History, Place, and Racial Self-Representation in 21^st Century America

Abstract

How is a person’s racial self-representation related to the race history of the place in which he or she lives? We use Census Bureau data about race and ancestry to address this research question for two groups of people with mixed racial heritage: those reporting white and American Indian heritages, or reporting black and American Indian heritages. Links between history, place, and self-representation can be seen in geographic clustering for each race/ancestry response combination. We use multinomial logistic regression models to predict individuals’ race/ancestry responses (e.g., white with American Indian ancestry versus white and American Indian races) using measures of local race history and the area’s contemporary racial composition. Multivariate results highlight the relationship between a person’s identity claims and the history of the area, net of contemporary area composition. Future research should attend to the history of the place as a potential contributor to contemporary patterns.

1. Introduction

How is a person’s own story – his biography – related to the history of the society in which he lives? In his famous description of the “sociological imagination,” C. Wright Mills (1959) described the great promise of sociology as its ability to find links between biography and history. William H. Sewell Jr. (2005) amplified Mills’ proposition, saying that to fully understand an element of our current society one must know its historical context. Relatedly, geographers Agnew and Duncan (1989) draw on Foucault (1980:70) to call for increased inclusion of geographical place as an important concept in social science studies. In this research, we apply these ideas to draw connections between a place’s race relations history and the racial self-identifications of those living within that place.¹

Life course sociologists have fruitfully explored ways in which individuals’ life stories are impacted not only by their age, but also by period and cohort membership, which determine the historical context in which they enter each stage of life (e.g., Elder 1998). However, sociologists often fail to consider implications of place-specific history as an additional piece of sociohistorical context shaping life stories. Each community has its own history of cooperative and competitive relationships between groups and individuals (e.g., Crowe 2012; Lowery 2010). Place-specific history influences the social structures and normative interactions that develop in the place (Macintyre et al. 2002; Messner et al. 2005). Structural symbolic interactionists (e.g., Stryker 2002, 2008) and ecological systems theorists (e.g., Bronfenbrenner 1977) see social structures and interactions as important to a person’s sense of self. Thus place-specific history is likely to influence identity.

Racial identification of people with mixed racial heritage is relatively open to influence by nuanced patterns in interactions (Khanna and Johnson 2010; Rockquemore and Brunsma 2002; Song 2003), and the content of these interactions is likely to vary across places with differing racial histories (Brunsma 2006; Cornell and Hartmann 2007; Kana’iaupuni and Liebler 2005; Liebler 2010b; Lowery 2010). Yet previous research on racial identity and self-representation has neglected to include substantial measures of place history, instead using a few (if any) broad measures of the current racial context of the place. In this research, we conceptualize local race history as a structural framework in which people develop, maintain, and present their identities. We promote the idea that a consideration of local history can help researchers understand the personal identities that develop in the area.

For any person, the race(s) he or she ends up identifying with has real implications. A person’s race can impact their interaction experiences, interests, opportunities, health, and wealth, among other things. A person of white and American Indian heritage, for example, might identify as racially white, as racially American Indian, or as a person of two races. If he identifies as white, he is likely to experience the social world as a member of the white majority and he may be relatively unlikely to join an American Indian activist organization or social group, to know details of an American Indian culture or language, or to ever racially identify as American Indian (Liebler 2010a; Liebler, Bhaskar, and Rastogi 2014; Snipp 1989; Waters 1990). Conversely, a very different biography would unfold if he identified as a minority group member and were treated as such by society.

We leverage modern data with unprecedented race detail to explore the potential power of place in influencing how a person sees him or herself. We focus on people whose heritage includes groups that have been in the U.S. the longest – American Indians or Alaska Natives (“American Indians” here), whites, and blacks.² More specifically, to study whether individuals’ racial self-identification choices reflect the local history of race, we select two groups of people with mixed racial heritage from the nationally representative Census 2000 and 2006–2008 American Community Survey (ACS) data. One group reports both white and American Indian background, either by marking one race and the other ancestry³ or by marking both races; we use the term “white-American Indians” to describe people in this group. The second group parallels the first and reports black and American Indian heritage; we label this group “black-American Indians.” The centuries of interactions between whites, blacks, and American Indians – some amicable and some much less so – have created deep and varied place-specific race-related histories that could influence local area patterns in personal identification decisions.

This study thus represents a step toward understanding how place-specific history affects personal identity formation, though our aim is to identify patterned associations rather than advance strong causal claims. We begin by introducing our conceptual model, discussing some historical race-specific policies and practices in the U.S., and describing two locales to illustrate the types of social histories that might influence patterns in race/ancestry identities developed in these places. Then, we present our data, analyses, and results. These include population pyramids and national residential patterns for each group of people in the study. We also present multinomial logistic regression analyses modeling whether a person of mixed heritage will be reported⁴ to the government as American Indian race, white race, black race, or multiple races.

2. Conceptual model and prior research

2.1. Conceptual model

What role do past and current race-related policies and practices play in helping people of mixed heritage decide (consciously or unconsciously) which race or races they will claim for themselves? As our conceptual model in Figure 1 shows, we see the race history of an area as an important contributor to the formation and maintenance of personal racial identity, which is reflected in answers to survey questions about race. In other words, racial self-representation is heavily influenced by components of a person’s life story (their biography), which is shaped by the places in which this story develops (and, in turn, shapes the next layer of place history).⁵ All aspects of our conceptual model can change over a person’s life course (Carter, Dyer, and Sharma 2007; Doyle and Kao 2007; Harris and Sim 2002; Hitlin et al. 2006; Penner and Saperstein 2008; Price 2012). Thus the conceptual model shown in Figure 1 should be seen as a snapshot of a fluid situation.

Figure 1.

Conceptual model of links between life story and place-specific history

In this study, we measure racial self-representation through responses to the Census Bureau’s race question and ancestry question (both are shown in Appendix A). Reporting group affiliation through the race question has been shown to be quite different from identifying only an ancestral link (Snipp 1989). Race responses are often a sign of relatively deep attachments to the group while ancestry responses have been repeatedly shown to reflect substantially weaker identification (Gullickson and Morning 2011; Liebler 2010a; Snipp 1989; Waters 1990). Based on this prior research, we expect that a person who reports white (or black) race and American Indian ancestry is less attached to their American Indian heritage than one who reports American Indian race and white (or black) ancestry. People who report multiple races may have a multiracial identity (Rockquemore and Brunsma 2002; Root 1990, 1996) but, as is true for all people, a point-in-time response does not necessarily indicate whether or when one of these race groups takes precedence in their life or their response choices at other times (Liebler, Bhaskar, and Rastogi 2014; Liebler et al. 2014; Mowen and Stansfield 2015).

Many facets of a person’s experiences affect how they see and present their race(s) and ancestral origins. These include, among other things: what the person looks like (Doyle and Kao 2007; Herman 2004; Khanna 2004), what they’ve been told and by whom (Brunsma 2005; Childs 2006; Franco and Franco 2015; Roth 2005), their social connections and cultural exposure (Khanna 2004; Liebler 2001, 2004; Miville et al 2005), what other people assume about them (Campbell and Troyer 2007; Rockquemore and Brunsma 2002; Penner and Saperstein 2008), and their experiences with discrimination (Miville et al 2005; Waters 1999). Each of these factors is moderated by other characteristics such as age, sex, socioeconomic status, nativity, and characteristics of others nearby (Alcoff 2006; Penner and Saperstein 2013; Porter et al. 2015; Root 1990, 1996).

Most aspects of a person’s life story have entangled relationships with racial identity development. A person with low income may be especially likely to identify himself as a racial minority (Liebler 2004) or be perceived by others as minority (Penner and Saperstein 2008), especially if he lives in an area with other minorities (Porter et al. 2015). Simultaneously, the low income may be caused by discriminatory treatment received because the person perceives himself (and is seen by others) as a racial minority. Thus, although our conceptual model and prior research identify many factors related to identity, we include only age, sex, and residence in one’s birth state in our empirical models focused on the relationship between history, place, and racial identity. We choose these characteristics because they are not likely to be caused by the racial identification that we are trying to predict. Also, life stage and gender are common master statuses that are particularly likely to affect self-conception (and thus racial self-presentation; Mills 1959).

We argue that the many formative influences on racial identification discussed above are themselves influenced by place-specific history such as legal and social boundaries of race, both past and present. This is consistent with several established ideas in sociology and psychology. Sociologists of race have shown that historic and contemporary social forces (re)create the socially-defined boundaries of racial groups – which people are seen as “real” members of the group and which are not (Alcoff 2006; Barth 1969; Cornell and Hartmann 2007; Liebler and DeRousse-Wu 2012; Lowery 2010; Omi and Winant 1994; Wimmer 2008). These race boundaries are expressed as “prevailing social norms” (Waters 1996:447). Social psychologists point out that identities are generally formed and expressed through interpersonal processes such as reflected appraisal (Cooley 1902; Goffman 1959; Khanna 2004, 2010; Miville et al 2005), and these processes happen in settings whose character should be taken into account (Agnew and Duncan 1989; Bronfenbrenner 1977).

Through its influence on prevailing social norms and processes like reflected appraisal, place-specific history is expected to guide individual residents' racial identification choices. Because place-specific forces of history affect many people simultaneously and they affect the social structures in which individuals act and interpret one-another, these forces have the potential to affect the identification of entire groups of people living in an area (see Neely and Samura 2011). Depending on the local race history, for example, one community of people may develop a common black identity, while elsewhere, a group with similar heritage may see themselves as American Indian (see Lowery 2010). The influence of place-specific history on racial identification might be unconscious for adolescents and young adults developing an identity, and perhaps more apparent to adults who move to an area with prevailing norms that differ from those in which they were raised. Adults may even modify their identification to fit with prevailing social norms.

Despite the importance of place to identity, prior research on the racial identification of biracial or multiracial individuals has neglected place-specific history as an empirically measured factor in the social construction of racial identity and/or racial self-representation (e.g., Brown, Hitlin and Elder 2006; Campbell 2007; Harris and Sim 2002; Holloway et al. 2009; Roth 2005; Xie and Goyette 1997). In the next section, we describe some race-specific policies related to place and identity occurring in the United States over the past several centuries.

2.2. Race-Specific Policies and Practices Related to Place

2.2.1. American Indians

Colonial and federal policies have dramatically affected the physical location of American Indians. For example, in the early 1800s, most southeastern American Indians were forcibly removed from the eastern side of the Mississippi River to areas in present-day Oklahoma (Perdue and Green 2007). In the late 1800s, the Indian Wars drew to a close and reservation lands were established in isolated areas in the western U.S., usually far from cities (Snipp 1989) and before federal allotment policies wrested much of this land from Indian control. By 1930, the American Indian population had begun to recover from its low point of about 275,000 at the turn of the century (Thornton 1987), but the population remained small. Assimilation programs, especially common in the early- and mid-20^th Century, worked to push Native people to identify as non-Native to the extent possible and extracted large tracts of land from American Indian tribes. In the 1950s, the Urban Indian Relocation Program, run by the Bureau of Indian Affairs, aimed to assimilate formerly rural American Indians by moving them to urban areas (Fixico 2000; Snipp 1989).

Today, land legally held by American Indian tribes is only a portion of the area that can be seen as a meaningful part of current “Indian Country.” In acknowledgement of this, the Census Bureau’s current measure of indigenous homelands indicates whether a census block is in a legal or statistical homeland area. Statistical areas are especially important in places such as Oklahoma or Alaska where tribes are active but do not hold land. The maps in Appendix B provide an illustration of changes in the American Indian tribal land base over the course of more than a century. We examine the association between residence in an American Indian homeland (variably defined) and identification in our multivariate analyses.

These and other programs and policies have influenced current American Indian group boundaries (Barth 1969) and personal identities. People of American Indian heritage who live in communities with a historically strong American Indian presence are especially likely to racially identify as American Indian. This may have multiple sources. Like other indigenous people, they may feel a connection to the land itself or to the tribal government and people on that land (Kana’iaupuni and Liebler 2005; Kana’iaupuni and Malone 2006; Liebler 2010b). They may feel personal connections to others who also identify as racially American Indian (Liebler 2001). And the race group boundaries between American Indians and others in the area may be especially defended through stereotypes and segregation, as well as pro-Indian cultural experiences. The prominent presence of monoracially-identified American Indians in an area may promote similar identity formation by others, and may also be a sign that racial boundaries are carefully enforced. Likewise, in places with historic connections to American Indians but little contemporary experience, it may be most socially acceptable to present American Indian heritage as an ancestry.

2.2.2. Whites

Whites have been essentially unlimited in their residential locations in the United States. Numerous federal and local policies sought to give full freedom of movement to US-born whites, especially whites hailing from Northern and Western Europe. Because of these and other privileges, whites are able to develop their racial and ethnic identities without severe constraint (Alba 1990; Gans 1979; Waters 1990), sometimes treating their American Indian heritage as part of a symbolic or situational identity (Gans 1979; Harris and Sim 2002; Okamura 1981; Waters 1990).

2.2.3. Blacks

American blacks have been severely restricted in their racial identifications in ways that vary across place within the U.S. Over the centuries, innumerable policies and practices (both legally sanctioned and illegal) have forced blacks to live in segregated areas of cities and towns (Crowe 2012; Massey and Denton 1993). As part of the enforcement of slavery and Jim Crow laws, areas of the former slave-holding south have historically imposed and enforced strict definitions of who is black (Davis 2001). This “one drop rule” is thought to socially require that part-African-American people identify as racially black (Davis 2001; Khanna 2010). Our data supports the notion that this “rule” has some power even today, as there are almost 30 times as many people reporting that they are racially black with American Indian ancestry (weighted N=522,607) as there are people reporting American Indian race with black ancestry (weighted N=16,226). The continued influence of this historic norm, however, is expected to vary by place.

More recent place-specific race history may also be related to black racial identification. For example, in the late 1960s and early 1970s, race riots protesting the unequal treatment of blacks were socially focused on black-white intergroup relations and stratification. The immediate and longer-term legacy of these protests may be to encourage black-American Indians to racially identify as black (and white-American Indians to identify as white) because of the importance of the black-white divide.

2.2.4. Multiple races

Recent history may be especially likely to impact people’s decisions to report multiple races, which was not an option on the census or ACS until 2000. Interracial marriage has become more common and is increasingly in the public eye (Johnson and Kreider 2013; Lee and Edmonston 2005; Root 1996). These marriages cluster in some areas more than others (see Appendix C for a map; Park 1937; Wright et al. 2003). Where intermarriage is more common there are more people with mixed racial heritage (i.e., the children of these marriages) and also the presence of these families and individuals brings the conversation about racial identification into the present (Lee and Bean 2007). This visibility of multiracial people in a community may in turn encourage others to acknowledge their own mixed-race background, even if it is from an earlier generation.

3. Two examples of links between place-specific history and racial identification

In this section we illustrate ways in which local policies, practices, and demographic composition seem to affect the racial identification choices of mixed heritage black-American Indians in two states: Hawaii and Oklahoma.

3.1. Multiracial military personnel in Hawaii

On the Hawaiian island of Oahu, three of the seven PUMAs⁶ hold high proportions of people reporting both American Indian race and black race. At the same time, however, Oahu has few single-race blacks and few single-race American Indians, so these individuals are probably not the result of recent, local interracial marriages. Instead, they are likely to be members of the military or their families; the vast majority of Hawaiian military bases are within these three PUMAs. To understand why there are so many biracial American Indian-blacks in Oahu, we look to Hawaii’s history of multiracial identity movements, (e.g., the ‘hapa’ movement as described in Bernstein and De la Cruz 2009) and distinct racial ideologies (Edles 2004; Jolly 2007; Reed 2001) in interaction with the personal biographies of military personnel and their families, most of whom migrated from the mainland.

People in Hawaii are especially multiracial due to generations of openness to interracial unions. They also tend to report mixed race when asked; 21% of people in Hawaii reported multiple races in 2000 while less than 3% of the total U.S. population did so (Jones 2005; U.S. Census Bureau 2010). Military families are also a relatively multiracial group; of people in our data who are in the military or who live with someone in the military, 21% report multiple races. Among those based in Hawaii, the multiracial percentage is nearly double this national average (41%).. Hawaii’s historically grounded emphasis on acknowledging mixed heritage seems to have molded these migrants from the mainland, encouraging them to acknowledge multiple heritages and present themselves as multiple-race. The race history of Oahu seems to have shaped the identities of individuals.⁷

3.2. Oklahoma’s Black American Indians

More so than in other places, blacks and American Indians have an intertwined history in Oklahoma. This history is complex, including two centuries of alliances, integration, segregation, and competition between blacks and American Indians (Krauthamer 2013; Wickett 2000). Since early European settlement in North America, free blacks and runaway slaves were absorbed into southeastern American Indian tribes (Katz 1997; Krauthamer 2013). The tribes were later encouraged to take up slavery, which some did (Perdue and Green 2007). In the early 1800s, southeastern tribes, along with their slaves and black members, were forcibly removed to Oklahoma territory (Perdue and Green 2007). Following the Civil War, some former slaves were granted full rights of tribal citizenship, including equal communal land ownership (Wickett 2000). In the following decades, many blacks moved to Oklahoma to escape southern race-based violence, forming more than 20 all-black towns in Oklahoma (Katz 1997; Miles and Holland 2006; Wickett 2000).

Perhaps because of Oklahoma’s local history of (often forced) American Indian migration and (frequently) cooperative relationships between blacks and American Indians, there is a particularly wide variety of racial self-representations among people of black and American Indian heritage living in Oklahoma. While black-American Indian responses are rare across the country, all three race/ancestry responses are substantially more common in Oklahoma. The response of American Indian with black ancestry is 2.4 times as common in Oklahoma (0.0139%) as it is nationally (0.0058%), and biracial American Indian and black is nearly four times as common (0.3064% versus 0.0781%). Black with American Indian ancestry is about 50% more common in Oklahoma than average (0.2560% versus 0.1751%). This suggests that all three ways of representing black-American Indian heritage are socially viable in Oklahoma. This may be a result of the acknowledged joint presence of both blacks and American Indians throughout Oklahoma’s history, supporting social validation of an especially wide variety of personal racial identities.

4. Data, Sample Selection, Methods, and Measures

4.1. Data

The data that enable this study were not available in the twentieth century. Beginning in 2000, people living in the United States have been able to make a wide variety of racial, ethnic, and ancestral identity claims on federal forms, including the ability to report multiple races (Office of Management and Budget 1997).⁸ This detail has been released in the nationally representative datasets we use: publicly available microdata from the 2000 decennial Census and the 2006, 2007, and 2008 American Community Surveys (ACS)⁹ administered by the United States Census Bureau and available at http://usa.ipums.org/ (Ruggles et al. 2010). The Census 2000 microdata consist of a 1-in-20 sample of all households, and the ACS data from 2006–2008 are each nationally representative 1-in-100 samples of households in the United States. We pool the data from these four samples to improve the accuracy of estimates in our smaller dependent variable groups, especially American Indians with black ancestry. All of the data were collected by the same entity (the U.S. Census Bureau) and most of the questions were phrased identically. We have examined the data to identify differences across the years, and found very similar results in multivariate analyses using only Census 2000 data.¹⁰ We adjust the weights such that the pooled data can be seen as a nationally representative single-year snapshot of the U.S. population in the 2000s.

4.2. Sample selection

We include all individuals in the data whose race and ancestry responses fit one of the dependent variable categories listed below, with two qualifications. We exclude Latinos so that the range of racial identities under study is clearly defined; future research would do well to study the spatial context of mestizo identity development. We also exclude foreign-born people because their racial identities are likely to be influenced by the racial history, context, and definitions they experienced in their birth country (Rumbaut and Portes 2001; Waters 1999).

We considered also limiting the study to those who did not migrate (domestically) in the past one or five years, as we expect that people’s racial identities are influenced by the history and contemporary characteristics of everywhere they have lived. Thus, we expect that the relationship between race response and place-specific history will be attenuated for migrants compared to non-migrants. Unfortunately, the measure of migration differs substantially across the data sets we use and the restriction would substantially reduce our sample sizes. Instead we present (in Appendices D and E) multivariate model results two groups: (a) people living in their birth state and (b) those living outside their birth state. We also include results of significance tests for interactions from fully-interacted models. These results, discussed below, provide a basis for future research on how a person’s migration intersects with the impacts of place history on racial identification.

4.3. A note about place

We conceptualize “place” as a meaningful location which has a history particular to the area (see Agnew 1987 and Cresswell 2004). Operationalizing place in any study is a challenge because multiple levels of geography (e.g., neighborhood, city, and region) are simultaneously relevant (see Sampson, et al. 2002; Holloway et al. 2009). Also, the same place can have different meanings for different people (Manzo 2005) and the boundaries of the place might be conceptualized differently by different local groups (see Rundstrom 1993).

Our data are available for a very limited set of geographic scales, the smallest of which is a state-specific Public Use Microdata Area (or PUMA) with a population between 100,000 and 199,999 (see Siordia and Fox 2013) covering small areas in cities but very large rural areas. State partners help the Census Bureau define PUMA boundaries. Because PUMA boundaries do not necessarily delineate socially meaningful areas they are imperfect measures of “place.” PUMAs are nevertheless useful for a first look at the relationship between local race history and personal racial identification. With over 2,000 PUMAs in the U.S., these geographic areas encompass spaces large enough to contain multiple relevant events and small enough that these events might plausibly be considered “local.”

4.4. Methods

After showing population pyramids and maps to describe our data, we use multinomial logistic regression models to predict the race/ancestry responses of people with mixed white-American Indian or black-American Indian heritage. Multinomial logistic regression does not enforce an assumption that our dependent variable is ordinal, yet gives results and model fit that are substantively similar to generalized ordered logit and stereotype models.

Our independent variables represent various levels of analysis: person, PUMA, SuperPUMA (described in section 4.6.1.6.), and state. We adjust the standard errors for clustering at the PUMA level because PUMAs are nested within SuperPUMAs and states. In the regression models we deflate the sampling weights by dividing each original weight by the sample mean to get a relative weight that reflects the underlying sample size and generates suitable standard errors; we use weighted data for all analyses shown here.

4.5. Dependent variables: race and ancestry responses

There are two ways to report white, black, and/or American Indian heritage in these surveys. One is to mark one or more of the boxes in the prominent race question, and the other is to write in a response to the more obscure question about ancestry (see Appendix A and footnote 4). Perhaps surprisingly, the change in 2000 to allow multiple race responses did not spur many Americans of mixed heritage to report multiple races (Farley 2004, 2008; Morning 2000). Instead, most mixed heritage reports have been in the form of a single race and a different ancestry. In fact, many people with mixed heritage probably neglect to report less salient heritages in either question. Because of this, our sample is biased toward people who are at least somewhat attached to, or aware of, their mixed background. Despite these limitations and consistent with other researchers (Brown et al. 2006; Harris and Sim 2002), we see our dependent variable measures as good but not perfect reflections of the person’s racial identity at the moment of measurement.

The Census Bureau coded the first two write-in responses to the ancestry question. It is challenging to code ancestry responses into race categories but it has been done in prominent research (e.g., Goldstein and Morning 2000). Our American Indian ancestry category includes people who report any American Indian or Alaska Native tribal group or who wrote “American Indian” or “Native American” or “Indian” as their first or second ancestry response. Our white ancestry category includes people who report a European country or who wrote “white” or “Caucasian.” Our black ancestry category includes people who report a sub-Saharan African country or who wrote “Afro-American” or “African-American.” In our sample selection and coding, we include respondents who report multiple ancestries (e.g., Chinese and white ancestry), but exclude people with additional race responses (e.g., Chinese race).

Our three-category dependent variable for people of white and American Indian heritage is: (W1) American Indian race with white ancestry (unweighted N=16,638); (W2) white race and American Indian race (unweighted N=101,864); or (W3) white race with American Indian ancestry (unweighted N=487,462). Similarly, our dependent variable for people of black and American Indian heritage is: (B1) American Indian race with black ancestry (unweighted N=1,170); (B2) black race and American Indian race (unweighted N=14,490); or (B3) black race with American Indian ancestry (unweighted N=33,597). Weighted Ns, by sex, are reported in Figure 2.

Figure 2. Population Pyramids (% in group) for People with American Indian Heritage, by Race/Ancestry Report.

Source data: Census 2000 5% PUMS combined with American Community Survey data from 2006, 2007, and 2008, not including Latinos or the foreign born (accessed through IPUMS.org). Statistics are weighted to represent these US populations in the early 2000s.

4.6. Independent variables

We present descriptive statistics relating our independent variables to our three groups of white-American Indians in Table 1, and parallel statistics for black-American Indians in Table 2. If a measure were to vary little across columns in Tables 1 and 2, this may indicate that the history of the area is not related to constituents’ race responses. Variation across columns, on the other hand, might indicate a correlation between the histories and contemporary characteristics of an area and the ways individuals translate their mixed heritage into race/ancestry responses.

Table 1.

Characteristics of People with White and American Indian Heritage

		% in group with this characteristic
	W-AI total % with characteristic	AI race & W anc. (W1)	W & AI races (W2)	W race & AI anc. (W3)
Past characteristics of the area
% in former Confederate slave state	40.2%	22.3%	24.6%	44.0%
% in former non-Confederate slave state	8.3%	3.9%	5.3%	9.1%
% in PUMA that held AI land in 1875	15.9%	32.7%	26.2%	13.3%
% in PUMA that held AI land in 1930	7.6%	18.2%	13.1%	6.2%
% in PUMA that held AI legal or statistical land in 2000	24.9%	43.6%	35.2%	22.2%
% in PUMA that held AI land in 1875, 1930, AND 2000	6.6%	16.8%	11.9%	5.2%
% in PUMA that held AI land in 1875, 1930, OR 2000	19.5%	27.9%	24.7%	18.1%
% in PUMA with no AI land in 1875, 1930, or 2000	73.9%	55.2%	63.5%	76.7%
% in SuperPUMA with race riot in 1960s	45.2%	34.6%	38.5%	47.0%
Current characteristics of the area
Mean % of PUMA pop. with non-AI race & AI ancestry	4.8%	5.7%	5.4%	4.7%
Mean % of PUMA pop. with AI race & AI ancestry only	1.0%	2.4%	1.9%	0.8%
Mean % of PUMA pop. reporting more than one race	2.2%	2.9%	2.9%	2.1%
Characteristics of the individual
Boy aged 0–18	13.3%	14.8%	15.0%	12.9%
Girl aged 0–18	12.9%	14.2%	14.5%	12.6%
Young man aged 19–35	12.1%	11.7%	11.2%	12.3%
Young woman aged 19–35	13.6%	12.6%	11.3%	14.1%
Man aged 36–64	18.5%	19.1%	19.0%	18.4%
Woman aged 36–64	22.0%	21.2%	20.5%	22.4%
Elder man aged 65 or older	3.1%	2.9%	3.7%	3.0%
Elder woman aged 65 or older	4.4%	3.5%	4.8%	4.4%
Living in state of birth	64.6%	56.6%	62.1%	65.4%
Unweighted N	605,964	16,638	101,864	487,462
Weighted N	7,558,368	203,928	1,244,999	6,109,441
Row percent	100.0%	2.7%	16.5%	80.8%

Notes: Data are Census 2000 5% PUMS combined with American Community Survey data from 2006, 2007, and 2008 (accessed through IPUMS.org). All statistics are weighted unless otherwise noted. Sample excludes Latinos and people born outside the United States.

Table 2.

Characteristics of People with Black and American Indian Heritage

		% in group with this characteristic
	B-AI total % with characteristic	AI race & B anc. (W1)	B & AI races (W2)	B race & AI anc. (W3)
Past characteristics of the area
% in former Confederate slave state	38.5%	37.5%	28.1%	42.9%
% in former non-Confederate slave state	8.8%	7.9%	7.1%	9.5%
% in PUMA that held AI land in 1875	6.0%	6.6%	9.5%	4.5%
% in PUMA that held AI land in 1930	2.6%	3.9%	4.1%	1.9%
% in PUMA that held AI legal or statistical land in 2000	15.7%	15.5%	17.8%	14.8%
% in PUMA that held AI land in 1875, 1930, AND 2000	2.1%	3.0%	3.5%	1.5%
% in PUMA that held AI land in 1875, 1930, OR 2000	13.9%	12.8%	14.7%	13.6%
% in PUMA with no AI land in 1875, 1930, or 2000	84.1%	84.2%	81.8%	85.0%
% in SuperPUMA with race riot in 1960s	63.4%	60.3%	60.3%	64.8%
Current characteristics of the area
Mean % of PUMA pop. with non-AI race & AI ancestry	3.4%	3.4%	3.5%	3.4%
Mean % of PUMA pop. with AI race & AI ancestry only	0.7%	0.9%	1.0%	0.6%
Mean % of PUMA pop. reporting more than one race	2.5%	2.5%	2.9%	2.3%
Characteristics of the individual
Boy aged 0–18	13.3%	15.2%	15.4%	12.3%
Girl aged 0–18	13.4%	19.1%	14.8%	12.6%
Young man aged 19–35	12.0%	10.7%	11.9%	12.0%
Young woman aged 19–35	14.2%	12.0%	13.1%	14.8%
Man aged 36–64	15.8%	14.8%	16.1%	15.7%
Woman aged 36–64	23.1%	20.3%	21.1%	23.9%
Elder man aged 65 or older	2.7%	2.9%	2.5%	2.8%
Elder woman aged 65 or older	5.6%	5.1%	5.2%	5.8%
Living in state of birth	62.5%	63.7%	59.7%	63.6%
Unweighted N	49,257	1,170	14,490	33,597
Weighted N	755,529	16,226	216,696	522,607
Row percent	100.0%	2.2%	28.7%	69.2%

Notes: Data are Census 2000 5% PUMS combined with American Community Survey data from 2006, 2007, and 2008 (accessed through IPUMS.org). All statistics are weighted unless otherwise noted. Sample excludes Latinos and people born outside the United States.

4.6.1. Past characteristics of the area

The focal independent variables in our multivariate models are measures of historical events or characteristics of the local area. We see these as proxies for historical forces impacting local norms surrounding race.¹¹

4.6.1.1. Confederate slave state

This bivariate indicator marks whether the respondent lives in a former Confederate slave-holding state: AL, AR, FL, GA, LA, MS, NC, SC, TN, TX, VA, or WV. About 40% of the people in our data live in one of these states.

4.6.1.2. Non-Confederate slave state

Four states (DE, KY, MD, and MO) and Washington DC had legal slavery before 1861 but did not join the Confederacy at any point during the Civil War. We expect the racial boundaries in these places to be slightly less strict than those in the Confederate states. About 8% of the people in our data live in one of these states or Washington DC.

4.6.1.3. American Indian homeland in 1875

Using ArcGIS software, we overlaid a map of Indian lands in 1875 (Paullin 1932) onto a map of PUMAs in 2000 (the result is shown in Appendix B). We consider a person in our data to live in an 1875 Indian homeland if they live in a PUMA that has any overlap with the Indian land. 16% of white-American Indians and 6% of black-American Indians live in one of these areas. We combine this measure with our two other homeland measures (see 4.6.1.4. and 4.6.1.5.) in the models to indicate whether the PUMA was a homeland at all three time points, at some of the time points, or at none of the time points.

4.6.1.4. American Indian homeland in 1930

To capture a more recent moment when Indian lands and identities were under siege, we created a measure of whether the individual lived in an area that was an American Indian homeland in 1930. To identify American Indian lands in 1930, we again used ArcGIS software to overlay a map of Indian lands (Paullin 1932; map is shown in Appendix B) onto a map of PUMAs in 2000. Few people in our study live in areas that were American Indian land in 1930.

4.6.1.5. American Indian legal or statistical area in 2000

This measure indicates whether a PUMA contains any census block that was considered an American Indian or Alaska Native legal or statistical area in 2000 according to the Census Bureau’s classification, which is shown in Appendix B (see Liebler 2010b for more on this measure).

4.6.1.6. Race riots in the 1960s

Spilerman (1970, 1971) and Carter (1983, 1986) collected information about race 784 riots that occurred between 1964 and 1971 (mostly 1967–1969). We identified the 2000 SuperPUMA (contiguous area of 400,000+ people within a state) of each riot listed in the Spilerman-Carter data and created a bivariate measure indicating whether there was at least one riot in the SuperPUMA during this period. 45% of white-American Indians and 63% of black-American Indians in our data live in one of these SuperPUMAs.

4.6.2. Contemporary race and ancestry responses of others in the area

The contemporary race and ancestry responses of other in the area may reveal the currently socially acceptable ways of presenting American Indian heritage in that area. A rare identity may be particularly difficult to maintain because of the need for outsider recognition of the identity as part of the identity maintenance (e.g., the Looking Glass Self; Cooley 1902).

4.6.2.1. American Indian as only an ancestry

We include a variable indicating the percentage of the total population of the PUMA who report American Indian ancestry but not American Indian race. On average, our white-American Indian respondents live in PUMAs in which 4.8% of residents report American Indian ancestry but only non-American Indian race(s); for black-American Indian respondents, the average is 3.4%.

4.6.2.2. Monoracial American Indians

We include a measure of the percent of people in the PUMA who reported one race (American Indian) and only American Indian ancestry (or no ancestry response). People in our sample who report American Indian as a race live near substantially more of these monoracially-identified American Indians.

4.6.2.3. Multiracial people

To measure the predominance of multiracial people in the area, we use the percent of people in the PUMA who reported more than one race. On average, about 2% of the local population reported more than one race. This measure is highly correlated with measures of the preponderance of interracial marriages in the PUMA and so we do not include a separate measure for interracial marriages. In Appendix C we show maps of the spatial distribution of monoracial American Indians marriages, white-American Indian interracial marriages, and black-American Indian interracial marriages.

4.6.3. Individual-level characteristics: age, sex, and residence in state of birth

For our population pyramids, we use standard ten-year age groups, by sex, to show the population distribution (see Figure 2 below). In the multivariate models, we code age into four life-stage-related categories (0–18, 19–35, 36–64, and 65 or older) and combine it with the bivariate measure of sex because we see the potential for patterns that are simultaneously age- and sex-specific. In the multivariate models, we also include an indicator for whether the respondent is living in their birth state, as some populations may be more likely to migrate than others and relocation may influence identification.

5. Results

5.1. Age and sex variations in racial identification

We provide a first look at the composition of each race/ancestry group by presenting six population pyramids in Figure 2. We constructed these with weighted data, so the case counts are estimates for population sizes in the early 2000s. The population pyramids’ uneven patterns reveal that the way that a person racially identifies is at least partially determined by social forces that vary by sex and by age, period, and/or cohort. Our multivariate analyses, below, explore whether and how race and ancestry responses also vary with place-specific race history.

Identity and racial identification can vary substantially over the life course and for different cohorts. Most of the six population pyramids show relatively few young children (with the exception of American Indians with black ancestry). And, among white or black people with American Indian ancestry there are suspiciously few teens compared to the number of people in their 40s; this response pattern (reporting an American Indian ancestry) seems to be more attractive in middle adulthood. It is possible that individuals change their answers over their life course as they engage in age-graded norms about interest in family history or life-stage-related patterns in willingness to answer write-in questions (like the ancestry question). It could also be due to differences in identification choices across cohorts; if successive cohorts of people are exposed to different racial milieus during the years of identity development and family formation, cohort-specific patterns in race and ancestry responses may emerge.

The population pyramids also show that in the early 2000s there were about one half-million more white-American Indian women than white-American Indian men, and about 95,000 more black-American Indian women than men. In fact, there are more women than men reporting each of the six mixed heritage responses explored in this study; women are especially overrepresented among those reporting white or black race with American Indian ancestry. The lopsided sex ratios are too large to be entirely due to sex differences in birth or survival and may reflect gendered patterns in interest in (or willingness to report) American Indian ancestry. In supplementary analyses (available on request) we found that among whites (especially white women) coresidence with children is correlated with reporting American Indian ancestry; perhaps this identity is heightened when a child is present. No pattern was found by marital status. In parallel analyses of blacks, the relationship between having children in the household and reporting American Indian ancestry was much less pronounced and consistent. This supports the idea that flexibility in ancestry responses is a white privilege (see Waters 1990).

5.2. Locations of white-American Indians and black-American Indians

People live near others who are like them for a variety of reasons including social, cultural, or political affinities (Kana’iaupuni and Liebler 2005; Wright et al. 2003). Generations of physical clustering of family members affects the composition of the local area. Social policies and practices, including imitation (Baller and Richardson 2002), also affect everyone in an area such that people in that area become more similar to one-another.

We show maps of the relative locations of the white-American Indian groups in Figure 3 and the black-American Indian groups in Figure 4. We echo prior work (Johnston, Poulsen, and Forrest 2009) by coloring the geographic areas in the maps based on whether the relative population size of the group in that area is different from the national average. The PUMA is shaded light gray if the population of interest is under-represented (below 50% of the national average), dark gray if it is over-represented (above 150% of the national average), and medium gray if the difference from the national average is not large (between 50% and 150% of the national average). This method of coding the group-specific population density of areas is impervious to population size within the area and highlights meaningful spatial variation in America’s racial landscape.

Figure 3. Spatial Distribution of Locations of People with White and American Indian Heritage, by Race/Ancestry Report.

Source data: Census 2000 5% PUMS combined with American Community Survey data from 2006, 2007, and 2008, not including Latinos or the foreign born (accessed through IPUMS.org). All statistics are weighted to represent these US populations in the early 2000s. Maps by Minnesota Population Center.

Figure 4. Spatial Distribution of People with Black and American Indian Heritage, by Race/Ancestry Report.

Source data: Census 2000 5% PUMS combined with American Community Survey data from 2006, 2007, and 2008, not including Latinos or the foreign born (accessed through IPUMS.org). All statistics are weighted to represent these US populations in the early 2000s. Maps by Minnesota Population Center.

The three maps in Figure 3 illustrate strong locational patterns in the race/ancestry reports of white-American Indians. The top map shows that white-American Indians who live in areas near American Indian homelands in the north, west, and southwest commonly report American Indian as their only race, probably reflecting the history of forced isolation and strong racial boundaries in these areas (Snipp 2007). As shown in the middle map in Figure 3, people who report both white and American Indian races live near the Pacific coast, on the northern plains, in the southwest, or near Oklahoma – places where American Indians are especially likely to meet and marry whites (see map of interracial marriages in Appendix C; also see Wright et al. 2003). In the bottom map, we see that white-American Indians who live in the south, Appalachia, Texas, or east of the Mississippi River (areas where most tribes were forcibly removed almost two centuries ago) commonly report their race as white and their ancestry as American Indian. At the same time, reporting white race and American Indian ancestry is particularly uncommon among white-American Indians living in the northern midwest or in the mountain region.

Few black-American Indian people report American Indian race and black ancestry; we show their relative locations in the top map of Figure 4. A response of American Indian with black ancestry is relatively common in some areas of the south (areas of North Carolina, Georgia, and Mississippi), Montana, Arizona, and southern California. Most of these are places with long-standing (if small) American Indian populations and have a relatively high incidence of black-American Indian intermarriage (see Appendix C). A more common response for black-American Indians is to report both of these heritages as races. This is most prevalent in Oklahoma, Montana, North Carolina, and in some mid-Atlantic areas (middle map in Figure 4) -- areas that also have relatively high rates of black-American Indian intermarriage. Black-American Indians living in the Deep South and mid-Atlantic areas most often identify as American Indian ancestry but not American Indian race (bottom map, Figure 4). This result is consistent with the history of these areas: most tribes were forced to leave centuries ago, so many southern blacks and whites who report American Indian heritage are likely acknowledging ancestors from many generations in the past. Also, the “one-drop rule” of black racial identification was imposed through laws and social norms in this area. Past strict enforcement of social, physical, and emotional boundaries between blacks and whites in the South may discourage southerners from reporting multiple races (Gullickson and Morning 2011; Holloway, et al. 2009). Again we see signs that the history of the community influences the biographies of individuals.

5.3 Multivariate models predicting race responses

We present five multinomial logistic regression models predicting the race/ancestry response of white-American Indians in Table 3 and five models predicting the race/ancestry response of black-American Indians in Table 4. Estimates in each table reflect the relative risk that the person is reported as a single race with another ancestry (e.g., white race and American Indian ancestry) instead of two races (e.g., American Indian and white).¹²

Table 3.

Multinomial Logistic Regression Models Predicting the Race/Ancestry Responses of People with White and American Indian Heritage

	1: Area in the past		2: Area currently		3: In past & currently		4: Individal level only		5: Full Model
	AI race & W anc. (W1)	W race & AI anc. (W3)	AI race & W anc. (W1)	W race & AI anc. (W3)	AI race & W anc. (W1)	W race & AI anc. (W3)	AI race & W anc. (W1)	W race & AI anc. (W3)	AI race & W anc. (W1)	W race & AI anc. (W3)
	RRR	RRR	RRR	RRR	RRR	RRR	RRR	RRR	RRR	RRR
Past characteristics of the area
Confederate and slaveholding	0.94	2.43 ***			1.02	1.62 ***			1.00	1.63 ***
Slaveholding, not Confederate	0.78 *	2.19 ***			0.91	1.35 **			0.90	1.36 **
AI homeland in 1875, 1930, and 2000	1.54 ***	0.57 ***			1.61 ***	0.73 ***			1.60 ***	0.74 ***
AI homeland in 1875, 1930, or 2000	1.28 ***	0.66 ***			1.40 ***	0.78 ***			1.39 ***	0.78 ***
No AI homeland in 1875, 1930, or 2000	--	--			--	--			--	--
Race riot in SuperPUMA in late 1960s	0.94	1.17 **			0.95	1.07			0.96	1.06
Current characteristics of the area
% in PUMA: non-AI race & AI ancestry			0.94 *	1.15 ***	0.92 ***	1.11 ***			0.93 ***	1.10 ***
% in PUMA: AI race & AI anc. only			1.09 **	0.78 ***	1.10 ***	0.85 ***			1.09 **	0.85 ***
% in PUMA: reporting 2+ races			0.99	0.77 ***	0.98 *	0.82 **			0.97 *	0.82 **
Characteristics of the individual
Boy aged 0–18							1.05	0.85 ***	1.02	0.89 ***
Girl aged 0–18							1.03	0.86 ***	1.01	0.89 ***
Young man aged 19–35							1.06	1.12 ***	1.07	1.14 ***
Young woman aged 19–35							1.12 **	1.27 ***	1.13 **	1.28 ***
Man aged 36–64							--	--	--	--
Woman aged 36–64							1.02	1.13 ***	1.02	1.12 ***
Elder man aged 65 or older							0.77 ***	0.83 ***	0.76 ***	0.83 ***
Elder woman aged 65 or older							0.71 ***	0.95	0.70 ***	0.94 **
Living in state of birth							0.78 ***	1.19 **	0.80 ***	1.10 **
Constant	0.15 ***	3.67 ***	0.20 ***	6.32 ***	0.19 ***	5.19 ***	0.19 ***	4.31 ***	0.22 ***	4.73 ***
Log Pseudolikelihood	−323,489		−319,877		−317,082		−336,204		−316,221
df	10		6		16		16		32
Pseudo R2	0.042		0.052		0.061		0.004		0.063

Notes: Data are Census 2000 5% PUMS combined with American Community Survey data from 2006, 2007, and 2008 (accessed through IPUMS.org). All statistics are weighted proportionally. In each model, standard errors have been adjusted to take into account the non-independence of observations within PUMAs. Note that Latinos and foreign born persons are not included in these models. RRR (relative risk ratio) shows the relative risk of reporting American Indian race and white ancestry (W1), or white race and American Indian Ancestry (W3), instead of reporting both as races (W2).

^*significant at p <= .05;

^**significant at p <= .01;

^***significant at p <= .001.

Table 4.

Multinomial Logistic Regression Models Predicting the Race/Ancestry Responses of People with Black and American Indian Heritage

	1: Area in the past		2: Area currently		3: In past & currently		4: Individal level only		5: Full Model
	AI race & B anc. (B1)	B race & AI anc. (B3)	AI race & B anc. (B1)	B race & AI anc. (B3)	AI race & B anc. (B1)	B race & AI anc. (B3)	AI race & B anc. (B1)	B race & AI anc. (B3)	AI race & B anc. (B1)	B race & AI anc. (B3)
	RRR	RRR	RRR	RRR	RRR	RRR	RRR	RRR	RRR	RRR
Past characteristics of the area
Confederate and slaveholding	1.63 ***	2.07 ***			1.60 ***	1.52 ***			1.61 ***	1.53 ***
Slaveholding, not Confederate	1.30 *	1.68 ***			1.27	1.35 *			1.30 *	1.38 **
AI homeland in 1875, 1930, and 2000	0.96	0.57 ***			0.94	0.80			0.94	0.83
AI homeland in 1875, 1930, or 2000	0.78 *	0.80			0.90	0.88			0.89	0.88
No AI homeland in 1875, 1930, or 2000	--	--			--	--			--	--
Race riot in SuperPUMA in late 1960s	0.97 ***	1.12 *			0.93	1.05			0.92	1.04
Current characteristics of the area
% in PUMA: non-AI race & AI ancestry			0.95 *	1.13 ***	0.92 **	1.09 **			0.92 **	1.09 **
% in PUMA: AI race & AI anc. only			1.06 *	0.77 ***	1.10 **	0.84 *			1.11 ***	0.84 *
% in PUMA: reporting 2+ races			0.87 ***	0.86 ***	0.92 **	0.90 **			0.92 **	0.90 **
Characteristics of the individual
Boy aged 0–18							1.03	0.77 ***	1.02	0.78 ***
Girl aged 0–18							1.35 *	0.82 ***	1.35 *	0.84 **
Young man aged 19–35							0.96	1.01	0.95	1.01
Young woman aged 19–35							0.97	1.12 *	0.95	1.10 *
Man aged 36–64							--	--	--	--
Woman aged 36–64							1.04	1.15 ***	1.02	1.15 **
Elder man aged 65 or older							1.29	1.20 **	1.29	1.21 **
Elder woman aged 65 or older							1.08	1.16 **	1.06	1.16 *
Living in state of birth							1.16	1.27 ***	1.15	1.20 ***
Constant	0.07 ***	1.74 ***	0.12 ***	2.84 ***	0.10 ***	2.28 ***	0.06 ***	2.09 ***	0.09 ***	2.01 ***
Log Pseudolikelihood	−40,585		−40,415		−40,248		−41,225		−40,093
df	10		6		16		16		32
Pseudo R2	0.020		0.024		0.028		0.005		0.032

Notes: Data are Census 2000 5% PUMS combined with American Community Survey data from 2006, 2007, and 2008 (accessed through IPUMS.org). All statistics are weighted proportionally. In each model, standard errors have been adjusted to take into account the non-independence of observations within PUMAs. Note that Latinos and foreign born persons are not included in these models. RRR (relative risk ratio) represents the relative risk of reporting American Indian race and black ancestry (B1), or black race and American Indian Ancestry (B3), instead of reporting both as races (B2).

^*significant at p <= .05;

^**significant at p <= .01;

^***significant at p <= .001.

5.3.1 Past characteristics of the area

We find that the race history of an area is associated with the race/ancestry responses of white-American Indians (Model 1, Table 3); 4.2% of the variation in the dependent variable can be explained by the measured race history of the area. Our measures indicate that in areas where race boundaries have historically been strictly defined, white-American Indians tend to self-present as single-race white. For example, those who live in former slave-holding states (Confederate or non-Confederate) are over twice as likely to report white single race as they are to report both white and American Indian as races. Similarly, living in an area that experienced race riots in the late 1960s is associated with responding as single-race white. In areas where the race history is more focused on American Indians, an American Indian race response is more common. In particular, those who live in areas with a consistent historical population of American Indians have a high risk of reporting single-race American Indian and a low risk of reporting American Indian as only an ancestry, relative to reporting both heritages as races.

Past characteristics of the area are less predictive of the responses of black-American Indians, explaining about half as much of the variance in the dependent variable (Model 1, Table 4). Black-American Indians living in areas where the historical race story focuses on blacks versus non-blacks–former slaveholding states in particular–are especially likely to be reported as one race, even if this means claiming single-race American Indian. Like the white-American Indians, black-American Indians who live in areas with historically consistent American Indian lands have substantially lower odds of an American Indian ancestry response; they instead tend to report both black and American Indian races.

5.3.2. Current characteristics of the area

The intersection of place and racial identification is also evident when considering characteristics of others currently living in the area as predictors of individuals’ race/ancestry responses (Model 2 in Tables 3 and 4). White-American Indians and black-American Indians both have a tendency to have a race/ancestry response similar to others in their PUMAs. For example, in areas where relatively many other people report American Indian as an ancestry (but not American Indian race), the relative risk of reporting American Indian heritage as an ancestry rather than a race increases among both white- and black-American Indians. Similarly, in PUMAs with more people who report multiple races, our sample members tend to do the same.

When both past and current characteristics of the area are taken into account simultaneously, as in Model 3 of Tables 3 and 4, we see that both retain their associations with race/ancestry responses. Together, these measures of past and current characteristics of the place explain 6.1% of variation in white-American Indians’ race responses and 2.8% of variation in black-American Indians’ race responses. The local social world, which is built on its place-specific history of policies and practices, is related to the construction of local individuals’ identities and racial identifications.

5.3.3. Individual characteristics

Few master status characteristics clearly precede racial identification; we include only age and sex (having already removed Latinos and foreign born people from our data), though other person-specific characteristics surely impact the choice of how to identify. We also include a dichotomous variable indicating whether the respondent lives in their birth state. Although some of these measures are significantly related to racial identification (as shown in Model 4 of Tables 3 and 4), the model fit is extremely poor. Less than half a percent of the variation in the dependent variable is explained by these variables.

5.3.4. Full models

In the final models (Model 5 in Tables 3 and 4), we include past and current characteristics of the area, as well as age, sex, and residence in one’s birth state. Results in Model 5 for both white-American Indians and black-American Indians are similar to results in Model 3. We continue to see a notable significant relationship between the race-specific history of the place and the racial identification of the individual, even taking into account an individual’s sex, age group, residence in birth state, and current racial context. Because racial identification is closely linked to life stories (biographies), these results showing a relationship between place-specific history and racial identification provide support for Mills’ (1959) claim that history (in this case place-specific race history) affects biography.

5.3.5. Residence in birth state

We expect a person with a long-term connection to a place to be consciously or unconsciously tied to the history of that place, and for this to be reflected in their racial identification. As an additional check on the expected relationship between place-specific history and race responses, we conducted analyses separately for those in their birth state and those living elsewhere (see Appendix D for white-American Indian models and Appendix E for black-American Indian models) We also ran fully interacted models to assist in comparisons; the significance levels of the interaction terms are embedded in the appendix tables.

Consistent with our expectations, the models in Appendices D and E show a stronger relationship between place history and race response for those living in their birth states than for those who live elsewhere. Almost 9% of the variance in the white-American Indian dependent variable is explained among those who live in their birth state, relative to 3.3% of variance explained in the model of white-American Indians living elsewhere. The model is less powerful in explaining variance in the black-American Indian dependent variable, but 5.0% of this variance is explained for those living in their birth states compared to only 1.5% of variance explained for black-American Indians living elsewhere. Future research using information about a lifetime of locations would likely uncover the ways in which local history matters for identity differently by life stage and depth of connection to the area. Whether or not a person lives in his or her birth state is certainly an imperfect measure of extensive residence, yet even this rough cut suggests that long term residence matters for the processes under study.

6. Discussion

In this research, we have mapped and analyzed patterns in race/ancestry responses to elucidate relationships between place-specific history and personal racial identification among nationally-representative samples of U.S.-born non-Latino mixed-heritage white-American Indians and black-American Indians. Our maps illustrate locational patterns in racial identity claims – claims which not only mirror contemporary patterns in intermarriage, but are also related to past place-specific events and policies. Our multivariate results reveal a substantial association between place-specific history and racial identification, net of contemporary racial composition and basic individual characteristics. For example, black-American Indians tend to identify racially as black if they live in an area with a strong history of tension between blacks and other groups (as measured by slavery and race riots). White-American Indians living in communities with a strong history of tribal presence are significantly more likely to identify as single-race American Indian. In supplementary analyses, we have found that location-specific factors are more strongly associated with racial identification for those who live in their birth state, suggesting that length of residence or connection to place strengthens the relationship between place history and racial identification.

This research represents a first-step in demonstrating the relationship between place-related history and racial identification; our results are meant to be descriptive and we make no strong causal claims. Nevertheless, several pieces of evidence are suggestive of a causal relationship. The population pyramids provide evidence that race claims are different for men and women and vary across age groups more so than is plausible based only on births and deaths. Also, the fact that members of military households based in Hawaii are much more likely to identify as multiracial than military families on the mainland suggests that the local norms have influenced some people to report their multiple heritages as races. Future qualitative research could better elucidate the precise mechanisms through which historical race patterns come to be associated with or affect current residents’ racial identifications.

Decades of sociological research has established that personal development is influenced by local social structures (e.g., Bronfenbrenner 1977; Stryker 2002, 2008), that historical roots underlie contemporary social forms (Gieryn 2000; Mills 1959; Sewell 2005) and that social boundaries are formed and enforced differently across time and place (Barth 1969; Cornell and Hartmann 2007; Omi and Winant 1994). We have built on these insights by highlighting ways in which the race history of a place is connected with the current identity claims of people who live there. We find support for Mills’ (1959) idea that history affects biography, demonstrating this through its association with racial identification. Thus, we agree with Gieryn (2000:466), who writes: “place is not merely a setting or backdrop, but an agentic player in the game – a force with detectable and independent effects on social life.” We present our results with the hope that future investigations of society’s effects on individuals will include efforts to understand the sociohistorical context of the place.

Appendix A. Race and ancestry questions asked in Census 2000 and the American Community Survey (2008 ACS shown)

Appendix B. Location of American Indian Lands in 1875, 1930, and 2000

Sources for 1875 and 1930: Paullin, Charles Oscar. 1932. The Atlas of the Historical Geography of the United States. Washington, D.C., New York: Published jointly by Carnegie Institution of Washington and the American Geographical Society of New York. Source for American Indian Homelands in 2000: US Census cartographic boundary files. Homelands in 2000 include both legal and statistically meaningful American Indian, Alaska Native, and Native Hawaiian entities. See http://www.census.gov/geo/www/cob/na_metadata.html#gad. Maps by Minnesota Population Center.

Appendix C. Spatial Distrubution of Marriages between Single-Race People

Source data: Census 2000 5% PUMS combined with American Community Survey data from 2006, 2007, and 2008, not including Latinos or the foreign born (accessed through IPUMS.org). All statistics are weighted to represent these US populations in the early 2000s. Maps by Minnesota Population Center.

Appendix D. Multinomial Logistic Regression Models Predicting Race/Ancestry Responses of People with White and American Indian Heritage, by Residence in Birth State

	(1)		(2)		(3)
	Full model: Only those living in birth state		Full model: Only those living outside of birth state		Interaction model: Significance of differences between model (1) and model (2)
	AI race & W anc. (W1)	W race & AI anc. (W3)	AI race & W anc. (W1)	W race & AI anc. (W3)	AI race & W anc. (W1)	W race & AI anc. (W3)
	RRR	RRR	RRR	RRR	p-value	p-value
Past characteristics of the area
Confederate and slaveholding	0.88	1.63 ***	1.08	1.57 ***	0.019	0.440
Slaveholding, not Confederate	0.83	1.31 **	0.95	1.31 **	0.365	0.953
AI homeland in 1875, 1930, and 2000	2.03 ***	0.70 ***	1.19 *	0.83 **	<0.001	0.010
AI homeland in 1875, 1930, or 2000	1.68 ***	0.76 ***	1.15 **	0.86 **	<0.001	0.013
No AI homeland in 1875, 1930, or 2000	--	--	--	--	--
Race riot in SuperPUMA in late 1960s	0.94	1.07	0.98	1.06	0.503	0.735
Current characteristics of the area
% in PUMA: non-AI race & AI ancestry	0.92 **	1.16 ***	0.92 ***	1.06 ***	0.971	<0.001
% in PUMA: AI race & AI anc. only	1.10 **	0.79 ***	1.09 ***	0.92 ***	0.786	<0.001
% in PUMA: reporting more than one race	0.97	0.77 ***	0.97 **	0.88 **	0.862	<0.001
Characteristics of the individual
Boy aged 0–18	0.90 *	0.92 **	1.25 **	0.86 ***	<0.001	0.067
Girl aged 0–18	0.89	0.94 *	1.22 **	0.84 ***	0.001	0.009
Young man aged 19–35	1.00	1.18 ***	1.12 *	1.10 ***	0.132	0.035
Young woman aged 19–35	1.09	1.31 ***	1.15 *	1.27 ***	0.520	0.387
Man aged 36–64	--	--	--	--	--
Woman aged 36–64	1.02	1.13 ***	1.03	1.11 ***	0.867	0.353
Elder man aged 65 or older	0.83 *	0.77 ***	0.72 ***	0.86 ***	0.191	0.036
Elder woman aged 65 or older	0.78 **	0.84 ***	0.66 ***	1.01	0.160	<0.001
Constant	0.17 ***	5.36 ***	0.24 ***	4.24 ***
Log Pseudolikelihood	−191,084		−123,154
df	30		30
Pseudo R2	0.089		0.033

Notes: Data are Census 2000 5% PUMS combined with American Community Survey data from 2006, 2007, and 2008 (accessed through IPUMS.org). All statistics are weighted proportionally. In each model, standard errors have been adjusted to take into account the non-independence of observations within PUMAs. Note that Latinos and foreign born persons are not included in these models. RRR (relative risk ratio) represents the relative risk of reporting American Indian race and white/black ancestry (W1 and B1), or white/black race and American Indian Ancestry (W3 or B3), instead of reporting both as races (W2 or B2).

^*significant at p <= .05;

^**significant at p <= .01;

^***significant at p <= .001.

Appendix E. Multinomial Logistic Regression Models Predicting Race/Ancestry Responses of People with Black and American Indian Heritage, by Residence in Birth State

	(1)		(2)		(3)
	Full model: Only those living in birth state		Full model: Only those living outside of birth state		Interaction model: Significance of differences between model (1) and model (2)
	AI race & B anc. (B1)	B race & AI anc. (B3)	AI race & B anc. (B1)	B race & AI anc. (B3)	AI race & B anc. (B1)	B race & AI anc. (B3)
	RRR	RRR	RRR	RRR	p-value	p-value
Past characteristics of the area
Confederate and slaveholding	1.82 **	1.64 ***	1.24	1.31 **	0.117	0.010
Slaveholding, not Confederate	1.49 *	1.56 **	1.06	1.18	0.239	0.068
AI homeland in 1875, 1930, and 2000	0.70	0.71	1.21	0.91	0.280	0.224
AI homeland in 1875, 1930, or 2000	0.76 *	0.85	1.12	0.93	0.082	0.374
No AI homeland in 1875, 1930, or 2000	--	--	--	--	--	--
Race riot in SuperPUMA in late 1960s	0.83	1.02	1.08	1.04	0.122	0.778
Current characteristics of the area
% in PUMA: non-AI race & AI ancestry	0.92 *	1.13 ***	0.94	1.05 *	0.698	0.007
% in PUMA: AI race & AI anc. only	1.11 **	0.77 **	1.08	0.93 *	0.603	0.001
% in PUMA: reporting more than one race	0.88 *	0.86 ***	0.96	0.94 ***	0.234	0.010
Characteristics of the individual
Boy aged 0–18	0.96	0.90	1.26	0.62 ***	0.403	0.001
Girl aged 0–18	1.27	0.97	1.70 *	0.69 ***	0.324	0.002
Young man aged 19–35	0.96	1.09	0.95	0.97	0.973	0.265
Young woman aged 19–35	0.96	1.30 ***	0.99	0.94	0.905	<0.001
Man aged 36–64	--	--	--	--	--	--
Woman aged 36–64	0.98	1.22 ***	1.11	1.08	0.545	0.113
Elder man aged 65 or older	1.49	1.07	1.08	1.22 *	0.482	0.371
Elder woman aged 65 or older	0.82	1.03	1.26	1.17 *	0.250	0.215
Constant	0.12 ***	2.31 ***	0.07 ***	2.15 ***
Log Pseudolikelihood	−24,234		−15,648
df	30		30
Pseudo R2	0.050		0.015

Notes: Data are Census 2000 5% PUMS combined with American Community Survey data from 2006, 2007, and 2008 (accessed through IPUMS.org). All statistics are weighted proportionally. In each model, standard errors have been adjusted to take into account the non-independence of observations within PUMAs. Note that Latinos and foreign born persons are not included in these models. RRR (relative risk ratio) represents the relative risk of reporting American Indian race and white/black ancestry (W1 and B1), or white/black race and American Indian Ancestry (W3 or B3), instead of reporting both as races (W2 or B2).

^*significant at p <= .05;

^**significant at p <= .01;

^***significant at p <= .001.