Place, scale and the racial claims made for multiracial children in the 1990 US Census

Abstract

Multiracial children embody ambiguities inherent in racial categorization and expose fictions of discrete races. Nevertheless, parents of multiracial children were asked for the 1990 US Census to report a single race for their offspring. Using confidential 1990 Census micro-data, we investigate the choices parents made for the three most common racially mixed household types (Asian-white, black-white and Latino-white) in twelve large metropolitan areas. We find that context affects the reporting of children's racial identity. We examine these effects with models that incorporate three spatial scales: households, neighbourhoods and metropolitan areas. Model estimates reveal that racial claims made by parents of Latino- and Asian-white (but not black-white) children varied significantly across metropolitan area. A neighbourhood's proportion white increased the probability that parents reported their children as white, while a neighbourhood's racial diversity increased the probability that black-white parents claimed a non-white race (black or ‘other’) for their children.

Keywords:

• Racial claims
• mixed-race households
• multiracial children
• neighbourhood
• scale
• racial identity

Acknowledgements

This research was made possible by a grant from the Russell Sage Foundation. We thank Rebecca Acosta and the California Census Research Data Center for assistance with the data and Serin Houston for her research support and valuable editorial comments. Previous versions of this paper were presented at the Race, Ethnicity and Place Conference, Howard University, September 2004, and at a seminar at the Baldy Center for Law and Social Policy, SUNY Buffalo, November 2004. We thank the Baldy seminar participants for their input and especially Catherine Cornbleth, Amy Kedron, Charles Lamb and Judy Scales-Trent for their commentaries.

This paper reports the results of research and analysis undertaken while the authors were conducting a study approved by the Center for Economic Studies at the US Census Bureau. It has undergone a Census Bureau review more limited in scope than that given to official Census Bureau publications. Research results and conclusions expressed are those of the authors and do not necessarily indicate concurrence by the Census Bureau. It has been screened to ensure that no confidential information is revealed.

Notes

1. This layering will always be imperfect because, as Goldberg observes, the census form ‘lags behind the complex negotiations of identities and (self-) identifications in everyday experience, even as it serves in part to shape and fix those identities and identifications’ (1997, pp. 32–3).

2. Most US research that measures contextual effects on parents’ racial claims for their children has relied on large geographic units (Public Use Microsample Areas) containing a minimum population of 100,000. Our analysis is the first to use considerably smaller census tracts that contain between 4,000 and 8,000 people, because the Census Bureau provided us with access to confidential information on individuals in households by census tract of residence for the 1990 Census. While considerable debate remains in the social science literature about the most appropriate scale of neighbourhood cartography (e.g. Martin 2003), and we could have used smaller block groups, census tracts are most suitable for two reasons. One, our broader project engages the existing racial residential segregation literature – a literature that predominantly uses census tracts. Two, despite having access to the full one-in-six sample of long-form respondents, our sample sizes remain small. Census tracts seem to us an appropriate compromise between the detail that a smaller geographic unit might provide and the precedent of previous literature and our data demands.

3. We follow demographic convention in treating all respondents that indicate Latino or Hispanic ancestry as a distinct racial category, regardless of their response to the race question. In so doing we do not assume away the multiple identities subsumed with this meta-category (e.g. those based on national origin, nativity, class and political ideology, among many others). Moreover, we explicitly reject the notion that the terms ‘Latino’ or ‘Hispanic’ mark a meaningful category of ethnic identity. Rather, we see the creation of pan-ethnic Latino/Hispanic identity in the US as a process of racial formation. Also, noting ongoing terminological debates, we use the terms Latino and Hispanic interchangeably.

4. Our larger project includes racial mixing between non-white adults. Because the children from these and other mixed-race pairings were numerically infrequent, we cannot publish such data as we would risk disclosure of information deemed confidential by Census Bureau.

5. Even when women retain the names with which they were born, children still most often carry their father's family name.

6. We assume that parents fill out the census questionnaire in households with young children.

7. We identify mixed-race households as those where the race of the adult male and adult female (where the relationship to the household head is reported as married or partner) differ across the standard race categories used in the 1990 Census (white, black, Asian, Indian, other). We treat all respondents who indicate Latino or Hispanic ancestry as a distinct racial category, regardless of their response to the race question (see note 3 for additional discussion).

8. We configured the sample to mitigate endogeneity between residential location decisions and parents’ views of their children's racial identity. The age restriction improves our assumptions that parents fill out the census form. It also reduces the chance that parents may declare racial identity as children see themselves, which is more likely in older children and reflects non-residential settings. We initially restricted the sample to households that did not move between 1985 and 1990 to eliminate recent residential moves made with children in mind. This restriction, however, removed over two-thirds of the sample. We kept recent movers in the sample and controlled for their residential mobility with a variable in the models.

9. The standard entropy diversity measure for each tract is , where k indexes racial groups, j indexes census tracts and t indexes total tract population. The maximum value of Ej is obtained when tract j's population is evenly divided between the k racial groups. Since the number of racial groups limits the magnitude of the maximum value, we include a scaling constant s so that Ej ranges between 0 and 1. We calculated Ej based on individuals in the six standard racial groups (Latino, white, black, Asian, Native American and other).