Originally Posted by NILP
By John R. Logan (November 19, 2012)
Since 1980 the Census Bureau has acknowledged the growing size of the country’s Hispanic population by asking people to identify as Hispanic or non-Hispanic in a separate question from the traditional one on race. These two questions provide the only full enumeration of Americans by race/ethnicity. Other questions (such as ancestry, country of birth, and parents’ country of birth) have always been asked only from a sample of people.
The Bureau now is considering a recommendation to combine the race and Hispanic origins items into a single question. One rationale is that the current questions lead to high “item nonresponse.” Between 0.6% and 1.2% of persons do not respond to a combined question. Nonresponse to the two-question format is higher at 4-5% because people often answer only one of the two questions, though only around 1% of people fail to answer at least one of them. Is this a problem?
The Bureau corrects for nonresponse by “borrowing” information from “similar” people in the household or neighborhood. This was done in the 2011 American Community Survey for 8% of respondents on occupation, 5-8% on health insurance coverage, and 16% on wage income. In comparison to these variables, nonresponse on race/Hispanic origin is a minor concern. I suspect that statisticians focus on nonresponse because it is among the few things they can measure. But just because we can measure it does not mean that it is important. I am much more worried that we have to make guesses about so many Americans’ jobs, incomes, and health insurance.
Another rationale is that a large share of Hispanics selects “some other race” rather than white, black, or another race category. For example, they may write in “Mexican” on the race question. The Bureau thinks of this as misreporting because “some other race” is not a race category as defined by the Office of Management and Budget. It was included in the expectation of being a very small residual category.
My view is that the Census stumbled onto a very important phenomenon. Now we know that many Hispanic Americans do not think of themselves in terms of black and white. We have also discovered that a majority do select a race category, most white, few black. Racial identification also varies widely across Hispanic groups – quite low shares of Cubans report as black, for example, compared to quite high shares of Dominicans. The pattern would not be so clear if Hispanics did not think of themselves in terms of race.
We live in a country with many disparities and boundaries organized by social categories that are hard to define but are nonetheless very real in people’s lives. We acknowledge the stakes when we define some people as members of protected classes and when we take racial and ethnic composition into account in organizing political representation at the Congressional, state, and local level. The Census Bureau’s most critical function is to provide information about American society and how it is changing. We rely on data about race and Hispanic origin to inform us about progress toward equal opportunity or about potentially disparate racial/ethnic impacts of policy decisions.
It does not make sense to base decisions on how we collect these data on small variations in response rates or on how policy makers at OMB would like us to understand race. The key criterion must be how well the data inform us about the patterns and complexities and changes over time in the nature of America’s color lines.
John R. Loganis Professor of Sociology at Brown University and director of Russell Sage Foundation’s US2010 project on recent trends in American society. He can be reached at email@example.com.