Diversity is in the Eye of the Beholder
This feature is part of an ongoing RSF blog series, Work in Progress, which highlights some of the research of our current class of Visiting Scholars.
With factors such as increased immigration and interracial unions propelling racial and ethnic diversity in the U.S., many have predicted that the nation will become “majority minority” in a few decades’ time. Yet, some researchers, such as former Visiting Scholar Richard Alba (CUNY Graduate Center), have argued that the U.S. is likely to remain majority-white as racial boundaries shift and more groups are incorporated into the mainstream. In other words, our idea of diversity today is contingent upon our society’s perception of who “counts” as white.
Perceptions of diversity also deeply inform how we view our environments at the individual level. Visiting Scholar Cara Wong (University of Illinois at Urbana-Champaign) is currently studying individuals’ ideas about the racial and ethnic diversity of their neighborhoods. Using a new map‐drawing measure of people’s “local communities” and multiple survey datasets, Wong and her colleagues are exploring how individuals’ perceptions of the racial makeup of their locales affect their intergroup attitudes.
In an interview with the Foundation, Wong explained how the social sciences have traditionally examined people’s neighborhoods, and discussed how further investigation of people’s perceptions of race and diversity can help provide new frameworks for more effective housing policies.
Q. Your current research examines the gap between people's "objective" neighborhood contexts and their perceptions of those contexts, focusing in particular on race and ethnicity. What kinds of problems do social scientists face when they attempt to analyze people's environments, and how does studying perceptions add a new dimension to research on racial inequality?
Wong: When scholars study the effects of place on attitudes and behavior, they have to confront theoretical, logistical, and statistical challenges. They need to measure the “right” context, the one that best captures the concept and mechanisms believed to be at work. In other words, when social scientists explore “neighborhood effects,” they want to understand how people’s neighborhoods affect them, above and beyond their individual characteristics (for example, do the voting decisions of a given African American change depending on whether she is living among many whites or blacks?) The theoretical problems arise in trying to figure out exactly what counts as her neighborhood. For a given behavior and theoretical problem, should a scholar define “neighborhood” as a Census block group, a voting precinct, an area defined by major traffic thoroughfares, or something else altogether?
The logistical challenge is finding the data for the contextual unit that social science theories say should matter. Perhaps we believe Census blocks should matter, but we cannot obtain racial data at that level of specificity—surveys oftentimes restrict our ability to identify respondents to that small of a unit for reasons of confidentiality. So, we satisfice and use what we can. Here is where we have to confront statistical challenges. There are certain well-known problems, whereby relationships between variables can change and even flip signs, depending on the level of analysis. In other words, if we were to look at the relationship between Republican voting and the size of the senior citizen population in a city, county, or zip code, we might reach very different conclusions depending on which geographic unit we used to measure context. If we can only get data at one geographic level, how do we know our finding reflects some causal process by which place constrains or changes the mind of an individual rather than some artifact caused by aggregation? Or, if I have data at two levels and the results change depending on which unit I use, how should I interpret the difference? Maybe the change has a theoretical explanation, but it could also be a statistical artifact. And, this problem is exacerbated when people are not distributed randomly across geography (which is hardly ever the case).
Studying people’s perceptions of their environments helps us avoid the statistical challenges while enabling us to understand better the mechanisms made explicit in our theories, particularly when the outcomes of interest are political attitudes. By bringing “context” down to the individual level and allowing each individual to carry his or her own context (however defined), we have no statistical aggregation problems. There are times when contexts matter at levels above the individual, regardless of whether we see them; pollution in the air can affect my chances of developing asthma, irrespective of my ability to observe the particulates. However, when we talk about our environments leading us to feel threatened, we assume people are first noticing the places where they live, and that these observations then cause them to feel fear. This assumption was never carefully tested.
For our work, we developed a map-drawing measure of context, where we first ask people to draw their local community on a map (allowing them to zoom in or out as much as they would like), and then ask them to describe that community in terms of its racial/ethnic, partisan, and economic diversity. That way, we are measuring the environment that is most relevant, central, and salient to an individual; the measure is not modifiable or arbitrary, helping us avoid one major statistical problem; and we measure what an individual sees (and possibly fears or celebrates), and can thus compare people’s perceptions with what government bureaucracies tell us what exist in an area.
Q. You and your colleagues conducted a study to measure Toronto residents' perceptions of the diversity of their communities. How did you assess and measure these perceptions, and how did they compare to Census/survey data? What effects did the perception of higher neighborhood diversity have on respondents?
We conducted an online survey of English-speaking Canadians, and we collected data from about 7,000 individuals living throughout the country. Participants drew their local communities and described them, and they also described one other geographic context that we named and showed them on a map. (They were shown one of 6 administrative units in which they lived, ranging from a couple blocks—their Census Dissemination Area—to their country.) We could therefore compare individuals’ perceptions across multiple contexts, as well as perceptions of different individuals living within the same context.
In general, people overestimated the number of minorities and underestimated the number of whites living in all contexts in which they lived. What does this mean substantively? Imagine two white neighbors living in the same place: the one who thinks she lives among more minorities is more likely to think that others in her community do not share the same values, that they will not work together for a common good, and that she is unsafe in her neighborhood.
Q. How can policymakers better take into account people's perceptions when crafting neighborhood and housing initiatives?
The simplistic answer would be to say that policymakers should educate the public when they introduce new housing and zoning initiatives. Ignorance and stereotypes obviously will dampen the chances for success, especially if misperceptions lead people to make suboptimal decisions. Unfortunately, research on information processing and motivational biases suggests that it is very difficult to change people’s minds about national events and facts, especially when they think they understand an issue and already have a position about which they feel strongly. Nevertheless, we have to try. It is possible that information about local neighborhoods and cities is more concrete and verifiable by experience, and thus, the factual arguments may have a better chance of being believed, even if they conflict with prior beliefs and commitments. And, the goal of making accurate information more easily accessible and available is certainly more feasible than randomly assigning people where to live.