do self-identified Republicans, descriptive
lators than white constituents assess black Democratic
legislators. As a result, attitudes towards black Demo?
Missingness on the dependent measures
interest
cratic legislators of
appear
more sharply typically
polarized along ra?was
greatest among ideological conservatives,
Republicans,
the young,
cial lines than do attitudes
towards white Democrats.
those with low incomes, and those represented by members of
What accounts
for years,
the fact thatand
descriptively-repreCongress who are black, have served
many
have liberal
voting records. (These "predictions"
sented constituents
do notimply
seem more satisfied
a causal
with their
relationrepre?
ship.) And while it is possible sentatives
that there
is NI who
missingness
than do constituents
can not identify ra?in at
least one of the explanatory variables, income (9.86 percent miss?
cially with their MC? Bianco (1994) attributes voters'
ing), I identified other variables (e.g., gender, age, education, emdesire to be represented
by someone
like them
to a "raployment status, party identification)
available
in the
ANES
that
can be used to predict these values.
(Thus,
the
second
condition
tional" calculus
aimed
at securing
favorable
policy out- is
not met.) I used the results from these series of regressions to emcomes. The relationship in Table 1 between shared party
affiliation
and favorable evaluations is consistent with
(MCAR) and to construct the imputation models.
such an explanation. That constituents continue to favor
pirically verify that the data were not missing completely at random
4 The EMis approach to multiple imputation was implemented uslegislators ofData
the same
race even after taking
party
into
ing AMELIA: A Program for Missing
(Honaker
et al.
1999).
It
was not feasible to construct a single
model
that included
allallow
of the
account suggests
that partisanship
alone may not
relevant variables (both analysis variables and variables to be used
for sharp inferences about a member's policy concerns.
only in the imputation process) because of memory constraints;
As
noted earlier, there are substantial racial differences in
AMELIA repeatedly crashed when the data matrix exceeded 7435
observations x 25 variables. Instead,
I scores
constructed
two separate
the mean ADA
of Democratic legislators,
with
models, one that included all variables relevant to the member-con-
black Democrats considerably more liberal than white
stituent dependent measures and one that included all variables rel?
Democrats.5
Some researchers
have was
argued significant
that black
evant to the congressional approval
measures.
There
overlap between these models. The multiple imputation models
5 Constituents
differ on whether
they perceive racial differences in
were run separately for white and
for black
respondents.
Each imputation process generated
complete
datasets
(20
the ideologicalfive
profiles of
Democratic legislators.
In seven of the
datasets total). Schafer and Olsen
(1998)
note that
multiple
impu?
ten ANES
surveys, respondents
were asked
to place their incumtation "can be highly efficient bents
even
small
In many
on a for
seven-point
ideology values
scale, rangingof
fromm.
1 (extremely
applications, just 3-5 imputations
are
sufficient
toWhite
obtain
excellent
liberal)
to 7 (extremely
conservative).
constituents
place
results" (Schafer and Olsen 1998,
548).
Using
Rubin's
white Southern
Democrats
1.09 points
higher (more(1987)
conservative) formula for calculating the efficiency
of scale
estimates
based
m outimpu?
on the ideology
than they do black
Southern on
Democrats;
tations, I calculated that with
m
= 5 white
I achieve
about
91 percent
side
the South,
Democrats score.
70 points higher
on the
efficiency.
scale than do black Democratic legislators. Each of these racial
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
CLAUDINE GAY
722
Table 1 Attitudes and Behavior Toward Democratic Representatives
Note: Table entries are the combined descriptive statistics across five multiply imputed datasets. Too few respondents sel
licans to be included in the table. The values listed in column one ("MC as Resource") capture the percentage of respo
bent would by "very helpful." "L" indicates a statistically significant difference (p<.05) between the evaluations of W
Black Democratic Legislators by the respondent group. "R" indicates a statistically significant (Racial) difference (p<.
Black Democratic legislators by white Democratic constituents and black Democratic constituents. "P" indicates a statistica
ference (p<.05) between the evaluations of Black Democratic legislators by white Democratic constituents and white R
scripts identify only those relationships that are statistically significant in each of five imputed datasets.
politicians' crossover appeal is handicapped primarily
by
of Congress,
each model includes interaction terms be?
the perception that they are too liberal (Sigelman
et al. ADA score and the respondent's level
tween the member's
1995). Gilliam (1996), in his study of Los Angeles,
found
of political
knowledge. The likelihood that a constituent
that white attitudes towards a black mayor will
were
heavily
take
into account a member's voting record and
influenced by perceptions of ideological compatibility,
bring it to beara in her assessment of that member is con?
result consistent with Sigelman et al.'s (1995)
hypothesis.
ditional
on how politically well informed that constituent
To determine whether the relationships observed
in Bartels 1996). More knowledgeable con?
is (Zaller 1992;
Table 1 reflect the rational calculations of policy-minded
stituents are more likely to be aware of their members'
constituents, I estimated a series of equations
that and, if they are motivated by policy
legislative records
included measures of ideological placement
and particoncerns,
more likely to link legislative behavior and atti?
models also include interaction terms between
sanship for members and constituents, as tudes.
well The
as other
relevant member characteristics and individual demo-
the race of the member and the ideological self-place-
graphic variables. In addition to the race of the member ment of the constituent. If assessments of black Demo?
crats were affected by the perception that they are more
liberal than their white counterparts, I would expect
differences is statistically significant. Furthermore, when regress-
ing the perceived ideology measure on both the race and actualmore liberal respondents to judge black representatives
ADA score of the incumbent legislator, with separate models for more favorably than do more conservative respondents.
each region (i.e., South-Non South), I find that black Democrats
The models also take into account the number of
are perceived as more liberal than whites (1.03 points in the South,
.54 points elsewhere) even after controlling for their actual ideo?years a member has served in the House. Longer tenure
logical profiles. By contrast, among black constituents, there is noincreases the probability that shared values and interests
statistically significant difference in the ideological placement ofbetween members and constituents will be demonstrated
white and black Democratic legislators, regardless of region.
Perceived ideology is not included in the subsequent regres-and perceived, providing a basis for more favorable
sion analysis. The perceived ideology item was not asked in 1984, evaluations. Over time, legislators may be able to build a
1988, or 1992. In the seven ANES surveys that included the item,reputation for trust, even if they did not come into office
the nonresponse rate was high: only 51.6 percent of the respon?
with that reputation ready made. As noted earlier, con?
dents represented by Democratic legislators answered this question. (By comparison, the response rates on the six dependent stituents represented by black Democrats are repre?
measures ranged from 60.3 percent to 80.9 percent.) Rather than sented, on average, by a legislator who has served only six
lose thousands of cases from 1984,1988, and 1992,1 chose an ide?
years, compared to the eight year average for the con?
ology measure for which I have complete information. The ADA
stituents of white Democrats. The assessment of black
scores have the additional advantage of being both objective and
clearly exogenous from the dependent variables.
Democrats may be a product, in part, of their relatively
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
THE EFFECT OF DESCRIPTIVE REPRESENTATION 723
short tenure and, as a result,
their
limited
to
This
result
is at opportunity
odds with earl
win over skeptics through their
performance
in office.
can-American
politicians
are
Finally, each model controls
for several
individualthe perception
that
they are
conflict
were toassessments
blame for whi
level characteristics that might
influence
of
receptivity
to Black
Democrati
Democratic legislators, including
gender,
education,
age,
income, and southern residence.
An
interaction
term be?
pect to
observe
ideological
dif
tween southern residence and
the
race of the
member
of
their
assessments
of
these legi
This is not
to research
say that has
ide
Congress is also included. Because
prior
shown that issues of race whites'
are particularly
assessments
salient
of Democ
in th
and black Democrats
particu
South, I would expect the relationship
between in
the
race o
job approval,
black D
a member of Congress and overall
her evaluation
by constituents
to be more pronounced in
clearly
that region
hurt by
than
their
elsewhere
comparat
one example,
legislators
with h
Table 2 presents the coefficients
from the
equations.
lower
feeling thermometer
Even controlling for the
ideological
differences rat
crease in the MC's ADA
score is associated withas
a halfamong white and black Democratic
legislators,
well as
point decline differences
on the thermometer scale.
For politically
for ideological and demographic
among
con?
well-informed
whites, ain
member's
ideological
stituents, the relationships
depicted
Table
2 profile
for has
white
little
effect on whether in
they are
likely initial
to recall her efforts
are similar to the patterns
observed
the
crosson
behalf
of
the
district
or
consider
her
a
resource
in
the whit
tabulation. While partisan affiliation affects how
resolution of a constituent's
problem.7 But for less
wellconstituents evaluate Democratic
legislators,
on
every
constituentsless
(i.e., those
with medium to low
measure they assess black informed
Democrats
favorably
than
scores
on themodel
political knowledge
measure), the
higher the
they do similar white MCs.
The
predicts
that
ADA
scores typical of black Democratic
MCs average
result in
average white Democratic
constituent,
where
i
less favorable
assessments.of the independen
defined by the mean value
on each
the member-constituent
relationship
varies
variables, represented by aAlthough
black
Democratic
MC
is 15.
with the incumbent
legislator's
length of
service, black
percentage points (27.6 percent
versus
43.1
percent)
les
relatively
short tenure27.7
cannot fully
account
likely to consider her MC Democrats'
"very
helpful"
percentage
for
the
less
favorable
evaluations
of
white
constituents.
In recal
points (51.3 percent versus 79 percent) less likely to
fact, percentage
white constituents, though points
less likely to recall
the ef? per
her MCs efforts, and 23.6
(31.9
of relatively
new legislators,
are more likely to of
apcent versus 55.6 percent)forts
less
likely
to approve
her
of their
job performance
and to consider
them po?
MCs job performance thanprove
the
same
average
constituent
tential
resources.
I
also
tested
a
model
that
included
an feelin
represented by a white Democratic MC.6 On the
term between theassign
race of the representative
thermometer scale, white interaction
constituents
black mem?
and the
length ofthose
service. The of
coefficient
on the interac?
bers of Congress scores that
trail
white
Demo?
tion term was statistically insignificant: there is no evicrats by 11 points.
that the effect
race varies based on the
legislator's
Unlike past research, Idence
find
no ofevidence
that
the
length of
service?or,
another way, that experience
effect of the legislator's race
on
theputassessment
of in
Demo
office
assumes added importanceideological
when white constitu?selfcratic members varies with
constituent's
ents are not represented
able to identify racially with their
identification. Liberal whites
byMC.
African
Whereas whites'
attitudes vary
with
American MCs are no more
likely
tosystematically
assess
these
race,
the attitudes
of African Americans
are seemingly
legislators favorably than
are
moderate
or conservativ
unaffected by their listed
ability to identify
with their
whites. In none of the equations
inracially
Table
2 (for
representatives.
No significant differences
whites) are the interactionelected
effects
between
black exist
repre?
among black constituents
in their assessments
of Demo?
sentation and individual ideology
statistically
significant.
cratic legislators based on their ability to identify racially
with the member of Congress. On every measure, the
6 Mean values used in the calculation ofthe predicted probabilities
coefficient on black representation is statistically in?
Tenure =10.8 years; ADA Score = 66.4; Age = 45 years; Education =
significant.
3.75 (between high school [3]
and some post-graduate [4]); In
come = 2.89 (between 17 and 33 percentile [2] and 34-67 percen
tile [3]). For the categorical variables, the predicted probabilitie
assume the constituent is a non-Southern
woman,
self-identified
7 The total effect of legislator ideology
for well-informed
whites is
calculated
as the sum of the coefficient on
"ADA Score"
and the co?
as a Democrat and ideological
conservative,
and
with
a medium
level of political knowledge.
efficient on the interaction term "ADA Score x Hi Info."
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
CLAUDINE GAY
724
>^ ^ ^ ^ ^ ^ LO h- ^
CDCMLOh-CD^CMCDO
COLOCOCXDCOOLOh-LO
CO
C0 CD LO
00CDCDCMCM00O00CM
T-^tOOJ-r-CM-T-T-
g
1o
I
LO
LO
CD
CO CO 00 OJ t- o
O O O O) h- "3"
O O O ^ LO CD
NCMCO^tCDLONCDCD
CDCDLO^CDOCD^-iCJ^-COCMOJOOt-CO
^ ^ ^ ^ ^ ^
O O O CD CD C\J
O O O h- ^ LO
h-CM^h-OOO^^
^^OO^COt-CD^O
C0Wt-OWOt-O,J
LO iLO tco o
i? C0 C0 "3" CM LO
O O O h- t- 00
O O O C0 "3" -*fr
OO^CMCM-i-OOOOO
LO CM
h- O
CD O
O CD h- O CM LO
O i? O LO i? CM
O O O ^t C0 i-
CD C\J
00 t^" O
2
r r r
?
O
LO
^ OO _ - 00 ? "" ? ^
CO
1
CD
' C\J
C\J C0 C\J C\J
C\J
LO
(0
u
0>
cd a>
o CM
LO CD
' \' \' ' ' '
^-OOOCDCDLOOOLO
CD C0
O CD
LO CD
Q_
E
OC
<D
=f
CMCMOh-T?CMCOCMhC0L0C0CT)h?~OC0C0i?
i-i-i-OOOOO---
? -2
</> o
O C0 CD
CD CD LO
CM CM CM
cc ?-
8?
^ LO
C0
CD ^ CM ^ ^ ^
i- i- i- LO CM CD
O O O O) <J) LO
h- C0
CD O
CM
C0
^
CM O
CD O
g
(0
cd
*3
as-
CO
+??
c
<D
CO
?
o
E
<D
Q
CD^CM^^^^^CM
II
II
C0 CM i? Is- LO C0
O O O ^ ^ CM
O O O LO ^" ^i-
lOt- OOOOOOOh-h-O
CMi-i-O00OCMC0-?-
LO 1? 1? LO C0 O
o o o co ^- CO
O O O C0 O LO
NCO^WOONOOin
CMCDCDLOCMOOI-CD
LOCMO-i-CMOOOO
1" f \' ' \' '
Q.
o
CDCO^^i-COLOi-O
LO^OinT-NCDCDCO
I
i
m>
2
^ ^ ^t cd
CM i- O
LO i- tO O O
T
co
<D
0)
^
iONO)(Dt-NW(DO
00^L0C0i-'?-CDOhOOCDNNCDOWCOCD
i
\' ' \' \' ' ' ' \'
cd _?
"? t=!
O C0
?
y -5>
u
<D
CC
CT>
CM
CO
LO
CM
I
LO O
^- o
1- 00
CT> 100 O
f
CM CM 1- C0 LO CT>
O O O ^ CD 00
O O O LO LO C0
CM "3" 00 C0 C0 CM
1- 1- O Is- 00 C0
O O O O C0 CD
\' ' ' \' \' '
^NCDCONCOLOCDCJ
CMCMCMCDCDOCMC000
t-t-t-OOOOOO
CM00Oi-CM^O00i00h-Oi-h-i-OOiCO1-OO1-O1-OO
CD
C0
co o
O
O
hC0
CD ?^
i
.2
co
CO ^
?B ~?
CD ?
CO
"S
- co
lo ?
CD Q
0>
<o o
CM CM
^f O
1- O
1- 1- 1- LO CD C0
O O O 00 O LO
O O O t- CM 1-
00OCMOCT)i-i-LO00
^-lo^-^-cmot--?-^OOOOOOOOO
00 ^f
^f O
UO O
CM CM 1- h- CM C0
O O O CM ^J" 1O O O CM 1? C0
OCD^t-OOOOT-OOiNCOU)00)t-Ot-cO
1-1-0000000
|cnT
C0 ^f hOOO
nr ?O CM LO
CO CT> ^t
N CO CO
LO CD
C0 t^f 00
h- OO
1?
o >
II
<
^ -S
C0 ^
~ 8
"O .co
T3
C0 =
^ CD
O
1- CM co
CM
in
o
C0
*-?
</>
c
o
O
o
o
? 2
o ?
o a-
sz sz sz
co w w
. . C0
CD CD CD
co ?
O cm
_J oc
? a>
CD 1=
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
o o
12 19 19
OOO
* S>
<= 2
^ co
55 ! O
THE EFFECT OF DESCRIPTIVE REPRESENTATION 725
What appears in Table 1 to
be a clear preference
for
constituents
the ability
to id
descriptive representation member
on the part
of African
Ameri
of Congress
has
no
member-constituent
relationsh
cans is, in reality, a function
ofthe ideological difference
Further,
the effect
of race
that separate white and black
Democratic
legislators.
It i
sessments does not
vary at a
ideology that drives African-Americans'
assessments
Democratic MCs. With the
As exception
demonstrated
of overall
by the job
insign
ap
interaction
between
the race
of the
MC and individual
proval, black constituents
who
are
at
least
moderatel
well informed about politics
ideology,
and
self-identified
public
liberalsaffairs
are no more likely
more
to fa?
fa?
vorably assess
black representatives
than
are black mod- provorably assess Democrats with
more
liberal
political
erates andwith
conservatives.
files. Only black constituents
low levels of political
But if theare
evidencenot
suggests that
only white to
constitu?
knowledge?constituents who
likely
know
ents put a premium
on racerecord?do
when evaluating a member
much about their legislator's
voting
not tak
of Congress, then
what explains thewhen
patterns of particuinto account their members'
ideology
making
larized contacting observed in Table
1, where Africanjudgments about them. Interestingly,
however,
lack o
American
behavior
varies more sharply
with the
race of
political information does
not
heighten
the
salience
the
MC
than
does
white
behavior?
To
test
whether
the re?
race. In a series of regressions that added interactio
lationship
and descriptive represen?
terms between the race of
the between
MC contacting
and individual
politi?
tation
arises from
policy concerns
alone, intera
I
cal knowledge to the initial
set
of constituents'
predictors,
the
regressed the
contacting item on the same
set of prediction effects were consistently
statistically
insignificant
tors used inthat
Table 2. The
results are
listed in Table 3.more to
That is, there was no evidence
race
matters
Policy concernsknowledge
clearly influence the likelihood
that
constituents without the political
necessary
t
a
white
or
black
constituent
has
contacted
a
Democratic
root their assessments in a member's actual votin
the results
from Table 2,informe
black
record. Quite simply, for legislator.
even Similar
the to
most
poorly
Table 3 Predicting Contacting of Democratic Representatives
Whites
Blacks
Variables_(Logit)_(Logit)_
Constant -2.96 (.245)*** -4.12 (.782)***
.980 (.396)**
.043 (.012)***
-.009 (.007)
.011 (.004)***
.007 (.003)**
-.437 (.406)
-.685 (.525)
.434 (.478)
-.349 (.348)
-.793 (.671)
.720 (.313)**
.349 (.300)
.007 (.196)
.010 (.006)
.084 (.079)
.255 (.113)*
.190 (.348)
1506
-493.18
Note: Coefficient and standard error (in parentheses) estimates, and goodness of fit measures are com?
bined statistical results across five multiply imputed datasets. *p<.05, **p<.02, ***p<.01.
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
726
CLAUDINE
GAY
constituents
who
valued
about
are
by
at
w
politics
are
more
sence
of
des
able
overall
assessments
and
to
doubt
over
the
ease
and
liberal
voting
re
more
usefulness ofknowledge
contacting. While Canon (1999) finds that
political
the
tact
liberal
members
black legislators
differ significantly in the amount of at-
members,
tention
focused on the white when
and African-American
even
count.
comparison,
communities
in their districts, with many representa?
or
to
By
t
moderate
tives activelylevels
seeking to balance the needs of
of multiple
p
contact constituencies,
members
there are black legislators who largely wi
ig-
After
among
controlling
nore their white constituents, choosing to practicefo
a
Democratic
"politics of difference" where priorities are defined
leg
in
Congress
constituent
has
the number of black congressional representatives,
black
age
continues
to
strictly racial terms. Until the recent post-1992
surge in
black
legislator
MC
than
a
c
Democratic
which brought with it greater diversity in representa-co
is
almost
twi
tional styles,
the "politics of difference" was more
preva-
athanblack
lent
the "politics of commonality,"cons
the approach
Democrat
(16.7
adopted by many of the percent
younger generation of black
age
House
members.
constituent,
a
white
legislator Black
is
11.4
perce
legislators'
extra attention to the
needs of the
percent)
white
exist
race
lower
African American community under
may resonate with black
representation.
constituents. These efforts (e.g., the establishment of a
F
among
white
or
downtown district
office with a predominantly
black
b
based staff)
on
may have theconstitue
effect of making particularized con?
Regardlesstacting
of
individua
appear less difficult.
However, it is interesting to
likely
a
to
contact
a
legisl
note that while descriptive representation
may ease par?
legislator
with
whom
ticularized
contacting for African
Americans, it does not
results
support
increase the perceived utility of Mansb
contacting: black con?
representation
stituents are as likely to consider white legislators poten-
scriptive
tion
between
represent
tially "very helpful"
as they are black legislators.
Thus,
her
a
constituent's
A second possible explanation for the residual effect
a
member
of
Congre
of race on the member-constituent
relationship is that it
with
that
sequences
that
legislator
in
reflects the racial biases held by whites and African
c
extend
bot
Americans. Prior research has foundto
racial prejudice
to
define
the
membe
be an important
determinant of
white Americans' atti?
rican
Americans,
is
tudes toward black political candidates. it
Perhaps white
at?
o
elected
ence
The
representatives
titudes and behavior toward black incumbent legislators
for
descriptive
derive from similar sentiments. Conversely, a combina-re
effect
of
race
t
tion of racial
group consciousness
and a sense ofin
comfort
tionship
as
may
be
a
with same-race elected
officials may
accountprod
for African
accessibility.
Americans' greater willingness to White
contact black legisla?
that
the
first
tors. However, judging priority
by the asymmetry in the relative
African-American
effect of descriptive representation on whitecom
and black
tatives
are constituents,
not
available
the predisposition
towards favoring mem?
ance
of
two-way
com
bers of one's own race appears to be stronger
among
whites than among blacks.
8
Mean
values
in
the
To test used
whether the effects of
race arise from
Tenure
=
10;
ADA
Score
=
extrapolicy
concerns regarding accessibility
or from
pretween
high
school
[3]
and
dispositions
that are difficult
to change, perce
I reestimated the
(between
17
and
33
the
categorical
equations in Tables variables,
2 and 3, adding to the original set of
constituent
is
a
southern
predictors a measure of social distance, a dummy vari?
and
ideological
conservati
able forconservative
legislators elected in 1992 or later, and interac?
identifies
as
dium
level
of
political
tion terms
between black representation
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
kno
and
each of
THE EFFECT OF DESCRIPTIVE REPRESENTATION 727
these new variables.9 Social
distance is measured
as
member-constituent
relatio
difference between the should
feeling
thermometer
score
be
more favorable
tha
older
generations
of black
MC
signed to the respondent's
own
racial group,
e.g.,
tacting
constituen
white feeling thermometer
score by
forblack
a white
responde
different
balanceracial
of repre
and the score she assigns
to the opposite
gro
among the
younger
e.g., the black feeling thermometer
score.10
If generatio
affective
Canon's
(1999) for
resea
tachments to one's ownislators.
racial group
account
t
of
commonality"
does
not
n
preference for descriptive representation, then white c
tion
to the
needsas
of
stituents who feel relatively
close
to blacks
a black
group c
w
of contacting
a
respond more favorablylikelihood
to black MCs
than will wh
who feel more distant. At
the based
extreme,
for those
whi
vary
on whether
the
leg
after
1992
(i.e.,
the
coefficien
who feel as close to blacks as they do to whites, th
should be no difference in
the attitudes
behavior
to
tween
the 1992 or
dummy
varia
lator should
statistically
in
wards Democratic legislators
based be
on
the legislato
rican
Americans
equate
race (i.e., the coefficient on
the dummy
variable
for bl
characteristic
of the "politics
representation should be statistically
insignificant).
If A
I should
rican Americans' greater accessibility,
willingness to then
contact
black l
islators has a similar affective
then elected
black consti
black basis,
legislators
befor
1992
or
later.
The
results
are
ents who feel no closer to blacks than they do to whit
attachment
to me
(i.e., a social distance score Affective
of zero) should
be as likely
group
plays a real
but limit
contact a white Democratic
legislator
as they
wou
white attitudes towards black
white constituents
who feel
If, however, the race coefficients
are capturing
un-
black legislator.
they do to
blacks,
the effect
measured extrapolicy concerns,
then
assessments
approval
and favorability
rat
black legislators should vary
depending
on whether
whites
who feel
close
to
MC is among the younger
generation
of as
black
offi
whites.
When
social
distanc
holders, elected since 1992,
who have
been
more
likely
white constituent
rates
a blac
practice the "politics of commonality."11
For
whites,
(7.8 + .14*100) feeling the
9I also tested models that controlled for whether the congress
white Democratic legislator; w
district was majority-minority or majority-white. The coeffic
the
is reduced
on the district composition tance,
variables
(a gap
dummy
variable64
anp
interaction term between theshort,
dummy
variable
and the toward
race of
white
attitudes
legislator) were consistently statistically insignificant.
isting tensions in black-wh
10 The two feeling thermometer
from 0fully
to 100;exp
the
do items
not, range
however,
ferences, -100 to 100. Median social distance for white resp
whites for descriptive repres
dents is 10, indicating that the median white respondent feel
constituents
who
feel blacks
close
degrees warmer towards other
whites than
towards
group. Median social distance lower
for black
respondents
0, indic
job
approval is
and
fav
ing that the median black respondent
feelslegislators,
as close to whites
Democratic
but
group as she does to other blacks.
on
of indicator
recalling
This social distance measure
isthe
the likelihood
best available
affective attachments. However,
a more
effective
of the
tors,
believing
thattest
these
leg
pothesis that the preference for
descriptive
representation
is a
ful"
if contacted,
or actually
flection of racial predispositions would include the more soph
gardless of how white consti
cated measures typically used in the research on racial attit
group,
there
is a significant
d
and policy preferences: simple
prejudice,
"symbolic"
or "mode
racism, general measures of out-group
hostility,
and racial gr
ber-constituent
relationships
consciousness (Bobo and Kluegel
1993;
Pia
islators
and Sniderman
those with and
white
1993; Dawson 1994; Kinder and Sanders 1996). Unfortunat
only
one
these measures are not available In
in the
ANES.
case, the like
would be "very helpful" if co
nA "post-1992" dummy variable is a blunt tool for difTerentia
of differences
among
among legislators and testingdence
for extrapolicy
concerns.
Howev
more in-depth study ofthe representational
styles
of black
leg
evaluations of
black
legislat
tors is beyond the scope of this research. (Canon [1999] provi
Since I
found
no relationship
useful framework for such a12
study.)
The
measure
employed b
h
attitudes
among
blackthe
con
allows for a conservative test and
of the
hypothesis
outlined;
f
present
a new set for
of results
th
ing of a significant effect suggests
possibilities
futurefor
resear
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
728
CLAUDINE GAY
Table 4 Predicting Attitudes and Behavior Toward Democratic Representatives
Note: Coefficient and standard error (in parentheses) estimates, and goodness of fit measures are combined statistic
imputed datasets. "MC as Resource" is coded 1(Not Very Helpful), 2(Depends), 3(Somewhat Helpful), 4(Very Helpful)
tor
elected
after 1992 (26.5 percent). Thus, on at lea
legislator was elected before or after 1992.
White
con?
dimension,
the "politics of commonality" resonat
stituents are more likely to consider the black
legislators
elected since 1992 "very helpful" than they are
vorably
the legisla?
with white constituents, perhaps by addres
tors elected prior to 1992, or even white Democratic
their extrapolicy
leg?
concerns about accessibility In gen
however,
there is little evidence that white constit
islators. Whereas the average white constituent
is 35.6
percent likely to consider a black Democrat
distinguish
elected in
among black legislators on the basis of
1992 or later to be "very helpful," she is only
sentational
about 27
style.
percent likely to similarly assess a black Democrat
Theelected
tendency among black constituents to co
before 1992 (26.4 percent) or a white Democratic
blacklegislalegislators at higher rates than they contact w
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
THE EFFECT OF DESCRIPTIVE REPRESENTATION 729
Democrats does not vary in
with
social
distance
or with g
local government.
In fact,
Abney and Hutcheson
erational differences between
legislators.
Regard
(1981) found black
trust in city government
to be more closely
of a constituent's affective
attachment
associated
with group identificationto
with other
the incumbentAfric
mayor thanto
it wascontact
with perceptions of
in the delivAmericans, she is more likely
a equity
black
legislat
Further, black constituents
ery of public
areservices.
no less
If public opinion
likely
regarding
toCon?
contac
gress1992
is responsive
to variations
the identification
of
black legislators elected in
or
laterin than
legislator
racial groups
with years
incumbent legislators
of different equa
elected before 1992, all else
(e.g.,
of service)
Thus, while it is clear that
assign
races, thenconstituents
white constituents represented
by white legis?some
lators should be more supportive of Congresswhites
than white
value to descriptive representation?with
sh
constituents for
represented
by black legislators. Black
con?
ing a consistent preference
legislators
with
wh
stituents
who are
represented shouldclear
be
they can identify racially
?it
isdescriptively
not entirely
w
accounts for the trust invested
in same-race
elected of
more likely to approve
of the performance of Congress
than black
constituents represented
by white legislators.
cials. After controlling for
ideological
differences
and
To test this hypothesis,
I regressed
the congressional
aping the best available measures
to take
into
account
co
proval item on and
the same set
of predictors used in
Table 2.
stituents' extrapolicy concerns
affective
attachme
I also included two dummy
variables indicating
whether attit
the race of a legislator continues
to affect
white
and the behavior of both blacks and whites. The racial
Democrats or Republicans controlled the House and
dynamic underlying the member-constituent relation? Senate (partisan control of only one chamber is the
ship is driven by a set of concerns or predispositions not omitted category), and four interaction terms between
easily captured in this analysis. It may be that the dy? the party identification of the constituent and partisan
namic has a strong affective basis but one rooted in a control of the Congress. The results are listed in Table 5.
more overt out-group hostility that is not adequately
Rather than the hypothesized "spiral of trust," the ef?
measured by social distance. Alternatively, even the black fects of descriptive representation in Congress are conlegislators who practice the "politics of commonality," fined to the member-constituent relationship. Public apwhile perceived as more "helpful" to whites than the leg? proval of Congress as an institution is unrelated to
islators who practice the "politics of difference," may be citizens' ability to identify racially with their legislators.
failing to address all of white constituents' extrapolicy As indicated by the statistically insignificant coefficients
concerns. Perhaps black legislators as a group, regardless
on black representation and its interaction terms, white
of their outreach to the white community, put more ef-
constituents represented by white Democrats are no
fort into making themselves available to black constitu?
more likely to approve of the performance of Congress
ents than do white Democrats. This could explain the than are white constituents represented by black Demo?
differences in contacting among black constituents; con-
crats. For most African-American constituents, opinion
versely, white attitudes and behavior may stem from re- on Congress is similarly unaffected by the race of the leg?
sentment over what appears as preferential treatment. islators who serve them. It is only among the 38 percent
Whatever the source of the effect, a constituent's ability of African Americans who self-identify as ideologically
to identify racially with her member of Congress has im-
conservative that the race of the incumbent legislator in-
portant implications for her subsequent relationship fluences attitudes towards Congress. The average black
with that elected official.
conservative represented by a white Democrat serving in
a Democrat-controlled Congress is 2.8 percentage points
(37.1 percent versus 34.3 percent) more likely to favor?
Descriptive Representation and
Congressional Approval
To explore further the effects of constituents' ability to
ably assess the Congress than a similar constituent repre?
sented by a black legislator. Compared to the size of the
measured effects reported in Tables 2 and 3, the influence
of race on congressional approval among black conservatives is substantively insignificant.13 There is little
identify racially with their representatives, I turned from
13There is also no clear explanation for the direction of the effect.
the attitudes and behavior that define the member-con?
stituent relationship to perceptions of Congress as an in?
stitution. Studies of minority office-holding at the local
level have found that trust in city government is closely
related to citizens' attitudes towards individual local po?
litical actors and to perceptions of group representation
Perhaps it stems from constituents' assumptions about the relative
influence of white and black legislators within the institution. If
black legislators?trusted by liberal and conservative African
Americans alike?are perceived to be on the margins of the insti?
tution (whether for reasons of ideology or due to racism), their
constituents may be less likely to draw favorable inferences about
Congress as a whole.
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
CLAUDINE GAY
730
Table 5 Predicting Congressional Approval Among Constituents
of Democratic Representatives
Note: Coefficient and standard error (in parentheses) estimates, and goodness of fit measures are combined statistical
results across five multiply imputed datasets. *p<.05, **p<.02, ***p<.01.
evidence that the effects of descriptive representation
more than exone in 435 to affect change in an instit
tend beyond the dyadic relationship between
members
and to
satisfy the programmatic standards (e.g., ex
tions about the manner of doing business) by
and constituents.
The lack of significant spillover effects Congress
attributable
is evaluated. For a constituent, the race o
to descriptive representation, while at oddsislator
with research
may speak volumes about her priorities and
on local politics, is consistent with research
suggesting
sibility,
factors that can influence the member-co
that constituents routinely distinguish between
Congress
ent relationship
and can endear an individual leg
and its members. Such distinctions may be to
understandher constituents, while offering no guarantees
able in light of the limited influence of a single
member or the outputs of a legislative pr
the efficiency
ofthe legislature. The disconnect between the
policy
and
which
the
preferences of hundreds of political
extrapolicy concerns that lead constituents
tobe
contact
must
taken into account. A mayor, as a city's exe
and, in the case of whites, favor legislatorsis
with
in awhom
position to affect the direction of local g
they racially identify and the relative stability
of congres?
ment;
the significance attached to racial change
sional approval may derive from a realization
that it is
takes
former
more likely to impact evaluations of the
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
THE EFFECT OF DESCRIPTIVE REPRESENTATION 731
blacks, themay
attitudes and
behavior that deflne
the mem?a sin
By comparison, constituents
recognize
that
ber-constituent
relationship
continue
to differ
based on a sign
black legislator is limited
in her
ability
to
affect
constituent's
ability to identify racially with
her represencant change in congressional
performance
and,
thus, s
tative. The
preferenceCongress.
for same-race elected officials?
nificant change in attitudes
about
particularly among whites, less so among blacks?is
clear; however, the reasons behind this preference are less
Conclusion
transparent. (In fact, neither extrapolicy concerns, as
measured, or affective attachments can explain the Afri?
can-American preference for contacting black legisla?
tors.) Perhaps
these preferences arise from other
Empirical research on black
representation
incon?
Cong
cerns regarding
the efforts
or character of elected
has emphasized the material
gains
associated
with b
representatives.
Abney and Hutcheson,
noting the associ
close
electoral success over the
"intangible
goods"
correlation between
political trust
and perceptions
of leg
with citizens' ability to identify
racially
with
their
honesty, goods,"
conclude that public
officials "may some
be agents of sch
tors. Yet these "intangible
which
change simply
as atrust
result of the
images inclusi
they
have theorized to includeopinion
feelings
of
and
project,"
regardlessfunctioning
ofthe policies they pursueof
(1981,a rep
sentiments critical to the
proper
Alternatively, these preferences
may arise fromenoug
unsentative democracy, are100).
politically
significant
measuredWith
predispositions
among
constituents,
suchthe
as
warrant closer examination.
this
in
mind,
p
out-group
as opposed
to just
affective in-group
ceding analysis considered
thehostility
effect
of
descriptive
rep
attachments. With
better data, these theories
may hold
sentation on the relationships
between
citizens,
leg
tors, and the Congress.
promise for future research.
The effect
of descriptive
representation
is limited
not
Descriptive representation
plays
a real
albeit
limited
only insofar
as it influences just the member-constituent
role in shaping constituents'
perceptions
of their gov
relationship
also in that
it appears to matter more exp
ment. While there is no
sign but
that
constituents
consistently for white constituents
than for
blacks. What
single legislator to fundamentally
alter
how
Cong
initially appeared
to be a preference
for descriptive repre?
does business, a constituent's
ability
to identify
racia
with her MC influencessentation
theamong
attitudes
African Americansand
is largelybehavior
explained
by the ideological
differences that separate white
and
define the member-constituent
relationship.
White
co
legislators.
the end, it is only
the white
levstituents are more likely black
to Democratic
recall
the In
efforts
of
els of particularized
contacting
that differ
islators, more likely to approve
of
their
jobsignificantly
performa
based
on whether
a black constituent
is represented
by a
and rate them favorably,
and
more
likely
to view
the
or white white
legislator.
legislators as resources. black
Both
and African-Am
The
asymmetry
between
and African legislato
Ameri?
can constituents are more inclined
to whites
contact
cans in the
preference for descriptive representation,
who share their racial group
membership.
while consistentelected
with earlier studies
showing lower reflect
rates
The preference for same-race
officials
of calculations
crossover voting among whites
(Bullock
1984; Willmore than the rational
of
policy-min
1990),ideological
is at odds with the concerns
expressed by po?
constituents, although iams
the
differences
observers who openly legislators
worry about African Ameri?
separate white and black litical
Democratic
do aff
willingness
to "think beyond
the desire forthat
black
constituents' evaluations.cans'
The
findings
suggest
ex
faces and black solidarity"
(Swain 1995,
210).accessibi
Swain lapolicy concerns such as skepticism
about
the
ments
that
"white
liberal
Democrats
who
view
themof black legislators may play a role in white constitue
selves as members
the allies of African of
Americans
cannot always
limited receptivity to black
Congress;
bl
count on black
support" (Swain
1995, 216). Although
legislators who adopt a more
racially
inclusive
represe
white liberal
are less likely
than whites
their black
tional style are more likely
to Democrats
be viewed
by
as
sources. Affective attachments
of one's o
colleagues in the to
House members
to hear from black constituents,
the results
here
suggest that it is white constituents
and
racial group also influence
the
preference
for same-r
not African Americans who
place the
greater premium do
legislators, with white constituents
who
generally
on race.
feel close to blacks as a group
being less likely to app
For scholars concerned with issues of minori
of or favorably rate black legislators.
resentation,into
these findings
underscore
the need
However, even after taking
account
ideologic
broadly about
the consequences
of black office-h
differences and constituents'
extrapolicy
concerns,
as
as the social distance that exists between whites and
taking into account the ways in which it sign
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms
732
CLAUDINE
GAY
Durr, Robert H., John
B. Gilmour, and Christina Wolbrecht.
influences
citizens'
1997.
"Explaining
Congressional
Approval." American Jourlitical actors. Too often these influences have been disnalof Political Science 41 (1): 175-207.
missed as matters of individual psychology without clear
political implications. However, the attitudes associated
Fenno Jr., Richard F. 1978. Home Style: House Members In Their
Distrkts. Boston: Little, Brown.
with black office-holding are significant for what they in- Gilliam, Frank. 1996. "Exploring Minority Empowerment:
dicate about the quality of representation and the ability
Symbolic Politics, Governing Coalitions, and Traces of Po?
litical Style in Los Angeles." American Journal of Political Sci?
of black representatives to build effective relationships
ence 40 (1): 56-81.
with white constituents who, even in majority-minority
Groseclose,
Tim, Steven D. Levitt, and James M. Snyder, Jr.
districts, can account for more than one-third of a
1999. "Comparing Interest Group Scores Across Times and
district's population. Furthermore, the behavior associ?
Chambers: Adjusted ADA Scores for the U.S. Congress."
ated with black representation may have important im?
American Political Science Review 93 (1): 33-49.
plications for the issues of policy responsiveness of con- Hibbing, John R., and Elizabeth Theiss-Morse. 1995. Congress
cern to many social scientists. Because representatives
As Public Enemy. New York: Cambridge University Press.
attend to the participant community, the level of congru- Honaker, James, Anne Joseph, Gary King, Kenneth Scheve, and
Naunihal Singh. 1999. Amelia: A Programfor Missing Data.
ence between the policy preferences of constituents and
Cambridge: Harvard University.
the legislative priorities of members depends in part on
Howell, Susan, and Deborah Fagan. 1988. "Race and Trust in
the willingness of constituents to communicate their
Government." Public Opinion Quarterly 52 (3): 343-50.
preferences through all modes of political participation.
Kinder, Donald R., and Lynn Sanders. 1996. Divided by Color:
Conditions that undermine the willingness to reach out
Racial Politics and Democratic Ideals in the American Repub-
to legislators may also undermine the quality of policy
lic. Chicago: University of Chicago Press.
representation. Conversely, conditions that facilitate ver- King, Gary, James Honaker, Anne Joseph, and Kenneth Scheve.
2001. "Analyzing Incomplete Political Science Data: An Altical communication between constituents and legisla?
tors may strengthen the link between constituent prefer?
ences and legislative behavior. Thus, by appreciating the
effects of descriptive representation on attitudes and be?
ternative Algorithm for Multiple Imputation." American Po?
litical Science Review 95 (1): 49-69.
Lublin, David. 1997. The Paradox of Representation. Princeton:
Princeton University Press.
havior, we may be able to better understand its policy Mansbridge, Jane. 1999. "Should Blacks Represent Blacks and
consequences as well.
Women Represent Women? A Contingeht 'Yes.'" Journal of
Politics 61 (3): 628-57.
National Advisory Commission on Civil Disorders. 1968. Report ofthe National Advisory Commission Civil Disorders.
References
New York: Bantam.
Reeves, Keith. 1997. Voting Hopes or Fears? New York: Oxford
University Press.
Abney, Glen, and John D. Hutcheson. 1981. "Race, Representa?
Rubin, Donald B. 1987. Multiple Imputation for Nonresponse in
tion and Trust." Public Opinion Quarterly 45 (1): 91-101.
Surveys. New York: J. Wiley and Sons.
Bartels, Larry M. 1996. "Uninformed Voters: Information Ef?
fects in Presidential Elections." American Journal ofPolitical Schafer, Joseph L. and Maren K. Olsen. 1998. "Multiple Impu?
tation for Multivariate Missing-Data Problems: A Data
Science40 (\): 194-230.
Bianco, William T. 1994. Trust. Ann Arbor: University of Michigan Press.
Analyst's Perspective." Multivariate Behavioral Research 33
(4): 545-71
Sigelman, Carol, Lee Sigelman, Barbara J. Walkosz, Michael
Bobo, Lawrence, and Franklin Gilliam. 1990. "Race,
Nitz. 1995. "Black Candidates, White Voters: Understanding
Sociopolitical Participation, and Black Empowerment."
Racial Bias in Political Perceptions." American Journal of Po?
American Political Science Review 84 (2): 377-93.
litical Science 39 (1): 243-65.
Bobo, Lawrence, and James R. Kluegel. 1993. "Opposition to
Sniderman, Paul M., and Thomas Piazza. 1993. The Scar of
Race-Targeting: Self-Interest, Stratification Ideology, or Ra?
Race. Cambridge: Belknap Press of Harvard University.
cial Attitudes?" American Sociological Review 58 (4): 443-64.
Swain, Carol. 1995. Black Faces, Black Interest. Cambridge:
Bullock, Charles S., III. 1984. "Racial Crossover Voting and the
Harvard University Press.
Election of Black Officials." Journal of Politics 46 (1): 238Williams, Linda F. 1990. "White/Black Perceptions of the
51.
Electability of Black Political Candidates." National Political
Cain, Bruce, John Ferejohn, and Morris Fiorina. 1987. The Personal Vote. Cambridge: Harvard University Press.
Canon, David. 1999. Race, Redistricting and Representation.
Chicago: University of Chicago Press.
Conyers, James, and Walter L. Wallace. 1976. Black Elected Offi?
cials. New York: Russell Sage Foundation.
Dawson, Michael C. 1994. Behind the Mule. Princeton:
Science Review 2: 45-64.
Williams, Melissa S. 1998. Voice, Trust, andMemory. Princeton:
Princeton University Press.
Zaller, John R.1992. The Nature and Origins ofMass Opinion.
Cambridge: Cambridge University Press.
Princeton University Press.
This content downloaded from
130.86.101.91 on Fri, 16 Dec 2022 13:26:14 UTC
All use subject to https://about.jstor.org/terms