Sociology Mind
2011. Vol.1, No.4, 183-191
Copyright © 2011 SciRes. DOI:10.4236/sm.2011.14024
The Politics of Data:Uncovering Whiteness in Conventional
Social Policy and Social Work Research
Ann Curry-Stevens1, Amanda Cross-Hemmer2, Nichole Maher3, Julia Meier4
1School of Social Work, Portland State University, Portland, USA;
2Center for Improvement of Services to Children & Families, Portland State University, Portland, USA;
3Native American Youth and Family Center, Portland, USA:
4Coalition of Communities of Color, Portland, USA.
Email: currya@pdx.edu
Received April 29th, 2011; revised June 21st, 2011; accepted August 2nd, 2011.
The implementation of a robust community based participatory research (CBPR) study in Multnomah County,
Oregon, has detailed broad and deep racial disparities across 27 institutions and systems. The process of this re-
search has led to the identification of numerous practices that misrepresent and negate the experiences and very
identity of communities of color. The research draws from engagement with numerous databases from the Cen-
sus Bureau, the Bureau of Labor Statistics, the Substance Abuse and Mental Health Services Administration,
and various administrative databases. The core issues at hand are population undercounts, understudy of the
unique characteristics of these communities, inaccuracies in how data are codified and analyzed, and data col-
lection efforts that are infused with white centrism and a colorblindness that renders issues minimized and the
experiences of communities of color obscured. Collectively, we analyze this experience to suggest that much
conventional policy research while wrapped in a cloak of objectivity is in fact a reproduction of whiteness that
renders communities of color invisible, marginalized and misunderstood. The impact of these practices is to ex-
tend whiteness into the arena of policy research, and correspondingly extend dynamics of oppression and white
centrism. The paper profiles each area of the policy research process that reflects and reinscribes whiteness and
concludes with an articulation the reach of such conventional practice and outlines an avenue to reduce the in-
fluence of whiteness.
Keywords: Communities of Color, Measurement, P olicy Advocacy, Policy Research , Racial Disparities,
Whiteness.
Introduction
Whiteness is recognized to pervade many areas of social
work practice (Baines, 2007; Dominelli, 2002, Hick, Fook, &
Pozzuto, 2005). While we understand that research endeavors
are typically ripe with dynamics of oppression and privilege
(Lawrence-Lightfoot, 1993; Brown & Strega, 2005), there are
few insights to date that profile how social work research in the
policy arena is similarly ripe with power issues and, as we will
show, embedded in the research tools themselves. Beginning in
the UK, the efforts of Wilson (2001) illuminate the deficiencies
of a stance of neutrality in database construction. Her work
uncovers the biases within how those who engage in large sur-
vey construction are prone to reproducing dominant discourses
concerning the elderly. Wilson’s work challenges the field to
build routine practices to unveil biases in how experience is
theorized, database tools constructed and then analyzed. In the
context of race and racial identity, the efforts of Zuberi &
Bonilla-Silva (2008) provide an array of evocative considera-
tions for how to discern research conventions that reproduce
whiteness.
Defining whiteness as the process of valuing white identities
over those of color gives us insight into the mechanics through
which white privilege is reproduced. Whiteness, in our applica-
tion, is both the process and the outcomes of systems, institu-
tions, practices, behaviors, and discourses that differentially act
on whites and people of color. Such are acts of privilege and
simultaneously those of oppression, even when not intended.
The intention of policy makers (and in this case, database de-
velopers, administrators, and researchers) is inconsequential to
assessing the whiteness framework for these policy practices;
instead it is the impact that is of significance.
Repeated encounters with policy practitioners and research-
ers typically reveal that these white-dominated professionals
have not considered the implications of their practices. A lack
of awareness is the predominant dynamic. Secondly, in the
research team, we have become aware of the complexity of
these issues due to the urgings of our partners with the Coali-
tion of Communities of Color. Our catalysts are thus people of
color. The mainstream policy and research communities typi-
cally do not have such nor accountabilities to bring such issues
to the foreground of their attention. The dynamic here is thus
one of neglect. Another factor is that advocates within commu-
nities of color are typically underfunded and overworked. The
members of the Coalition know many of these issues experien-
tially, but have little time to advocate for changing policy re-
search practices as they are busy dealing with more concrete
needs of their communities. Finally, we assess that these issues
are sometimes borne of intention. In several instances, the data
surrounding disparities and inequities have been hidden by the
institutions that advance such practices.
While analysis of whiteness does not depend on the intention
with which research practices were conducted, but rather their
outcome, we have, however, found instances where there is
intention to obscure the experiences of our communities. The
most profound of these experiences is the Census Bureau’s own
upward revision of its 2000 numbers (US Census Bureau,
2004). Two major studies followed this release, both of which
A. CURRY-STEVENS ET AL.
184
were not adopted by Congress. It takes an act of Congress to
accept Census figures and the best interpretation of the political
environment is that the domination of Republicans ensured that
these issues did not come to a vote. Two vested interests were,
and continue to be in play. The first is that significant funds are
tied to Census numbers and heightened numbers result in larger
federal expenditures. A study by National Economic Consult-
ing (2001) documents that undercounting in the Census was
approximately 3.3 million (or 1.18%) and it set the net financial
value of these undercounts at $478 million over 10 years.
Additionally, Census figures have far-reaching consequences
in federal policy. Population counts are used to determine the
number of appointments in the House, for drawing Congres-
sional districts within states and local governments and school
district boundaries. These issues are directly tied to political
voice and the access that local communities have for political
representation.
The direction of the undercounting bias has a significant re-
gional impact. Undercounting occurs when people move fre-
quently, when they refuse to answer the forms, when they do
not have phones and when they do not have good enough flu-
ency in official languages to answer forms. These problems
exist most frequently within poor inner cities, and are more
likely to undercount children (National Economic Consulting,
2001). The composite impact of all these factors is likely ex-
perienced most dramatically among communities of color. The
impact of fixing these figures is to grant greater political repre-
sentation to these communities—and the implicit bias becomes
an explicit one—one of political choice to ignore the Census
Bureau’s upward revisions.
Most US policy practitioners rely on official data sets such as
the Census and the American Community Survey to document
important dimensions of the experiences of communities of
color. These are generated by conventional practices within
mainstream society, and as we will illustrate, that are imbued
with notions of colorblindness and innocuous understandings of
the influence of white privilege. Policy researchers use these
databases to document how social conditions challenge com-
munities of color. But these practices have failed to detail, and
perhaps even notice, the ways in which the typical datasets of
policy research are replete with whiteness. Bringing a critical
lens to the foreground builds insights into research practices
that continue to marginalize and render invisible communities
of color.
Over the last two years, a team of researchers and commu-
nity practitioners have been involved in a community-based
participatory research project (CBPR) in Multnomah County,
Oregon with the goal of documenting the disparities facing
communities of color in the region. The partnership is between
researchers at the School of Social Work at Portland State Uni-
versity and the Coalition of Communities of Color. A total of
28 different institutions and systems have been reviewed with
significant results uncovered to document inequities that run
broad and deep across all dimensions of life experiences
(Curry-Stevens, Cross-Hemmer, & Coalition of Communities
of Color, 2010). The findings reveal substantial and worsening
disparities, with distinctly toxic conditions facing local com-
munities of color. Disproportionality reaches from economic
and education institutions, to social service institutions such as
child welfare and justice, to programs that have intended to
ameliorate disparities such as protected government contracting
initiatives. Quite uniformly, all show marked disparities and
typically are worse than the national averages for such dispari-
ties, and worse as well than King County to the north of us,
where Seattle is based. Despite the progressive identity of the
region, communities of color are experiencing previously un-
told inequities and in the vast majority of the measures, the
situation is worsening. Since the report was released in May
2010, the Coalition has been engaged in advocacy efforts to
change the policy environment that beleaguers communities of
color. While the goal of the research has been to document the
disparities, a corollary of this research has been to uncover
shortcomings of available data, and make recommendations for
revised practices. These discoveries form the basis for this pa-
per.
This paper will contextualize the discoveries within domi-
nant discourse and white privilege, explain why this work mat-
ters, and then illustrate several decision points within the policy
research process. We primarily center our comments on the
array of population surveys on which the field relies, but also
stretch into some standard research practices at the stages of
analysis and dissemination. Our work is organized into the
following steps: data form design (and how race and ethnicity
questions are asked), data collection method, coding of data,
representation and dissemination of data, and issues related to
dominant discourses about race and ethnicity. The affects of the
discovered whiteness are laid out and solutions suggested.
Dominant Discourse and White Privilege
All features of this research experience are best explained by
a critical framework that centers issues of power and inequities.
These research practices are infused with white centrism and a
colorblindness that renders issues facing communities of color
minimized and populations themselves close to invisible. In the
funding arena, there is a kiss of death for having a small “n.”
Our work illustrates that both the small “n” and all other visi-
bility issues are exacerbated by conventional policy research
practices.
Complicating this dynamic is that many within communities
of color are reticent to self-identify as a person of color. Histo-
ries of genocide, persecution and cultural obliteration create
fear of the state. Even if this experience is in another country,
such attitudes flow into relationships with the state in the USA.
In addition, the USA has not been exemplary in wielding its
power with communities of color. Internment of Japanese
Americans, genocide of Native Americans, seizure of Native
American children and their placement in residential schools,
and the enslavement of African Americans are all state-legis-
lated policies which creates fear and distrust that extends to
tools of the state such as official surveys.
This history, coupled with self-deprecation of one’s minority
racial identity, is likely to create an impetus for individuals to
deny their identification as a person of color. Younger people
of color are less likely to deny their racial identity as social
movements have sought to increase their pride of identity and
resistance practices include affirming one’s minority status. The
impact of these shifts in discourse are illustrated in the local
pattern of the size of the Native American community. The
growth of the community counts in Multnomah County in
Census 1990 to 2000 of 6967 people (a jump of 118%) cannot
be explained by migration and birth rates. Instead, the commu-
nity perceives that the spike in identificating as Native Ameri-
can is due to heightened willingness to identify as Native
American.
Why This Work Matters
Before sharing our findings, we must briefly profile the im-
A. CURRY-STEVENS ET AL. 185
portance of accuracy in documenting our communities of color.
Why does this matter so much? While it might appear to be
“simply” an issue of what Fraser (2003) calls the “politics of
recognition,” the depths of this issue stretch extensively into the
lives and fabric of society among communities of color. The
consequences are many: since “dollars follow numbers,” com-
munities of color have lesser claim to society’s collective re-
sources (primarily government budgets) when they have
smaller numbers. Current estimates are that $1439/year is fore-
gone for each person not counted by the Census Bureau
(Reamer & Carpenter, 2010). Without such numbers, little
claim exists. Additionally, numbers serve to leverage influence.
Consider the policy-making table: when groups have sizeable
numbers, their representatives are often invited to consultations
and sometimes to participate in making decisions. When they
do not, they are rarely invited to join. While the conditions of
joining are often tokenistic, at least policy makers realize they
do not have enough legitimacy to make decisions for underrep-
resented groups. An easy synthesis of these issues is that larger
numbers equals power, frequently translatable into dollars.
Survey Design Issues and How Race and
Ethnicity Questions Are Asked
Let us begin by first asking why questions of race and eth-
nicity are asked. At its most basic level, such questions are
asked so as to enable the policy making community to assess
racial progress and the impacts of affirmative action policies
(Humes & Hogan, 2009; Young, 2000). The categories con-
structed to ask such questions give us the dimensions of the
disaggregation possible within the dataset. Probing more deeply,
the questions of race and ethnicity seek to let researchers assess
moral and political dimensions of the experiences of people of
color, specifically aiming to assess whether the policy envi-
ronment and related laws are successful in “redress of past in-
justice and/or bring about future equality” (Alcoff, 2000: p.
171). Classifications are thus intended to accurately capture
more than simply some typology of identity and difference –
instead their purpose is to “establish useful distinctions that
might elucidate our political realities and moral responsibili-
ties” (Alcoff, 2000: p.171).
The social construction of race is dynamic, changing across
time. While race and racial identifiers hold the possibility for
studying the current state of racism in the USA, they are prob-
lematic for many reasons. The first is that dimensions of racism
are not experienced similarly across a racial group, such as
“African American.” Here, those clearly from a lineage of
slavery are combined with others who have escaped such dam-
age, although Alcoff (2007) surfaces the possibility that a heri-
tage of diaspora serves to render all Africans similarly harmed
by racism. Secondly, the term masks additional features that
serve to buffer or exacerbate experiences of racism. The racial
identifier “black” combines all Africans regardless of the
lengths of time they have been in the country such as those here
for generations, eclipsing the forms of immigration under
which one arrived in the country (which might range from
refugees to business class immigrants carrying tremendous
wealth into the country). We also combine those with varying
degrees of “blackness” which serves to influence the extent of
anti-black racism experienced. Corlett (2007) illustrates how
the category “black” fails to identify the group of persons who
deserve reparations from US slavery and thus constricts options
for tracking the impacts of policies such as affirmative action.
The salience of racial identities thus has an ethnic, geo-
graphic and temporal context. Racial identifiers are dynamic
constructs, and each has its distinct strengths and weaknesses.
Let us now share our reflections on the tensions created with
the Census identifiers for our communities of color, of particu-
lar salience since the majority of databases we examined used
the conventions in the Census and its surveys. To facilitate this
discussion, we reproduce the two race and ethnicity questions
contained within Census 2000 and used again in 2010. The vast
majority of databases we examined used these same questions
in their own administrative databases.
To begin, Latino identity is separated from racial identity. If
one follows the directions by the Census Bureau in their train-
ing initiatives with canvassers, they advise Latinos to identify
themselves as a distinct ethnicity, and then suggest to respon-
dents the choice of defining themselves in the racial categories
of “white” or “some other race,” with a preference to encourage
them to select “some other race.” Our first concern is that the
very construction of this tool requires Latinos to be “othered”
and subsequently both implicitly and explicitly suggesting they
are outsiders to the USA. Sue (2010) articulates that exclusion-
ary practices such as this occur on a daily basis among commu-
nities of color with devastating impacts on health, well being
and economic progress. The alternative is to identify oneself as
“white” which is an inaccurate naming and results in a distor-
tion of both white and Latino experiences. The persistent dy-
namic of “othering” racial minorities is one of implicitly
claiming the superiority of those who are named accurately.
This method of identifying Latinos is an act of marginalization
and its corollary of imperialism is advanced (drawing from
Young’s 1990 framework of the five faces of oppression).
When data practices require a community of color to identify
themselves as “some other race,” the very act of completing the
survey reinforces one’s social exclusion from US society, and
simultaneously inscribes the inclusion of Whites, revealing the
normative impact of such a construction of identity.
The history of the decision to exclude Latinos as a separate
race is a political decision, with an expansive set of resources
available to track the impacts and wisdom of this decision (in-
cluding Gracia & Greiff, 2000; Gracia, 2007). A synopsis of
this challenge is that the decision to exclude Latinos as having a
distinct race is a reflection of the success of many Latinos in
their enfranchisement struggles and their social, political and
economic gains in the USA, primarily in the nation’s northeast
region. This is far, however, from a universal achievement as
Latinos in the south, west and northwest remain deeply margin-
alized and socially excluded from mainstream society. At the
national level, 47% of Latinos identify their race as white,
while 53% of the same community do not, with the breakdown
of this “some other race” category being that 42% identify their
race as Latino (the remainder is a combination of more com-
plex identities with the majority being black and the second
largest being Native American) (Alcoff, 2000; Lafoya, 2004).
In total, 97% of the “some other race” categor y is L atino.
When research is conducted at the microfile data level, and
disaggregated by ancestry, a class and nationality divide
emerges. Tafoya (2004) illustrates that 85% of Cubans identify
as white, while only 16% of those from the Dominican Repub-
lic identify as white. Mexicans, who are the largest of the La-
tino communities, have a divided identity with 47% identifying
as white. In our region, Latinos prefer to be identified as people
of color, as there is little evidence that they have benefited from
whiteness and have been largely subjugated to colonization.
Indeed, most Latinos share this legacy, but are personally tied
to this history in varying ways. Tafoya (2004) illuminates fea-
tures of these differences among the Latino community. Latinos
A. CURRY-STEVENS ET AL.
186
who identify as white are the wealthiest, best educated, best
employed of the Latino community, and live in closest prox-
imity to white communities.
While we recognize that the Latino identification issue re-
mains contentious, we report that our local Latino community
wishes to be seen as a distinct race and to be defined as a com-
munity of color. We also highlight that the convention of nam-
ing Latino identity as an ethnicity as opposed to a race is deeply
problematic since it obscures the challenges facing the Latino
community, as well as hiding the advantages experienced by
the white community. Essentially, Latino people “disappear” in
the racial identifiers and this construction of identity serves to
minimize our ability to understand the racial inequities the
community faces. Our interpretation of this dynamic is that the
process of separating Latinos from other communities of color
is a convoluted and misguided attempt to affirm the marginal
gains that have been experienced by limited portions of the
Latino community. The net impact is to erode our ability to
examine the racial inequities experienced by the community
(for creating this community as ethnic creates abysmal research
challenges to allow for accurate analysis of community experi-
ences) and it extends and perhaps entrenches a divide in the
solidarity needed among communities of color to resist dis-
crimination and raci a l ineq uities.
The racial identifying question for Native Americans is prob-
lematic as it appears from the flow of this question that one
needs to be tribally enrolled to answer this question, which is
not intended. This signals to the community to not list in this
section as a separate race if one is not enrolled. We also believe
that the historic relations of colonization and particularly of
tribal termination narrows the community’s likelihood of se-
lecting this identifier. Concerns exist among our Native Ameri-
can community that this process simultaneously engages in two
troubling practices: the first being that it ties Native identity to
enrollment, which for many is a source of deep pain. Tribal
enrollment can be very difficult, particularly for the thousands
of Native Americans who have been apprehended by the state
either through the residential boarding schools or through the
child welfare system who have had ties with their communities
involuntarily severed. The second impact is that by asking Na-
tive Americans to identify only one tribe in this section, the
state is requiring that Natives afford one tribe primacy over
others with which we may also be affiliated. Such is not Native
culture and this does harm to our communities through the very
construction of racial identifiers in the Census and other data
collection forms that draw from Census conventions.
Among the African American community, no supplemental
identifiers are permitted for African Americans, obscuring their
ancestry and origin as included in the Asian details. In our re-
gion, the needs of different African American groups diverge,
with those of recent African ancestry seeking identification as a
group with specific needs. Among Native Americans, reducing
one’s identity to just one tribe causes an excessive curtailing of
culture. In addition, the terminology for African American that
includes “Negro” is offensive to many. While it is understood
that there is an age variance here, with older African Americans
preferring the term “Negro,” many African Americans deem it
offensive as it is cconsidered a racial slur by leaders of the civil
rights movement, and the term has since been abandoned for its
association with slavery and segregation (Kiviat, 2010). One
African American woman told us she found it to be “code for
the other N-word that identifies us.” Offending respondents is
one way to make community membe rs not c omple te t he survey ,
and, by impact, an act of racism.
For the Asian/Pacific Islander community, greater complex-
ity of racial identifiers exists. There remains, however, a pattern
of “othering” those communities not listed in the list.
Turning now to other systems for identifying race and eth-
nicity, we find that while the Census Bureau has stayed away
from the “multiracial” category of identity as an avenue to han-
dle identities when one holds more than one race, some other
databases have not (the Substance Abuse and Mental Health
Services Administration (SAMSHA) and Oregon’s Kindergar-
ten Teacher Survey are examples). These surveys do not permit
multiple identifiers to be used, presumably trying to circumvent
the later challenge of how to code and represent such persons.
As researchers and communities members seeking to heighten
the visibility of communities of color, and support self-deter-
mination, this practice rolls back our progress. While seemingly
an advance (for there have been many community members
who want to affirm their multiracial identities), the impact is to
obliterate the specificity of these data. For example, SAM-
HSA’s National Survey on Drug Use and Health, 2009, uses a
“2 or more races” category that is 8% of the total number of
respondents in the survey. We examined the findings of this
survey to ascertain the dynamics of alcohol use within the Na-
tive American community. The data show that whites now sur-
pass Native Americans in their binge use of alcohol at rates that
are 11.7% higher (24.8% compared with 22.2%). But an addi-
tional 24.1% of respondents indicating high alcohol use are
“multiracial.” If we add this Figure 1, or even a fraction of it, to
the Native American category, Native Americans quickly sur-
pass whites in their binge alcohol use. How to disaggregate
these data is not information provided with the survey report.
To determine how many of this 24.1% belong in the Native
measure, we would reference the USA ACS data for those Na-
tive Americans who identify themselves as having more than
one race. This calculation results in 32.8% of the total multira-
cial identity belonging to Native Americans. Applying this
figure to our 24.1% rate, we should recalculate the alcohol use
Figure 1.
Census Bureau’s ethnicity and racial identifier s.
A. CURRY-STEVENS ET AL. 187
rate for Native Americans to be: 32.8% of 24.1% ( = 7.9%)
additional, which equals 30.1% binge drinking levels.
Thus Native American binge drinking levels continue to be
higher than that of Native Americans, though if one were to
consider only the reported data for each of Native and Whites, a
misleading interpretation would re s ul t.
In another example, the Oregon Department of Education in
2006 changed the racial categories for its survey of Kindergar-
ten teachers on the developmental benchmarks of incoming
students and the type of early education experiences they had
had. This category served to mask disparities that existed in the
differential experiences of children of color, when compared
with the developmental achievements of white children. In its
2008 version of this survey, the Oregon Department of Educa-
tion eliminated its use of this racial category.
What quickly becomes clear is that multiracial identifiers
obscure and render invisible the experiences of specific com-
munities of color through an array of practices which inaccu-
rately ask people of color to define their own race and ethnicity.
These data are then used to identify disparities and inequities,
and inaccurate data (particularly undercounting, but also the
complex problem created by holding Latino identity as an eth-
nicity) serves to reduce the visibility of disparities among
communities of color. The construction of racial identifiers in
the Census serves as well to “other” our communities in harm-
ful ways, and typically inscribes identity in ways that sustain
marginalization and powerlessness, while simultaneously con-
tributing to the elevation of white US society. While the mar-
ginalization process is harmful, the differentia l impact on white
communities and communities of color makes the impact worse,
for deprivation is the consequence of these practices, and po-
larization of deprivation (with excess accruing to the white
community) deepens the harms experienced.
Data Collection Methods
Turning now to data collection methods, we highlight the is-
sue of who completes the race and ethnicity identifiers for
community members. The core concern is that when others
complete these identifiers for people of color, there will typi-
cally be errors that stem from one’s appearance. While we have
heard stories of Census canvassers doing this on behalf of re-
spondents, this is not official practice. In all cases, they are to
ask. But comfort with the questions and a lack of appreciation
for the significance of the issue results in uneven practices. For
Census 2010, communities of color in Multnomah County
made official requests to train canvassers on this issue. This
request was denied as it is official practice to not allow such
training, despite the support of our local Complete Count com-
mittee.
Two additional examples are provided. The first is in the
health arena where death rates serve as the basic measure of the
impact of mortality on a population. When measuring racial
disparities, mortality data from death certificates supply the
numerator for death rates, while census population estimates
provide the denominator. The primary source of mortality data
in the United States are death certificates. On death certificates,
race reporting is typically the responsibility of a funeral director.
Funeral directors often must rely on personal observation to
make this determination, or alternately, must gather this infor-
mation from whoever accompanies the body at the time of
death. Understandably, such questions are avoided wherever
possible, with appearance thus being estimated. This means that
people of color may be identified as belonging to another group
on their death certificate.
Race and ethnicity specific death rates are used to calculate
all-cause and cause-specific mortality differences between the
racial and ethnic subgroups in the United States.
Studies have found that self-report (on Census or other sur-
vey) and death certificate proxy report of race and ethnicity
matching varies significantly by racial or ethnic grouping, and
that this disparity has persisted over time. The most recent re-
search on this confirmed that the error rate in this reporting is as
high as 42% among Native Americans, and to a lesser but still
significant group among the Asian/Pacific Islander community,
at approximately 12% (Centers for Disease Control and Pre-
vention, 2008). This confirmed earlier studies revealing the
same pattern with data dating as far back as 1979 (Frost & Shy,
1980; Stehr-Green, Bettles, & Robertson, 2002). This inaccu-
racy pervades all ages of deceased people from infants in hos-
pitals to elderly in their own homes.
The impact of these errors is to define large numbers of
deaths as white when in fact they are people of color, an error
that serves to underestimate the mortality rates among commu-
nities of color. Correction for death certificate misclassification
can make a large difference to both age-specific and age-ad-
justed death rates: for the Native American population, correct-
ing for misclassification makes the age-adjusted death rate
climb from 85% to 111% of that of the white population. This
changed a relative large Native American-to-white mortality
advantage to a relatively large disadvantage.
Death rate mortality data is used to highlight health dispari-
ties and drives initiatives aimed at improving the public’s
health. Incongruence between race and ethnicity classification
on numerators and denominators of death rates has the potential
to bias race- and ethnicity-specific mortality differentials,
which in turn impacts policy and funding allocations. The case
above is striking—imagine the funding and policy conse-
quences flowing from an assumption of a mortality advantage
where in reality a negative disparity exists.
For example, imagine the impact that 42% misclassification
on death certificates might have on the resources flowing to the
Native American community. To provide health care for this
increasingly urban area, the Indian Health Service (IHS) awards
contracts and grants to 34 nonprofit agencies located in major
metropolitan areas across the United States. As a group, Urban
Indian Health Organizations have minimal technological infra-
structure with no shared standardized data system that can be
used to provide a collective description of the target popula-
tions they serve (Castor et al., 2006). This raises the stakes and
need for other data sources to provide accurate, reliable epide-
miological information about the population. Such information
is necessary for these organizations to allocate their resources
effectively, customize health care services, implement program
evaluations, and launch policy initiatives. High rates of racial
misclassification in mainstream health-related data make it
harder for these organizations to advocate for increased funding
or to improve organizational efficiency.
Our second example of the errors involved when others de-
fine the race and ethnicity of people of color comes from state
child welfare administrative data. Filled out by child welfare
workers, and vulnerable to their discomfort in even asking the
question, there is a huge “unknown” racial identity among chil-
dren in foster care in Oregon. Children with racial/ethnic des-
ignations of race “unknown” represented 12.8% of the foster
care population. Of children still in care during this study’s 6
month analysis period, 42.8% of these “race unknown” children
A. CURRY-STEVENS ET AL.
188
had been in foster care 1 year or longer (Cahn, et al., 2009).
This means that the state does not know the race of these chil-
dren, even though they have had guardianship responsibilities
over their lives for at least a year.
An additional difficulty exists within data collection methods.
There are language limitations in data collection. While most
contemporary forms are available in Spanish and Span-
ish-speaking translators and interviewers are typically available,
there is a significant population that speaks neither English nor
Spanish. In Multnomah County, 19% of residents speak a lan-
guage other than English at home. Of these, half speak English
“less than very well.” This pattern is similar for the nation-wide
data. Only 6% of residents (about one-third of the non-English
homes) speak Spanish at home, leaving roughly 6.5% of indi-
viduals unable to communicate adequately in English or Span-
ish for the purposes of a survey. It is highly likely that the vast
majority of this population are people of color, meaning that
their experiences are routinely omitted for major data collection
processes. While there is a measure called “linguistic isolation”
that is substantially lower than these data, this measure depends
on the presence of a youth over 13 years of age to interpret for
older non-English speakers. Many survey topics salient to iden-
tifying disparities are not appropriately filtered through a youth
in the family. The net impact of language limitations is that
databases will have more accurate and larger samples of those
in white and English-speaking communities—an example of
whiteness, which is not intended but nevertheless is the impact.
Coding of Data
When data reports on the experiences of families and house-
holds (as opposed to individual level data), there is an embed-
ded undercounting of communities of color. The identity of the
“head of household” determines the identity for the entire fam-
ily or household. When families hold a shared identity, this is
not a problem. For mixed race families, however, mixed race
identities of children disappear. If we assume that there are
equivalent numbers of mixed race couples with white men as
head of the household and men of color as head of the house-
hold, then there is a proportionate under counting going on.
Notice, however, that their children will all be mixed race and
all will be children of color. Thus the net effect of this bias will
be to undercount the population of communities of color.
The second issue in data coding flows from the federal De-
partment of Education which has mandated that all reporting to
the federal government code all Latino persons who are also
identified as other races solely as Latino (US Department of
Education, 2007). This means that a student who is both Latino
and Native American will be coded only as Latino. In this sense,
the federal government has dictated that Latino identity is to
“trump” all other racial identities. This dynamic is most sig-
nificant for the Native American community that has the largest
overlap with Latino communities. While we appreciate efforts
to redress overcounting, and to have population counts clearly
add up to 100%, this policy will render other communities of
color significantly less visible. Local practices of more details
and equitable reporting are not prohibited, but cost constraints
may compel local boards of education and state agencies to
follow suit.
Representation of Data
This dimension of the research process addresses the com-
munication of data in the public arena. There are four issues in
this section: the aggregation of community experiences into
“average” reports, the separation of Hispanic communities from
other racial groups, the inappropriate aggregation that “hides”
community specifics, and the suppression of data for various
purposes.
One pronounced issue is the failure of research practices to
routinely disaggregate important data by race. While it is possi-
ble to track patterns in this area, we instead share our experi-
ence as indicators of these failings: time after time, too little
data was available on disparities as we aimed to collect these
findings for our research reports. Examples include incomes,
unemployment (a pervasive problem even at the state level),
wealth and expenditure data, child welfare (not corrected lo-
cally until late 2009), student discipline levels (still not in the
public arena), social assistance recipient numbers, health issues
prevalence rates, and Head Start service numbers.
Our second issue flows from the data representation of Lati-
nos as a non-racial group. Such grouping causes findings to be
awkward to comprehend, and implies accuracies that do not in
fact exist. In Cahn et al (2009), the disproportionality data from
local child welfare practices highlights how reports become
cumbersome and even impossible to work with. Readers may
be familiar with this format that places racial figures that total
100% and then post Latino identities below that total, with a
separate figure for their population. When these results are
viewed in the report, it appears as though Latinos experience
very low levels of foster care—and seem to have a “halo” effect
in how they are treated in child welfare. If, instead, we redid the
calculations of Hispanics as a part of the total racial grouping,
these numbers in fact reveal that there is a proportionate level
of Latino children in foster care. Recalculating these numbers
has been possible as the research was also conducted in our
university department, meaning we had access to the original
counts—a feature rarely possible in research. An additional
impact is that this separation of the Latino community high-
lights their separation from other communities of color, and by
impact, narrows and even precludes the solidarity that exist
between these groups.
In essence, the pra ctice of separating out the Latino commu-
nity serves to let researchers off the hook in ensuring that num-
bers accurately represent the size of communities of color. If
conventional practices of showing Latino figures after total race
data has been established were changed, the visibility of com-
munities of color in service figures would be both more accu-
rate but also more alarming. The net impact here is to diminish
the visibility of Latino children in child welfare. The additional
consequence is that we cannot conduct a consolidated evalua-
tion of disproportionality for children of color as an entirety.
This, in fact, is quite a tragedy as it precludes the possibility of
solidarity among communities of color.
The third representation issue facing communities of color is
the practice of inappropriately aggregating data for varied
communities. While there are numerous examples, a contem-
porary practice in government contracting practices is particu-
larly egregious. Most governments have policies to ensure that
their contracts are awarded equitably to groups who face barri-
ers in competing in open markets created by historic and cur-
rent experiences of marginalization and discrimination. These
groups are typically women, new small businesses and minor-
ity-owned businesses (our area of focus). In our region, Port-
land’s metropolitan government aggregates all three to report
on the success of this protected contracting program. The data
is not available to let us see the effectiveness of this program in
A. CURRY-STEVENS ET AL. 189
directing funds towars minority-owned businesses. This amal-
gamated reporting practice exists at all levels of government,
with the exception of the City of Portland. Digging deeper, we
find that only 4% of contracts under $100,000 are awarded to
minority-owned businesses, and 0% of contracts over $100,000
are similarly awarded. Thus this practice of inappropriate ag-
gregation serves to hide the inadequacy of the department and
the policy to create an affirmative funding environment to help
minority-owned businesses flourish. These failings have just
been profiled in the media (Har, 2010) and are likely to catalyze
some policy review.
The final dimension of faulty representation issues is the
suppression of data that is deemed to have “insufficient” num-
bers to disclose. While this issue has already been shared as a
catalytic moment for the research team at the start of this pro-
ject, its implications are far-reaching. The Census Bureau re-
quires 20,000 members in its American Community Survey
(ACS) counts to report, but this same contraint does not exist
for the findings of the Census itself, as it is not a representative
sample, but rather a full count. The decision to drop the the
long form, and increasingly rely on the ACS for population
characteristics, when compiled with lack of sampling size
growth, means that deterioration of available data is embedded
in the Census Bureau’s policies. This means that amalgama-
tions of data will be increasingly featured instead of disaggre-
gated data. This is of concern today and we accentuate that it
will be of increasing concern in the years to come. We face this
issue today in the Bureau of Labor Statistics’ reporting on un-
employment data. Insufficient numbers are avaiable to report
on the race of unemployed Americans, and despite the “aver-
age” rate being available for the state of Oregon, nothing is
available by race and ethnicity.
Accessibility of Data
Significant barriers exist to accessing racially-specific data.
While the centralized databases of the Census Bureau are
available free-of-charge, fees are incurred to access supple-
mental data. One example is the civil service hiring practices in
the City of Portland. Fees are charged to make requests on the
equity patterns of the workforce. In addition, special requests
for Oregon’s Department of Education data to provide more
details on their equity practices (such as special education data
disaggregated by race and ethnicity) take time and money. In
addition, special runs of Census Bureau data is costly, as con-
siderable expertise is needed to work with the microfile data.
While costs and expertise can sometimes be addressed by aca-
demic researchers, community practitioners rarely have access
to such capacities.
This biggest problem, however, is that each piece of data
needs to be requested, often many times. Advocates need to
agitate for access and this can use up valuable political capital
that the community would prefer to retain for systems change
work.
We want to impress upon all institutions that make data
available to adopt research practices that makes transparent the
experiences of communities of color. Routine disaggregation of
data by race and ethnicity is essential. It is unfortunate that at
the onset of the 21st century in the USA that we still have to
make the argument that not all Americans are the same, and
that communities of color have life experiences that are pro-
foundly unequal to whites. Aggregated data pretends this is not
true and must be rejected.
Effects of Whiteness in Databases
The most basic element of undercounts is that each person
missed in the Census population counts represents a loss of
$1,439/year to the region (Reamer & Carpenter, 2010). Under-
counting costs our community. In addition, these undercounted
numbers are used not only to dictate federal and state levels of
funding, but supplementally used by local governments and
foundations to help them stratify funding allocation decisions
for culturally-specific communities and services. Losing num-
bers is equated with losing financial resources.
Census figures are the gold standard for population counts.
Despite the emergence of the ACS and the intention for it to
increasingly inform community-level research, the population
counts in the ACS are deemed “second best” as they are the
results of a survey count as opposed to a complete census count.
While we are beginning in mid-2011 to see the data from Cen-
sus 2010, these data no longer include detailed experiences as
the long form was dropped for 2010. For these details, we need
to turn to ACS. The reach of Census is, however, expansive as
these counts will be used to stratify and weight the figures of
the ACS. Thus the impact of undercounting or miscoding
communities of color is to largely pass on this error to the ACS.
It also serves to stratify every other mainstream, administrative
and institutional sampling method for the next ten years (until
the next Census is administered).
Yet, the influence of Census and ACS figures stretch well
beyond mainstream databases. Consider how researchers use
these figures. For every piece of research that has a demo-
graphic component, such as a sample size, the method to strat-
ify these samples is informed by the most recent Census figures
or ACS. In addition, one assesses whether a community under
study is representative of the wider population, is again,
through Census and ACS. This means that when communities
of color are undercounted or miscoded, this is reproduced
throughout all such research. This means that insufficient num-
bers of people of color are included in research projects, and the
official status of such projects is that they are “representative”
and thus generalizable.
Additional problems with this reliance on the Census and
ACS population counts is that they are used for two very im-
portant routine practices. In social work, service providers look
at the profile of their service users and ascertain whether or not
these numbers reflect the community where services are based.
If these community figures do not match service user profiles,
then barriers to accessing services are believed to exist. The
second practice is that most employers who are either legislated
to ensure their hiring practices are equitable. To ascertain
whether the organization has employment barriers, the organi-
zation’s workforce is compared with the local community
population statistics. If people of color are undercounted
throughout the country in Census and ACS, then the work-
forces across the nation are similarly too small, even though
they may be deemed to be free of employment barriers. In es-
sence, too small official figures mean that the bar for hiring is
set too low.
Turning now to research that aims to understand disparities,
the reference points for the “relative rate index” and the “dis-
proportionality ratio” (or index) are, again, the ACS or Census
figures. The very existence of disproportionality becomes fact
based on an erroneous dataset. To remedy this situation, we
want to alert researchers to the fact that the Census Bureau
figures undercount communities of color, and reliance on the
A. CURRY-STEVENS ET AL.
190
“alone” or even the “alone or in combination” figures will un-
dercount their figures. The direction of this bias, is however, in
the alternate direction. Undercounting the size of the Native
American community in the denominator position (while the
local data will be poised in the numerator position) will actually
overdetermine the existance of disproportionality. Expanding
this problem are difficulties with how the numerator is calcu-
lated, given that every survey uses either the Census race and
ethnicity identifiers, or a method that is also problematice, such
as having workers complete the form and are reluctant to ask
the questions.
Solutions
Solutions to whiteness need to be advanced. The first is that
the voice of communities of color need to be centered in these
developments. It is this knowledge that needs to be held central
to the development of alternatives. To some extent, this strug-
gle is a post-modern challenge over meaning-making of com-
munities’ experiences, who gets to define them and how
knowledge is created. This is ultimately “the validation of local
knowledges” and an opening for “new systems of meaning
making to affect social reality” (Wilson, 2005: p. 5). The bene-
fits of such visibility may appear to be in advancing recognition,
yet as described in this paper are in fact profoundly material
and social, as social capital is tied to social capital which in turn
frames possibility for inclusion and influence. Without suffi-
ciently robust policy research practices that result in increased
visibility, possibilities are foreclosed and the white dominant
discourse continues unabated.
The work of the Coalition of Communities of Color has be-
gun to articulate alternative practices. While such efforts are in
early stages, we draw attention to the following principles for
construction of improved data collection tools. First, people of
color should be actively encouraged to identify their race and
origin accurately and complexly. Second, racial designations
should be amalgamated to be “race or origin” so as to be inclu-
sive and to capture identity more fully and without practices of
“othering” respondents. Third, Latinos should be included as an
equivalent community among other communities of color.
Fourth, individuals should be allowed to self-designate their
identifies, having major groupings pre-named, with additional
open spaces for supplemental identities. These categories
should be developed in consultation with communities of color
so as to reflect local conditions which are dynamic. Fifth, prac-
tice that allows multiple designations to be defined should con-
tinue. Sixth, the multi-racial category should be omitted as an
identifier due to its potential to obscure the experiences of our
communities of color. Instead of the multi-racial designation,
two supplemental questions can be asked: do you identify as a
person of color; and if you had to identify as only one race,
what race would you like used? Seventh, wherever possible,
data collection tools should be administered by those who share
the same race as those completing the form, and in their local
language wherever possible. Eighth, all contracts, subcontracts
and grants should require compliance with, and reporting of,
these same practices. Ninth, disaggregated data should be
available to the community and readily accessible by the gen-
eral public.
We believe that these principles will assist in opening con-
versations with policy researchers in numerous contexts. A
final principle is to ensure that these dialogues occur with
communities of color at the opening of the review and reform
process, instead of latter stages when the scope of their possible
involvement is curtailed.
Conclusion
The interactions of the policy research practices featured in
this article serve a troubling pattern that render our communi-
ties of color undercounted, misinterpreted, and misrepresented.
The extent to which our communities of color are visible, ac-
curately counted and clearly portrayed has a significant impact
on both recognition and redistribution.
The continued dynamic of centering the “average” experi-
ence as universal is an act of cultural imperialism as the white
experience is presumed to speak for all. The interplay of invisi-
bility and marginality, including the dynamic of promoting
separation between Latinos and other communities of color,
serve to preclude possibilities for solidarity and collective re-
sistance to disparities and inequities among communities of
color.
The function of whiteness as embedded in these database is-
sues is to exacerbate the oppression of communities of color.
While we have examined, so far, the technicalities in these
policy research practices, the outcomes of these research prac-
tices are far-reaching.
Researchers and policy practitioners need to understand the
shortcomings of their data and the decisions informed by them.
Ensuring that communities of color are able to lay claim to
collective resources and accurate representation is integral to
modern-day enfranchisement struggles, particularly those that
concern racial inequities and disparities. The influence of the
conventional policy-related databases is broad and deep. Un-
derstanding the whiteness embedded in the ways in which race
and ethnicity are asked, and the dominant discourse within
which these policy tools are situated serve to reproduce white-
ness. Such practices are accentuated when data from these
sources are used for additional research practices, policy mak-
ing and funding decisions.
References
Alcoff, L. (2000). Is latina/o identity a racial identity? In J. Gracia, & P.
DeGreiff (Eds.) Hispanics/latinos in the U.S.: Ethnicity, race and
rights (pp. 23-44). New York: Routledge Press.
Baines, D. (2007). Doing anti-oppressive practice: Building transfor-
mative politicized social work. Halifax, NS: Fernwood.
Bell, J., & Ridolfi, L. (2008). Adoration of the question: Reflections on
the failure to reduce racial and ethnic disparities in the juvenile jus-
tice system. San Francisco , CA: W. Haywood Burns Institute.
Brown, L., & Strega, S. (2005). Research as resistance. Toronto, ON:
Canadian Scholars Press International.
Buery, R. (1998). GOP census politics. The Nation, 267, 6-7.
Cahn, K., Miller, K., Bender, R., Cross-Hemmer, A., Feyerher m, B., &
White, J. (2009). What we know about racial disproportionality and
disparity in Oregon’s child welfare system: Decision point analysis
quantitative report. Portland, OR: Portland State University.
Centers for Disease Control and Prevention (2008). The validity of race
and Hispanic origin reporting on death certificates in the United
States. Vital and Health Statistics, 2, 1-32.
Curry-Stevens, A., Cross-Hemmer, A., & Coalition of Communities of
Color (2010). Communities of color in Multnomah County: An unset-
tling profile. Portland, OR: Portland State University.
Delgado, R., & Stefancic, J. (1997). Critical white studies: Looking
behind the mirror. Philadelphia, PA: Temple.
Dominelli, L. (2002). Anti-oppressive social work theory and practice.
New York: Palgrave McMillan.
Fraser, N. (2003). Social justice in the age of identity politics: Redis-
A. CURRY-STEVENS ET AL. 191
tribution, recognition and participation. In N. Fraser, & A. Honneth
(Eds.), Redistribution or recognition? A political-philosophical ex-
change (pp. 7-109). New York: Verso.
Frost, F., & Shy, K. (1980). Racial differences between linked birth and
infant death records in Washington State. American Journal of Pub-
lic Health, 70, 974-976. doi:10.2105/AJPH.70.9.974
Gracia, J., (Ed.) (2007). Race or ethnicity? On Black and Latino iden-
tity. Ithaca, NY: Cornell University Press.
Gracia, J., & Greiff, P. (Eds.) (2000). Hispanics/Latinos in the United
States: Ethnicity, race and righ t s. New York: Routledge
Har, J. (2010). Minority program helps whites. Oregonian, 6 January, p.
B1.
Humes, K., & Hogan, H. (2009). Measurement of race and ethnicity in
a changing, multicultural America. Race and Social Problems, 1,
111-131. doi:10.1007/s12552-009-9011-5
Kiviat, B. (2010). Should the Census be asking people if they are negro?
New York: Time Magazine.
http://www.time.com/time/nation/article/0,8599,1955923,00.html
Lawrence-Lightfoot, S. (1994). I’ve known rivers: Lives of loss and
liberation. New York: Addison -Wesley.
McFadden K., & McShane L. (2010). Use of word Negro on 2010
census forms raises memories of Jim Crow. New York: New York
Daily News.
http://www.nydailynews.com/news/2010/01/06/2010-01-06_census_
negro_issue_use_of_word_on_forms_raises_hackles_memories_of_
jim_crow.html
Mullaly, B. (2002). Challenging oppression: A critical social work
approach. Don Mills, ON: Oxford.
National Economic Consulting (2001). Effect of census 2000 under-
count on federal funding to states and selected counties, 2002-2012.
Washington, DC: Price Waterhouse Coopers.
Navarro, M. (2003). Going beyond black and white, Hispanics in Cen-
sus pick “other.” In P. Rothenberg (Ed.), Race, class and gender in
the United States (pp. 214-218). New York: Worth.
Oregon Department of Education (2010). 2008 Oregon kindergarten
readiness survey report: Readiness to learn. Salem, OR: Oregon
Department of Education.
Reamer, A., & Carpenter, R. (2010). Counting for dollars: The role of
the decennial census in the distribution of federal funds. Washington
DC: Brookings Institution.
Stehr-Green, P., Bettles, J., & Robertson, L. (2002). Effect of ra-
cial/ethnic misclassification of American Indians and Alaskan Na-
tives on Washington State death certificates, 1989-1997. American
Journal of Public Health, 92, 443-444. doi:10.2105/AJPH.92.3.443
Substance Abuse and Mental Health Services Administration (2009).
Results from the 2008 national survey on drug use and health: Na-
tional findings (Office of Applied Studies, NSDUH Series H-36,
HHS Publication No. SMA 09-4434). Rockville, MD.
Sue, D. (2010). Microaggressions in everyday life: Race, gender and
sexual orientation. Hoboken, NJ: Wiley.
Tafoya, S. (2004). Shades of belonging. Washington, DC: Pew His-
panic Center.
Urban League of Portland (2009). The state of Black Oregon. Portland,
OR: Urban League of Portland.
United Way of the Columbia-Willamette (2007). Community needs
assessment 2007. Portland, OR: United Way of the Columbia-Wil-
lamette.
US Census Bureau (2004a). Accuracy and coverage evaluation of
Census 2000: Design and methodology. Washington, DC: US De-
partment of Commerce.
US Census Bureau (2004b). Meeting 21st century demographic data
needs: Implementing the American Community Survey. Suitland, MD:
US Department of Commerce.
http://www.census.gov/acs/www/AdvMeth/acs_census/creports/Rep
ort08.pdf
US Department of Education (2007). Final guidance on maintaining,
collecting and reporting racial and ethnic data to the US Department
of Education. Washington, DC: US Department of Education.
http://www.ed.gov/legislation/FedRegister/other/2007-4/101907c.ht
ml
Wilson, E. (2005). Reinventing liberatory practice: How do we work
with groups of which we are not a part? International Conference on
Engaging Communities—Con ference Papers, Queensland, Australia.
http://www.engagingcommunities2005.org/abstracts/Wilson-Erin-fin
al.pdf
Wilson, G. (2001). Conceptual frameworks and emancipator research in
social gerontology. Ageing and Society, 21, 471-487.
doi:10.1017/S0144686X01008315
Young, I. (2000). Structure, difference and Hispanic/Latino claims of
justice. In J. Gracia, & P. Greiff (Eds.) Hispanics/Latinos in the
United States: Ethnicity, race and rights (pp. 147-165). New York:
Routledge.
Young, I., 1990. Justice and the politics of difference. Princeton, New
Jersey: Princeton University Press.
Zuberi, T., & Bonilla-Silva, E., (Eds.) (2008). White logic, white meth-
ods: Racism and methodology. Lanham, MD: Rowman & Littlefield.