bringing the person back in: boundaries, perceptions, and ... · bringing the person back in:...

Bringing the Person Back In: Boundaries,Perceptions, and the Measurement ofRacial Context

Cara Wong University of Illinois

Jake Bowers University of Illinois

Tarah Williams University of Illinois

Katherine Drake Simmons Pew Research Center

Place is sometimes vague or undefined in studies of context, and scholars use a range of Census units to measure‘‘context.’’ In this article, we borrow from Parsons and Shils to offer a conceptualization of context. Thisconceptualization, and a recognition of both Lippmann’s pseudoenvironments and the statistical Modifiable Areal UnitProblem, lead us to a new measurement strategy. We propose a map-based measure to capture how ordinary peopleuse information about their environments to make decisions about politics. Respondents draw their contexts onmaps—deciding the boundaries of their relevant environments—and describe their perceptions of the demographicmake-up of these contexts. The evidence is clear: ‘‘pictures in our heads’’ do not resemble governmental administrativeunits in shape or content. By ‘‘bringing the person back in’’ to the measurement of context, we are able to marrypsychological theories of information processing with sociological theories of racial threat.

The study of individuals interacting with theirenvironments has a long history (Lewin 1936)and a vibrant presence today in the large

literature on ‘‘context effects.’’ The results are fascinat-ing, but practical matters may hinder continuedprogress. In particular, scholars could make the follow-ing observations about the last 75 years of this research:(1) we use contextual measures that are readilyavailable; (2) different measures of context generatedifferent results; (3) people’s perceptions of theirenvironment do not resemble governmental units;and (4) people define their environments differently.Taken together, these critical observations suggest thatwhen we use administratively drawn geographic unitsas ‘‘context,’’ we risk confusion of multiple types.1

Given these observations, why would scholarscontinue to use data collected at the level of an admin-istrative unit as measures of people’s contexts almostexclusively? There are at least two reasons for this

persistence: many scholars do not have the resourcesto collect alternative data, and more importantly, wedo not know how to determine whether a unit isthe ‘‘right’’ one. But, we often require our studentsembarking on research projects to answer two ques-tions: Are the concepts and measures clear and mean-ingful? And, how does the proposed design relate tothe ideal? The observations made above imply that itwould be difficult to answer these fundamental ques-tions. To remedy this problem, we offer a theoreticalfoundation for contextual effects, propose a datacollection methodology commensurate with it, andpresent illustrative data as proof of concept.

Our new map-based measure of context ought tohelp future researchers answer both questions ofmeasurement—what kind of geographic units oughtto be most relevant for individual attitudes andbehavior?—as well as questions of mechanism—howdoes place get into the heads of people, and how do

The Journal of Politics, Vol. 74, No. 4, October 2012, Pp. 1153–1170 doi:10.1017/S0022381612000552

! Southern Political Science Association, 2012 ISSN 0022-3816

1We thank the Interdisciplinary Institute at the University of Michigan for funding the pilot study, and the University of Illinois Centeron Democracy in a Multiracial Society for a fellowship for Wong. An online appendix for this article is available at https://journals.cambridge.org/jop containing supplemental analyses. Data and supporting materials necessary to reproduce the numericalresults in the paper are available at www.carawong.org.

1153

This content downloaded from 130.126.162.126 on Mon, 19 Jun 2017 17:00:10 UTCAll use subject to http://about.jstor.org/terms

people represent places in their minds? The oper-ationalization of context needs to capture the manydifferent ways that the macro-social demography ofa community is linked to the political psychology ofits residents. Therefore, we ask respondents to drawtheir contexts on maps—wherever they see theboundaries of their communities—and then describethe content of these contexts. With our new measureof context, we examine whether people’s ‘‘pseudoen-vironments’’ reflect the ‘‘objective’’ context acrossmultiple units of analysis and discuss the distancebetween common practice and our conceptualizationof context. We also ask why levels of misperceptionmay vary at different geographies and across differentattributes of an area, raising new questions about themechanisms by which contexts affect individuals.

The results help deepen and extend our knowledgeabout how context matters in politics, showing thatgovernmental administrative units do not define theenvironments relevant to individuals, and that peopledo not ‘‘see’’ what the Census ‘‘sees.’’ Census numbersmay be related to political judgments, but only bystudying how these facts become beliefs can we un-derstand more fully the mechanisms by which placeaffects politics. By measuring an environment that isrelevant to the individual, researchers can also avoid aserious methodological problem—the modifiable arealunit problem—that could explain the inconsistency ofresults in research on racial threat and contact.

After a quick summary of some of the meanings ofcontext and contextual outcomes, we explain why aclear conceptualization will help us address some ofthe observations made about previous research. Wethen describe why there are problems in measurementat the level of both the aggregate contextual unit aswell as the individual. Finally, we use a pilot study toillustrate the use of new ways to approach the study ofcontext and compare the effects of different measures.

Situating Research on Context

The places where people live can affect them in amyriad of ways because context can be both objec-tive and subjective, and the outcomes can be phys-ical or psychological (Sampson, Morenoff, andGannon-Rowley 2002).2 For example, in the case of

pollution, environmental allergens in the air breathedby individuals have a direct physical effect on theirlungs and health. This ‘‘context effect’’ requires noknowledge on the part of its residents about theenvironment; one does not have to see and recognizesmog to be affected by it. Similarly, living in a poor areacan minimize people’s ability to engage in politics dueto limited access to information and politicians’ atten-tion, all without entering into these residents’ calcu-lations about whether to participate in politics; theymay not realize resources are missing, but the contextsin which they live can still influence their behavior.

People’s surroundings can also have less directeffects, particularly when the outcomes are psycho-logical in nature. For many important politicalphenomena, people must be cognizant of theirenvironment for it to have an impact. Living in apoor area, for example, can also lead individuals tofeel threatened; living in an area with many racialoutgroup members can lead to fear for one’s job,identity, or political voice (Blalock 1967; Taylor1998). These perceived threats are the main concernof political scientists focused on racial context (see,for only a few examples, Glaser 1994; Key 1949;Oliver 2010). However, if these environmental char-acteristics lead to feelings of threat, someone—elites,if not also ordinary residents—must be cognizant ofthe context (Lazarsfeld, Berelson, and Gaudet 1968).After all, fear is an emotional response to a perceivedthreat, and the brain acts as the filter for recognizingoutgroup members as ‘‘pollutants’’ or threats.

Just as the outcomes of context can be bothphysical and psychological, context itself can beobjective—defined by gates, walls, and borders—aswell as subjective (Wong 2010). Political sciencemeasurement of racial context tends to emphasizethe former. When we conceptualize the environmentas a container (of recognizable and commonly shareddimensions), it is sufficient to pick the size of thecontainer, measure its contents, and test for relation-ships between said content and our outcome ofinterest, whether that is a propensity to developasthma or prefer a particular policy. The researcherwill still need to justify the choice of the container sizefor statistical and theoretical reasons discussed later.But, if ‘‘context’’ is conceptualized as an ‘‘objective’’container, it is unnecessary to ask residents if they areaware of the container and its contents.

However, while one’s environment may serve as aphysical container within which individuals exist andinteract, ‘‘context’’ can also refer to the environmentin which people believe they live; the ‘‘pictures intheir heads’’ or their ‘‘pseudoenvironments’’ are the

2In this paper, we set aside explicit considerations of personalnetworks of acquaintances and focus on the perceptual mecha-nisms of impersonal interactions only. We do discuss how somemeasures of impersonal interactions may reflect interpersonalmechanisms and focus on ‘‘geographic’’ rather than ‘‘socialcontext’’ (Huckfeldt and Sprague 1987, 1995).

1154 cara wong et al.


reality with which people engage, regardless of theaccuracy or distortion of these perceptions (Lippmann1991). While it is a truism in social science researchthat race is a social construct, we want to emphasizethat people’s racial contexts are also social con-structions, as well as seemingly objective attributes ofa physical reality. Researchers have learned muchfrom studies of phenomena like white flight, tippingpoints, and racial discrimination in housing, insur-ance provision, and medical care, all of whichemphasize the importance of how individuals per-ceive, construct, and evaluate the race and ethnicityof those they observe and with whom they interact(see, for example, Yinger 1997). Recognition of thesemultiple types of outcomes and perceptions of con-text is important because it highlights the differentmechanisms by which context can affect politics, justas adding frustration-aggression displacement the-ory to realistic group conflict theory emphasizeshow ‘‘perceived’’ shortfalls can matter as much as‘‘objective’’ deprivations (LeVine and Campbell 1971).

Conceptualization of Context

While political scientists have written about howenvironments matter for politics for almost a century,geographic ‘‘context’’ is very rarely defined in con-temporary articles, perhaps because its ordinary lan-guage usage is shared by researchers. It seems almostsuperfluous to state explicitly that ‘‘context’’ is theplace, area, or environment in which people live.However, these general definitions serve more todelineate what Adcock and Collier (2001, 530) wouldcall the ‘‘background concept’’ (including all possiblemeanings associated with the concept) rather than the‘‘systematized concept’’ (the specific formulationadopted by a researcher). This lack of specificity leavesscholars with little guidance as they move fromconcept to operationalization, making the dangerof operationalism very vivid (Blalock and Blalock1968). For example, context simply becomes percentblack or ethnic diversity in a county.

The concept of context should convey what unitof analysis might be most appropriate. Fortunately,there is no need to reinvent the wheel, nor is it nec-essary to develop an idiosyncratic conceptualization;instead, we use a classic social science definition. InToward a General Theory of Action, Parsons and Shilsdescribe three components of their theory: ‘‘actors,a situation of action, and the orientation of the actorto that situation’’ (1951, 56). The situation of action

is that part of the external world which means some-thing to the actor whose behavior is being analyzed. Itis only part of the whole realm of objects that might beseen. Specifically, it is that part to which the actor isoriented and in which the actor acts. (56)

We believe that Parsons and Shils’s ‘‘situation ofaction’’ is the context to which scholars refer whenthey talk about context having an effect on individuals’political attitudes and actions. Note that this conceptencompasses containers; a container is just a ‘‘part ofthe external world’’ toward which people are assumeduniformly oriented, in terms of both its boundariesand contents. In other words, using the Parsons andShils definition does not preclude the instances inwhich objective characteristics of a fixed containeraffect individuals’ attitudes and actions directly.

According to this conceptualization, a contexteffect can encompass the effects of communication.For example, one could imagine an individual whoreceives all of his information about his context,accurate or not, from a demagogue who warns ofencroaching outgroup members. (The individualcould be visually impaired or blithely unobservantof his environment.) If the demagogue mobilizesvoters using threats posed by outgroups withoutreference to a geographic area—possibly stressing aviolation of ideology or values—then this would notbe a contextual effect. In contrast, if informationgiven about his surroundings then affects this indi-vidual’s attitudes and actions, we would argue thatthere is both a contextual and communication effect.

Measurement of the Contextual Unit

Political scientists almost always justify their choiceof contextual unit on theoretical grounds. For exam-ple, because individuals rarely live and work in thesame census tract—and the effects of racial context inone’s life overall, not just one’s neighborhood, mayhave political effects—county may be the ideal con-textual unit of analysis despite its extreme hetero-geneity (Branton and Jones 2005). Alternatively,because racial context may be understood as theinteraction between ingroup and outgroup mem-bers, the unit of analysis should be small—like theneighborhood—in order to capture actual contact be-tween individuals and significant social relationships(Gay 2004). Or, because the outgroup is defined atthe national level (i.e., immigrants), the appropriatecontext is the country (Quillian 1995). Political elitesat the national level may also prime a local context

boundaries, perceptions, and the measurement of racial context 1155


that is, or is becoming more, diverse (Hopkins 2010).And, scholars argue that multiple contexts, macro-and microenvironments, may interact and should beconsidered simultaneously; this raises the additionalquestion of which of the many possible combinationsof contexts to consider (Key 1949; Liu 2001). As Oliverand Mendelberg explain, ‘‘Identifying a context’sboundaries is essential for understanding its potentialeffects’’ (2000, 577).

While the different theoretical arguments justi-fying area size are reasonable prima facie, a majorreason for why a particular unit may be chosen instudies of racial context is that of practical necessity.As Glaser describes,

. . . the social scientist’s choice of neighborhood,precinct, county or state is often arbitrary, determinedby what is available in a dataset, and there is noguarantee that he/she is really capturing racial balancein a way that is salient to respondents. (2003, 608)

Because few studies are designed explicitly to addressquestions of racial context, scholars are constrainedby the units available in a given dataset.3 Many studies,like the General Social Surveys for example, do notautomatically identify respondents at multiple geo-graphic levels in order to protect the anonymity ofsurvey respondents; one by-product is that scholars areoften limited to asserting that the contextual unit theyuse is ideal rather than testing and proving it. Andunfortunately, the many studies of racial context havenot converged to a consensus about the ideal con-textual unit. In their article, Tam Cho and Baer (2011)provide an excellent summary of many examples ofthe research on racial context; across 34 studies, thegeographic units chosen vary from prison cellsto countries. And, even when the same contextualunit of analysis is chosen, they show the results cansupport contact theory, threat theory, or even bothsimultaneously. They also point out that these dispa-rate findings could arise because of a statistical artifact.

In response to the cacophony of units and results,Tam Cho and Baer recommend that more than oneunit should be used in analyses, partly as a robustnesscheck, but mostly to be aware of the potentialconflicting results that can arise from the ‘‘modifiableareal unit problem’’ (MAUP). The MAUP is a phe-nomenon well-documented by geographers, wherebyrelationships between variables at one level canchange when studied at a different level of aggregation;the areal units chosen are ‘‘modifiable’’ or arbitrary

(Achen and Shively 1995; Cho and Manski 2008). TheMAUP is actually composed of two problems con-cerning scale and aggregation. Gehlke and Biehl (1934)find that larger units lead to larger correlation co-efficients, even when the correlation at the individuallevel across units is constant, showing that scale canaffect statistical inference. Wong explains:

. . . a general characteristic of the scale effect is tosmooth out extreme values so that the range of thevalues is narrower . . . If one follows the logic thatmore spatially aggregated data are less variable, andthis logic is extended to analyze correlation betweenvariables, it is not difficult to come to the conclusionthat data at the higher aggregation levels will likelyhave higher correlation than more spatially disag-gregated data. (2009, 5)

The aggregation effect is easily understood if oneconsiders the problems of redistricting after a census;even with the constraint that districts must haveroughly the same number of residents, there are mul-tiple possible maps that can be drawn. Districtsdrawn by individuals with different political leanings,for example, rarely lead to the same number ofRepublicans and Democrats elected.

Openshaw and Taylor (1979) show both thescale and aggregation effects of the MAUP using therelationship between the percentages of Republicanvoters and of elderly voters in Iowa. They comparedthe correlations in the state’s 99 counties as well as inall possible combinations of these counties into largerdistricts. They were able to find correlations rangingfrom -.97 to 1.99, with no clear pattern between thespacial characteristics of the districts and the varia-tion in the coefficients.

The effects of the MAUP are less severe if theaggregation is done in a noncontiguous or spatiallyrandom way or if the variable of interest is randomlydistributed. However, racial groups in the UnitedStates are not randomly distributed. And, given thattracts are composed of contiguous block groups,counties of contiguous tracts, and so on, the aggrega-tion effect of the MAUP may be even more severe inpolitical science research on racial context. Furtherresearch has shown that the MAUP affects multivariateregression, Poisson regression, multilevel models, spatialinteraction models, and spatial autocorrelation statis-tics, along with simpler statistics like the mean, variance,and correlation coefficient (Gotway and Young 2002).Scholars have proposed complex statistical solutions forthe MAUP, but each has its own limitations, set ofassumptions, and critics (King 1997; Wong 2009).

Not many political scientists have looked atcontextual effects at multiple levels, and almost all

3Although more recent studies have tended to make moregeographic identifiers available, this was a common problemwith older surveys.



explain discrepancies across levels in substantive terms,rather than noting the MAUP (Baybeck 2006; Oliverand Mendelberg 2000). For example, an argumentcould be made that racial diversity has an effect at themetropolitan level and not at the neighborhood levelbecause threat is only experienced at the larger,aggregate levels (Oliver 2010). It is rare for scholarsto test whether their findings are consistent acrossmultiple levels, which Wong (2009) argues is theminimum standard in handling the MAUP (seeHopkins 2010, for example). An equally good explan-ation for differences in coefficients across scales is theMAUP. We have no way of knowing whether thesedifferences arise because of the explained substantiverationales or whether the differences are statisticalartifacts of the MAUP. Even if one takes into accountthe effects of nearest neighboring contextual units,there is still no way to adjudicate between onescholar’s argument that census tract is the best unitof analysis and another’s that city is superior.

Openshaw (1984) explains that the ‘‘simplestsolution’’ to the MAUP has been to ignore it andthat ‘‘the general absence of comparative studies mayhave helped to disguise the extent to which zone-dependent regularities are being uncovered’’ (31).While simple statistical solutions to the MAUP areelusive and ignoring it is problematic, we proposehere a way of sidestepping the MAUP. The choice ofunit should not be limited by the constraints of asurvey (e.g., geographic identifiers gathered for re-spondents), and it should have geographical mean-ing. Census data, for example, are reported formodifiable areal units (blocks, tracts, etc.), followingcriteria determined by political and logistical consid-erations rather than because they are ‘‘natural areas’’with intrinsic geographical meaning (Hatt 1946).4 Ifwe think of the Parsons and Shils conceptualizationof context and the questions that motivate scholarsof racial threat, the real question of interest is howpeople react to their surroundings. Such a conceptu-alization naturally emphasizes new possibilities foroperationalizing context: given this definition, whatis the ideal context to be studied? The ‘‘situation ofaction’’ should be measured at levels that individuals‘‘see,’’ and it should be allowed to vary by individual.The ‘‘external world’’ that matters may also vary frompolicy to policy.

In the new measures we develop and test, we askindividuals to define their own relevant contexts;

rather than assume that administrative units are placeswith meaning to individuals, we ask our respondents todraw their contexts on maps. The level of interest doesnot have to be less aggregated than the level of dataavailable: we care about the context to which people arereacting, and we can gather individual-level data onboth the context seen (i.e., its size, boundaries, andcontent) and the reactions to it. With this new measure,we are also able to answer the question of how theboundaries of people’s communities match up toadministrative units used in most previous racialcontext work in political science. If we think contextis a ‘‘situation for action’’ and want to understand thepolitical outcomes of individual orientations to thesituation, then we need to measure context at the levelof the individual. If we do so, we also avoid the MAUP:we will know that our statistical summaries reflectdifferences across individuals, not artifacts of aggrega-tion and scale.

Measurement of Racial Threat as aPsychological Process

If context matters for individual-level psychologicalphenomena like threat and ethnocentrism, it mustmatter via perceptions. That people observe andunderstand their contexts is assumed by most re-search on this topic in political science; the standardpractice is to use Census data as the measure ofdiversity, and the interpretation is that individualsobserve qualities of their locales and these obser-vations may affect their opinions and actions. Wetend to assume that people’s perceptions are similarboth to Census numbers and to each other’s.

However, since the 1940s, scholars of publicopinion have learned a great deal about why Censusnumbers may not be good measures of people’s‘‘pseudoenvironments.’’ The American public’s levelof political knowledge is often surprisingly low(Converse 1964; Delli Carpini and Keeter 1997).Ordinary citizens also make incorrect inferences basedon personal experiences or recent, salient events dueto perceptual biases (Ross 1978). When it comes to theracial/ethnic make-up of the United States, for exam-ple, surveys have shown that Americans greatly over-estimate the numbers of minorities in the country(Highton and Wolfinger 1991; Nadeau, Niemi, andLevine 1993; Wong 2007). Furthermore, individuals’misperceptions influence policy preferences (Gilens2000, 2005; Hochschild 2001), and simply givingpeople factual information does not ‘‘correct’’ their

4In other fields, scholars have already shown that institutionallydefined boundaries are not the most relevant for a wide range ofoutcomes of interest (Aitken and Prosser 1990).



policy opinions (Kuklinski et al. 2000). Nevertheless,the issue of misperception or misinformation haslargely been missing in the political science researchon racial context (but see Alba, Rumbaut, and Marotz2005; Chiricos, McEntire, and Gertz 2001).

Inaccurate perception of racial demographics canbe compounded by other misperceptions. Whilesome scholars have equated ‘‘context’’ with racial/ethnic diversity, the content of a context obviouslyextends beyond this one dimension. Research onracial threat has also incorporated the effects ofsocioeconomic and partisan context along with racialcontext. The findings indicate that such attributes caninteract or that socioeconomic or partisan contextmay be even more relevant for people’s racialattitudes than the racial/ethnic diversity of wherepeople live, individuals’ personal income, or their ownpartisan identity (Branton and Jones 2005; Campbell,Wong, and Citrin 2006; Oliver and Mendelberg 2000).

However, the same problem of possible misper-ceptions arises for economic and partisan contexts.For example, the research on the effect of the econ-omy on vote choice has supplied us with ampleevidence that ‘‘it’s the economy, stupid’’ should bechanged to reflect that people’s perceptions of theeconomy are what, in fact, spur votes (Hetherington1996). While broken windows are easy to observe,other indicators of the average socioeconomic statusof one’s neighbors can be less visible than race; un-less someone is wearing an old college sweatshirt, forexample, it is difficult to discern at a distance if one’sneighbor is a college graduate or not. Similarly, par-tisanship is not something most people wear on theirsleeve (or car or front yard; however, see Baybeckand McClurg 2005); while individuals are oftenegocentrically biased to think other people think likethem—and therefore assume a high level of homo-geneity in their surroundings—a Republican livingin a liberal college town could think himself undersiege.5 Using our new measure of context, we can showthe similarities and differences of misperceptions aboutthe racial make-up, socioeconomic well-being, andpartisanship of one’s surroundings.6

We are not focused on misperceptions qua mis-perceptions; the extent of the misperceptions simply

provides evidence that we need a better understandingof the intervening steps linking Census facts about anarea to perceptions of the demographic characteristicsof that same area to feelings of threat and policypreferences. The objective indicators are, of course,important, but we need to know about people’s‘‘pseudoenvironments’’ as well. We want to be ableto answer, for example, the question of whether twopeople who live in a block group with the samepercent outgroup—but believe the diversity varies agreat deal—will feel the same level of racial threat.

Application of our NewMeasurement

To illustrate the utility of our proposed conceptual-ization and measurement strategy, we use data from apilot survey that we conducted of 62 black and whiteindividuals living in one Midwestern county in 2004.7

The in-person survey included a map-drawing meas-ure of context, and the research design allows us tobring the individual back into the conceptualizationand measurement of context. (See Appendix A formore details.) While the particular maps drawn byour respondents are not meant to be generalizable tothe nation, our measure of context is broadly appli-cable, as are the questions our measure raises aboutstandard practices.

Survey Design

To create a measure of personally relevant places thatoperationalizes our conceptualization of context, wedeveloped a map-drawing addition to a traditionalpolitical science survey.8 So, in addition to answeringa number of survey questions, the respondents werealso asked to refer to a few maps. They were firstshown two maps—one centered on the block groupin which their house was located and one encompass-ing all of the county from which the block groupswere sampled—and were asked to draw on either

5However, reactions to status as a numerical minority may notwork in similar ways for partisanship and race (Bledsoe et al.1995; Finifter 1974).

6Even if data for multiple contextual variables come from thesame source—like the U.S. Census—the causal mechanismslinking these different types of context to political judgmentsare not necessarily identical.

7Overall, the racial demographics of the county are quite similarto the nation’s. In 2000, the U.S. was 75% white and 12% black;the county in which we conducted our study was 77% white and12% black.

8We build on the work on mental mapping pioneered by Lynchand which has continued in geography and sociology (Coultonet al. 2001; Lynch 1973; Matei, Ball-Rokeach, and Qiu 2001).



map the area that made up their ‘‘local community.’’9

The approximate location of their house was indi-cated on each map, and the maps included bothblock group boundaries as well as nearby labeledstreets. In order to measure respondents’ perceptionsof the content of their contexts, we then asked aboutperceptions of the proportion of blacks and whites,Democrats and Republicans, and unemployed livingin that community.10 Later, they were shown a mapwith their block group highlighted and were again askedto describe the racial, partisan, and economic break-down of the block group. They were also asked abouttheir perceptions of these factors at the national level.Thus, we were able to gauge perceptions of multiplepolitically relevant characteristics of three geographiccontexts, and because some of the contexts were shared,we could make comparisons both within and acrossindividuals of two different racial groups.

Does the Government DefinePseudoenvironments?

We wanted to allow individuals to define for them-selves (using maps) what they mean by ‘‘community,’’rather than assume that how context affects politicaljudgments must occur within administrative unitsdefined by government officials. While we weredesigning the map-drawing questions, we had contra-dictory predictions. On the one hand, we wereskeptical that people knew the locations of theboundaries of any government-designated units inwhich they resided. On the other hand, we thought

that respondents might be guided by the cuesprovided and follow the bold lines presented on themap, which designated block group boundaries.

Almost two-thirds of the respondents (n 5 36)chose to draw their community on the smaller mapscentered around their block group (see Figure 1(a)for examples), while the rest chose the larger one(23).11 Even given the limited scope of the pen andpaper maps, it is clear that people’s communities are(1) different from governmental administrative unitsand (2) idiosyncratic and not commonly shared visions.Regardless of the size of their ‘‘local community,’’none of the respondents’ drawings followed blockgroup lines neatly. A couple of the maps had non-contiguous areas marked, some marked areas as smallas a single street, while others encompassed the entirelarge map of the two cities within the county. No twomaps were identical, even for residents living in thesame block group, and they ranged in size from smallerthan a block to larger than two cities together.12 Thereis no clear pattern by race, income, or education as towhat size of community was drawn. Furthermore, thereis no simple relationship between size of communitydrawn and ethnocentrism. More ethnocentric individ-uals could choose a smaller community—limiting theirinclusion of outgroup members in the community—but there is no clear pattern across a range of racialattitudes that this is the case.13

Later in the survey, respondents were shownthe ‘‘small’’ map again, this time with their blockgroup highlighted. They were told that the borders onthe map showed the boundaries of what the Census

9The map-drawing task was one of the first in the survey, so therespondents were not primed to think about particular issues orcommunities by other survey questions. We also provided nodefinition of ‘‘local community,’’ both to avoid the respondentssecond-guessing what the researchers might want and also toavoid cueing any particular level of aggregation. Granted, thelargest map that could be drawn was at the county level, althoughrespondents could also respond that they did not think of their‘‘local community’’ in this way. While it would have been ideal toknow how perceptions of their communities changed dependingon the issue, for a pilot study we were interested in getting a moregeneral sense of how people would define their local community.In future research, we plan to prime respondents to think ofdifferent policies or geographies before the map drawing. Theterminology was chosen to emphasize a politically and personallyrelevant grouping of people (a ‘‘community’’) that is spatiallyinterdependent (‘‘local’’).

10While there are many possible measures of economic context, wechose to use percent unemployed. The average education level ofresidents in an area may be a more reliable measure thanunemployment (Huckfeldt 1986), but we believe that if perceptionsare driven in part by observational learning, then residents’ employ-ment status may be more visible to others than their diplomas.

11We were concerned that respondents might find the map-drawingtask overly difficult, but our worries appeared to be unfounded.Only one respondent did not draw on the maps, explaining that shedid not think of her ‘‘local community’’ in these ways. Tworespondents who were visually impaired were able to describe themajor streets and landmarks that defined the borders of their localcommunities. All other respondents were able to draw their localcommunities on the maps. Of course, the ability to draw maps is notthe same as a clear understanding of maps; nevertheless, respondentsdo not need to be cartographers drawing navigable maps in order tocommunicate mental representations of their contexts.

12Each community map was traced into ARCGIS to determine itssize and how it related to Census block groups.

13We need to keep in mind that even though the size of thecommunities drawn in the pilot varied widely, we believe therange would have been even greater if there had been more thantwo choices of maps offered. Given social psychological researchabout anchoring, it is very likely that people chose communitiesthat would fit on one of these two maps and that if they were notso restricted to these particular spatial resolutions, the size of thecommunity may have been even larger for some respondents,extending beyond the county in which they live.



Bureau defines as a ‘‘block group.’’ Respondents werethen asked whether they thought the highlighted areacaptured what they thought of as their neighborhood.Sixty-one percent agreed that their block group was areasonable representation of their neighborhood. Thispercentage may be biased upward for two reasons:suggestibility—with a credible government institutionas the source—and acquiescence bias. Even taking intoaccount that the response could be inflated, almost4 out of 10 respondents did not think of their blockgroup as a close approximation of their neighborhood.Other research has also shown that people’s percep-tions of their neighborhood vary greatly (Huckfeldt1979; Sastry, Pebley, and Zonta 2002). This all servesas evidence that ‘‘neighborhood’’ is no more a comm-only shared container than is ‘‘community’’ and that

the variance across maps is not because ‘‘community’’is a particularly idiosyncratic term.14

Do You See What I See?

A common assumption in the research on racialcontext is that people’s beliefs align with the facts

FIGURE 1 Examples of Map-Based Measures of Context: ‘‘Local Communities’’ Drawn on Block Groupand County Maps

Cross

W I 94E I 94

Hewitt

Washtenaw

Aad

ms

aH

im

tlon

Grant

2nd

Whittier

1st

Harriet

Pearl

Monroe

Roosevelt

FerrisBurns

Emmet

Forest

Was

ihnot gn

Congress

Merrill

Ellsworth

Madison

Valley

Jefferson

Sum

mti

Bal

r ald

Owka

ood

Elm N

ormal

Midvale

Packard

Penirr

Harding

Olive

Elder

Mansfield

Smith

Taft

McKinley

Kirk

Jones

Michigan Hur

on

Ow

enda

le

Wallace Hill

Hawkins

Pittman

College

Hart

Woods

LathersColony

James L Hart

Arbor

Cou

rtlan

d

Sherman

Sheridan

Wilson

Buffalo

Fairf

ield

Cor

nell

Worden

Draper

Fairview

Frederick

Hamilton/W I 94

Carriage

Warner

Perry

Firwood

Orchard

Stevens

Mid

dleAveline

Ivy

Berk

ley

Florence

Westmoorland

Manor

Mar

ino

Catherine

Oxford

Lowell

Ro

esda

le

Ani

wsroth

Stratford

Burton

Anna

Woodward

Her

itage

Pleasant

Collegewood

Do

gul

sa

Lumber

Leaf

Meadow Woods

Belle

vue

Arcade

Elbridge

Hia

t aw

ha

Armstrong Goo

dman

Gouchnour

Peyton

Put

anm

Linden

Grassland

Driscoll

Highland

Huron/W I 94

Warner

Aveline

Huron

Sherman

aW

lalce Pearl

Madison

2nd

Hawkins

Mansfield

Tyler

Clark

Ecorse

Ford

Michigan

Holmes

E I 94

Nash

W I 94

Parkwoo

d

Wiard

AirportUS 12O

hoi

ForestHar

ris

Russell

Allen

Ri

gde

Foley

Gra

nd

Des

t oo

Ore

gno

Lamay

Fox

Loi r

Jerome

Hydramatic

ShareWildwood

diM

way

Kasn

sa

Hyaes

Hunter

neguEe

oD

rset

Evel

yn

Cal

der

Division

Harry

De

novhs

ier

McG

rego

r

Woo

dlaw

n

Andrea

Che

vrol

et

Geer

nwal n

Bailey

Buick

Prescott

Jeff

McC

ar nt ey

Mary Catherine

Spencer

Eileen

Pasadena

Conway

Ecorse

Service

Drive

Main

Ware

Blossom

Dak

ota

Woodg

el n

uO

tre

Appleridge

Wat

osn

liG

l

Dub

ei

Ona

ndoga

Oak

Miam

i

Seneca

N Wiard/Tyler

Zeph

yr

Northern

Byron

Dawn

Park

aG

tes

Davis

Oswego

Centennial

Raw

sonv

ille

Redleaf

N Wiard/W US 12

Bramble

W US 12/S W

iard

Riley

Don

s

Willow Run/W US 12

W US 12

Airp

ort

nIdu

stria

l

Cross

Emreso

n

W U

S 12

/Willo

w R

un

Delaware

E US 12/Willow Run

E US 12

Todd

Olds

Clark Service Drive

Stu

edba

ker

Duncan

Tadd

i

Village

N Wiar

d/E US 12

S Wiard/Airport

E US 12/S Wiard

Avis

Frontage

Ypsi

Laurel

BomberCarol Ann

Cedarcliff

Tyler/N Wiard

Cay

uga

Peachcrest

Willow Run/E US 12

Hudson

Sunnyglen

June

West

Ger

enbr

ire

Brooktree

saEt

McCarthy

Lakewood

Old Michigan

Dodge

Broadmoor

Nev

ada

Tang

elw

ood

Engi

neer

gni

US 12 rai

Wd

Tyler

Kans

as

Tyler

Des

oto

Mai

n

Wi

r ad

Michigan

Wiard

Harris

E I 94

W I 94

Huro

n

Michigan

dn2

1st

Harriet

Ferris

Monroe

aH

lim

ton

Madison

Adam

s

Jefferson

James L Hart

Elder

Taft

Kirk

Jones

Mansfield

Wallace Hill

Hawkins

Bell

Catherine

Was

hign

not

Buffalo

Hart

Woods

Lathers

Arbor

Congress

Worden

SpringFrederick

Hamilton/W I 94

Whittaker/E I 94

Wilson

WarnerWoodward

Peryr

Firwood

Brooks

Orchard

Mid

dle

E I 94/Whittaker

Sum

mit

Aveline

Ivy

Watling

Manor

W I 94/Hamilton

Ains

wor

th

Burton

Wat

erNorm

al

Casler

Pleasant

Lumber

Leaf

Small

Kramer

Armstrong

Goo

dman

Woody

Short

Timberlane

Linden

Grassland

Highland

Huron/W

I 94

Hamilton/E I 94

Frederick

2nd

Hawkins

Madison

Aveline

Mansfield

Warner

Plat

t

Packard

E I 94

W I 94

Washtenaw

Huro

n

Yost

Verle

Esay

Fern

woo

d

Stadium

Pitts

field

Oakwood

Lorraine

Cre

ek

Pakrw

ood

Mcnahe

ster Eli

Sto

en

cSoohl

Norwood

iKeb

mrly

Pistt

i vew

Edgewood

Redwood

La S

alle

Burton

Sharon

Medford

Tnru

ebrry

Hal

lOverridge

Ridge

Cah

mle

sr

Elm

woo

d

Emebr

Dorchester

Birch HollowBe

llwoo

d

Independence

S US

23

Gladstone

Arlington

sEes

x

Ham

pshir

e

Frieze

Sprin

bgro

ok

Eisenhower

Marshall

Col

ony

Glenwood

Beacon

Bedford

Map

lew

ood

Ndr o

man

Canterbury

iLllian

Exmoor

oR

seadle

NU

S 23

Camelot

Service

ahC

ring

C ro

ssBaylis

Towner

Eden

Terhune

Winchell

Powell

Crestland

Cradi

nla

Kenilworth

Sheridan

Woo

dcre

ek

Williamsburg

Car

em

l

iH

kone

cM

oC

mb

Pine

rce

ts

Wolverine

Champagne

Carhart

Butternut

Saint F

rancis

Dar

row

Whitew

ood

Saint Aubin

Brandywine

Ne

deha

m

Hem

lock

Cra

bnro

ok

Vict

oria

Chelsea

Newcastle

Lookout

A asilC

riag

Cumberland

Boulder

Meadowside

Radc

liffe

Columbia

Dw

giht

Richard

La Fere

Yorks

hire

Belvidere

Jeanne

Adare

Nottingham

Roon The Ben

Glo

uces

ter

Ticknor

Logan

Ponds

Calumet

Monument

hC

erry

rTee

Warwick

Shadford

Amelia

Stonehaven

Lindsay

Shady

Revere

Braeburn

Woo

dman

or

Cascade

Englewood

Homestead Com

mons

Bristol

Emer

ald

Salem

Weeburn

Pheasant Run

CaCanny

Metroview

Algebe

Donegal

Burlingame

Nature Cove

Brian

Old Hickory

Stratton

Ward

Kilbrennan

Roc

klan

d

Gallway

Wolverine

Hikone

Pow

ell

Saint Aubin

Ali

as C

raig

Service

Fernwood

14We expect that in areas where neighborhoods are named,isolated, or gated, agreement about the boundaries of theneighborhood may be greater; however, the clarity of theseborders varies a great deal across the nation, and one cannotassume that knowledge in one area means knowledge in others.Furthermore, while other Census units may be more politicallymeaningful than block group—like city—there is no evidencethat reality and perception would overlap more for city than forblock group. Again, there is a great deal of heterogeneity acrossthe United States.



about the demographic make-up of the areas in whichthey live. We examine whether individuals are indeedaccurate when it comes to describing their contexts.15

Greater Misperception of Racial Context at LargerAggregations. We predicted that respondents wouldbe more accurate in their perceptions of their localcommunity than the nation, given their personalexperiences. Figure 2(a) shows two scatterplots, com-paring respondents’ perceptions of blacks and whitesin their block groups with the Census percentages forthe respective groups in the same areas. As can be seenfrom the plot on the left, the fitted smoothed curve isflatter than a 45-degree line, indicating that respond-ents’ perceptions of the percentage of blacks living intheir block group is greater than that reported by theCensus. Conversely, their estimates of the number ofwhites are smaller than the objective numbers. Even ifthe situation were a container, orientations towardthe container are not uniform and are not accurate.

FIGURE 1 Continued

I 94 E

I 94 W

WILLIS RD

US HWY 12

GEDDES RDI 94 BUS

TEXTILE RD

US

HW

Y 23

CAR

PEN

TER

RD

US HWY 12 BUS

STATE HWY 14

STON

Y CR

EEK

RD

WH

ITTAKER R

D

STATE HWY 153

S GROVE ST

PONTIAC TRL

PRO

SPECT R

D

W WILLIS RD

PACKARD ST

S ST

ATE

ST

NIXO

N R

D

E CLARK RD

GO

TFRED

SON

RD

MILLER AVE

PLYMOUTH RD

HURON PKW

Y

W HURON RIVER DR

WARREN RD

WASHTENAW AVE

FULLER ST

MAC

ON

RD

SALINE MILAN RD

S ST

ATE

HW

Y

S M

AIN

ST

N DI

XBO

RO R

D

W LIBERTY ST

ELLSWORTH RD

MAP

LE RD

ANN ARBOR SALINE RD

E ELLSWORTH RD

RID

GE R

D

GEDDES AVE

E CROSS ST

ECORSE RD

EISENHOWER PKWY

ANN ARBOR RD W

W CLARK RD

PAULINE BLVD

E STADIUM BLVD

S M

APLE

RD

HOLMES RD

HURON RIVER DR

TYLER RD

GR

EEN

RD

DEXTER RD

SALINE ANN ARBOR RD

HU

RO

N S

T

AIRPORT DR

MAIN ST

BROADWAY ST

MILLER RD

US

HW

Y 23

BU

S

SPRING ST

E WILLIS RD

S CONGRESS ST

LOH

R R

D

EAR

HAR

T R

D

JACKSON AVE

BARTON DR

MACARTHUR BLVD

S W

AG

NER

RD

MO

NR

OE ST

WASHTENAW AVE

I 94

BUS

US HWY 12

US H

WY 23

US

HW

Y 2

3

US HWY 23

ECORSE RD

STATE HWY 14

STATE HWY 14

US HWY 12

US

H2 Y

W3

PLYMOUTH RD

I 94 E

I 94 W

WILLIS RD

US HWY 12

GEDDES RDI 94 BUS

TEXTILE RD

U SH

W

Y23

CA

PR

TNE

ER R

D

US HWY 12 BUS

STATE HWY 14

STON

Y CR

EEK

RD

WH

ITTAKER R

D

STATE HWY 153

S GROVE ST

PONTIAC TRL

PRO

SPECT R

D

W WILLIS RD

PACKARD ST

S TS

ATE

ST

NIXO

N R

D

E CLARK RD

GO

TFRED

SON

RD

MILLER AVE

PLYMOUTH RD

HURON PKW

Y

W HURON RIVER DR

WARREN RD

WASHTENAW AVE

FULLER ST

MAC

ON

RD

SALINE MILAN RD

S ATS

TEH

WY

M SA

NI S

T

N DI

XBO

RO R

D

W LIBERTY ST

ELLSWORTH RD

MAPLE R

D

ANN ARBOR SALINE RD

E ELLSWORTH RD

RID

GE R

D

GEDDES AVE

E CROSS ST

ECORSE RD

EISENHOWER PKWY

ANN ARBOR RD W

W CLARK RD

PAULINE BLVD

E STADIUM BLVD

S M

APLE

RD

HOLMES RD

HURON RIVER DR

TYLER RD

GE

RNE

R D

DEXTER RD

SALINE ANN ARBOR RD

HU

RO

N S

T

AIRPORT DR

MAIN ST

BROADWAY ST

MILLER RD

US

HW

Y 23

BU

S

SPRING ST

E WILLIS RD

S CONGRESS ST

LOH

R R

D

EARA

HR

TR

D

JACKSON AVE

BARTON DR

MACARTHUR BLVD

S W

AGN

ED

R R

MO

NR

OE ST

STATE HWY 14

US HWY 12

ECORSE RD

WASHTENAW AVE

I 94

BUS

US H

WY 23

US

HW

Y 2

3

US

HW

2 Y

3

PLYMOUTH RD

US HWY 12

STATE HWY 14

US HWY 23

I 94 E

I 94 W

WILLIS RD

US HWY 12

GEDDES RDI 94 BUS

TEXTILE RD

US

HW

Y 2

3

CAR

PEN

TER

RD

US HWY 12 BUS

STATE HWY 14

STON

Y CR

EEK

RD

WH

ITTAKER R

D

STATE HWY 153

S GROVE ST

PONTIAC TRL

PRO

SPECT R

D

W WILLIS RD

PACKARD ST

S ST

ATE

ST

NIXO

N R

D

E CLARK RD

GO

TFRED

SON

RD

MILLER AVE

PLYMOUTH RD

HURON PKW

Y

W HURON RIVER DR

WARREN RD

WASHTENAW AVE

FULLER ST

MAC

ON

RD

SALINE MILAN RD

S ST

ATE

HW

Y

S M

AIN

ST

N DI

XBO

RO R

D

W LIBERTY ST

ELLSWORTH RD

MAPLE R

D

ANN ARBOR SALINE RD

E ELLSWORTH RD

RID

GE R

D

GEDDES AVE

E CROSS ST

ECORSE RD

EISENHOWER PKWY

ANN ARBOR RD W

W CLARK RD

PAULINE BLVD

E STADIUM BLVD

S M

APLE

RD

HOLMES RD

HURON RIVER DR

TYLER RD

GR

ED

R NE

DEXTER RD

SALINE ANN ARBOR RD

HU

RO

N S

T

AIRPORT DR

MAIN ST

BROADWAY ST

MILLER RD

US

HW

Y 23

BU

S

SPRING ST

E WILLIS RD

S CONGRESS ST

LOH

R R

D

EAR

HAR

T R

D

JACKSON AVE

BARTON DR

MACARTHUR BLVD

S W

AGN

ER

RD

MO

NR

OE ST

US HWY 12

STATE HWY 14

US H

WY 23

US HWY 12

US

HW

Y 2

3

PLYMOUTH RD

WASHTENAW AVE

US

HW

Y 2

3

STATE HWY 14

I 94

BUS

US HWY 23

ECORSE RD

I 94 E

I 94 W

WILLIS RD

US HWY 12

GEDDES RDI 94 BUS

TEXTILE RD

STATE HWY 14

PONTIAC TRL

US HWY 12 BUS

STATE HWY 153GO

TFRED

SON

RD

S GROVE ST

PRO

SPECT R

D

W WILLIS RD

PRA

CEN

TER

RD

US H

WY 23

WH

ITT

KAER

RD

PACKARD ST

S ST

ATE

ST

NIXO

NR

D

E CLARK RD

N TERRITORIAL RD

STON

Y CR

EEK

RD

MILLER AVE

W HURON RIVER DR

N TERRITORIAL RD E

PLYMOUTH RD

HURON PKW

Y

WARREN RD

WASHTENAW AVE

FULLER ST

JOY RD

S ST

ATE

HW

Y

M SNI A

TS

N DI

XBO

RO R

D

W LIBERTY ST

ELLSWORTH RD

MAPLE R

D

ANN ARBOR SALINE RD

E ELLSWORTH RD

RID

GE R

D

GEDDES AVE

E CROSS ST

ECORSE RD

EISENHOWER PKWY

ANN ARBOR RD W

W CLARK RD

PAULINE BLVD

E STADIUM BLVD

LIBERTY RD S MAPLE R

D

HOLMES RD

HURON RIVER DR

TYLER RD

GE

REN

RD

DEXTER RD

SALINE ANN ARBOR RD

HU

RO

N S

T

AIRPORT DR

MAIN ST

BROADWAY ST

MILLER RD

US

HW

Y 23

BU

S

SPRING ST

E WILLIS RD

S CONGRESS ST

RH

OL

DR

EAR

HAR

T R

D

JACKSON AVE

BARTON DR

MACARTHUR BLVD

S W

NGA

ER

RD

MO

NR

OE ST

STATE HWY 14

PLYMOUTH RD

ECORSE RDU

S H

W Y23

WASHTENAW AVE

STATE HWY 14

US HWY 12

JOY RD

US

HW

Y2 3

US HWY 12

US HWY 23

I 94

BUS

US

HW

Y

23

US H

WY 23

15For the analyses presented here, we use as a point ofcomparison for respondents’ ‘‘local community’’ the average ofany block group that is included in their drawing. For example, ifa ‘‘community’’ was drawn to overlap with three different blockgroups, then the ‘‘objective’’ point of comparison for racialcontext used is the Census-reported percentages of whites andblacks for those three block groups together. We considered acouple other alternative measures, but none seemed superior.Using only block groups that were entirely enclosed within a‘‘community’’ is problematic because some communities did notinclude a single entire block group. We also considered calculat-ing the fraction of the area of a block group contained within a‘‘community’’ and including only that proportion in the demo-graphics; however, this assumes that individuals are evenlydistributed across a block group, which is clearly incorrect.



While there is an obvious relationship between objectiveindicators and subjective perceptions, Lippmann’spseudoenvironment is not identical to the world de-scribed by Census numbers.

Because of the sampling design, we are also ableto look at the perceptions of white and blackrespondents living in the same block group; whileone could hypothesize that neighbors may have

FIGURE 2 ‘‘Objective’’ versus Perceived Racial, Economic, and Partisan Context At theBlock Group Level



shared spatial experiences and similar perceptions ofthe neighborhood in which they live (Stipak andHensler 1983), one might also hypothesize that blackand white respondents have different experiencesliving in the same area (Kwan 2000). Scholars havefound, for example, that while both whites and blacksare willing to live in integrated neighborhoods, theirdefinitions of ‘‘integration’’ are markedly different(Clark 1991). Comparing white and black respond-ents who live within the same block group, we findthat blacks’ estimates of the proportion of blacksliving in their block group was, on average, 12 per-centage points lower than their white neighbors’estimates, and their estimates of the proportion ofwhites was 13 percentage points higher than that oftheir white neighbors. Within racial-groups averagedifferences range from 1 to 5%. However, whites andblacks both tend to misperceive their racial contextsand in the same directions, overestimating the per-centage of blacks and underestimating that of whites.

Perceptions of racial context in subjectivelydefined ‘‘communities’’ are similar to those at theblock group level, with some variation. At this geo-graphic level (see Appendix B), the line for percentblack appears even flatter and the percent whitesteeper than those in Figure 2(a), indicating thatperceptions of the numbers of blacks and whites areeven more distorted when people evaluate their own‘‘community’’ compared to block group percep-tions. Respondents do not know their own drawn‘‘communities’’ better than their Census block group,and misperceptions are even greater at the nationallevel. Furthermore, while black and white neighborshave different pictures of the same block group inwhich they live, black and white respondents in oursurvey share similar visions of the nation; blacks’estimates of the proportion of blacks living in thenation were only 3 percentage points lower on aver-age than whites’ estimates.16

While the percentages of blacks are overestimatedand that of whites underestimated, an interestingpattern emerges from comparing these different geo-graphic units: perceptions of racial context becomemore accurate at more localized levels. In other words,while respondents tended to overestimate the percent-age of blacks in the nation by 18 percentage points—thinking the nation is 30% black instead of 12%—theoverestimate at the block group level was only 8 per-

centage points.17 The story of growing levels of mis-perception as the ‘‘pseudoenvironment’’ increases in sizebecomes more complicated, however, when we lookat different types of context.

Raising New Questions About Mechanisms withEconomic and Partisan Context. We start with re-spondents’ beliefs about the economic and partisan con-texts of their block groups. Figure 2(b) and Figure 2(c)present plots for perceptions of the percent unem-ployed, Democrat, and Republican at the block grouplevel, compared with Census figures and vote returnsfor 2004. In contrast to the racial content of one’scontext, the relatively flat line in Figure 2(b) shows thatrespondents are even more misinformed about the levelof unemployment in their block group than thepercentage of blacks and whites, with very littlerelationship between objective and subjective economiccontexts.18 On average, respondents overestimated thepercentage of unemployed residents by 16 percentagepoints, although the misperceptions ranged from anoverestimate of 82% to an underestimate of 10%.Respondents’ perceptions of the partisan make-up oftheir contexts also seem to bear only a passingresemblance to the objective measures of partisancontext.19 On average, respondents underestimatedthe percentage of Democrats by 6 percentage pointsand overestimated the percentage of Republicans by6 percentage points So, while there is little relationshipbetween perceptions and objective indicators of socio-economic and partisan contexts, respondents are moreaccurate, on average, in their estimates of living inpredominantly Democratic block groups than in theirbeliefs that about one-out-of-every-six residents oftheir block group is unemployed.

Misperception of the economic context of the‘‘local community’’ is similar to that at the block

16We do not think these results are merely due to innumeracy(see Appendix C).

17These national results are replicated almost identically in the2000 General Social Survey. In the GSS, respondents’ estimates ofthe proportion of whites and blacks were 59% and 31%,respectively. In our pilot, the corresponding numbers were 58%and 30%. This similarity lends support to the idea that theselevels of misperception in our pilot are not unique to our sample.

18For ‘‘objective’’ data about unemployment rates, we use datafrom the 2000 Census. For the block groups in our sample,unemployment ranged from 0 to 14%. In contrast, Baybeck andMcClurg (2005) find that respondents are fairly well-informedabout the economic and partisan status of their neighborhoods.However, their measure of ‘‘objective’’ context was the aggre-gated self-report of respondents in the neighborhood.

19For the ‘‘objective’’ basis of comparison of partisanship, we usethe 2004 presidential vote. Because vote returns are aggregated atthe precinct level, we average the vote for any precincts thatoverlap with a block group’s boundaries to create its ‘‘objective’’partisanship. The block groups in our sample ranged from 62 to96% Democrat, and the county as a whole voted 64% Democrat.



group level, with an average overestimate of 15 per-centage points and very little overall relationshipbetween perceptions and objective measures. Percep-tions of partisan context at the ‘‘local community’’level are also similar to that at the block group level:respondents underestimated the proportion of Dem-ocrats by 6 percentage points and overestimated thatof Republicans by 3 percentage points on average, buthand-in-hand with this greater accuracy is very littlerelationship between the perceived partisan make-upof a community and objective measures of partisan-ship for that same area (see Appendix B). A similarpattern of misperception appears at the national levelas well. Misperceptions about unemployment in theUnited States are, on average, an overestimate of15 percentage points. So, in contrast to perceptionsabout racial context, people’s visions of the economiccontext are distorted to the same extent across allthree levels, whether they are thinking of their blockgroup, community, or nation. There was also a lackof variation across levels for partisan context: at thenational level, there was again only a low level ofmisperception on average; respondents overestimatedthe proportion of Democrats by less than 5 percent-age points and underestimated that of Republicans byabout 5 percentage points.

Comparing results across the three types ofcontext—racial, partisan, and economic—across thethree levels of context—block group, ‘‘local com-munity,’’ and nation—it is clear that (1) there is littlevariation in average misperceptions of unemploy-ment and partisanship from the local to nationallevel, and (2) perceptions of economic and racialcontext are more distorted than those of partisancontext on average (see Appendix D for a table sum-marizing these results).

Sources of Varying Misperceptions

Why does our new measurement strategy revealdiffering levels of accuracy about racial context acrossgeographic units? People may simply have a bettervision of their block group and can more easily envi-sion 100 people, for example, than they can pictureabout 300 million Americans. Another explanation isthat people learn about their various environments indifferent ways.

In our pilot, we asked all respondents to tell us, inan open-ended format, the source(s) of informationfor their perceptions at the local, block group, andnational level. The responses differ greatly between

where one gathers knowledge about the nation andwhere one gains information about one’s local com-munity and block group, with the responses aboutthe latter two blending together (see Appendix E fordetails). Personal experience, families, and friendsplay a much larger role in knowledge about people’scommunity and block group; as shown in previousresearch, everyday observation was cited by at leasthalf of our sample as a source of information abouttheir ‘‘local community’’ and block group, and per-sonal social networks provided information to abouta quarter (Gilliam, Valentino, and Beckmann 2002).In contrast, media is the primary source of informa-tion about the nation; over two-thirds of the re-sponses referred to media, and few mentioned anyother source for their knowledge about the UnitedStates as a whole. Exaggerated numbers are mostprominent in national ‘‘pictures in people’s heads,’’which suggests the important role of the media increating perceptions of racial context and feelings ofthreat.

One possible explanation for perceptions ofeconomic context is that such beliefs—regardless oflevel—are mainly based on media reports. Becausefew news sources would ever discuss unemploymentat levels smaller than a respondent’s city, he or shemost likely is not exposed to systematic informationabout unemployment at the block group level.Inferences across all levels could be drawn from thesame source of information, with an assumption ofuniformity across levels.

When it comes to understanding the accuracy ofperceptions of partisan context across all three levels,on average, one possible explanation is the role ofmedia and elections in learning. Stories about elec-tion returns may highlight partisan surroundings in adramatic way for individuals, unmatched by anyevent that would highlight their racial and economiccontexts as clearly, regularly, and officially. Of course,one obvious problem with the explanation that mediacoverage can lead to greater accuracy is the fact thatthe news media regularly presents stories about thenation’s unemployment rate and Census results, yetrespondents were much more misinformed aboutthese facts relative to facts about partisanship at thenational level. Neighbors’ partisanship is unlikely tochange rapidly from year to year, so it may be easierfor respondents to learn about their partisan contextthan the more volatile unemployment rate overtime—assuming residential stability—but neighbors’races are even less likely to fluctuate than partyloyalties. There is another plausible explanation: incontrast to economic and racial context, the partisan



composition of the nation (and likely many com-munities and block groups) is more evenly distrib-uted between Democrats and Republicans. Therefore,if a respondent took a guess around 50%, he or she ismore likely to appear ‘‘accurate’’ about partisan con-text at the national level, but grossly incorrect abouteconomic and racial contexts. Research also showsthat people are more likely to misperceive rare eventsthan common ones. However, if one looks moreclosely at the contexts in our sample, this explanationis also problematic: the county from which we drewour sample leans Democratic, and the means andmedians of the distributions in both block groupsand the local communities drawn, ranged between75 and 80% Democratic. Obviously, future researchis needed to gain a better understanding of howpeople develop their perceptions of different typesand sizes of context.

Old versus New Measures of Context

While this article has focused on measurement, werealize that some may ask whether it is worth theextra effort to measure perceptions of context ratherthan use administrative numbers. The first answer isthat we cannot know for sure whether results thatdiffer using administrative numbers across multiplelevels of aggregation arise as statistical artifacts fromthe MAUP unless we measure ‘‘context’’ at the levelof the individual. A second answer is that measuringperceptions gives us the potential to understandbetter the mechanisms by which context affectspolitics; for example, we can observe if racial contextaffects people like particulates in the air (i.e., affect-ing outcomes, regardless of subjects’ awareness oftheir environment) or if it works via Lippmann’spseudoenvironments.

Some might suggest that we could use Censusnumbers to approximate the psychological measuresthat we might prefer. Or, put more critically, mightdifficult, novel, and messy psychological measuresmerely proxy for clean and easy Census numbers? Inthis section we engage with these questions usingan example focused on the relationship between‘‘racial context’’ and racial attitudes among whiterespondents; our pilot study offers the opportunity tocompare the effects of (1) Census numbers in aCensus geography, (2) Census numbers in a perceivedgeography, (3) perceived numbers in a Census geog-raphy, and (4) perceived numbers in a perceivedgeography.

We start with the simple relationship betweenpercent black in a Census block group and responsesto a racial resentment scale (Kinder and Sanders1996).20 On average, do whites living among moreblacks report higher racial resentment? We replicateswhat has become the standard in larger and morerepresentative surveys: the leftmost line in Figure 3shows that, in our data, whites who live with manyblacks in their block group tend to be more resentfulthan whites who live near relatively few blacks. In ourparticular data, we see that the 66% interval for adifference of 10 percentage points in percent black ina block group ranges from about 0.03 to about 0.07points of difference in the resentment scale: the scaleruns from 0 to 1, s.d. of .2. In the absence of othermeasures of context, we cannot know whether thisrelationship is masking some aggregation and/orzoning artifact (i.e., the MAUP), nor do we knowwhether this relationship has the psychological con-tent that many theories of context would use thisanalysis to assess.

What if we were interested in the differencesbetween people based not on what the Census reportsabout a Census unit, but about what the Censusreports for a personally relevant unit? The second linefrom the left in Figure 3 shows that Census numbersfor percent black in subjective maps relate to racialresentment more or less as strongly and in the expecteddirection as using Census numbers and boundaries,even if the diversity of maps increases the width ofthe interval: the center of the interval is in the .04 to.10 range. Because we were able to produce anindividual-level, nonmodifiable measure of content,we could thus reassure ourselves that the analysisusing Census boundaries was not misleading.21

While the map-drawing measure helps us avoidthe MAUP, it does not tell us if Census numbers aregood proxies for perceptions, or more generally,much about the mechanism by which environmentsaffect people’s political decisionmaking. We thereforeturn to what we believe is the next step in the mech-anism that can lead to threat: perceptions. We com-pare the effects of subjective measures—measureswhere we asked white respondents to report on the

20See Appendix F for information about the scale and details ofthe analyses that follow.

21In these analyses we do not adjust the estimates for anycovariates. Here, the goal is not to disentangle the effects ofcontext from the many and complex possible sources of it (suchas selection effects), but rather to compare context based differ-ences in racial resentment across measures of context (each ofwhich may have its own interesting relationship with backgroundcovariates such as education and socialization).



proportions of blacks living either in the boundariesof a context drawn by the Census or the boundariesdrawn by the respondent. Would such subjectivereports have explanatory power beyond that providedby objective context? We matched respondents to oneanother, creating pairs on the basis of objectivecontext (using either Census numbers in the blockgroup or in their hand drawn community).22

If the subjective reports have little to add beyondobjective numbers, then within pairs matched nearlyexactly on objective numbers, we should see littledifference in subjective reports and, more impor-tantly, differences in subjective reports should not

relate to substantively relevant outcomes (and shouldmainly be seen as noise). The rightmost lines inFigure 3 suggest that this is not the case: subjectivecontext is nearly as strongly related to racial resent-ment as is objective context even holding objectivecontext almost exactly constant. Specifically, thepoint estimates and centers of the posteriors arisingfrom the multilevel models estimated conditional onthe matching show that differences in subjectiveperceptions of percent black in a block group areassociated with differences of about .05 on the racialresentment scale.23 A similar result is evident when itcomes to perceptions of the subjective boundaries(the line labeled ‘‘subjective community’’ in Figure 3).Even within pairs in which both respondents’ mapscontained almost the same percent black, the whiterespondent who perceived more blacks tended toreport more racially resentful attitudes than theperson who perceived fewer black people in their‘‘local community.’’24

So, what we have shown here suggests that(1) our pilot is consistent with past research, suchthat objective context predicts racial attitudes; (2) anyof the three methods to side-step the MAUP shows arelationship with racial attitudes (and, in our partic-ular case, helps confirm that the sign on the relation-ship between objective block group percent blackand racial resentment in our pilot study is not likelya statistical artifact); and (3) even within pairs ofwhite respondents more or less identical on ob-jective Census numbers, the person perceiving moreblacks tended to be the person with higher racial

FIGURE 3 Effects of Racial Context on RacialResentment

22This matching is not the usual form used in political science inwhich members of a treatment group are matched to members ofa control group (also called ‘‘bipartite’’ matching to indicate thatthe matching is in two parts). We used nonbipartite matching, inwhich every unit is a potential pair with every other unit. Thealgorithm then assigns units to pairs so as to minimize the totaldifferences on the matching scale (here, percent black in theblock group). For more on nonbipartite matching, see Lu et al.(2001, 2011) and Rosenbaum (2010). See Appendix F for moreexplanation about how to use and assess nonbipartite matching.

23The uncertainty around this estimate (shown as the verticallines in the ‘‘subjective block group’’ line on the figure) is widerthan the corresponding objective block group analysis: a prioriwe did not know whether to expect that the loss of degrees offreedom from conditioning on pairs would decrease our pre-cision or whether the increased variation arising from thesubjective block group reports would increase our precision orwhether the conditioning itself would increase precision byadding explanatory power (given the preceding analysis in whichthe objective block group numbers did predict the outcome).

24Notice that we did not control for education or other covariatesin these analyses. When we did this in analyses not reported here(using matching on mahalanobis distances created from ranktransformed education, income, gender, and length of residencein a community, and/or using said distances to penalize thematchings on objective context and/or adding those covariatesinto the linear models as controls), we noticed weaker effectsalbeit a similar pattern. This supports the points made here: ifsubjective context is a real entity, then it ought to relate stronglyto education and other personal attributes that would makedifferent people perceive their contexts differently. And, objectivecontext effects ought to depend on such characteristics too, both(a) indirectly as they matter for perceptions and (b) directly assuch characteristics are proxies for the processes driving residentialsegregation.



resentment. That is, in terms of the mechanism bywhich the environment affects political beliefs, welearnt that perceptions of the context matter forattitudes, above and beyond objective characteristics.Parsons and Shils’s ‘‘situation of action’’ matters, andobjective numbers should not simply be used asproxies for people’s perceptions, even if they maymatter in other ways.

Discussion

One question raised by the finding that mispercep-tion of racial context is greater at larger units is howbest to interpret previous findings that context ispolitically, economically, and culturally threateningat larger units but not at smaller ones. For example,most of the research that finds support for the racialthreat hypothesis measures context at the level ofcounty, metropolitan area, state, region, or country;in contrast, the research that has found support forthe contact theory tends to measure context atsmaller units, like census tract or zip code. In thepast, this difference was interpreted as evidence thatcontact at the smaller units would diminish ethno-centrism, while diversity at the larger aggregate levelsmade such intergroup contact more likely. Diversitywithout contact, in contrast, would lead to feelings ofthreat (Stein, Post, and Rinden 2000).25 Anotherinterpretation, however, is that these differences inoutcomes across levels are simply the result of theMAUP, and we should draw no substantive conclu-sions about differing perceptions of threat acrosslevels. A third interpretation, given our findings, isthat overestimates of the size of outgroups at largercontextual units relative to misperceptions at smallerunits is what leads to a greater sense of threat in themore aggregated levels. In other words, if individualshad a more accurate picture of their surroundings,they might not feel a sense of racial threat at themetropolitan or county level, for example. This doesnot preclude the idea that personal interactions with

outgroup members diminish ethnocentrism, but itdoes raise more questions as to why racial contextrarely poses a ‘‘threat’’ at the more localized level inprevious research.

We also raise the question of how one shouldinterpret perceptions by different racial or ethnicgroups. While the bulk of research on racial threathas focused on the attitudes and actions of whiteAmericans, recent research has expanded to includethe effects of context on the political judgmentsof racial/ethnic minorities (Barreto, Segura, andWoods 2004; Gay 2004; Glaser 2003). The theoreticalarguments advanced for all groups are similar—predominantly group conflict or contact theory—although the specifics are left a bit vague as to whygroups of such disparate sizes should behave in similarways. It does seem peculiar to argue that whiterespondents who compose a majority in a county areas ‘‘threatened’’ by blacks as black respondents in thatsame county are by whites. After all, if one considersthe Black Belt studied by Key, the threat of lynching forblacks was much more real and vivid than the threat oflosing political power was for whites. Another com-plication is added by the fact that previous researchhas shown that Americans of all races overestimate thenumbers of minorities in the United States andunderestimate the number of whites in the country(Wong 2007). What has not been mentioned is thatthe motivations attributed to these inaccuracies cannotbe the same across groups: if whites, for example,exaggerate the numbers of blacks due to a sense ofthreat, then why do blacks also exaggerate the numbersof blacks? Threat cannot explain why an individualoverestimates the numbers of both her outgroup andingroup.

These misperceptions also have policy effects.The literature on white flight discusses the fact thatwhites are unhappy and flee when blacks move intotheir neighborhoods. However, Harris (2001) hasfound that blacks are also threatened by other blacksmoving into their neighborhoods, a threat thatcannot be motivated by fear of a racial outgroup.It is possible that economic threat motivates bothgroups’ responses, or they may have differentimpetuses; teasing out the meaning depends onunderstanding how different groups perceive andinterpret the same environments. When scholarsstudy one group at a time, it is easier to findplausible theoretical explanations for the empiricalfindings; it is only by comparing groups that oneis confronted by the more complicated picture ofhow perceptions of context can affect politicaljudgments.

25We find in our pilot study that contact with outgroup membershas no noticeable effect on people’s perceptions of their contexts;for example, whites who have contact with blacks at work, attheir place of worship, or in their circle of friends are not more orless accurate in their perceptions of how many blacks live in theirblock group or community, compared to whites whose daily livesare more segregated. In other words, in our pilot, actualinteractions with outgroup members do not lead to either greateraccuracy or misinformation. So, while contact may diminishethnocentrism (Allport 1979), it may not do so via greaterawareness of one’s surroundings.



The housing literature is also relevant for anotherreason to the analyses presented here. According toSchelling’s argument about tipping points, smallpreferences at the individual level can become mag-nified at the aggregate level, such that slight prefer-ences for a homogeneous neighborhood lead to thedramatic levels of residential racial segregation thatexists in this country (Schelling 2006). Furthermore,as Bruch and Mare (2006) show in their analysis ofSchelling’s model, the existence of a threshold is keyto explaining how slight differences in preferencescould lead, in a theoretical model, to an ‘‘apartheidAmerica.’’ While the strikingly low threshold ob-served in real life may reflect the ‘‘objective’’ status ofthe neighborhood’s make-up, it is also possible thatthe neighborhood is perceived as more diverse than itis in reality and that the threshold for white flight inpeople’s minds is higher than the one observed. Inother words, someone might be willing to live in aneighborhood that is one-third black, but because heoverestimates the percentage of blacks living in hisneighborhood, he ends up moving away well beforehis real threshold has been reached. While thepractical implications may be the same in terms ofpolicy interventions to prevent flight and hyper-segregation, there may be a role for public educationabout the demographics of one’s surroundings thatcould at least slow down white flight.

Conclusion: Attitudes arenot Asthma

Medical researchers who study the effects of pollu-tants on the development of asthma measure air-borne particulates in a geographic region and thepresence of asthma and other respiratory symptomsfor the population living in that area. When politicalscientists study the effects of racial context on ethno-centrism and policy opinions, their studies mimic thepollution studies: outgroup members are the threat-ening allergens and attitudes are the asthma. How-ever, in general, context does not surreptitiouslyaffect one’s attitudes and actions like smog caninvisibly penetrate one’s lungs, even if perceptionsof it can lead to fear (which in turn can lead to apanoply of unconscious reactions). Parsons andShils include this perceptual stage—Lippmann’s‘‘pseudoenvironments’’—in their conceptualizationof context, and our map-based measure of contextis valid for this conceptualization. Our operational-ization of context allows us to ask questions about

respondents’ self-drawn communities as comparedto places that group people by administrative fiat,thus side-stepping the MAUP. We also do not needto presume that ordinary citizens are fully in-formed consumers of government statistics, nordo we need to assume that people are exchangeableagents who see and react to their contexts in thesame ways. When we find contextual effects usingconvenient, institutionally drawn boundaries, it isimpossible to determine if these effects are overlyconservative or liberal, or whether they are stat-istical artifacts, since they most likely vary a greatdeal across individuals, across subgroups, andacross geographic units.

Our first attempt at following our guiding re-search principles from concept to questionnairemeasure and design suggests a need to ‘‘bring theperson back in.’’ Furthermore, by examining differentaspects of people’s contexts in our pilot study, we areable to contrast the varied ways that facts about theenvironments in which people live are converted intobeliefs. We find that the mechanisms by which racial,socioeconomic, and partisan contexts are perceivedand affect political attitudes and actions do notappear to be the same.

If pictures fixed in people’s heads do not matchCensus pictures, the practical implications extendbeyond the confines of citizens’ minds or the votingbooth. The ‘‘fear of crime’’ literature in sociology hasexplained that personal and altruistic fear—regardlessof accuracy—leads to purchases (e.g., guns), behavioralchanges (e.g., not going out at night), and abandon-ment of locations (e.g., parks and industrial areas;Warr and Ellison 2000). Political scientists need tounderstand whether perceptions of community het-erogeneity and interracial competition have equallyserious consequences for political actions andoutcomes.

Acknowledgments

We thank Andrea Benjamin, Alana Hackshaw, andLaura Schram for their excellent research assistanceon the survey. Thanks for excellent suggestions,critiques, and comments on the overall researchproject from colleagues and seminar participants atIllinois, Michigan, Harvard, Wisconsin, the 2011Chicago Area Behavior Workshop, and many anon-ymous reviewers for the political science, geography,and sociology panels of the National Science Foun-dation. We particularly want to thank Wendy TamCho, Dennis Chong, Jim Kuklinski, Jeff Mondak,



Eric Oliver, and Spencer Piston for their carefulreading of this manuscript.

References

Achen, C. H., and W. P. Shively. 1995. Cross-Level Inference.Chicago: University of Chicago Press.

Adcock, R., and D. Collier. 2001. ‘‘Measurement Validity: A SharedStandard for Qualitative and Quantitative Research.’’ AmericanPolitical Science Review 95 (03): 529–46.

Aitken, S. C., and R. Prosser. 1990. ‘‘Residents’spatial knowledgeof neighborhood continuity and form.’’ Geographical Analysis22 (4): 301–25.

Alba, R., R. G. Rumbaut, and K. Marotz. 2005. ‘‘Distorted Nation:Perceptions of Racial/Ethnic Group Sizes and Attitudes towardImmigrants and Other Minorities.’’ Social Forces 84 (2): 901–19.

Allport, G. W. 1979. The Nature of Prejudice. Cambridge, MA:Persevs Books.

Barreto, Matt A., Gary M. Segura, and Nathan D. Woods. 2004.‘‘The Mobilizing Effect of Majority-Minority Districts on LatinoTurnout.’’ American Political Science Review 98 (01): 65–75.

Baybeck, B. 2006. ‘‘Sorting Out the Competing Effects of RacialContext.’’ Journal of Politics 68 (2): 86–96.

Baybeck, B., and S. D. McClurg. 2005. ‘‘What Do They Know andHow Do They Know It?’’ American Politics Research 33 (4):492–520.

Blalock, H. M. 1967. Toward a Theory of Minority-groupRelations. Boston, MA: Wiley.

Blalock, Hubert M, and Ann B. Blalock. 1968. Methology in SocialResearch. New York: McGraw-Hill.

Bledsoe, T., S. Welch, L. Sigelman, and M. Combs. 1995. ‘‘Resi-dential context and racial solidarity among African Americans.’’American Journal of Political Science 39 (2): 434–58.

Branton, R. P., and B. S. Jones. 2005. ‘‘Reexamining racialattitudes: The conditional relationship between diversity andsocioeconomic environment.’’ American Journal of PoliticalScience 49 (2): 359–72.

Bruch, E. E., and R. D. Mare. 2006. ‘‘Neighborhood choice andneighborhood change.’’ American Journal of Sociology 112 (3):667–709.

Campbell, A. L., C. Wong, and J. Citrin. 2006. ‘‘‘‘Racial Threat’’,Partisan Climate, and Direct Democracy: Contextual Effectsin Three California Initiatives.’’ Political Behavior 28 (2): 129–50.

Chiricos, T., R. McEntire, and M. Gertz. 2001. ‘‘Perceived racialand ethnic composition of neighborhood and perceived riskof crime.’’ Social Problems 48 (3): 322–40.

Cho, W.K.T., and C.F. Manski. 2008. Cross Level/EcologicalInference. In Oxford Handbook of Political Methodology, ed.H. Brady, D. Collier, and J. Box-Steffensmeier. Oxford, UK:Oxford University Press, 530–69.

Clark, W.A.V. 1991. ‘‘Residential preferences and neighborhoodracial segregation: a test of the Schelling segregation model.’’Demography 28 (1): 1–19.

Converse, P.E. 1964. ‘‘The nature of belief systems in masspolitics.’’ In Ideology and Discontent, ed. D.E. Apter. New York:Free Press, 206–61.

Coulton, C.J., J. Korbin, T. Chan, and M. Su. 2001. ‘‘Mappingresidents’ perceptions of neighborhood boundaries: a meth-odological note.’’ American Journal of Community Psychology29 (2): 371–83.

Delli Carpini, M.X.D., and S. Keeter. 1997. What AmericansKnow about Politics and Why it Matters. New Haven, CT: YaleUniversity Press.

Finifter, A.W. 1974. ‘‘The friendship group as a protectiveenvironment for political deviants.’’ The American PoliticalScience Review 68 (2): 607–25.

Gay, Claudine. 2004. ‘‘Putting Race in Context: Identifying theEnvironmental Determinants of Black Racial Atittudes.’’American Political Science Review 98 (4): 547–62.

Gehlke, C.E., and H. Biehl. 1934. ‘‘Certain Effects of GroupingUpon the Size of the Correlation Coefficient in Census TractMaterial.’’ Journal of the American Statistical Association,Supplement 29: 169–70.

Gilens, M. 2000. Why Americans Hate Welfare: Race, Media, andthe Politics of Antipoverty Policy. University of Chicago Press.

Gilens, M. 2005. ‘‘Political ignorance and collective policy prefer-ences.’’ American Political Science Review 95 (02): 379–96.

Gilliam, F. D., N. A. Valentino, and M. N. Beckmann. 2002.‘‘Where you live and what you watch: The impact of racialproximity and local television news on attitudes about raceand crime.’’ Political Research Quarterly 55 (4): 755–80.

Glaser, J. M. 1994. ‘‘Back to the black belt: Racial environmentand white racial attitudes in the South.’’ The Journal of Politics56 (1): 21–41.

Glaser, J. M. 2003. ‘‘Social context and inter-group politicalattitudes: Experiments in group conflict theory.’’ BritishJournal of Political Science 33 (4): 607–20.

Gotway, Carol A., and Linda J. Young. 2002. ‘‘CombiningIncompatible Spatial Data.’’ Journal of the American StatisticalAssociation 97 (458): 632–48.

Harris, D. R. 2001. ‘‘Why are whites and blacks averse to blackneighbors?’’ Social Science Research 30 (1): 100–16.

Hatt, P. 1946. ‘‘The concept of natural area.’’ American SociologicalReview 11 (4): 423–27.

Hetherington, M. J. 1996. ‘‘The media’s role in forming voters’national economic evaluations in 1992.’’ American Journal ofPolitical Science 40 (2): 372–95.

Highton, B., and R. E. Wolfinger. 1991. ‘‘Estimating the Size ofMinority Groups.’’ Technical report Memo to Board ofOverseers, National Election Studies.

Hochschild, J. L. 2001. ‘‘Where you stand depends on what yousee: Connections among values, perceptions of fact, andpolitical prescriptions.’’ In Citizens and Politics: Perspectivesfrom Political Psychology, ed. J. H. Kuklinski. Cambridge:Cambridge University Press, 313–40.

Hopkins, D. J. 2010. ‘‘Politicized places: Explaining where andwhen immigrants provoke local opposition.’’ American Polit-ical Science Review 104 (01): 40–60.

Huckfeldt, R., and J. Sprague. 1987. ‘‘Networks in context: Thesocial flow of political information.’’ The American PoliticalScience Review 81 (4): 1197–216.

Huckfeldt, R. R. 1979. ‘‘Political participation and the neighbor-hood social context.’’ American Journal of Political Science23 (3): 579–92.

Huckfeldt, R. R., and J. Sprague. 1995. Citizens, Politics, andSocial Communication: Information and Influence in an Elec-tion Campaign. Cambridge: Cambridge University Press.

Key, V. 1949. Southern Politics in State and Nation. Knoxville:University of Tennessee Press.

Kinder, D. R. and L. M. Sanders. 1996. Divided by Color: RacialPolitics and Democratic Ideals. Chicago: University of ChicagoPress.



King, Gary. 1997. A Solution to the Ecological Inference Problem.Princeton, NJ: Princeton University Press.

Kuklinski, J. H., P. J. Quirk, J. Jerit, D. Schwieder, and F. Robert.2000. ‘‘Misinformation and the Currency of DemocraticCitizenship.’’ Journal of Politics 62 (3): 790–816.

Kwan, M. P. 2000. ‘‘Interactive geovisualization of activity-travelpatterns using three-dimensional geographical informationsystems: a methodological exploration with a large data set.’’Transportation Research Part C: Emerging Technologies 8 (1-6):185–203.

Lazarsfeld, Paul F., Bernard Berelson, and Hazel Gaudet. 1968.The People’s Choice. New York: Columbia University Press.

LeVine, R. A., and D. T. Campbell. 1971. Ethnocentrism. Boston,MA: J. Wiley.

Lewin, Kurt. 1936. Principles of Topological Psychology. New York:McGraw-Hill.

Lippmann, W. 1991. Public Opinion. New Brunswick, NJ: Trans-action Publishers.

Liu, B. 2001. ‘‘Racial contexts and white interests: beyond blackthreat and racial tolerance.’’ Political Behavior 23 (2): 157–80.

Lu, B., E. Zanutto, R. Hornik, and P. R. Rosenbaum. 2001.‘‘Matching with doses in an observational study of a mediacampaign against drug abuse.’’ Journal of the AmericanStatistical Association 96 (456): 1245–53.

Lu, B., R. Greevy, X. Xu, and C. Beck. 2011. ‘‘Optimal nonbipartitematching and its statistical applications.’’ The AmericanStatistician 65 (1): 21–30.

Lynch, K. 1973. The Image of the City. Bathmore, MD: MIT press.

Matei, S., S.J. Ball-Rokeach, and J.L. Qiu. 2001. ‘‘Fear andmisperception of Los Angeles urban space.’’ CommunicationResearch 28 (4): 429–63.

Nadeau, R., R.G. Niemi, and J. Levine. 1993. ‘‘Innumeracy aboutminority populations.’’ Public Opinion Quarterly 57 (3): 332–47.

Oliver, J.E. 2010. The Paradoxes of Integration: Race, Neighbor-hood, and Civic Life in Multiethnic America. Chicago: Uni-versity of Chicago Press.

Oliver, J.E., and T. Mendelberg. 2000. ‘‘Reconsidering theenvironmental determinants of white racial attitudes.’’American Journal of Political Science 44 (3): 574–89.

Openshaw, S. 1984. The Modifiable Areal Unit Problem. Conceptsand Techniques in Modern Geography No. 38. Norwich, UK:Geo Books.

Openshaw, S., and P.J. Taylor. 1979. ‘‘A Million or So Correla-tion Coefficients: Three Experiments on the Modifiable Arealunit Problem.’’ In Statistical Applications in the SpatialSciences, ed. N. Wrigley. London: Pion Limited, 127–44.

Parsons, Talcott, and Edward A. Shils. 1951. Toward a GeneralTheory of Action. Cambridge, MA: Harvard University Press.

Quillian, L. 1995. ‘‘Prejudice as a response to perceived group threat:Population composition and anti-immigrant and racial preju-dice in Europe.’’ American Sociological Review 60 (4): 586–611.

Rosenbaum, P. R. 2010. Design of observational studies. NewYork: Springer Verlag.

Ross, L. 1978. ‘‘The intuitive psychologist and his shortcomings:Distortions in the attribution process.’’ In Advances inExperimental Social Psychology (vol. 10), ed. L. Berkowitz,New York: Academic press, 172–220.

Sampson, R. J., J. D. Morenoff, and T. Gannon-Rowley. 2002.‘‘Assessing neighborhood effects: Social Processes and NewDirections in Research.’’ Annual Review of sociology 28 (1):443–78.

Sastry, N., A. R. Pebley, and M. Zonta. 2002. ‘‘Neighborhooddefinitions and the spatial dimension of daily life in LosAngeles.’’ California Center for Population Research. Onlineworking paper series, University of California, Los Angeles.

Schelling, T. C. 2006. Micromotives and Macrobehavior. New York:W.W. Norton & Company.

Stein, R. M., S. S. Post, and A. L. Rinden. 2000. ‘‘Reconcilingcontext and contact effects on racial attitudes.’’ PoliticalResearch Quarterly 53 (2): 285–303.

Stipak, B., and C. Hensler. 1983. ‘‘Effect of neighborhood racialand socioeconomic composition on urban residents’ evaluationsof their neighborhoods.’’ Social Indicators Research 12 (3):311–20.

Tam Cho, W.K., and N. Baer. 2011. ‘‘Environmental Determi-nants of Racial Attitudes Redux: The Critical DecisionsRelated to Operationalizing Context.’’ American PoliticsResearch 39 (2): 414–36.

Taylor, M. C. 1998. ‘‘How white attitudes vary with the racialcomposition of local populations: Numbers count.’’ AmericanSociological Review 63: 512–35.

Warr, M., and C.G. Ellison. 2000. ‘‘Rethinking Social Reactionsto Crime: Personal and Altruistic Fear in Family Households.’’American Journal of Sociology 106 (3): 551–78.

Wong, C. J. 2007. ‘‘Little’’ and ‘‘Big’’ Pictures in Our Heads:Race, Local Context, and Innumeracy about Racial Groups inthe United States.’’ Public Opinion Quarterly 71 (3): 392.

Wong, C. J. 2010. Boundaries of Obligation in American Politics:Geographic, National, and Racial Communities. Cambridge:Cambridge University Press.

Wong, D. 2009. ‘‘The modifiable areal unit problem (MAUP).’’The SAGE Handbook of Spatial Analysis. Los Angeles: SAGEPublications. 105–23.

Yinger, J. 1997. Closed Doors, Opportunities Lost: The ContinuingCosts of Housing Discrimination. Russell Sage FoundationPublications.

Cara Wong is Assistant Professor of Polit-ical Science at the University of Illinois, Urbana-Champaign, Urbana, IL 61801.

Jake Bowers is Assistant Professor of Polit-ical Science at the University of Illinois, Urbana-Champaign, Urbana, IL 61801.

Tarah Williams is a Ph.D. Candidate in Polit-ical Science at the University of Illinois, Urbana-Champaign, Urbana, IL 61801.

Katherine Drake Simmons is a Research Asso-ciate at the Pew Research Center, Washington, D.C.20036.



bringing the person back in: boundaries, perceptions, and ... · bringing the person back in:...

Documents