Abstract
Aligning census ethnoracial categories with America’s changing demography is a never-ending task and becomes more difficult when identity claims are rationales for altering categories. We examine four current problems: (1) the Census Bureau projects a population more nonwhite than white by midcentury—social demographers document trends pointing to a different racial future; (2) the census inadequately measures second- and third-generation Americans, limiting the nation’s understanding of why some immigrant groups are “racialized” while others are “whitened”; (3) on health, education, and employment, there is more intrarace than between-race variability, which is better measured for Asians and Hispanics than it is for whites and blacks; and (4) consistency in racial self-identification is stronger for whites, blacks, and Asians than for Hispanics, Native Americans, and biracial groups, lowering the reliability of race data. These measurement problems weaken policy choices relevant to economic growth, social justice, immigrant assimilation, government reforms, and an enlightened public.
Keywords
America’s racial classification can be traced to 1735, when Carl Linnaeus, the father of taxonomy, briefly shifted his attention from flora and fauna—he was a botanist—and fearlessly divided the human species into four subspecies: Americanus, Asiaticus, Africanus, Europeaeus. A few decades later, the classification was slightly modified by a German doctor, Johan Blumenbach, in his influential On the Natural Varieties of Mankind. He divided the Asians into the Mongolian and Malaysian; he also took care to rank-order the races, Europeans being superior as the most civilized, and Africans inferior and uncivilized.
These five races—though not the rank-order—structure America’s racial numbers today. Few are happy with this straitjacket, and for a half century the U.S. Census Bureau has been adding fixes big and small to better align eighteenth-century race science to changing social and demographic realities and public uses. It added an ethnic category—Hispanic—though that tripped over the other category, which, nonsensically, then became the fastest growing race in the country. 1 It invited Native American Indians to write in their tribal affiliation, which mixed up a civil status (membership in a federally recognized tribe) with culture (community belonging). It made extensive use of national origin subcategories for two races—Asians and Pacific Islanders—but neither whites nor blacks have subcategories. The latter two races are uniquely denoted by color, which, of course, has never been measured by the census. I doubt that it ever will be.
The “fix the race question” pace picked up as we entered the twenty-first century. One change was, on the face of it, more tinkering than transforming. It was in the 2000 census that Blumenbach’s Mongolian and Malaysian subspecies were finally counted as separate races in the American census. The story merits retelling. In the mid-1990s, Hawaii’s Senator Daniel K. Akaka successfully argued that the Native Hawaiians and Pacific Islanders (NHPI) should be a race separate from the Asian category. He cited statistics demonstrating that the NHPI were disadvantaged compared to the larger Asian population; he wanted these disadvantages separately tabulated and government programs targeted accordingly. In this Akaka was emphasizing the social justice reasoning so central to the civil rights era a half century earlier. Akaka, however, was not finished. In congressional testimony, he also argued that the current classification “denies us our identity as indigenous peoples” (U.S. House 1998, 263, italics added). Self-esteem rationales were not entirely new to the census (Schor 2005), but never had they led to such a radical change—adding to the census a new primary race category. Akaka’s bold identity claim kept intact the eighteenth-century race science category scheme, but pushed aside its rationale. Races are not biologically meaningful; they are what groups of people claim—or reject as the case might be—as their rightful identity.
It was a short step to an even more radical assertion. The first census of the new millennium made clear that treating us as if we all belonged to pure mono-races was badly mistaken. Racial mixture occurs, in numbers large enough to merit counting. 2 This was officially announced in an understated option on the 2000 census form: respondents could, if they wished, “mark one or more” of the five primary races. This option is traced to the 1968 Supreme Court ruling that interracial marriage was constitutionally protected, which of course led to biracial children. Again, this was hardly new, but a long-standing demographic reality the census would finally take into account. In 2000, mark-one-or-more entered census history with these children in mind. It had a bumpy beginning.
The National Association for the Advancement of Colored People (NAACP) insisted that because race statistics existed for “the enforcement of civil rights laws,” the move to multiracial census categories would reduce the size of discrete minority races, “diluting benefits to which they are entitled as a protected class under civil rights laws and under the Constitution itself” (Williams 2006, 308). In a show of racial solidarity, Asian and Hispanic advocacy groups agreed that the census categories were not called on “to provide vehicles for self-identification” (Williams 2006, 308).
But self-identification and self-esteem were exactly what Senator Akaka argued for, and what multiracial advocates sought. The Association of Multiethnic Americans demanded “choice in the matter of who we are, just like any other community. We are not saying that we are a solution to civil rights laws or civil rights injustices of the past.” It is ironic that “our people are being asked to correct by virtue of how we define ourselves all of the past injustices [toward] other groups of people” (U.S. House 1998, 383).
This rationale notwithstanding, the Census Bureau is not in the self-esteem business. It does not produce identity races; it produces statistical races. The Bureau’s website emphasizes that statistical races are used “to assess fairness in employment practices, meet legislative redistricting requirements by knowing the racial make-up of the voting age population, learn who may not be receiving medical services, determine disparities in health and environmental risks” and other similar purposes, as required by ten key executive agencies.
Nowhere on the Bureau’s website, or those of the ten executive agencies, will you find attention to self-esteem, recognition, or identity expression. Instead, you will find programmatic reasons for statistical races, linked to concerns about disparities and discrimination, but also to city planning, transportation policy, and related government tasks and commercial uses—all drawing on census race and ethnic statistics. Stated differently, Akaka started with disparities and tagged on identity. The Office of Management and Budget (OMB) ran statistics on the former, concluding that the evidence justified NHPI as a separate race. Had Akaka offered only a self-esteem rationale, the OMB and Census Bureau would not have known how to measure it. NHPI would not today be a separate race.
But that was two decades ago. Do we not now have a broader array of identity claims and a richer understanding of, especially, the multiracial population? Yes, and much that is relevant is on display in this volume, starting with the fact that a multiracial emphasis turns out to have unexpected political consequences.
To understand those consequences, I briefly trace familiar history, dating to America’s founding fathers. The War of Independence they led was justified as a war against tyranny and a proclamation of liberty and equality. This posed a tricky question: How could a new nation reject tyranny but simultaneously impose it on the indigenous Native Americans in forceful removal from their homelands and, even more comprehensively, on the African slave? The answer—impose a citizenship test in the new republic, conveniently tethered to who was civilized and who was not. The Europeans wrote the rules, declaring themselves civilized and fit for citizenship, but of course not the uncivilized Indian and African. A white superiority narrative was born (Parkinson 2017). It haunts our history, very often finding in the census a convenient tool: a Jim Crow apartheid regime, a whites-only melting pot, second-class citizenship for Asian and Hispanic labor, continuing in current times in gerrymandering and voter suppression, and, today, unexpectedly, in the revival of white nationalism that shows traces of the eighteenth-century fixation on white superiority.
The civil rights era took on the white superiority narrative when the census, previously an aid to exclusion, shifted 180 degrees to serve as a tool of racial inclusion. Its most telling early achievement was statistical proportionality—groups matter proportionate to their share of the total population. This idea was anticipated by the constitutionally protected decadal census reapportionment process, but in 1787 none could have imagined its mid-twentieth-century application.
Statistical proportionality began its official life in the Kerner Commission Report (National Advisory Commission on Civil Disorders 1968), a response to the mid-1960s epidemic of urban race riots. The commission introduced an idea new to public discourse—unintentional discrimination, the kind that flows not from a racially prejudiced employer or realtor but from how the labor and housings markets are structured. Soon the phrase institutional racism caught on, and from that it was a short step to preferential hiring. Institutional discrimination required institutional solutions (Knowles and Prewitt 1969). Statistical proportionality slipped into government policy and public usage in the late 1960s, later named affirmative action or, today, the widely established practice of diversity hiring. The white superiority narrative faded from public usage, its rejection further hastened by a demographic transformation as America again opened its door to immigration, this time overwhelmingly from non-European regions. This gave the nation a new multicultural narrative.
The census, still anchored to an eighteenth-century classification and heavily constrained by path dependency, struggled to match the new demography and discourse. Mark-one-or-more was an awkward response to the hundreds of national origins, ethnic, and language groups confused about what census box to tick. Mark-one-or-more applied to only the five primary races. The part white and part Asian could find a census home, but not the part white and part Hispanic. But change was still in the air. A few years after the partial fix represented by mark-one-or-more in 2000, the Census Bureau went at it again. It mounted a research project on racial categories unprecedented in its ambition and scale—comparing seventeen different ways to count races and ethnicity, with two control panels; adding in fifteen experimental treatments and focus groups; as well as assembling expert panels and presenting in dozens of professional settings.
The research led to various mode, wording, and framing improvements and then, much more consequentially, to two ambitious category recommendations: (1) the 2020 census should introduce a race category that the botanist and the doctor had missed—Middle-Eastern North African (MENA); and (2) it should merge what had been a separate Hispanic question into a single ethnoracial classification with the primary races. Mark-one-or-more stays in place, though now applied to seven primary categories (if MENA is accepted). The census form also offers space to write in specific ancestry, ethnicity, tribe, or national origin identities for each of the basic categories. As of this writing, these recommendations are before the statistical policy office of the OMB, the controlling executive agency. 3
Are the proposed 2020 changes an improvement? Yes, is the general answer in this volume. Does this mean that the country will finally have a coherent set of categories? Not exactly. Problems persist and are pointed out in the articles that follow. The reader will find many sentences with some variant of “the census should….” with specific recommendations as well as sweeping complaints. Telles, for example, says that the Census Bureau is “out of step with popularly held notions of race and ethnicity.” If the Census Bureau accepted all the recommendations found in this volume, would it then be in-step, would we have a coherent set of categories? Not likely. As Lopez (2005) observes, incoherent census categories are inevitable because they “arise out of (fundamentally irrational) social practices” (p. 50).
Repeated effort to fix the census classification system is welcome, but it will never completely fix what is inescapably an incoherent classification scheme. There is a racial reality independent of the reality captured in the census; and as academic research points out, that reality is more complex, nuanced, changeable than what census statistics capture. This is not for lack of intelligence, imagination, or effort in the Census Bureau—all present in abundance. It is because the Bureau is not a free agent. Its purpose is decided elsewhere, starting with the first Article of the U.S. Constitution. Its past, present, and future cannot escape the “fundamentally irrational” task imposed in law and practice—count America’s races.
Academic research is not so constrained. Consider a simple example. Social scientists find implicit bias and racial stereotyping, and find that phenotypic attributes are in play. Skin color matters. Perhaps it would be useful if the American census measured color (Brazil does). This will not happen. Academic research, with its experiments, in-depth interviews, and tailored surveys, will always probe corners not reachable by the census.
This volume, under four headings, reviews research that draws primarily on census data but also brings other data into the picture. The back and forth between census measures and academic unpacking, aided by other measures, inevitably reveals measurement flaws, some that are, at least in principle (first two), fixable; others less so (second two).
The majority-minority future now being much discussed, with significant political consequences, is highly likely to be more distant than has been asserted. Under reasonable assumptions, it may be postponed indefinitely as assimilation “whitens” some nationality groups. It did so for the once alien Irish, Italian, and Polish “races” and is now under way, if very selectively, among Asians (especially the Chinese, Japanese, and Korean-Americans), Hispanics (especially from the Caribbean Islands), Native Americans (especially those living in cities rather than on reservations), well-educated African immigrants who started arriving in the 1970s, and multiracials.
The inadequately measured generational differences, especially where place-of-birth gets tangled up with the dynamics of racializing immigrants, that is, where color-line and nativity-line overlap. This is a nontrivial matter; in 2016, nearly one in five Americans were immigrants (43.7 million) or first generation (16.6 million).
The difficulty of assessing and responding to disparities and inequalities because within-group variability (detected by comparing the national origin groups that belong to one or another primary race group) is greater in magnitude than that between the primary race groups, which are generally used in policy design and evaluation.
The low reliability of ethnoracial statistics, across settings and over time, when compared to other population characteristics measured by the census, playing havoc with trend lines.
The Majority-Minority Projection
In 2015, the Census Bureau projected that three decades hence America will be majority-minority, defined as fewer whites than nonwhites, as is already the case of babies being born. A Census Bureau Report—Non-Hispanic Whites May No Longer Comprise over 50 Percent of the U.S. Population by 2044—compared non-Hispanic whites (European decent) to all other racial groups, including white Hispanics and all multiracial respondents (U.S. Census Bureau 2015). The report generated a media storm.
Hold on, said demographers and sociologists, several of the best represented in this volume. This projection is misleading, dangerously so as reflected in the angry white vote demanding that “we want our country back.” Alba commented on the census projection in the journal American Prospect, tellingly titled: “The Likely Persistence of a White Majority: How Census Bureau Statistics Have Mislead Thinking about the American Future” (Alba 2016). The journal editor asked me to provide a comment, from the perspective of an ex-director of the Census Bureau. I wrote, “Richard Alba’s analysis is a service to the country. I write to urge the Census Bureau and its various oversight agencies and committees to take his message seriously” (Prewitt 2016). 4 What Alba characterized as misleading thinking, and what I described as his service to the country, was projecting a nonwhite majority that was not supported by sociological theory based on more nuanced analysis of the full range of census data (discussed in more detail in this volume). This, of course, was before the nation experienced a presidential campaign that aggressively labeled immigration as a threat, and an aftermath of the election that strikes many as hospitable to angry white nationalism. Certainly the phrase “majority-minority” is now common in national political discourse, with consequences examined in a Russell Sage workshop, and now continued in this volume of The ANNALS. 5
I take a detour to place the majority-minority phrase in a more general context. It is well established that official statistics shape social realities. Bourdieu writes that struggles over racial definitions and classification are “struggles over the monopoly of the power to make people see and believe, to get them to know and recognize, to impose the legitimate definition of the divisions of the social world and, thereby, to make and unmake groups” (Bourdieu 1991, 306, emphasis in original).
Bourdieu’s influential line of reasoning is persuasive but is not the entire story. Certainly the making and unmaking of groups are located in power centers, but they are increasingly also found in how people self-identify on the census forms. The population itself can make and unmake groups. Alba reports that interracial marriage is predominantly between white and nonwhite, and results in “social identities, affiliations, and contexts of Americans … [as] on the whole closer to those of whites than to those of minorities” (italics added). Moreover, and of great importance, there is nothing in the conduct of the American census preventing these “closer to whites” from ticking only the white box. Because this is happening in substantial numbers, the arrival of a majority-minority America is continually being postponed.
Alba adds that partly black persons are an exception, either remaining multirace or identifying as black only. There is a prominent example, underscoring that inviting identity expression in census practice has consequences for the making and unmaking of groups.
The Obama rule
When President Obama returned his census form in 2010, the New York Times reported: It is official: Barack Obama is the nation’s first black president. A White House spokesman confirmed that Mr. Obama, the son of a black father from Kenya and a white mother from Kansas, checked African American on the 2010 census questionnaire. (Roberts and Baker 2010)
The phrase, “It is official,” is hyperbole. Ticking a census box does not make anything about an individual “official”—it is just a tick on a form that is then aggregated with millions of other ticks to give an estimate, in this case, of the size of the African American population group. Other ticks, biracial black and white ticks, for instance, in that same census give us an estimate of the size of a mixed-race population group—now, we see in Obama’s census decision, one person fewer than it might have been (and the black race one person more).
The Census Bureau did not say of Obama “we know this person to have a white mother and African father, making him (and his daughters) biracial and will so be recorded statistically.” No such thing happened or was even contemplated. This was not timidity because Obama was president. His census self-identification stands because, in the census, you are in the category you choose, or in the case of children, is chosen for you. If you are in some other category a decade later, and millions are, you have changed your mind. Racial identity in the census is not whether a taxi stops for you, or what is on your birth certificate, or what your grandparents thought they were. It is a tick in a box. Call that the Obama rule, and recognize that statistical races are every bit as real as socially constructed or identity races. And in the arena of public policy, only statistical races are real, and we cannot be surprised if they then are forcefully deployed in political battles over those policies.
If Obama was signaling that he feels closer to his African than to his European heritage, he is joined by many African Americans. Alba presents relevant data, as he does for the fact that other mixtures, especially white-Asians and white-Hispanics, tend to relate more easily to their European than their minority heritage. We see that similar patterns occur when mixed-race parents, for various reasons, describe their children as monoracial.
Myers and Levy carefully unpack the different ways that census race counts can be presented depending on how the multirace count is handled; one version, no less plausible than the majority-minority projection, indicates a white majority well into the century and perhaps indefinitely if children of white-Hispanic or white-Asian parents migrate to the white-only census option.
But that story has been drowned out by the majority-minority media story, now treating the presumed decline of the white population to minority status as an inevitable and official fact. For example, more than a third of the white population now describe their racial identity as extremely or very important to them (American National Election Study 2016). We do not have trend data on that number, but it is unlikely that “my white identity is important to me” response would have been so numerous two decades ago.
The current beliefs and behaviors of white Americans are affected. As reported by Craig, Rucker, and Richeson, there is “clear evidence that white Americans … experience the impending ‘majority-minority’ shift as a threat to their dominant (social, economic, political, and cultural) status.” Whites who believe they will lose their place “at the top” of various status hierarchies are drawn to “politically conservative policy positions, including on policies most relevant to societal racial equity (e.g., affirmative action, immigration policy, harsh criminal justice policies).”
The emergence of a politically assertive white identity occurs to the extent that the binary white/nonwhite narrative gets traction. Bean, however, emphasizes that this narrative and the new forms of identity politics it spawns are inconsistent with powerful social-demographic transformations simultaneously under way. He reports on the social mobility of various nonwhite groups, whose members “have been gaining in average education and income and are experiencing less residential segregation from whites, often as a result of geographic mobility.”
This finding underpins a narrative counter to the binary majority-minority framing. “Rather than emphasizing that newcomers are essentially people of color whose mobility is limited by discriminatory treatment,” which supports anxiety narratives and underscores the weakening of social cohesion, there is substantial evidence on the positive impact of diversity, including ways in which it strengthens social cohesion.
For Bean and others, the most effective way to move away from the white-nonwhite picture of America’s population and its majority-minority framing is to emphasize what is going on within the primary race groups. There we find a more nuanced and differentiated population, which we consider in Section III.
A takeaway lesson from Section I. The majority-minority projection is problematic in two respects: it will not happen by 2044, and it will possibly be postponed indefinitely. It is based on a binary framework of whites versus all other Americans, which seriously distorts the demographic realities and the lived experience of millions of Americans.
A caution about this takeaway. Of course a Census Bureau report did not cause an insurgent white racism, which dates to the early Crusades targeting Jews, and was then deepened by the Atlantic slave trade and colonization (Fredrickson 2002). Its twenty-first-century political traction stems from the anti-immigrant sentiment that swept across Europe and the United States. In brief, white nationalism and the specific majority-minority framing has multiple sources, all of which predate a 2015 Census Bureau Report to which we have drawn your attention. With that caution emphasized, I also underscore that political discourse has never hesitated to lift statistically grounded phrases out of context and to deploy them to shape and even deliberately mislead public opinion. We are in a heightened phase of such political practice. The final takeaway lesson is that it matters if America measures races, and then, of course, how the government decides what those races are. It matters because law and policy are not about an abstraction called race but are about races as they are made intelligible and acquire their numerical size in our statistical system. When we politically ask why black men are jailed at extraordinarily high rates, whether undocumented Mexican laborers are taking jobs away from working-class whites, or whether Asians have become the model minority in America, we start from a count of jailed blacks, the comparative employment patterns of Mexicans and whites, and Asian educational achievements. When our political questions are shaped by how many of which races are doing what, and when policies addressing those conditions follow, we should worry about whether the “how many” and the “which races” tell us what we need to know about what is going on in our polity, economy, and society. (Prewitt 2013, 7–8)
The Color Line, the Nativity Line
Why should the Palatine Boors [Germans] be suffered to swarm into our Settlements, and by herding together establish their Language and Manners to the Exclusion of ours? Why should Pennsylvania, founded by the English, become a colony of Aliens, who will shortly be so numerous as to Germanize us instead of us Anglifying them, and will never adopt our Language or Customs, any more than they can acquire our Complexion. (Labaree 1961, 234)
Thus anxiously pondered Benjamin Franklin, 266 years ago. Obviously the alien immigrant has long been with us, always with a complexion darker than the superior English.
When in 1903 Du Bois famously wrote that “the problem of the 20th century is the problem of the color-line,” Jim Crow white racism was in full stride. He correctly anticipated that the white-black color line would persist, even as the then-immigrant populations began to melt, losing their alien identities as German, Irish, Polish, or Italians. Intermarrying across nationality and religious lines, these outsiders Franklin found so threatening, would “whiten.”
And what of our times? The color-line has not disappeared but it is permeable, the one-drop rule was discarded midcentury, and interracial marriages and biracial children increase their number with each census. What about the nativity-line? The nation needs to know if assimilation of the alien proceeds today, as it did for the Palatine Boors, when Germanizing the Anglos (from Christmas trees to soccer) and Anglifying the Germans (from Thanksgiving to baseball) jointly strengthened the nation.
We know less than we can and should about the pace of assimilation today, especially as it differs across the many national origin and linguistic groups that constitute today’s population. Readers will know the reason—“the absence of reliable data on the children of immigrants remains the single greatest weakness in the U.S. statistical system,” writes Massey. The resurgence of mass immigration starting in the 1970s was itself captured in a place of birth question, but did not statistically survive as a second generation (where were your parents born?). We now are on the verge of an unmeasured third generation.
Yet the third generation—grandchildren of the post-1965 immigrants—“will write the next chapter in the contemporary American immigrant assimilation story,” writes Jimenez. Their “social, political, and economic fortunes will reveal the extent and kind of assimilation among the descendants of today’s largely non-European immigrants.” Tran, like Massey, emphasizes “ongoing data limitations” that have “created unique challenges for empirical assessments of second-generation decline or progress,” and also draws attention to the way in which today’s immigrants confront a color-line even as they navigate social and economic barriers confronting those born of foreign parents. We know from academic surveys that the nativity-line is selectively racialized, but we lack a nation-wide, detailed grasp of the scope and durability of this racialization. We are blindfolded, and open up a space for misleading exaggerations that have political consequences.
Duncan and Trejo, handicapped as well by the absence of census data, make do with the much more limited Current Population Survey (CPS) to tease out patterns of ethnic attrition, that is, the rate at which the children and grandchildren of immigrants cease to self-identify with the ethnicity of their immigrant ancestors. They report that for the third generation, “ethnic attrition predominately occurs in children with mixed parental origins.” This finding is of course an important indicator of whether today’s immigrants, like their nineteenth-century predecessors, are assimilating, and whether multiraciality hastens that assimilation.
I have written elsewhere (Prewitt 2013, 189–90), echoing Bean and many others, that assimilation can fail, leading to a xenophobic politics, a disillusioned and angry second generation, and heightened fears in white enclaves about the incursion of nonwhite “non-Americans.” Or it can succeed, leading to cultural, economic, and political achievements similar to those produced by the successful assimilation of white ethnics in the twentieth century. More likely, there will be a mixed pattern. Some nonwhite population groups are assimilating into an America broadly understood as multiracial. But not every group is learning English, graduating from high school, naturalizing, finding a job, and buying a home at the same rate. There are second- and even third-generation Americans being left behind, often being racialized in the process. The scope and distribution of this disparity determines what kind of America is in store for us.
Realistic policy and law—on school conditions, job training, language acquisition, and homeownership—can tilt the country away from failure and toward success. Accurate statistics cannot in themselves ensure the right kind of policy and law, but the primary race framework hides variation, and is a weak basis on which to understand, let alone manage, the demographic transformation now unfolding in America.
A takeaway lesson from Section II. To again cite Massey: “the absence of reliable data on the children of immigrants remains the single greatest weakness in the U.S. statistical system.” We apply Du Bois’s classic prediction of the twentieth century to twenty-first-century conditions: the problem of the twenty-first century is the problem of the color line as it intersects and interacts with the nativity line. The census is less help than it can be to our understanding of this dynamic.
Variability within the Primary Race Categories
Though variability within America’s broad race categories is a long-standing research interest, its relevance to public understanding and to policy was muted in the melting pot decades, when European nationality groups gradually assimilated, adding numeric strength to the white population and reinforcing the white/black color line. The second wave of immigration, 1970s to today, forced more attention to the nativity line—where were you, your parents, your grandparents born?—and the obvious growth of a multiracial and multiethnic population. It is in these population groups that we find substantial “within-race” variability.
Porter and Snipp, for example, note that only a few Asian countries—China, Japan, South Korea—supply the high percentage of technology workers immigrating with H1(B) visas, whose language skills, education levels, and career opportunities significantly differ from Vietnamese or Laotian immigrants. Today’s immigrants from Ghana or Ethiopia differ from, and in some cases try to avoid contact with, African Americans whose ancestors arrived as slaves. Telles’s complaint, cited above, that the Census Bureau is “out of step with popularly held notions of race and ethnicity,” is developed in his analysis of Latinos, where he emphasizes a racial hierarchy based on “phenotypic and color gradients.” This hierarchy leads some to treat their “blackness as an identity to be avoided.”
A less familiar within-group variation is described by Liebler, who reports that self-identification as American Indian-Alaskan Native (AIAN) is conditioned by any number of factors, including their tribal affiliation, whether they are responding to the term ancestry or race, and their level of trust in government (and its census-takers).
Scholars emphasize that Asian Americans are significantly more likely to identify with their national origin than the “Asian race,” a pattern especially prevalent in the first generation. Moreover, as noted by Lee, Ramarkrishnam, and Wong, national origin groups vary in income, occupational status, education, and health. Consequently, data disaggregation is “a civil rights issue.” Preserving the higher number of subgroup census options for the Asian race is strongly urged by academics and advocacy groups.
With approximately 11 million unauthorized persons living in the country, there are numerous policy issues relevant to how they differ from the remainder of the population. Capps, Bachmeier, and Van Hook take advantage of multiple datasets: the CPS that in a once-a-year supplement asks a question on parental place of birth and the Survey of Income and Program Participation (SIPP) that asks a question on legal status. The American Community Survey (ACS) asks neither of these questions, but has respondents’ place of birth and other characteristics permitting statistical modeling that can be used to assign legal status to immigrants in this much larger survey. By combining legal status information from the SIPP with the greater statistical power of the larger ACS, researchers have been able to derive estimates of small unauthorized immigrant populations in certain locations. They have been able to model the characteristics of Deferred Action for Childhood Arrivals (DACA) participants—a group of unauthorized immigrants receiving protection from deportation under the Obama administration policy. 6 Another instance was Hurricane Harvey (August 2017) flooding large sections of Houston, where the number of unauthorized immigrants is approximately 17 percent of Houston’s population of four hundred thousand—estimates critical to disaster response and communication strategies.
A takeaway lesson from Section III. Policy tied to the primary race categories, whether five or seven, makes little difference, seriously misses what a more granular measurement system provides, one making greater use of national origin categories and immigration status. Particularly problematic is the absence of African American and white subcategories in the current census classification scheme. This results, for example, in forcing recent immigrants from Ethiopia, Ghana, and Nigeria into the all-inclusive black category, despite significant educational, employment, or family structure differences from the African Americans whose ancestors most certainly did not voluntarily migrate to America (Prewitt 2013, 165–66). It is equally important to unpack white nationalism. When and how does it vary by national origin, religion, social class, or geography, and with what consequences?
The Reliability Problem
America’s first census racial classification system, administered in 1790, produced counts of three races: the Europeans, soon to be recognized as citizens of the new nation; the Native Indians in the new thirteen states, separately counted and taxed, but not slated for citizenship; and the enslaved, counted as three-fifths of a person and totally excluded from civic life. It was taken for granted that this racial classification produced reliably consistent results. It was unimaginable that a large number of Americans might be recorded differently from one census to the next.
Much history separates that 1790 starting point from what we know to be true of today’s census, when we routinely note the shifts in how Americans record themselves from one census to another, and across different information platforms and settings. We acknowledge and accept high levels of inconsistent and unreliable data in our race statistics. This was vividly demonstrated in the first application of mark-one-or-more. Forty percent of those who marked more than one race in the 2000 census did not give the same answer to a Census Bureau follow-up survey conducted a year later. Many had become monorace. The reverse was also true, as respondents shifted from single race in 2000 to more-than-one a year later (Bennett 2003). There is now a large research literature on multirace as a flexible identity. The self-esteem rationale invites this flexibility. Race is not biological; it is attitudinal.
Or, as Bratter notes, at any given time, those “who choose to mark more than one box are a nonrandom subset of those who could do so” (italics added). Those “who could do so” constitute a denominator that is not captured in the census. Roth makes a similar observation when she calls for “determining the denominator,” that is, all the people who could identify in the multiracial group. However useful this might be for the research community, there is no realistic way the Census Bureau could determine that hypothetical denominator. Should the bureau have coded Obama as black and white so he would be in the correct denominator? If so, what of the unknown number of the multiracial who reported only one race in the 2010 census, but neglected to tell the New York Times they had done so.
How mixed-race couples identify their children adds a further level of complexity and confusion. Lichter and Qian find that a sizable share “of America’s children from mixed-race marriages are identified by their parents as monoracial.” More than 25 percent of the black-white and Asian-white couples do so. And there is negotiation by parents in deciding which race to favor, with various factors having an influence: the educated parent prevails or the composition of the neighborhood serves as a reference point. Overall, there is a drift in favor of whiteness, underscoring Alba’s doubts about the majority-minority future.
Compounding this problem is the tendency for people to make strategic choices. Deaux writes that people present different racial identities as their social settings shift. The extent to which this occurs varies from one racial group to another. We are beginning to understand that race groups differ in the strength of their respective classification norms. Whites, blacks, and Asians have comparatively stronger, more stable norms. Hispanics and Native Americans have weaker norms and, thus, are more likely to shift depending on context. The multiracial identifiers are least likely to give consistent responses across time or settings.
Morning and Saperstein note yet another inconsistency in the data. They start with the census report that America’s multirace population is approximately 2.5 percent, then, drawing on the Pew Survey of Multiracial Adults, warn that the 2.5 percent estimate is seriously misleading. It does not take into account the possibility of multiracial ancestors earlier in the family tree, before multiracial was recognized by the census. I offer a personal example. When I was five or so years old, on a visit to my paternal grandparents, an older cousin shared a hush-hush family secret; my grandmother was part Cherokee. If true (to a five-year-old she looked the part), my father was multiracial, as then am I, though both of us appear in the census record as white only—an identity that mark-one-or-more now invites me to reconsider. Morning and Saperstein write that the “multiracial population is not solely a ‘first-generation’ community, but rather a layered collective including second, third, and further generations.” America’s assimilation patterns and census practice have historically erased multirace ancestors. Morning and Saperstein, taking advantage of the Pew survey data, report a current multiracial rate of 18 percent. This, they write, is a more accurate multiracial denominator than the 2.5 percent, an argument that gives further support to the Bratter and Roth pieces.
A takeaway lesson from Section IV. There is fluidity in race statistics, prominently visible in the multirace data but not limited to that population group. Trend lines on health and employment disparities or analysis of the hardening/softening of racial boundaries or understanding rates and patterns of immigrant assimilation will vary in their accuracy as classification norms vary in their stability. The direction and scope of the bias are difficult to detect. It is another strong signal that our ancient classification categories do not adequately capture the social and demographic realities of the nation—a misalignment that frustrates the academic, misinforms the policy-maker, and confuses the public.
What Is Fixable?
The first and second conditions discussed—the majority-minority problem and the inadequacy of measuring generational differences—have easy technical fixes but face political and bureaucratic challenges. The majority-minority storyline is now too embedded in the surge of white nationalism to be easily erased. We can expect it eventually to fade if, as expected by demographic sociologists, there is a steady shift toward the white category. But that is at least a generation away. Adding parental place of birth to the ACS is technically easy. The barrier is the paper reduction act, which works to keep census questions to a minimum. Adding a question would probably require dropping a question now on the form, and every question has its constituency. 7
The third and fourth conditions—variability within and between different race and national origin labels and less than reliable ethnoracial statistics—are substantively very different from the first two. They are embedded in the census numbers and cannot be whisked away with a technical fix. But academic research using census and federal statistics more generally should be alert to what authors in this volume emphasize: within-group variability is high, often greater than between the primary races, and there are serious reliability issues in our race statistics.
Conclusion
Is the census racial classification doing its job? Obviously “yes” is the answer but, of course, with the always present qualifier: “there is room for improvement.” The Census Bureau launched its major 2010 research, on the promise that the results of that research would help to improve the accuracy and reliability of census statistics.
With improvement in mind, I published a book with the subtitle The Census and our Flawed Efforts to Classify Americans (Prewitt 2013). The Princeton University Press editor asked why not “failed” instead of “flawed.” Because, I replied, the census only fails if its numbers go unused, clearly not the case for any census-produced counts, least of all for its race and ethnicity numbers. What, then, to do about its flaws? The core of my argument is incremental change—take one step in 2020, another step a generation or so later, and yet another later in the century. 8
When we reach the endpoint, the census will be making less use of the eighteenth-century primary races, with the possible exception that the African American and Native American populations will still be better served by retaining their identity as such. They uniquely experienced centuries of punitive and cumulative disadvantage; it does not get fixed in decades, even when there is a will to do so.
The Hispanics, Asians, Native Hawaiian Pacific Islanders, and, if a separate race, the Middle-Eastern North Africans, will, I suggest, benefit from measurement more attuned to their variability and diversity based on national origin and linguistic differences. They also, more recently arrived, will assimilate differently from one national origin to another, a process that will occur across the rest of the century (maybe beyond, depending on immigration rates in the decades ahead).
My guides in reaching this conclusion are Michael Omi and Howard Winant, who ask whether America is best understood through a more cultural or more racial lens (Omi and Winant 1994). Critics of the cultural focus claim that it ignores the persistent color line that was and remains racist in its construction, including the American instinct to racialize any newly arrived nonwhite immigrant groups. The counterargument is that the racial lens hides too much about the highly varied way in which different immigrant groups have made a place for themselves in American society.
Census statistics are not a neutral bystander. Insofar as they are structured as Blumbachian primary races, they favor the more racial interpretation and, in their lack of detail about national origin and the importance of generational differences, they miss much of importance to a cultural interpretation. I want to loosen the grip of the primary race classification.
David Hollinger has made clear that this loosening need not come at the expense of robust racial measurement. The design can be sufficiently flexible that the analyst, advocate, and policy-maker can have it both ways (Hollinger 2005). The proposal now under review in the OMB is a major step in this direction. It allows data to be disaggregated by national origin/tribal affiliation. It also allows the data to be reaggregated into the primary races, be they five or seven. It also allows special purpose groups to be constituted for academic or policy purposes. For example, all national origin/tribal groups with limited education could be assembled—white Appalachians, inner-city African Americans, Hispanic migrant workers, Native American Indians on impoverished reservations, selected Asian nationalities. Today’s race borders need to be porous. This is happening in the lived experience of millions of Americans; it needs to happen in the statistics that try to capture the lived experience. The articles that follow make a compelling case for greater granularity and flexibility in federal statistics, led by the Census Bureau.
Footnotes
Notes
Kenneth Prewitt is the Carnegie Professor at Columbia University, President of the American Academy of Political and Social Science, past director of the Census Bureau, and author of What Is Your Race? The Census and Our Flawed Effort to Classify Americans (Princeton University Press 2013).