This article appeared in a journal published by Elsevier. The

This article appeared in a journal published by Elsevier. The attached

copy is furnished to the author for internal non-commercial research

and education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling or

licensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of the

article (e.g. in Word or Tex form) to their personal website or

institutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies are

encouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Choosing race: Multiracial ancestry and identiﬁcation

Aaron Gullickson

⇑

, Ann Morning

Sociology Department, University of Oregon, 719 PLC, Eugene, OR 97403, United States

Sociology Department, New York University, 295 Lafayette St., Room 4118, New York, NY 10012, United States

article info

Article history:

Received 24 August 2010

Available online 29 December 2010

Keywords:

Multiracial studies

Racial identiﬁcation

Racial classiﬁcation

abstract

Social scientists have become increasingly interested in the racial identiﬁcation choices of

multiracial individuals, partly as a result of the federal government’s new ‘‘check all that

apply’’ method of racial identiﬁcation. However, the majority of work to date has narrowly

deﬁned the population of multiracial individuals as the ‘‘biracial’’ children of single-race

parents. In this article, we use the open-ended ancestry questions on the 1990 and 2000

5% samples of the US Census to identify a multiracial population that is potentially broader

in its understanding of multiraciality. Relative to other studies, we ﬁnd stronger historical

continuity in the patterns of hypodescent and hyperdescent for part-black and part-

American Indian ancestry individuals respectively, while we ﬁnd that multiple-race

identiﬁcation is the modal category for those of part-Asian ancestry. We interpret this as

evidence of a new, more ﬂexible classiﬁcation regime for groups rooted in more recent

immigration. Our results suggest that future work on multiracial identiﬁcation must pay

closer attention to the varied histories of speciﬁc multiracial ancestry groups.

1. Introduction

Although the United States has been home to a signiﬁcant multiracial population since its founding, American scholarly

interest in the racial identity of mixed-race people is a fairly new phenomenon. This development is due in large part to the

federal government’s recent change in its ofﬁcial classiﬁcation system to allow individuals to identify with more than one

race (see Ofﬁce of Management and Budget, 1997). With multiple-race statistical data newly available, especially after

Census 2000, it became clear that millions of Americans choose to ‘‘mark one or more’’ races when given the opportunity.

This ﬁnding gave new impetus to social scientiﬁc research on identity formation.

In this article, we seek to explain patterns of racial self-identiﬁcation by multiracial people in the United States. Do they

prefer to select one race or several to describe themselves, and why? Using census data from 1990 and 2000, we identify a

mixed-race population by targeting adults who report having ancestry in more than one racial group. This approach offers

several advantages over other more common methods of identifying multiracial populations. Most importantly, this ap-

proach captures a multiracial population that is broader and potentially more historical in its understanding of multiraciality

than the post-Loving ‘‘biracial baby boom’’ often targeted in previous studies.

Our method identiﬁes patterns of multiracial self-classiﬁcation that have not been identiﬁed by researchers using other

techniques. In particular, we describe a striking divergence between the racial self-reporting of people with mixed black,

white, and/or American Indian ancestry on one hand, and those who report some Asian ancestry. We conclude that two

regimes of multiracial identiﬁcation are at work—one historical and the other contemporary—and that they may coexist

doi:10.1016/j.ssresearch.2010.12.010

⇑

Corresponding author. Fax: +1 541 346 5026.

E-mail address: [email protected] (A. Gullickson).

Loving v. Virginia was the 1967 Supreme Court decision that struck down all state laws prohibiting interracial marriage (Davis, 1991).

Social Science Research 40 (2011) 498–512

Contents lists available at ScienceDirect

Social Science Research

journal homepage: www.elsevier.com/locate/ssresearch

Author's personal copy

in the foreseeable future without one necessarily displacing the other. Moreover, we offer a series of potential explanations

for the understudied question of why Asian ancestry is more strongly associated with multiracial identiﬁcation.

The racial afﬁliations of mixed-race people offer insights into both macro-level historical trends in racial ideology and

boundary formation, and micro-level mechanisms that shape individual decisions about racial identiﬁcation. As we will

show, the identiﬁcation choices that individuals make today continue to be inﬂuenced by concepts of race that formed cen-

turies ago: ideas (or their absence) of the properties of races and the nature of hybridity still dictate to a considerable extent

how people conceive of their racial membership. Multiracial identiﬁcation may be even more signiﬁcant, however, for what

is implies about the future of our heterogeneous nation. If the rigid racial-assignment rules of the past are on the wane, can

we expect a more racially egalitarian society as well?

Before we proceed, we think it worthwhile to address a point of potential concern. Any discussion of ‘‘multiracial’’ people

implicitly risks essentializing and reifying race by implying that multiraciality results from the mixing of ‘‘pure’’ racial types.

This would directly contradict the accepted sociological understanding of race as a social construction. Instead, our notion of

‘‘racial ancestry’’ is meant to take advantage of the common folk understanding of race that ties it to ancestry as a matter of

inheritance, in order to compare two different forms of identiﬁcation. In contrast to many other studies that seek to deﬁne

some objective criteria for multiracial classiﬁcation (e.g. the child of parents belonging to different single races), our study

takes the subjective nature of racial identity seriously and simply attempts to compare two different subjective forms of

racial identiﬁcation, one implicitly embedded in a person’s ancestral identiﬁcation and the other, more explicitly, in their

response to the ofﬁcial race question.

2. Background

2.1. Historical patterns

Historically, the racial assignment of mixed-race people in the United States has primarily followed two conventions. The

better-known of these conventions is the ‘‘one-drop rule,’’ whereby an individual with any known African ancestry has tra-

ditionally been classiﬁed, both socially and ofﬁcially, as black (Davis, 1991; Hollinger, 2003). This custom follows the logic of

hypodescent, according to which mixed people are assigned to the lower-status of their (two or more) ancestral groups.

Although evidence of this practice can be found during the era of slavery, the one-drop rule became more heavily codiﬁed

in law after the Civil War, when Southern states enacted segregation measures that necessitated precise legal deﬁnitions of

who was black and who was white (Williamson, 1980; Davis, 1991). Although it is not clear how widespread the one-drop

principle remains, it certainly survives to some extent in the contemporary United States: the fact that a man whose mother

was white is routinely considered our ﬁrst black president testiﬁes to its enduring power.

Perhaps less well-known, but just as deeply rooted, is the historical convention whereby people of mixed white and

American Indian ancestry have frequently been racially classiﬁed as white. In this hyperdescent regime, American Indian

heritage is not considered to have the determining properties that African ancestry does, and so individuals who claim to

have indigenous American origins can—and indeed are likely to—classify themselves as white (Snipp, 1989). This outlook

reﬂects a history of belief that American Indian ‘‘blood’’ could and would be absorbed without detriment by the European

American population (Wolfe, 2001). The hypodescent/hyperdescent asymmetry was also apparent in the very different

instructions that census enumerators were given in the 19th and 20th centuries for classifying mixed-race people depending

on whether they were white and black or white and American Indian (Morning, 2003).

Not surprisingly, these two longstanding regimes of racial assignment arose to sort out the mixed offspring of the groups

that were most numerous in the United States’ early history. In short, our most prominent and rigid racial-classiﬁcation rules

reﬂect the history of what might be called our nation’s ‘‘original’’ races: European, African, and American Indian. Without

discounting either the long history of Asian immigration to the United States, or the early presence of some groups that

are now included under the ‘‘Hispanic’’ umbrella, we employ this ‘‘old’’ (or ‘‘original’’) vs. ‘‘new’’ distinction to reﬂect the rel-

ative demographic weight (and political recognition) of various racial groups at different points in our nation’s history.

Today, however, our racial landscape is very different from what it was when the ‘‘old’’ regimes of black hypodescent and

American Indian hyperdescent emerged. Hispanics now make up the largest minority group, and although the US govern-

ment does not recognize them as a race, they are increasingly being ‘‘racialized’’ in the public, the media, and the acad-

emy—that is, positioned as a large bloc akin to (and distinct from) whites and blacks. At the same time, the US

population of Asian descent is projected to multiply considerably in size over the coming decades (US Census Bureau,

2008), and individuals of Asian ancestry have played a prominent role in the multiracial movement, particularly through

college organizations.

Moreover, the growth of these ‘‘new’’ groups is not the only challenge facing existing regimes of racial assignment. Since

the 1960s, the salience of American Indian ancestry has clearly been on the rise, contributing to dramatic growth in the size

of the population enumerated as American Indian (Eschbach, 1993, 1995; Liebler, 2010). Finally, the rapid growth in black/

white interracial marriage (Gullickson, 2006) and the resulting growth in the black/white biracial population have led some

commentators to question the relevance of the one-drop rule in post-Civil Rights America.

These demographic and social developments pose an exciting challenge for the study of multiracial identiﬁcation in the

21st century. What ‘‘rules,’’ if any, guide the ways in which mixed-race people of varied backgrounds are identiﬁed and

A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

499

Author's personal copy

identify themselves today? Are the hypodescent and hyperdescent conventions of the past likely to remain in force, and if so,

do they apply to the United States’ relatively ‘‘new’’ races as well?

2.2. Contemporary decisions

Research on individual-level decisions about how to identify one’s self or children in racial terms has often involved qual-

itative ethnographic or interview studies (Korgen, 1998; Renn, 2004; Rockquemore and Brunsma, 2002; Twine, 1996), with a

liberal dose of autobiography (see Root, 1992, 1996). With the advent of ‘‘mark one or more race’’ data collection, however,

quantitative analyses of racial classiﬁcation on census forms and large-scale surveys are steadily increasing in number. These

might best be described as ‘‘identiﬁcation’’ studies, following Brunsma’s (2005, p. 1134) argument that ‘‘racial identity and

racial identiﬁcation represent two related yet distinctly different outcomes’’: ‘‘Identiﬁcation...refers to how multiracial peo-

ple racially classify and designate themselves on surveys and censuses.’’

Identiﬁcation studies for multiracial people have become prevalent in recent years but are difﬁcult to summarize for sev-

eral reasons. First, as we will discuss in more detail below, the methodology and data limitations of these studies vary con-

siderably. For example, many studies examine the identiﬁcation choices of individuals when they are forced to choose only

one race, while other studies examine data where individuals are allowed to check multiple races. Second, scholars have ap-

proached the topic of identiﬁcation from different ﬁelds with somewhat overlapping, but distinct, research agendas. For

example, some scholars approach the issue from the ﬁeld of child development, keenly focused on the individual experience

of identity formation (Herman, 2004; Hitlin et al., 2006), while other scholars have approached the issue from the ﬁeld of

racial boundary formation and its macro-level consequences for racial formation in the United States (Brunsma, 2005). Third,

studies vary considerably in terms of the variables they focus on—such as economic status (Xie and Goyette, 1997; Korgen,

2010) or appearance (Rockquemore and Brunsma, 2002)—when seeking to account for multiracial people’s self-identiﬁca-

tion. Finally, despite a growing number of studies, there is a clear lack of established theoretical expectations, competing

or otherwise, to drive the ﬁeld forward.

Probably the most common theme linking multiracial identiﬁcation studies is the question of whether the pattern of hyp-

odescent observed for part-black individuals in the past continues today and whether that pattern applies to other minority

groups as well. The evidence on the ﬁrst question is equivocal, partly because we lack an objective historical baseline from

which to decide whether the one-drop rule is or is not on the decline. For example, both Roth (2005) and Campbell (2006),

using different methodologies, argue that the proportion of part-black individuals who check more than one box, and even

their non-black race alone, is large enough that the one-drop rule cannot be thought of as dominant in these individuals’ self-

classiﬁcation practices. However, Qian (2004), despite similar proportions, argues that the one-drop rule remains dominant

for individuals with some black ancestry because they are far less likely to claim either both identities or their non-black

identities alone relative to other multiracial individuals.

The evidence on whether the pattern of hypodescent applies to other groups is clearer; partial Asian or Hispanic ancestry

is much less likely to be subjected to a ‘‘one-drop’’ rule. Part-Asian individuals have been consistently shown to vary more in

their self-classiﬁcation than part-black individuals do, and they are less likely to claim Asian ancestry alone than part-black

individuals are to describe themselves solely as black (Saenz et al., 1995; Xie and Goyette, 1997; Qian, 2004; Roth, 2005).

There is little work on individuals of part-Hispanic ancestry because of a data limitation in the way Hispanicity is identiﬁed

in most data. Nonetheless, the few studies that have examined part-Hispanic individuals have found signiﬁcant variability in

self-classiﬁcation (Brunsma, 2005).

Identiﬁcation studies have paid much less attention to whether the traditional regime of hyperdescent remains powerful

among those of part-American Indian ancestry, despite an active ﬁeld of research on American Indian racial classiﬁcation

overall (Garroutte, 2003; Nagel, 1996; Unrau, 1989; Wilson, 1992). In fact some scholars have extended the ‘‘test of hypo-

descent’’ to individuals of part-American Indian ancestry (Brunsma, 2005), seemingly ignoring a well-established literature

that documents exactly the opposite phenomenon for those of part-American Indian ancestry. In general, the placement of

part-American Indian individuals within multiracial identiﬁcation studies has been confused and problematic, largely as a

result of the way in which multiracial individuals are identiﬁed in these studies, as we will discuss in more detail below.

Other studies of American Indians do suggest a challenge to the regime of hyperdescent, however. It is well established

that since at least 1970, the American Indian population has grown much faster than it could have by natural increase alone,

suggesting that there has been a net inﬂux of individuals who have ‘‘switched’’ their racial identiﬁcation to American Indian

from something else (Eschbach, 1993, 1995; Liebler, 2010). Such switching implies a weakening of the pattern of hyperdes-

cent, as more individuals discard other possible racial identiﬁcations in favor of an American Indian one or, as of 2000, add

American Indian to existing racial identiﬁcations.

2.3. Who is multiracial? The methodological challenge of identifying the mixed-race population

Researchers who want to investigate the identiﬁcation of multiracial people all face a common problem: How to deﬁne

and measure the mixed-race population to be studied? As Harris and Sim (2002, p. 625) put it, surveys (including the decen-

nial census) each capture ‘‘a multiracial population, not the multiracial population’’ (emphases added). Morning (2000)

shows how ﬂexible the boundaries are by calculating that a deﬁnition of ‘‘multiracial’’ that included individuals with

500 A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

Author's personal copy

genealogically-distant mixture would put the share of mixed-race Americans around 40% of the total population, instead of

the roughly 2% ﬁgure to emerge from the 2000 census.

To identify mixed-race people in large statistical databases, researchers have primarily used three strategies. The most

direct approach is simply to rely on self-reports, that is, individuals’ responses indicating they are multiracial (e.g. Tafoya

et al., 2004). Although this technique is useful for analyzing the characteristics of self-identiﬁed multiracial people, it is inap-

propriate for exploring the varied racial labels that mixed-race people choose, since it includes only those who have opted to

report a multiracial identity. In other words, if these responses were to be used to analyze the race options taken by indi-

viduals, then the analysts would be guilty of sampling on the dependent variable.

Some studies try to work around this problem by analyzing the choices that self-reported multiracial individuals make

when they are also asked to indicate which single race best describes them. For example, Campbell (2006) uses the 1995

Current Population Survey Supplement on Race and Ethnicity to analyze which single race is preferred by respondents

who indicate a multiracial identity. This technique is particularly helpful for ‘‘bridging’’ methods that attempt to compare

results across datasets where some allow for multiracial reporting and some allow only single-race reporting. However, this

approach limits the racial identiﬁcation choices of multiracial respondents because it forces a single-race response, thereby

eliminating the option of multiracial identiﬁcation as their preferred choice.

Perhaps the most common strategy to date for identifying multiracial individuals has been based on the racial identities

that their parents report (Brunsma, 2005; Chew et al., 1989; Eschbach, 1995; Herman, 2004; Qian, 2004; Roth, 2005; Saenz

et al., 1995; Xie and Goyette, 1997). The advantage of this approach is that it provides an objective measure of multiracial

status that does not suffer from the problem of sampling on the dependent variable above. However, there are several nota-

ble drawbacks to this approach. First, parental information is commonly only recorded for parents who live in the child’s

household. Therefore, children living in single-parent or stepparent families are either excluded from analysis or potentially

misclassiﬁed. Second, information on parental race is usually gathered only for children younger than 18 years of age and

living with their parents. This poses two problems. One is that this technique is only able to identify the segment of the mul-

tiracial population that is under 18, ignoring several generations of older mixed-race Americans. The other problem is that

since children’s race is often reported by one of their parents, the resulting data does not represent multiracial individuals’

racial self-identiﬁcation, but rather, the labeling choices their parents make on their behalf. While such information is cer-

tainly valuable, it tells us more about older generations than it does about younger ones, and it says nothing about the iden-

tiﬁcation that mixed-race individuals choose for themselves, nor how those choices might change as children grow to

maturity and leave their parental home.

The ﬁnal shortcoming of this approach, at least in our view, is that it limits the deﬁnition of ‘‘multiracial’’ to ﬁrst-

generation children of interracial couples, also known as ‘‘biracial’’ offspring. This reduction reinforces the erroneous

notion that multiraciality is a new, post-Loving phenomenon in the United States (Spencer, 2006). More importantly

for scholars who analyze attitudes toward multiracial ancestry as a window onto broader race thinking, it eliminates

from study a crucial element of analysis: the major role that historical period plays in shaping the range of racial iden-

tities—if any—from which people feel they can choose. This blind spot is clearest in the case of individuals of part-

American Indian ancestry. Of the 2.4% of individuals who checked more than one box on the race question in Census

2000, 15.9% of them checked the American Indian and white boxes (Grieco and Cassidy, 2001). The American Indian/

white combination was the second most frequent multiple check-off option, after white/other. Almost certainly, most

of these individuals are not the product of one parent who identiﬁes only as white and one parent who identiﬁes only

as American Indian, but are rather drawing upon more historical knowledge about their family history in claiming both

European and American Indian ancestry. Yet most studies of multiracial identiﬁcation use the parental method to iden-

tify multiracial respondents and thus the only part-American Indian ancestry individuals in the sample are those where

one parent was identiﬁed as American Indian alone and the other as something else. In other words, the parental

method is incapable of telling us anything about what the census has enumerated as one of the most common forms

of multiraciality.

The third approach for identifying multiracial people is really a loose collection of strategies that rely on comparing

multiple responses to the race question in different contexts. For example, Harris and Sim (2002) take advantage of the

fact that the National Longitudinal Study of Adolescent Health (known as ‘‘Add Health’’) asked respondents for their race

in different ways (permitting multiple responses and forcing single-race choices) and locations (asking at home and at

school), and recorded parental race as well. As a result, students who reported one race at home and another at school

might be considered multiracial. Hitlin et al. (2006) also use Add Health data, but they search for race-reporting incon-

sistencies over time by comparing the Wave 1 (1994–1995) results to those of Wave 3 (2001–2002).

These methods

have the appeal of relying on individual’s self-reports, rather than their racial identiﬁcation by others. However, for the pur-

poses of examining the racial identiﬁcation choices of multiracial individuals, this approach has two important limitations.

First, even small numbers of misreports can swamp the ‘‘true’’ population of interest. Second, it is difﬁcult to determine

what racial identiﬁcation individuals prefer, precisely because the method of identifying multiracial individuals is to search

for discrepancies in their reporting.

Penner and Saperstein (2008) also examine race-reporting inconsistency over time, but their objective is not to identify a multiracial population.

A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

501

Author's personal copy

3. Methods and data

3.1. Using ancestry to identify the multiracial population

To avoid many of the drawbacks described above, we adopt Goldstein and Morning’s (2000) method for identifying the

multiracial population based on adults’ self-reported ancestry, regardless of generation.

Speciﬁcally, we use US census data

to identify individuals who reported ancestral descent from two or more groups that are commonly considered racially distinct.

The major innovation in our approach is that the identiﬁcation of multiracial individuals is entirely divorced from the race ques-

tion itself.

We argue that ancestry reports offer a special advantage over methods that rely on race responses. Race is a basic cog-

nitive dimension of social interaction in the United States that individuals use to interpret and organize their social world.

Ancestry, on the other hand, is a less sensitive concept that has far less salience in daily life. Furthermore, the routinized

nature of race reporting for most Americans often yields preset, single-race answers, rather than detailed descriptions of

one’s origins. In recent decades, the ancestry question appeared only on the long form of the US census, which was distrib-

uted to 1 in 6 households. (It has since been supplanted by the Census Bureau’s annual American Community Survey.) Due to

its open-ended nature and relative rarity, the ancestry question is more likely to generate spontaneous and novel responses.

Although the concepts of race and ancestry are distinct, they are not independent. Race is popularly understood in the

United States to be a matter of inheritance, and therefore a person’s responses to the ancestry question implicitly provide

a form of racial identiﬁcation, albeit not necessarily an identiﬁcation that is consistent with the racial categories available

on the US Census. It is this implicit correspondence between race and ancestry responses that we take advantage of here

to ask a straightforward question: conditional on acknowledging two ancestries that are commonly understood to belong

to different racial groups, how does a person answer the race question?

By probing ancestry responses for evidence of multiracial heritage that is distinct from the information provided on the

more familiar race question, we are able to address several shortcomings of other methods. First, our approach does not limit

the target group to those who explicitly select a mixed-race label, thus avoiding the problem of ‘‘selecting on the dependent

variable.’’ Second, unlike studies of biracial children, we are able to identify an adult multiracial population. By restricting

this population to heads of household, we can limit our analysis to individuals who are likely to be self-reporting, rather than

relying on the proxy identiﬁcation of parents or other relatives. Third, again in contrast to studies of biracial children, we are

able to identify a broader population that includes both individuals who are the children of parents who identify as belong to

different single races and individuals who recognize a long history of race mixing within their family tree.

Our approach is not without its own drawbacks and we are not advocating that other methods should be discarded in

favor of our own. As we noted earlier, different approaches will capture different types of multiracial populations. For

researchers particularly interested in the identity formation of the children of interracial parents, our approach is not par-

ticularly useful, because that particular group is not identiﬁable within our sample. On the other hand, for researchers inter-

ested in a more historical approach to understanding multiracial identiﬁcation, our method has the advantage of

incorporating individuals with a far more diffuse understanding of their identity than do other approaches. In particular,

our strategy is preferable to other methods for testing popular theories of hypodescent and hyperdescent, because it better

captures the relevant ‘‘at-risk’’ population: the population that is aware of and acknowledges ancestral descent from more

than one racial group, independent of whether they believe such ancestry makes them multiracial. Given the differences

between our methodology and those used by other researchers, we will compare our results to prior ﬁndings and consider

how they are shaped by these diverse approaches to sample construction.

3.2. Data and sample selection

Data come from the 5% samples of the 1990 and 2000 US Censuses available from the IPUMS project (Ruggles and Sobek,

1997). Using both the 1990 and 2000 Censuses allows us to compare results in the case where respondents were allowed to

only check one box in response to the race question to the case where respondents were allowed to check multiple boxes.

Typically, one member of the household ﬁlls out the census form for the entire household. Although there is no deﬁnitive

way to determine which person in the household responded for the entire household, the most likely candidate is the person

reporting as the head of household (HH). Therefore, we limit our analysis to 5,273,239 heads of household from our Census

2000 sample in order to more accurately capture self-identiﬁcation. This restriction limits our ability to generalize our

results across gender because female heads of household are a selective group.

We seek to analyze the race reporting of individuals whose ancestry responses indicate multiracial heritage. The ancestry

item is an open-ended question that asks ‘‘What is this person’s ancestry or ethnic origin?’’ and provides two lines for

respondents to write in their answers. To determine whether these write-in ancestries indicate multiple racial origins, we

must ‘‘racialize’’ them. Because both the terms ‘‘ancestry’’ and ‘‘ethnic origin’’ imply some information about race, we do

See also Eschbach et al. (1998), Lieberson and Waters (1988) and Liebler (2010).

We analyzed our data separately for men and women to determine how sensitive our results were to gender. Aside from a few moderate differences, the

distribution of race reporting is similar for men and women and our overall conclusions would be unchanged were we to exclude one or the other gender. These

results are available upon request.

502 A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

Author's personal copy

not believe that this approach misrepresents respondent’s answers. However, because the ancestry question is an open-

ended question, respondents are not restricted to an understanding of their racial ancestry that is necessarily consistent with

census racial categories. In fact, responses to the ancestry question are highly varied; the Census Bureau classiﬁed about 500

unique responses to the ancestry question in Census 2000 (Brittingham and de la Cruz, 2004). To simplify, we collapse this

large number of responses into a smaller set of racial ancestries, using Goldstein and Morning’s (2000) method. These racial

ancestries are the ﬁve different ofﬁcial race categories established by the Ofﬁce of Management and Budget’s (1997) Direc-

tive 15—i.e. white, black, American Indian, Asian, and Paciﬁc Islander—as well as a residual set of ancestries that either fall

outside these ofﬁcial categories or are racially uninterpretable. The unofﬁcial categories are Hispanic, Caribbean and South

American, Middle Eastern, and South Asian.

The racially uninterpretable categories are American

and Mixed.

Table 1 shows the distribution of household heads’ (HH) racial ancestry responses for both our Census 1990 and Census

2000 samples. Signiﬁcantly more HH’s reported no ancestry in 2000 than in 1990 (18.8% vs. 8.9%). Other than this issue, the

results are similar for both, so we focus on Census 2000. About 19% of HH’s did not report any ancestry, about 60% of HH’s

reported one ancestry, and 21% reported two ancestries. Among those HH’s who reported one ancestry, about 72% reported a

racial ancestry that corresponds to an ofﬁcial race category, while 15% reported an unofﬁcial racial ancestry, and the remain-

der (13%) reported a racially uninterpretable ancestry. In about 87% of the cases where HH’s reported two different ances-

tries, these ancestries came from the same ofﬁcially-recognized racial ancestry. This group is almost exclusively composed of

individuals reporting two different European ancestries. A very small fraction of individuals (0.2%) report two ancestries that

correspond to the same unofﬁcial racial ancestry. The ﬁnal three groups of HH’s all reported two different racial ancestries.

For this analysis, our focus is on individuals who identiﬁed with two different ofﬁcially-recognized racial ancestries, a cat-

egory that made up 1.8% of household heads in our Census 2000 sample and totaled 95,410 respondents in our Census 2000

sample and 106,977 respondents in our Census 1990 sample.

Our intent is not to reify the existing categories of racial distinction in the US by selecting only individuals who report

ofﬁcially-recognized racial ancestries. We choose these individuals because their responses to the ancestry question reveal

recognition of a multiracial ancestry that could be reported using the standard racial categories available in the US Census.

Our coding of racial ancestry certainly leads to some measurement error because some individuals who we believe are

providing a multiracial ancestry are not intending to do so. For example, an African-ancestry British immigrant to the United

States might report both an ‘‘English’’ and an ‘‘African American’’ ancestry. According to our approach, this individual would

be coded as reporting a multiracial ancestry, when they are actually using the ancestry response to report both a racial iden-

tity (black) and their national origins (English). While these sorts of measurement errors are unavoidable, we believe that the

number of such errors is likely to be small and concentrated among the foreign-born. Furthermore, the bias that is generated

with such measurement errors is fairly clear. These measurement errors will reduce the proportion of individuals with a

multiracial ancestry who report more than one race and increase the proportion of the same individuals who will report sin-

gle races.

It might seem unusual that we exclude individuals who report a Hispanic ancestry from our sample given the growth of

the Hispanic population in the US and the increasing academic and popular interest in Hispanic groups. This exclusion is

largely driven by a data limitation inherent in the way Hispanicity is recorded in the US Census. Because Hispanicity is asked

Table 1

Distribution of racial ancestry reporting for heads of household, Census 1990 and 2000. Source: IPUMS 5% samples of the 1990 and 2000 US Censuses.

Response 1990 2000

Number % Number %

No ancestry 410,143 8.9 992,042 18.8

One ancestry 2,945,582 64.1 3,179,084 60.3

Ofﬁcially-recognized race 2,412,937 81.9 2,298,142 72.3

Unofﬁcial race 279,191 9.5 460,245 14.5

Racially uninterpretable ancestry 253,454 8.6 420,697 13.2

Two ancestries 1,237,675 26.9 1,102,145 20.9

Same ofﬁcially-recognized race 1,094,807 88.5 957,691 86.9

Same unofﬁcial race 1121 0.1 1835 0.2

Different ofﬁcially recognized races 106,977 8.6 95,410 8.7

Different unofﬁcial races 696 0.1 2075 0.2

Ofﬁcially recognized and unofﬁcial 33,992 2.7 44,941 4.1

Notes: Results are unweighted. Ofﬁcially recognized race refers to the ﬁve racial categories recognized by the US Ofﬁce of Management and the Budget

(white, black, American Indian/Alaska Native, Asian, and Paciﬁc Islander).

Although ‘‘Asian Indian’’ is technically a racial category on the US Census, there is ample evidence that the racial identiﬁcation of South Asians in the US is

quite ambiguous and that, for most Americans, ‘‘Asian’’ means East Asian (Kibria 1996; Morning 2001; Shankar and Srikanth, 1998).

This includes individuals who reported a US State as an ancestry (i.e. ‘‘Texan’’) and individuals who report ‘‘Canadian’’ as an ancestry (but not ‘‘French

Canadian’’).

This category might seem to capture our interest in multiracial individuals, but given that 86% of people who report this ancestry reported being white-

alone in Census 2000, it is more likely that this response is primarily used by white individuals who can claim various and diverse European ancestries.

A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

503

Author's personal copy

as a separate question on the US Census, part-Hispanic ancestry respondents were never forced to make a single-race/multi-

ple race decision in completing their Census forms. They did not face the same options in confronting the Census form itself

as other respondents who indicate a multiracial ancestry. The appropriate parallel to our methodology here would be to ana-

lyze answers to the Hispanicity question among those individuals who report a part-Hispanic ancestry, but such an analysis

is beyond the scope of this article (see Emeka and Vallejo (2010) for an analysis of this type).

Within our sample of multiracial ancestry respondents, we have ten different multiracial ancestry groups. Table 2 shows

the number of HH’s within each group. Although the overall totals for each Census are around 100,000 respondents, the

white/American Indian group alone accounts for about 90% of the total sample in 1990 and 85% of the total sample in

2000. The sample sizes for the remaining groups are thus considerably smaller.

For purposes of our analysis we exclude part-Paciﬁc Islander-ancestry individuals and American Indian/Asian ancestry

individuals, leaving us with ﬁve multiracial ancestry groups (white/black, white/American Indian, white/Asian, black/Amer-

ican Indian, black/Asian). The results for Paciﬁc Islander-ancestry individuals overwhelmingly reﬂect the racial atmosphere

of Hawaii, which is considerably different from the racial atmosphere of the mainland United States. There is also a potential

methodological problem with identifying individuals of native Hawaiian ancestry. In some cases, individuals who report

‘‘Hawaiian’’ as their ancestry may have been indicating a native Hawaiian ancestry. However, state names were also reported

as ancestries for other states without any implicit ethnic identiﬁcation. In two cases, moreover, the part-Paciﬁc Islander

ancestry groups are too small to produce reliable statistics (as is the American Indian/Asian group). While we think exam-

ination of these understudied groups would be beneﬁcial in future studies, due to space constraints we limit our discussion

here to the ﬁve largest multiracial ancestry groups formed by the sharper racial boundaries of the mainland United States.

3.3. Methods

Our analysis proceeds in two steps. First, we analyze the distribution of race reporting for the ﬁve multiracial ancestry

groups. We analyze this distribution separately in the 2000 and 1990 samples, in order to compare the results when the mul-

tiracial option is not available with the results when the multiracial option is available. Second, we then compare these dis-

tributions of race reporting to distributions observed for the same multiracial group by scholars using other methods to

identify multiracial populations. Throughout the remainder of this paper, we use the weighted frequencies for both our

Census 1990 and Census 2000 samples to determine the distribution of race reporting.

For cases in the Census 2000 data where a respondent checked more than one box in response to the race question, we

treat such respondents as identifying as multiracial. In almost all cases, multiracial-ancestry individuals who selected more

than one box chose at least the two races that corresponded to their multiracial ancestry. The percentage of respondents who

selected more than one race but did not choose races consistent with their multiracial ancestry ranged from 4.1% for white/

black ancestry respondents to 0.6% for white/AIAN ancestry respondents. A substantial proportion of these individuals in all

cases checked one of their racial ancestries and ‘‘some other race.’’ We exclude these inconsistent multiracial respondents

from the numbers presented in the analysis below.

We primarily use graphical techniques to display our ﬁndings, but the full distributions of race reporting for each mul-

tiracial ancestry group (including ones not analyzed here) are available in Tables A1 and A2 of the data appendix, for 1990

and 2000 respectively.

Table 2

Distribution of multiracial ancestry heads of household by type, Census 1990 and 2000. Source: IPUMS 5% samples of the 1990 and00 US Censuses.

Combination 1990 2000

Number % Number %

White/American Indian 96,206 89.9 80,133 84.0

Black/American Indian 4015 3.8 5336 5.6

White/Asian 2651 2.5 4362 4.6

White/black 2062 1.9 2865 3.0

Black/Asian 207 0.2 433 0.5

American Indian/Asian 121 0.1 239 0.3

White/Paciﬁc Islander 764 0.7 881 0.9

Black/Paciﬁc Islander 15 0.0 27 0.0

American Indian/Paciﬁc Islander 51 0.0 59 0.1

Asian/Paciﬁc Islander 885 0.8 1075 1.1

Notes: Results are unweighted. References to ‘‘American Indian’’ incorporate Alaska Natives. Italics designate groups excluded from later analysis.

Although we do not examine the results for these groups here, we do show the distribution of race reporting for all groups in 1990 and 2000 in Appendix

Tables 1 and 2, respectively. Based on the results for other groups presented below, it appears that Paciﬁc Islander ancestry functions in some ways similar to

American Indian ancestry because very few part-Paciﬁc Islander-ancestry individuals identiﬁed as Paciﬁc Islander alone. At the same time, however, Paciﬁc

Islander ancestry also functions like Asian ancestry in that multiracial identiﬁcation is the modal category for all part-Paciﬁc Islander ancestry groups. Because

of the data issues raised above and the small sample size of some groups, however, we consider these results to be highly tentative.

504 A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

Author's personal copy

4. Results

4.1. Multiracial groups’ race reporting in 2000 and 1990

Fig. 1 shows the distribution of race reporting for each of our multiracial ancestry groups in Census 2000. The most obvi-

ous feature of these distributions is how different they are from one another. Despite the seeming heterogeneity, we argue

that the results follow three simple nested rules. First, part-Asian-ancestry groups are most likely to identify themselves as

multiracial. Multiple-race responses were the modal category for both white/Asian ancestry (49.2%) and black/Asian ances-

try (64.1%) respondents. Second, among those groups with no Asian ancestry, part-black ancestry groups are most likely to

identify themselves as black alone, consistent with a pattern of hypodescent. For white/black individuals, the modal category

is black alone (52.7%) and few respondents (6.6%) chose white alone. Similarly, the majority of black/American Indian ances-

try individuals identify themselves as black alone (67.7%). Third, the remaining group of white/American Indian ancestry

individuals is most likely to select white alone (85.1%), consistent with a pattern of hyperdescent.

What happens when multiracial identity is not an option? In Fig. 2, we show the distribution of race reporting in the

1990 census in which respondents were only allowed to identify with one race. For the three non-Asian-ancestry groups,

the patterns of hypodescent and hyperdescent observed in 2000 were similar in 1990: both white/black (82.1%) and black/

American Indian (93%) ancestry individuals overwhelmingly identify as black, while white/American Indian ancestry

Fig. 1. Distribution of race reporting by multiracial ancestry heads of household, Census 2000. Source: IPUMS 5% sample of the 2000 US Census. Notes: All

frequencies are weighted. ‘‘AIAN’’ stands for ‘‘American Indian and Alaska Native.’’

Fig. 2. Distribution of race reporting by multiracial ancestry heads of household, Census 1990. Source: IPUMS 5% sample of the 1990 US Census. Notes: All

frequencies are weighted. ‘‘AIAN’’ stands for ‘‘American Indian and Alaska Native.’’

A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

505

Author's personal copy

individuals overwhelmingly identify as white (94.0%). But the multiple-response option that was the modal response for

the two Asian-ancestry groups in 2000 was not available in 1990. In both of these cases, the majority of individuals iden-

tiﬁed with the non-Asian race: 55.9% of white/Asian respondents identiﬁed as white and 73.9% of black/Asian respondents

identiﬁed as black. In other words, under these restrictive conditions—similar to those that have prevailed for much of US

history—where multiracial recognition is disallowed, Asian heritage seems to function like American Indian ancestry.

Asians with African ancestry are subject to hypodescent, whereas those with European origins follow a hyperdescent pat-

tern. This ﬁnding lends support to DeGenova’s (2006) contention that in many crucial respects, the racial positioning of

American Indians has served as a template for the racialization of Asians (and Hispanics) in the United States (see also

Snipp, 2000, 2003).

Perhaps the most striking feature of the results for 1990 is the apparent lack of variability in identiﬁcation within each

ancestry group. In four of the ﬁve groups, the modal category accounts for 70% or more of the observations. Only in the case

of the white/Asian ancestry group does there appear to be less agreement on the ‘‘correct’’ category. Moreover, very few mul-

tiracial ancestry respondents rejected the standard racial categories available to them. Across all groups and on both the

1990 and 2000 censuses, the proportion of respondents either checking ‘‘other’’ or leaving the race question blank was very

small. In general these proportions were smaller in 2000 than in 1990, suggesting that the ability to identify with more than

one race in 2000 encouraged some multiracial-ancestry individuals who would have otherwise rejected the race question

to respond. Together, these results suggest a high degree of internalization of group conventions concerning racial

self-classiﬁcation under restrictive conditions, as well as acceptance of the norm of racial self-classiﬁcation.

4.2. Discussion of 1990 and 2000 multiracial reporting: a new regime for Asian ancestry

These three distinct regimes of racial identiﬁcation that we discern—that is, multiraciality for mixed-race people of Asian

descent, hypodescent for all others with African ancestry, and hyperdescent for the remaining white/American Indian

group—reﬂect different histories of racialization for people of Asian, African, European, and Native American origin. We ﬁnd

clear evidence of traditional one-drop classiﬁcation still applying to groups with black ancestry, as well as the legacy of

hyperdescent affecting the self-identiﬁcation of people who report American Indian and European origins. These patterns

are only fully detectable when using a method such as ours that captures individuals’ acknowledgement of historical (or gen-

erationally distant) multiracial ancestry.

In contrast, our ﬁndings on the racial identiﬁcation of Asian-ancestry groups show a striking departure from the long-

standing patterns of hypodescent and hyperdescent that have traditionally characterized the classiﬁcation of other multira-

cial groups in the United States. Unlike groups with some combination of white, black, and/or American Indian ancestry,

which prefer single-race labels even when—as on the 2000 census—multiple-race reporting is an option, part-Asian individ-

uals favor marking more than one race. We suggest several potential explanations, even though they are not testable with

our data.

One possible explanation is simply that not enough time has passed for rigid racial-assignment rules to have formed for

this group since Asian ancestry typically implies a generationally recent immigration to the United States; about 73% of Asian

Americans were ﬁrst or second generation in 1998–2000 (Yang, 2006). In that case, the current tendency toward multiracial

identiﬁcation for people of partial Asian descent may simply be a temporary stage in a process toward an eventual hyper- or

hypodescent regime.

Alternatively, it may be the case that Asian-origin population growth has taken place largely after the period in which

strict racial-categorization norms were most likely to be formed. The hypodescent and hyperdescent rules governing the off-

spring of what we have called the US’s ‘‘original’’ races were very much instruments of racial oppression that took shape in

an era of the most extreme political, economic, and social discrimination. The multiracial Asian population may simply have

come into being in a period in which such strict classiﬁcatory tools for racial governance no longer seemed as necessary. In

this case, we would not expect strong single-race assignment norms to form with respect to part-Asian people anytime in the

foreseeable future.

The relative generational recency of the Asian-American population may mean that a process of immigrant incorporation

and/or assimilation is at work, one that is quite distinct from the modes of incorporation of the United States’ African,

European, or American Indian populations. Although it is not clear why recent immigrant origins would predispose people

toward the recognition of multiple origins, it may be that migration entails an acknowledgement of and sense of connection

to two social worlds—i.e. the home country and the country of destination—that lends itself to the recognition of racial

mixture as well. This may be all the more true in our era of globalization, when technological advances in transportation

and communications make such transnational ties more feasible.

A corollary of the Asian-American population’s relative generational recency is that the part-Asian respondents in our

sample are more likely than others to be the biracial children of single-race parents as opposed to the descendants of inter-

racial unions further back in the family tree. Multiraciality may be more salient for such biracial offspring than it is for people

whose mixed-race heritage is more generationally distant.

We reiterate that these are speculative accounts whose merit we cannot test here. However, given the consistency of our

ﬁnding about part-Asian groups’ preference for multiracial identiﬁcation, juxtaposed with the scarcity of existing literature

explaining this phenomenon, we hope to offer some original contributions to an eventual empirical research agenda in this

area.

506 A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

Author's personal copy

4.3. Comparing multiracial groups’ race reporting across different analytic methods

We now turn to a comparison of our results with results obtained by other authors using different techniques to identify a

multiracial population. First we compare our Census 2000 results to the results of Roth (2005) and Brunsma (2005), both of

which used parental race to identify a multiracial population. With the Early Childhood Longitudinal Study (ECLS), Brunsma

had the advantage of being able to identify the race of biological parents even when they were non-resident, thanks to the

survey design of the ECLS. However, the children in his sample were younger and in a narrower age range (4–6 years of age)

than those in Roth’s Census 2000-based study.

Fig. 3 compares our results for Census 2000 to the results of Roth (2005) and Brunsma (2005). We exclude Brunsma’s

results for black/American Indian and black/Asian ancestry individuals, due to small sample sizes (17 and 30 children,

respectively). Despite the differences in design noted above, the results of Brunsma (2005) and Roth (2005) are quite similar

for all groups. What is equally remarkable is just how different our results are from both of these studies for white/black,

white/American Indian and black/American Indian ancestry individuals. For white/black as well as black/American Indian

individuals we ﬁnd a far stronger pattern of hypodescent. For example, according to both Brunsma and Roth, the modal

Fig. 3. Comparison of race reporting among Census 2000 respondents in our study, Roth (2005), and Brunsma (2005).

A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

507

Author's personal copy

category for black/white people is a multiracial designation, while in our sample, the modal category is black alone and

considerably fewer respondents are reported as either white alone or multiracial. Although a substantial number of respon-

dents in our sample (32.6%) still reported being multiracial, we believe that these results should serve as a caution on any

declarations that the ‘‘one-drop rule’’ is no longer operative for individuals of part-black ancestry in the United States.

Similarly, our results for white/American Indian ancestry individuals reﬂect a stronger pattern of hyperdescent than do

the results of Brunsma and Roth. We ﬁnd far more white/American Indian ancestry individuals who are identiﬁed as white

alone, and far fewer who are identiﬁed as multiracial or American Indian alone. The reason for this discrepancy is fairly clear.

The biracial children of parents where one parent identiﬁes as American Indian and the other as white are only a tiny fraction

of individuals who recognize a white/American Indian multiracial ancestry. The broader population of individuals who

recognize both white and American Indian ancestry are far more likely to follow the rules of hyperdescent and select white

only as their race. Put simply, analyzing biracial children is a poor method for understanding part-American Indian multi-

raciality. More generally, a focus on biracial children alone underestimates the continuing weight of traditional hypo- and

hyperdescent conventions in the United States.

Fig. 4. Comparison of race reporting among Census 1990 respondents in our study, Campbell (2006) and Qian (2004).

508 A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

Author's personal copy

It is perhaps not surprising then that our results for part-Asian ancestry individuals are similar to the results for the other

two studies. Unlike the other two groups, who have a long history in the United States, our method and the parental method

are largely congruent for the newer part-Asian ancestry population. Most of the part-Asian ancestry individuals that we

identify probably are the biracial children of white/Asian or black/Asian union, given the recentness of most Asian migration

to the United States.

We can also compare our 1990 census results to two studies that examined a forced single-race response. Qian (2004)

used a method similar to Roth (2005) to analyze the racial identiﬁcation of biracial children in Census 1990. Campbell

(2006) used the Race and Ethnicity Supplement of the 1995 Current Population Survey to identify multiracial respondents.

The Supplement ﬁelded four different panels, each of which posed questions about race and ethnicity in a slightly different

manner. However, at some point each panel asked respondents to select one race, and elsewhere gave them the opportunity

to report more than one race. Qian (2004) only examined the race reporting for part-white biracial children, while Campbell

(2006) only examined the race reporting of part-black respondents, and due to a small sample size, she excludes black/Asian

ancestry respondents from analysis.

Fig. 4 compares our results for Census 1990 to the results of Qian (2004) and Campbell (2006). Once again, our results are

more sensitive to the inﬂuence of traditional conventions on racial self-reporting. In our study, white/black individuals show

a stronger pattern of hypodescent, and white/American Indian individuals a stronger pattern of hyperdescent, than they do

in the research of Qian (2004) and Campbell (2006), while the results are similar for white/Asian individuals.

4.4. Discussion of methodological comparisons: the relevance of an ancestry-based approach

In comparing the results obtained using different methods for identifying the multiracial population, one might raise the

critique that our analysis of an ancestry-based sample is largely retrospective, while methods that focus on biracial children

are more forward-looking.

There are two reasons that we believe such a critique is misplaced. First, aside from those who checked ‘‘some other

race,’’ American Indian was the most frequent box on the race question to be checked in combination with other boxes

(Grieco and Cassidy, 2001). From a historical perspective, this result is not surprising because the logic of hyperdescent

has allowed acknowledged American Indian ancestry to diffuse across a broad population. Yet this population is not

well-represented among biracial children because the generational placement of multiraciality is relatively distant for

most individuals of part-American Indian ancestry. If, as DeGenova (2006) suggests, American Indians offer a template

for thinking about the racialization of ‘‘new’’ groups, then ignoring part-American Indian multiracial-ancestry individuals

creates a very large ‘‘blind spot’’ for understanding the future of racial boundaries in the United States (Snipp, 2000,

2003).

Second, studies of biracial children over-emphasize an unstable historical moment in time. The biracial ‘‘baby boom’’

has certainly led to substantial growth in the biracial population, and presumably future increases in intermarriage will

continue to contribute to its growth. However, as these biracial children grow to maturity, ﬁnd their own partners, and

begin to have children themselves, the salience of interracial mixing in the ‘‘ﬁrst’’ generation will decline and the family

genealogies of individuals who might be considered the inheritors of this biracial baby boom will become increasingly

complex and variable (Bratter, 2007). While no study can claim to accurately represent this future world, we do believe

that our method in some ways better captures a population comparable to this future population than do studies limited

to biracial children.

5. Conclusions

In response to the question of how mixed-race people in the United States prefer to identify themselves in racial terms,

our research reveals two sets of patterns. One can be referred to as the ‘‘historical’’ regime whereby multiracial people com-

bining European, African, and/or American Indian ancestry were assigned to monoracial groups. Although we describe it as a

historical convention, our data suggest it still has real purchase today. The second contemporary pattern of multiracial iden-

tiﬁcation we discern might be called simply the ‘‘new’’ or even ‘‘immigrant’’ regime whereby a multiple-race label is

selected. This convention seems to dominate among all the groups of partial Asian ancestry.

We believe that these very clear empirical ﬁndings have important implications for the study of multiracial identiﬁcation.

We can sum them up in three catch-phrases: ‘‘history matters,’’ ‘‘ancestry matters,’’ and ‘‘data matter.’’

5.1. History matters

In our attempts to explain different patterns of multiracial classiﬁcation across ancestry groups, we have emphasized the

role that traditional practices and norms play in guiding individual self-identiﬁcation decisions. In the cases of the long-

standing hypo- and hyperdescent customs, we have been able to draw on a well-established literature concerning the social,

economic, political and legal factors that gave rise to these regimes of single-race assignment. But in the case of our unex-

pectedly strong and consistent ﬁnding that part-Asian groups tend toward multiple-race identiﬁcation, we have had to

broaden the scope of historical and contemporary experiences and conditions that might play a role. In this connection,

A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

509

Author's personal copy

we have considered factors as diverse as immigration timing, the lived experience of migration, and exposure to period-spe-

ciﬁc forms of institutional racial discrimination. None of these hypotheses may prove to have any merit, but the evidence

that historical norms concerning the classiﬁcation of the United States’ ‘‘original’’ mixed-race groups still endure suggests

to us that when contemplating the identiﬁcation of part-Asian individuals, elements of the broader societal context, past

and present, must be taken into account.

5.2. Ancestry matters

A corollary of the claim that ‘‘history matters’’ is that the speciﬁcity of particular combinations of racial ancestry matter as

well in the study of contemporary mixed-race identiﬁcation. Because different groups have been racialized differently—that

is, conceived of in different ways, ascribed different characteristics, and subjected to varying forms of social stratiﬁcation—

different racial heritages signal distinct trajectories of inclusion and exclusion. As a result, no one-size-ﬁts-all template for

multiracial self-identiﬁcation is likely to apply across different mixed-ancestry groups.

5.3. Finally, data matter

As noted above, there is no single multiracial population; instead, different methods and assumptions produce different

portraits of multiracial America. Our research shows that selecting a population that is broader in the generational scope of

its multiracial ancestry reveals important patterns and characteristics that have been overlooked by other methods. We do

not argue that our strategy is the only one of value, but rather that it provides insights not provided by others. For example,

not only are the historical conventions of hyper- and hypodescent more evident in our sample, but their robustness throws

into sharper relief the new regime of multiple-race identiﬁcation that emerges among part-Asian groups. We believe that

this approach offers a more thorough and more historically-grounded examination of mixed-race identiﬁcation than studies

that examine the racial classiﬁcation of biracial children.

A frequent motive for studying the mixed-race population is the belief that its classiﬁcation is a harbinger of racial atti-

tudes—and demographic shifts—to come. The identiﬁcation patterns of multiracial people can be read as a signal of prevail-

ing or changing beliefs about the nature of racial differences in general and the boundaries of speciﬁc races in particular.

Hence the intense interest, for example, in whether or not the one-drop rule seems headed for extinction (e.g. Roth,

2005). The potential for shifts in our deﬁnitions of who belongs to which race in turn raises the prospect of changes in

the racial composition of our nation that are independent of the usual demographic variables of fertility, mortality, and

migration. New developments in the way we categorize blacks, whites, or any other group could have a signiﬁcant impact

on the demographic portrait that we draw of the nation. Recent predictions have included the scenario of a ‘‘beige majority’’

(Lind, 1998) where mixed-race people, whites, Asians, and Hispanics come to be understood as making up a kind-of-white,

‘‘beige’’ majority that is distinct from darker-skinned or ‘‘black’’ America. (See also Bonilla-Silva, 2004; Gans, 1979; Sanjek,

1994; Warren and Twine, 1997; Yancey, 2003 for variations on this theme.)

Our research suggests a cohabitation of historical single-race and contemporary multiple-race identiﬁcation regimes that

does not augur a burgeoning of the self-acknowledged multiracial population in the foreseeable future. Such predictions

however are notoriously difﬁcult to make. To give them more solid backing, social scientists will have to take a closer look

at how Americans think about their mixed-race ancestry: the kinds of backgrounds they count as multiracial, how they pre-

fer to describe that multiracial heritage, and the considerations they take into account when identifying themselves and

others in terms of race.

Appendix A

See Tables A1 and A2.

Table A1

Percentage distribution of race reporting for multiracial ancestry heads of household, Census 1990. Source: IPUMS 5% samples of the 1990 US Census.

Race Ancestry

White/Asian Black/Asian White/black Black/AIAN White/AIAN AIAN/Asian White/PI Black/PI Asian/PI AIAN/PI

White 55.9 0.5 11.5 1.0 94.0 35.0 45.8 0.0 7.2 35.3

Black 0.3 73.9 82.1 93.0 0.5 5.8 0.4 80.0 0.0 0.0

AIAN 0.5 0.0 1.0 4.3 4.9 27.5 0.0 0.0 0.4 23.5

Asian 40.8 20.8 0.0 0.1 0.0 29.2 44.7 13.3 81.5 29.4

Pac Islander 0.1 0.5 0.0 0.0 0.0 0.0 6.5 6.7 2.5 9.8

Other 0.5 1.0 4.1 0.7 0.2 1.7 0.5 0.0 1.1 0.0

None 1.8 3.4 1.3 0.8 0.3 0.8 2.1 0.0 6.9 2.0

Notes: Results are weighted. ‘‘AIAN’’ stands for ‘‘American Indian and Alaska Native.’’ Columns may not sum to 100% due to rounding error.

510 A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

Author's personal copy

References

Bonilla-Silva, Eduardo, 2004. From bi-racial to tri-racial: towards a new system of racial stratiﬁcation in the USA. Ethnic and Racial Studies 27 (6), 931–950.

Bratter, Jenifer, 2007. Will ‘multiracial’ survive to the next generation? The racial classiﬁcation of children of multiracial parents. Social Forces 86 (2), 821–

849.

Brittingham, Angela, de la Cruz, G. Patricia, 2004. Ancestry 2000. Census 2000 Brief C2KBR-35, June 2004. U.S. Census Bureau.

Brunsma, David L., 2005. Interracial families and the racial identiﬁcation of mixed-race children: evidence from the early childhood longitudinal study.

Social Forces 84, 1131–1157.

Campbell, Mary E., 2006. Thinking outside the (black) box: measuring black and multiracial identiﬁcation on surveys. Social Science Research 36, 921–944.

Chew, Kenneth S.Y., Eggebeen, David J., Uhlenberg, Peter R., 1989. American children in multiracial households. Sociological Perspectives 32 (1), 65–85.

Davis, Floyd James, 1991. Who is Black? One Nation’s Deﬁnition. Pennsylvania State University Press, University Park, PA.

DeGenova, Nicholas, 2006. Introduction: Latino and Asian Racial Formations at the Frontier of US Nationalism. In: DeGenova, Nicholas (Ed.), Racial

Transformations: Latinos, Asians Remaking the United States. Duke University Press, Durham, NC, pp. 1–20.

Emeka, Amon, Vallejo, Jody Aguis, 2010. I’m not Latino, But My Grandma Was: Non-Latino Identities among Latin American Descendents in the US. Paper

Presented at the USC Center for the Study of Immigrant Integration Faculty Seminar.

Eschbach, Karl, 1993. Changing identiﬁcation among American Indians and Alaska natives. Demography 30 (4), 635–652.

Eschbach, Karl, 1995. The enduring and vanishing American Indian: American Indian population growth and intermarriage in 1990. Ethnic and Racial

Studies 18 (1), 89–108.

Eschbach, Karl, Supple, Khalil, Snipp, C.Matthew, 1998. Changes in racial identiﬁcation and the educational attainment of American Indians, 1970–1990.

Demography 35 (1), 35–43.

Gans, Herbert, 1979. Symbolic ethnicity: the future of ethnic groups and cultures in America. Ethnic and Racial Studies 2 (1), 1–19.

Garroutte, Eva Marie, 2003. Real Indians: Identity and the Survival of Native America. University of California Press, Berkeley, CA.

Goldstein, Joshua R., Morning, Ann J., 2000. The multiple-race population of the United States: issues and estimates. Proceedings of the National Academy of

Sciences 97 (11), 6230–6235.

Grieco, Elizabeth M., Cassidy, Rachel C., 2001. Overview of Race and Hispanic Origin. Edited by US Census Bureau, US Department of Commerce,

Washington, DC.

Gullickson, Aaron, 2006. Black/white interracial marriage trends, 1850–2000. Journal of Family History 31 (3), 289–312.

Harris, David R., Sim, Jeremiah Joseph, 2002. Who is multiracial? Assessing the complexity of lived race. American Sociological Review 67 (August), 614–

627.

Herman, M., 2004. Forced to choose: some determinants of racial identiﬁcation among multiracial adolescents. Child Development 75 (3), 730–748.

Hitlin, Steven, Brown, J.Scott, Elder Jr., Glenn H., 2006. Racial self-categorization in adolescence: multiracial development and social pathways. Child

Development 77, 1298–1308.

Hollinger, David A., 2003. Amalgamation and hypodescent: the question of ethnoracial mixture in the history of the United States. The American Historical

Review 108 (5), 1363–1390.

Kibria, Nazli, 1996. Not Asian, black or white? Reﬂections on South Asian American racial identity. Amerasia Journal 22 (2(Spring)), 77–86.

Korgen, Kathleen Odell, 1998. From Black to Biracial: Transforming Racial Identity among Americans. Praeger, Westport, CT.

Korgen, Kathleen Odell, 2010. Multiracial Americans and Social Class: The Inﬂuence of Social Class on Racial Identity. Routledge, New York.

Lieberson, Stanley, Waters, Mary C., 1988. From Many Strands: Ethnic and Racial Groups in Contemporary America. Russell Sage Foundation, New York.

Liebler, Carolyn A., 2010. A group in ﬂux: multiracial American Indians and the social construction of race. In: Korgen, Kathleen Odell (Ed.), Multiracial

Americans and Social Class. Routledge Press, New York, pp. 131–144.

Lind, Michael, 1998. The Beige and the Black. The New York Times Magazine. pp. 38–39.

Morning, Ann, 2000. Who is multiracial? Deﬁnitions and decisions. Sociological Imagination 37 (4), 209–229.

Morning, Ann, 2001. The racial self-identiﬁcation of South Asians in the United States. Journal of Ethnic and Migration Studies 27 (1), 61–79.

Morning, Ann, 2003. New faces, old faces: counting the multiracial population past and present. In: DeBose, Herman, Winters, Loretta (Eds.), New Faces in a

Changing America: Multiracial Identity in the 21st Century. Sage, Thousand Oaks, CA, pp. 41–67.

Nagel, Joane, 1996. American Indian Ethnic Renewal: Red Power and the Resurgence of Identity and Culture. Oxford University Press, New York.

Ofﬁce of Management and Budget, 1997. Revisions to the Standards for the Classiﬁcation of Federal Data on Race and Ethnicity, Federal Register.

Washington, DC.

Penner, Andrew M., Saperstein, Aliya, 2008. How social status shapes race. Proceedings of the National Academy of Sciences 105 (50), 19628–19630.

Qian, Zhenchao, 2004. Options: racial/ethnic identiﬁcation of children of intermarried couples. Social Science Quarterly 85 (3), 746–766.

Renn, Kristen A., 2004. Mixed Race Students in College: The Ecology of Race, Identity, and Community on Campus. State University of New York Press,

Albany, NY.

Rockquemore, Kerry Ann, Brunsma, David L., 2002. Beyond Black: Biracial Identity in America. Sage, Thousand Oaks, CA.

Root, Maria P.P. (Ed.), 1992. Racially Mixed People in America. Sage Publications, Newbury Park, CA.

Root, Maria P.P. (Ed.), 1996. The Multiracial Experience: Racial Borders as the New Frontier. Sage Publications, Thousand Oaks, CA.

Table A2

Percentage distribution of race reporting for multiracial ancestry heads of household, Census 2000. Source: IPUMS 5% samples of the 1990 US Census.

Race Ancestry

White/

Asian

Black/

Asian

White/

black

Black/

AIAN

White/

AIAN

AIAN/

Asian

White/

Black/

Asian/

AIAN/

White 31.2 0.7 6.6 0.3 85.1 13.8 31.1 0.0 2.3 17.2

Black 0.1 31.8 52.7 67.7 0.6 7.9 0.0 33.3 0.0 6.9

AIAN 0.0 0.0 0.2 0.7 3.3 10.0 0.2 0.0 0.0 8.6

Asian 13.9 0.9 0.0 0.0 0.0 6.3 10.8 0.0 12.0 10.3

Pac Islander 0.1 0.0 0.0 0.0 0.0 0.0 7.5 0.0 1.3 1.7

Multiracial 49.2 64.1 32.6 28.7 9.9 52.7 47.3 66.7 82.0 41.4

Multiracial

(inconsistent)

2.3 1.6 4.1 2.0 0.6 9.2 2.2 0.0 0.3 10.3

Other 2.7 0.7 2.8 0.3 0.2 0.0 0.0 0.0 1.5 1.7

None 0.4 0.2 0.9 0.3 0.3 0.0 0.9 0.0 0.8 1.7

Notes: Results are weighted. ‘‘AIAN’’ stands for ‘‘American Indian and Alaska Native.’’ Respondents who check more than one box, but did not check at least

both of their reported racial ancestries were recorded as ‘‘inconsistent’’ multiracial reporters. Columns may not sum to 100% due to rounding error.

A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512

511

Author's personal copy

Roth, Wendy D., 2005. The end of the one-drop rule? Labeling of multiracial children in black intermarriages. Sociological Forum 20 (1), 35–67.

Ruggles, Steven, Sobek, Matthew, 1997. Integrated Public Use Microdata Series: Version 2.0. Historical Census Projects, University of Minnesota,

Minneapolis.

Saenz, Rogelio, Hwang, Sean-Shong, Aguirre, Benigno E., Anderson, Robert N., 1995. Persistence and change in Asian identity among children of intermarried

couples. Sociological Perspectives 38 (2), 175–194.

Sanjek, Roger, 1994. Intermarriage and the future of races in the United States. In: Gregory, Steven, Sanjek, Roger (Eds.), Race. Rutgers University Press, New

Brunswick, pp. 103–130.

Shankar, Lavina Dhingra, Srikanth, Rajini (Eds.), 1998. A Part, Yet Apart: South Asians in Asian America. Temple University Press, Philadelphia.

Snipp, Matthew, 1989. American Indians: The First of this Land. Russell Sage Foundation, New York.

Snipp, C. Matthew, 2000. Clues to the future of other groups from the American Indian experience. In: Presented at Conference on Multiraciality, September

22–23.

Snipp, C. Matthew, 2003. Racial measurement in the American census: past practices and implications for the future. Annual Review of Sociology 29

(August), 563–588.

Spencer, Rainier, 2006. Challenging Multiracial Identity. Lynne Rienner, Boulder, CO.

Tafoya, Sonya M., Johnson, Hans, Hill, Laura E., 2004. Who Chooses to Choose Two? Russell Sage Foundation and Population Reference Bureau, New York and

Washington, DC.

Twine, France Winddance, 1996. Brown skinned white girls: class, culture and the construction of white identity in suburban communities. Gender, Place

and Culture 3 (2), 205–224.

Unrau, Wiliam E., 1989. Mixed-Bloods and Tribal Dissolution: Charles Curtis and the Quest for Indian Identity. University Press of Kansas, Lawrence, KS.

US Census Bureau, 2008. Table 6. Percent of the Projected Population by Race and Hispanic Origin for the United States: 2010 to 2050.’’ Edited by Sex US

Interim Projections by Age, Race, and Hispanic Origin: 2000–2050, Population Division.

Warren, Jonathan W., Twine, France Winddance, 1997. White Americans, the new minority? Non-blacks and the ever-expanding boundaries of whiteness.

Journal of Black Studies 28 (2), 200–218.

Williamson, Joel, 1980. New People: Miscegenation and Mulattoes in the United States. Louisiana State University Press, Baton Rouge.

Wilson, Terry P., 1992. Blood quantum: native American mixed bloods. In: Root, Maria P.P. (Ed.), Racially Mixed People in America. Sage, Newbury Park, CA,

pp. 108–125.

Wolfe, Patrick, 2001. Land, labor, and difference: elementary structures of race. The American Historical Review 106 (3), 866–905.

Xie, Yu, Goyette, Kimberly, 1997. The racial identiﬁcation of biracial children with one Asian parent: evidence from the 1990 census. Social Forces 76 (2),

547–570.

Yancey, George, 2003. Who is White? Latinos, Asians, and the New Black/Nonblack Divide. Lynne Rienner Publishers, Boulder, CO and London.

Yang, Philip Q., 2006. Future prospects of Asian Americans. In: Pyong Gap Min (Ed.), Asian Americans: Contemporary Trends and Issues. Pine Forge Press,

Newbury Park, CA, pp. 292–316.

512 A. Gullickson, A. Morning / Social Science Research 40 (2011) 498–512