Prejudice and Politics: Symbolic Racism Versus Racial ... · Prejudice and Politics: Symbolic Racism Versus Racial Threats to the Good Life Donald R. Kinder Political Science Department,

Journal of Personality and Social Psychology1981, Vol. 40, No. 3, 414-431

Copyright 1981 by the American Psychological Association, Inc.0022-3514/81 /4003-0414S00.75

Prejudice and Politics: Symbolic Racism VersusRacial Threats to the Good Life

Donald R. KinderPolitical Science Department, Yale University

David O. SearsUniversity of California, Los Angeles

Although theories of prejudice have been extensively catalogued, empirical con-frontations between competing theories are surprisingly rare. The primary goalof the present research was to test two major theoretical approaches to prejudiceby whites against blacks: realistic group conflict theory, which emphasizes thetangible threats blacks might pose to whites' private lives; and a socioculturaltheory of prejudice termed symbolic racism, which emphasizes abstract, mor-alistic resentments of blacks, presumably traceable to preadult socialization. Themain dependent variable in our analysis is suburban whites' voting behavior intwo mayoral elections in Los Angeles, both strongly influenced by racial issues,that matched the same two candidates, one black and one white. In both elections,symbolic racism (sociocultural prejudice) was the major determinant of votingagainst the black candidate for people removed from possible personal threatsposed by blacks as well as for those at risk. Direct racial threats to whites' privatelives (to their jobs, their neighborhoods, their children's schooling, their families'safety) had little effect on either antiblack voting behavior or symbolic racism.The article closes by developing the implications of these results for theories ofprejudice and, more speculatively, for interpretations of the effects of voters'private lives on their political behavior.

Theories of racial prejudice suffer frombenign neglect. Although the theories them-selves have been extensively and ably cata-logued (most notably, by Allport, 1954; Ash-more & DelBoca, 1976; LeVine & Campbell,1972), empirical confrontations between al-ternative theories occupy surprisingly little

Portions of this material were originally presented bySears and Kinder to the Western Psychological Asso-ciation, Los Angeles, 1970. A more elaborate versionof this article won the 1978 Gordon Allport IntergroupRelations Prize from the Society for PsychologicalStudy of Social Issues. The data were collected as partof a larger project on black mayoral candidates underthe direction of Thomas Pettigrew, to whom we owedeep thanks. Robert Riley was an invaluable collabo-rator in early stages of the project. John Morgan, An-drew Walker, and Richard E. Whitney also contributedvaluable assistance. Comments on earlier drafts by FayeCrosby, Barbara Gutek, Richard R. Lau, John B.McConahay, M. Brewster Smith, Tom Tyler, James Q.Wilson, Raymond Wolfinger, and the Journal's anon-ymous referees have been particularly helpful. Supportwas provided by a Rockefeller Foundation grant to Pet-tigrew, and by National Science Foundation and U.S.Public Health Service grants to Sears.

Requests for reprints should be sent to Donald R.Kinder, Political Science Department, Yale University,New Haven, Connecticut 06520.

space in the prejudice literature. Some ob-servers even contend that such tests, al-though never common, are now actually onthe decline (Katz, 1976). The principal pur-pose of this article is to present a systematicempirical comparison between two majortheoretical approaches to understandingwhites' prejudice against blacks: realisticgroup conflict theory, which emphasizes thetangible threats blacks pose to whites' pri-vate lives, and a sociocultural theory of prej-udice, which emphasizes abstract, moralisticresentments of blacks traceable to preadultsocialization.

We are partly interested in the origins ofprejudice itself, but our primary focus in thisarticle is on prejudice as a political force. Itremains in this century, as it was in the last,a potent influence in many political choices.Black candidates for public office at all levelsof government are increasingly common;"forced busing" and affirmative action arehot issues in the courts and in school board,mayoral, congressional, and even referen-dum elections; and even seemingly nonracialissues such as the "tax revolt" and unem-ployment are often interpreted as having a

414

PREJUDICE AND POLITICS 415

strong racial component. So our main con-cern is with racial prejudice as a determinantof whites' political behavior.

In all these cases the signs of white resis-tance to change in the racial status quo areunmistakable, but how might such resistancebe understood in psychological terms? At theheart of one prominent hypothesis is racialthreat. That is, resistance may be traced atleast in part to whites' perception that blackspose real and tangible threats to their per-sonal lives. As blacks increasingly demanda larger share of the Good Life, whites reactwith predictable hostility. Thus it is said thatnorthern whites' enthusiasm for the civilrights movement waned when the movement"moved north," because they supposedly fa-vored it only when it did not disrupt theirown lives.

This line of reasoning is of course drawnfrom realistic group conflict theory (LeVine& Campbell, 1972). The origins of this kindof prejudice (and presumably its politicalforce), according to this theory, are to befound in the realities of direct competitionbetween blacks and whites for scarce re-sources. Competitive interdependence pro-duces the perception of threat, which in turnleads to hostility directed at members of thethreatening group. Social attitudes, in short,reflect private interests.

It is easy to imagine the areas of conflictthat might now be feeding whites' racial an-imosities. Many blacks want to move intopleasant suburban neighborhoods, most ofwhich are currently all white. They wantbetter jobs, which threatens to displace whiteworkers. They often want their children toattend integrated schools, which threatensto entail "forced busing" and racial mixingfor young children. And the rising threat ofviolent crime in our society is widely per-ceived as due to high rates of criminal ac-tivity among urban blacks. So a white voter'shostility toward blacks, and consequent po-litical behavior, might plausibly stem fromthe perception that blacks constitute a tan-gible threat to his or her own private life.

A somewhat more formal version of thisracial threat hypothesis can be generatedfrom the familiar means-ends formula forrational decision making. This normativeequation specifies that rationally derived at-

titudes or decisions should be the product ofthe value placed on some end state and theprobability of its attainment. In social psy-chology this is often described as the expec-tancy-value theory of decision making (e.g.,Edwards, 1954; Fishbein & Ajzen, 1975).It is easily applied to the case of racialthreat. The magnitude of racial threat to anindividual white should be the product ofaffect about some end state and expectancythat it may materialize (or, as it is perhapsbetter described in the racial threat context,vulnerability to it.) For example, the threatposed by blacks' possibly moving into a whiteperson's own neighborhood is the product of(a) the white person's evaluation of blacks'living in the neighborhood and (b) the prob-ability that blacks actually will move in.Maximum threat is posed by a very probableand intensely disliked end state. Threat isobviously much reduced if the end state ei-ther is not much disliked or is improbable.

Compelling as the racial threat hypothesismay seem, its empirical support turns out tobe surprisingly weak. Evidence that threatleads to prejudice (or antiblack political be-havior) has been provided by studies of threetypes: (a) those that employ aggregate cor-relations to show, for example, that amongsouthern whites prejudice covaries with therelative size of the nearby black populationor that groups presumed to be in direct eco-nomic competition with blacks (such asworking-class ethnics) express more preju-dice (Blalock, 1967; Key, 1949; Pettigrew& Cramer, 1959); (b) informal time-seriesobservations of change in group stereotypes,as for example, in the shifts in Americanattitudes toward the Japanese registered af-ter Pearl Harbor (this evidence is reviewedby Ashmore, 1970, and by Ehrlich, 1973);and (c) experimental research on interde-pendence, the exemplar being SheriPs(Sherif, Harvey, White, Hood & Sherif,1961) field experiment that varied the natureof interdependence between groups ofschoolboys at a summer camp. All in all, thefindings either are quite indirect; invite al-ternative interpretations; or, in the case ofthe experimental results, are of uncertaingenerality.

Our research is partly a response to theselimitations. It attempts to test the racial

416 DONALD R. KINDER AND DAVID O. SEARS

threat hypothesis directly, by using explicitmeasures of affect and vulnerability for eachof several potent areas of threat: neighbor-hood and social contact, economic compe-tition, busing of school children for racialintegration, and threats to personal safetyfrom black violence. In so doing, we hope atthe same time to be providing a sensitive testof realistic group conflict theory.

A second major approach to prejudiceattributes it to sociocultural learning. In thisview, children and adolescents acquire prej-udice along with other values and attitudesthat are normative in their social environ-ments. Conformity pressures, as well as theintrinsic strength of early-learned attitudes,promote the persistence of prejudice throughthe vicissitudes of later life. Realistic threatsmay come and go, but the solid core of prej-udice remains, no matter how anachronisticit may become. There is of course no ques-tion that the acquisition of prejudice beginsin childhood (e.g., Ashmore & DelBoca,1976; Maykovich, 1975; Middleton, 1976).Our particular purpose here is not to add tothis already well-documented case, but ratherto examine the version of sociocultural learn-ing with most political impact today.

In years gone by, it was easy to specifythe content of this early-learned prejudice;it centered on intentional and legitimate dis-crimination and segregation. However, overthe past 30 years, white opposition to equalopportunity has sharply declined. On votingrights, schools, public accomodations, hous-ing, and employment practices, segregation-ist sentiment has by now all but disappeared.White America has become, in principle atleast, racially egalitarian—a momentousand undeniably significant change (Greeley& Sheatsley, 1971; Taylor, Sheatsley, &Greeley, 1978). Since the explicitly segre-gationist, white supremacist view has all butdisappeared, it no longer can be a majorpolitical force.

What has replaced it, we suggest, is a newvariant that might be called symbolic rac-ism. This we define as a blend of antiblackaffect and the kind of traditional Americanmoral values embodied in the ProtestantEthic (Sears & Kinder, 1971; Sears &McConahay, 1973; McConahay & Hough,

1976). Symbolic racism represents a formof resistance to change in the racial statusquo based on moral feelings that blacks vi-olate such traditional American values asindividualism and self-reliance, the workethic, obedience, and discipline. Whites mayfeel that people should be rewarded on theirmerits, which in turn should be based onhard work and diligent service. Hence sym-bolic racism should find its most vociferousexpression on political issues that involve"unfair" government assistance to blacks:welfare ("welfare cheats could find work ifthey tried"); "reverse discrimination" andracial quotas ("blacks should not be givena status they have not earned"); "forced"busing ("whites have worked hard for theirneighborhoods, and for their neighborhoodschools"); or "free" abortions for the poor("if blacks behaved morally, they would notneed abortions").

If symbolic racism is rooted in deep-seatedfeelings of social morality and propriety andin early-learned racial fears and stereotypes,it may well have little to do with any tangibleand direct impact of racial issues on thewhite person's private life. The stereotypicalsymbols of blacks' violation of traditionalvalues, which are in the media and informalcommunication all the time, may be moreimportant than the realities of the more oc-casional actual damage blacks do to whites'own lives. So, symbolic racism may be, po-litically, the most potent vehicle for racialprejudice today, whereas racial threat towhites' personal lives may have little polit-ical effect and little role in the origins ofsymbolic racism.

In summary, our purpose is to investigatecontemporary white prejudice from two gen-eral theoretical points of view: realistic groupconflict theory, which emphasizes the im-portance of tangible racial threats to privatelife; and the symbolic racism version of thesociocultural perspective, which emphasizesgeneral and moralistic resentments of blacksand is more likely traceable to preadult so-cialization than to current racial threat.These two theoretical perspectives will betested in an explicitly political context. Al-though the study of attitudes for their ownsake has an honorable history, its value is


clearly enhanced if the attitudes can beshown to influence socially important out-comes, such as which officials hold powerand make policy. Consequently, our maindependent variable is voting behavior in twosuccessive pivotal elections that were to de-termine whether the nation's third largestcity would reject its incumbent white mayorfor its first black mayor. As we shall show,racial issues were of central significance inboth races. This was partly because the can-didates were of differing races and partlybecause race was deliberately made a majorcampaign issue. We shall test these theoret-ical alternatives in the following three mainways:

1. Do symbolic racism and racial threatsproduce antiblack voting behavior? Thisfirst cut analysis will assess the simple maineffects of symbolic racism and racial threats(with affects toward potential threats andvulnerability to those threats treated sepa-rately).

2. Does the expectancy-value version ofthe racial threat hypothesis improve its pre-diction of voting behavior? Specificially, donegative affects toward potential threatsproduce antiblack voting behavior moreamong the highly vulnerable than among theminimally vulnerable?

3. Is symbolic racism itself determined byracial threat, contrary to the socioculturalperspective outlined above?

Method

Subjects

Evidence for this investigation comes from personalinterviews with white suburbanites during the 1969 and1973 campaigns for the mayor's office in Los Angeles,California. Both contests featured the white conserva-tive incumbent, Samuel Yorty, challenged by the liberalblack city councilman, Thomas Bradley—a set of cir-cumstances that permitted us the rare luxury of a com-plete replication. On each occasion, Bradley won a plu-rality in the primary election, thereby narrowing a largerfield to two. Yorty won the general election in 1969,with 53% of the vote; in 1973, Bradley won, with 56%of the vote. In both years, the interviewing took placeafter the^primary and about 3 weeks prior to the generalelection. In 1969, 198 adults were interviewed by theNational Opinion Research Center, and in 1973, 239were interviewed by the Field Corporation.

The sample area was identical in both cases: the 1stand 12th council districts in the City of Los Angeles,

which cover the northern half of the San Fernando Val-ley. Neighborhoods were chosen with probabilities pro-portional to their population; age and sex quotas werethen imposed (for a fuller description of this mixed sam-pling design, see Sudman, 1976). The valley is the larg-est "bedroom suburb" within the city limits, with a pop-ulation of nearly one million. Hence our respondentswere like white suburbanites generally—mainly home-owners with good incomes and above average levels ofeducation (more than one third had attended college)—and most identified with the middle class. A majoritywere Protestant; the remainder were Catholic. Nearlyall were married, and most were parents.

This section of the valley, like most suburban com-munities, is rather distant from any major concentrationof blacks, preventing high levels of interracial contact,but is nevertheless close enough to yield widespread feel-ings of racial threat. The valley itself contains a smallblack and Chicane community, Pacoima, but as a wholewas only 1.6% black in 1970. It is separated from therest of the city by a low mountain range and is far fromthe major black residential area of the city (e.g., it isabout 20 miles from Watts, at the closest point). Aboutone third of the sample (36% in 1969 and 34% in 1973)reported virtually no contact with blacks ("are aroundblacks less than once a week"). Nevertheless, as will beseen, many valley residents felt substantial levels of ra-cial threat, perhaps, inter alia, because of the substantialblack population in the city as a whole (18% in 1970),the savage violence of the 1965 Watts Riot (see Searsand McConahay, 1973), and the then slowly but inex-orably progressing court case that ultimately (in 1977)would force desegregation of the area's schools.

MeasuresPersonal racial threats. According to realistic group

conflict theory, the politically effective component ofracism should be respondents' perceptions of threatsposed by blacks to their own personal lives. We assessedracial threat in four domains: (a) neighborhood deseg-regation and interracial social contact; (b) economiccompetition; (c) racial busing; and (d) black violence.The specific items are presented in Table 1, along withthe corresponding distribution of replies to each.

As indicated earlier, racial threat involves both affectand vulnerability (as we choose to call expectancy inthis context). Affect questions tapped the evaluativenature of respondents' reactions to a particular racialthreat if it were to materialize—for example, how theywould feel if their own neighborhoods were to be de-segregated. Vulnerability questions are of two sorts.Subjective vulnerability questions (confined for the mostpart to the 1973 survey) asked respondents to judge thelikelihood that a particular potential personal racialthreat would actually materialize—for example, thelikelihood that black families would soon be moving intotheir neighborhoods. As shown in Table 1, substantialnumbers of suburban respondents felt personally raciallythreatened, although the exact percentage varied con-siderably from domain to domain and from question toquestion. Objective vulnerability questions assessedwhether respondents were in objective situations thatmight make them more vulnerable to racial threats, such


Table 1Personal Racial Threats: Affect and Vulnerability

Correlation withPercent threatened" candidate preference11

Question 1969 1973 1969 1973

Neighborhood desegregation and interracialsocial contact

If a Negro/black family with about the sameincome and education as you moved nextdoor, would you mind it a lot, a little, ornot at all? (% a lot or a little) 36 30 .43** .48**

[How likely is it that] a few black familieswill move into this neighborhood? (% veryor somewhat likely) NA 80 NA .06

How strongly would you object if a memberof your family wanted to bring a Negro/black friend home to dinner? (% verystrongly, strongly, or slightly) 27 27 .50** .58**

Economic competitionAre you satisfied or dissatisfied with your

economic gains compared to those ofNegroes/blacks? (% dissatisfied) 12 18 .03 .31

Have the economic gains of Negroes/blacksbeen about the same, much greater than,greater than, or less than yours over thepast five years? (% greater) 35 29 .30 .30**

In the next few years, it is possible that youor members of your family will have morecontact with blacks than you do now. Howlikely is it that you will have a blacksupervisor at work? (% very or somewhatlikely) NA 48 NA -.18

In the next few years, how likely is it thatmembers of your family or friends will bedenied applications for jobs or promotionsbecause of preferential treatment tomembers of minority groups? (% very orsomewhat likely) NA 58 NA .12

Racial busingHow likely is it that black children will be

bused into the elementary schools of thisneighborhood? (% very or somewhatlikely) NA 46 NA .09

How likely is it that children from theelementary schools in this neighborhoodwill be bused to other parts of the city toachieve school desegregation? (% very orsomewhat likely) NA 36 NA .04

Have children (% with children not inparochial school) 75 NA .03 NA

Have children in public elementary school NA 36 NA .16

Have children in public high school NA 21 NA -.05


Table 1 (Continued)

Percent threatened"Correlation with

candidate preference6

Question 1969 1973 1969 1973

Black violenceHow likely is it that Negroes/blacks will

bring violence to this neighborhood?very or somewhat likely) 35 27 .27 .46**

Live close to Pacoima (% in closest one thirdof census tracts) 30 30 .41** -.09

Note. NA = not asked." The percent not threatened is the remainder of the sample, less about 5% to 10% "don't knows."b The entry is the gamma coefficient indexing the relationship between personal discontent and voting. Some itemshave been reflected so that in each case, a positive gamma means that the threatened are more likely to vote forYorty. In the 1969 sample, n ranges from 157 to 178; in 1973, n ranges from 181 to 232.* p < .05, by x2. ** p < .01, by x2.

as having children in the Los Angeles public schools(who might be affected by busing) or living close toPacoima (which might increase vulnerability to resi-dential desegregation or crimes committed by minori-ties).

Symbolic racism. The symbolic racism items areshown in Table 2. As indicated earlier, they are char-acterized by their abstraction, by their moral tone, andby the absence of any reference to the respondent's per-sonal situation. Almost all relevant items were used in1969. In 1973, in an effort to refine our measurementof symbolic racism, a factor analysis was conducted onthe more numerous symbolic racism items available.Two main factors emerged. One clearly reflected atti-tudes on racial busing: The five busing items loadedfrom .50 to .81 on it, and less than .20 on the otherfactor. The second factor was nearly as distinctive: Theremaining nine items loaded from .36 to .65 on it, andonly one of them loaded as high as .23 on the busingfactor. This dimension included a mixture of symbolicresentments, such as affirmative action, government fa-voritism, and black activism. As shown in Table 2, theitems with the highest loadings formed the basis of twoadditive scales in each survey: opposition to racial bus-ing and expressive racism. The scales are not identicalin both years, since new questions were added in 1973,but the core of items is the same. The median gammaacross items within the busing scale was .63 in 1969 and.65 in 1973; within the expressive racism scale, it was.43 and .38. Defined in this way, the two indexes rep-resent conceptually separable dimensions of symbolicracism, although as expected, they are correlated (in1969, Pearson r = .33; in 1973, r = .39).

Candidate preference. Respondents were asked di-rectly who they intended to vote for. In 1969, 45% re-ported that they intended to vote for Yorty, 43% forBradley, and 13% were undecided. In 1973, the corre-sponding percentages were 51% for Yorty, 41% forBradley, and 8% undecided.

In both surveys, respondents were also provided witha list of six trait adjectives and asked whether or not

each described the two candidates, completing the listfirst for Yorty and then for Bradley. Three of the traitswere positive (honest, intelligent, effective) and threenegative (underhanded, unpleasant, corrupt). An eval-uative index was formed for each candidate, reflectingthe number of positive traits attributed to him minusthe number of negative traits attributed (hence rangingfrom -3 to +3). An overall candidate preference mea-sure was then constructed by subtracting the Yorty in-dex from the Bradley index. In both 1969 and 1973,this measure ranged from —4, indicating a decided pref-erence for Yorty, to + 4, indicating a decided preferencefor Bradley.

The analysis of voting behavior we will present usesthe trait measure of candidate evaluation, rather thanvote intention, because a 9-point scale is more sensitivethan is a simple dichotomous choice. Furthermore, useof a dichotomous dependent variable in multiple regres-sion analysis violates standard Gauss-Markoff assump-tions (Wonnacott & Wonnacott, 1970), which posesdifficulties for estimation and significance testing. Thecandidate preference index is highly correlated with voteintention (in 1969, Gamma = .90; in 1973, Gamma =.88), of course, and the basic findings are the samewhichever measure is used.

Results

As indicated earlier, white opposition toequal opportunity for blacks has all but dis-appeared, according to national survey evi-dence (e.g., Taylor, Sheatsley, & Greeley,1978). Our samples resemble the rest of thewhite public in this respect. Despite Jensen-ism, few thought blacks less intelligent thanwhites (15% in 1969; 16% in 1973); despitecontroversy over "forced busing" for schoolintegration, almost no one supported sepa-rate schools for blacks (8% and 10%, re-


Table 2Symbolic Racism

Percent nonliberalCorrelation with

candidate preference

Question 1969 1973 1969 1973

Expressive racismDo you think that most Negroes/blacks who receive

money from welfare programs could get alongwithout it if they tried, or do they really need thehelp? (% could get along), 54 52 .69** .56**

Negroes/blacks shouldn't push themselves wherethey're not wanted (% agree), 63 64 .48** .44**

Because of past discrimination, it is sometimesnecessary to set up quotas for admission tocollege of minority group students (% disagree), NA 47 NA .39**

Do you think Los Angeles city officials pay more,less, or the same attention to a request orcomplaint from a black person as from a whiteperson? (% more), 26 34 .55** 130**

Of the groups on the card, are there any which youth ink have gained more than they are entitled to?(% choosing "black") 8 10 .40 .48*

It is wrong to set up quotas to admit black studentsto college who don't meet the usual standards (%agree) NA 73 NA .30

Over the past few years, blacks have got more thanthey deserve (% agree), 23 30 .40* .40**

In Los Angeles, would you say many, some, or onlya few blacks miss out on jobs or promotionsbecause of racial discrimination? (% only a fewor none) 34 36 .49** .32*

Opposition to busingBusing elementary school children to schools in

other parts of the city only harms their education(% agree), 54 62 .43** .25

In some cases it is best for children to attendelementary schools outside their neighborhood (%agree), 68 70 .45** .26

Are you in favor of or opposed to the busing ofchildren to achieve racial desegregation of thepublic schools? (% opposed), NA 86 NA .42*

If the Supreme Court ordered busing to achievedesegregation of the public schools, would you beopposed to it? (% yes), NA 79 NA .40*

If necessary, children should be bused to achieveracial desegregation of the public schools (%disagree), NA 83 NA .22*

Note. The percent liberal is the remainder of the sample, less about 5% to 10% "don't knows." Item wordings areapproximate; exact wordings are available upon request. Correlations are gamma. A subscript x indicates that theitem is included in the indices of expressive racism used in the subsequent analysis, and a subscript y that it isincluded in the antibusing index. NA = not asked.* p < .05, by x2. ** P < .01, by x2.


spectively). Consequently, the major politi-cal impact of racism must be carried byother forms of prejudice.

One obvious candidate is racial threat,because many respondents experiencedblacks as directly or potentially threateningtheir personal lives. As indicated in Table1, about half thought that they would bepersonally affected by affirmative action inthe near future. Eighty percent thoughtblack families would soon be moving intotheir neighborhoods. Almost a third felttheir neighborhoods vulnerable to black vi-olence, that blacks' economic gains had ex-ceeded their own, or that more intimatesocial contact with blacks would be ob-jectionable. So, to a considerable number ofthese white suburbanites, personal safety,economic status, and the sanctity of thehome and of the neighborhood all were insome jeopardy.

At the same time, whites also expressedmuch antiblack feeling on symbolic issues(see Table 2). A vast majority opposed bus-ing for racial integration. Nearly as manyresisted the idea of quotas for black collegestudents. From one third to two thirds be-lieved that blacks get more than they de-serve, push themselves where they are notwanted, milk welfare programs, or get moreattention from government than they should.Our first problem, then, is to determine therelative contribution of these two seeminglypotent forms of racial prejudice—racialthreat and symbolic racism—to white sub-urbanites' political choices.

Determinants of the Vote

Symbolic racism versus racial threat.We begin by simply comparing the bivariaterelationships between symbolic racism andpersonal racial threat on the one hand andcandidate preferences on the other. Suchcomparisons suggest that symbolic racismwas the more important determinant of vot-ing behavior in both elections, with the moreprejudiced supporting the white conserva-tive, Yorty. Symbolic racism had a meangamma with candidate preference of .41when all 21 items included in the two surveysare considered (shown in Table 2); racial

threat yielded a mean gamma of .21 over 20items (shown in Table 1).

It is likely that the symbolic racism andracial threat measures share some overlap-ping variance, because of their common ra-cial content. So to sort out more satisfac-torily their separate contributions, it isnecessary to turn to multiple regression anal-yses. The first set of these are done sepa-rately for the 1969 and 1973 data, assessingthe simple main effects of symbolic racismand racial threat. (The only difference be-tween the two was that the 1973 analysisincluded the vulnerability measures.) Theresults are shown in Table 3. They also dem-onstrate the greater importance of symbolicracism in several ways. The simplest is thatthe R2 contributed by symbolic racism isgreater than that contributed by personalracial threat when each is considered alone.Equation 1 for each election year considersonly symbolic racism and accounts for R2sof .223 and .153. Equation 2 considers onlyracial threat and yields less than half thatR2: .096 and .055.

However, this does not show the uniquecontributions of each factor. When both fac-tors are considered simultaneously, in Equa-tion 3, the symbolic racism terms are all sta-tistically significant (though busing was onlymarginally so in 1973, p< .10).' In contrast,only 2 of 12 racial threat terms were signif-icant in the 2 years. This shows that symbolicracism is a considerably more powerful de-terminant of voting behavior than racialthreat when both are considered simulta-neously.

The disadvantage of this analysis is thatit pits a series of individual racial threatitems against two symbolic racism scales andtherefore may stack the cards against thelatter. We next turned to the hierarchicalpossibilities of multiple regression analysisas a vehicle for guarding against this. Herethe question is: do all the several racialthreats to personal life, as a set, add signif-icantly to our ability to predict mayoral vot-ing above and beyond the contribution made

1 This is one instance in which the finding varied some-what depending on the dependent measure. In 1973, thebusing index had a stronger, more reliable effect on voteintention (p < .05) than on the candidate preferencemeasure presented above.


Table 3Effects of Symbolic Racism and Racial Threats on Candidate Preference

Measure

Expressive racismOpposition to busing

1969 («= 178)

Equation

1 2 3

Symbolic racism

.365** — .309**

.217** — .260**

1973 (« = 220)

Equation

1 2 3

.338** — .300**

.119 — .108

Racial threats

Neighborhood desegregation andinterracial social contact

Oppose neighborhood desegregation — .070 -.066 — .122 .083Dislike social contact with blacks — .124 .057 — .053 -.039Likelihood of neighborhood

desegregation — NA NA — -.038 -.030

Economic competitionEconomic resentment of blacks — .080 -.022 — .179* .113Likelihood of black supervisor — NA NA — -.103 -.062Likelihood of being affected by

affirmative action — NA NA — -.058 -.112

Racial busingChildren not in parochial school — .094 .136* — — —Children in public elementary school — — — — -.090 -.083Children in public high school — — — — -.039 -.046Likelihood of busing blacks in — NA NA — .057 .012Likelihood of busing whites out — NA NA — -.076 -.025

Black violenceCloseness to PacoimaFear of black violence

— .107— .148*

R2 values

.223 .096

.075

.160*

.253

— -.037— .133

.153 .055

-.040.112

.144

Additional variance contributed by:Symbolic racism (3-2) .157 .089Racial threat (3-1) .030 -.009

Note. Variables are coded so that a positive effect of either racism or threat will yield a positive beta. Entry is thestandardized regression coefficient (beta) for each predictor variable. R2 is adjusted for number of variables in theanalysis.* / > < .05. * * / > < . 0 1 .

by symbolic racism? The answer is no. In not accounted for by racial threat (16% andthis analysis we start by determining how 9% for the 2 years, as indicated by subtract-much racial threat adds to R2 once symbolic ing the R2 for Equation 2 from that forracism already has been included in the Equation 3). This procedure has the advan-equation—which is not very much (3% in tage of pooling all the contributions made1969, 0% in 1973; subtract the R2 for Equa- by the separate racial threat items into onetion 1 from that for Equation 3). In contrast, increment to R2. This comparison is in ourwhen this procedure is reversed, the power view the strongest and most compelling testof symbolic racism is maintained. Symbolic of the main effects of symbolic racism andracism accounts for considerable variance racial threat. It is apparent that symbolic


racism has powerful unique effects, whereasthose of racial threat are only at chancelevels.

As indicated earlier, even this analysistests only the most rudimentary version ofour hypotheses, since it assesses only themain effects of each set of variables. Beforemoving on to tests of more complex hy-potheses, however, we need to determine thatthese effects of symbolic racism are not ar-tifacts of some third variable or set of vari-ables. A most obvious possibility is demo-graphic factors, but these results werecompletely unaffected by the addition ofdemographic variables (age, education, andsex) to the regression analysis.

Political predispositions represent anothercategory of artifactual possibilities. We haveinterpreted the effects of symbolic racism asreflecting racial prejudice, and hence as re-vealing a predominantly racial basis for thechoice between the two candidates. But theycould conceivably instead reflect some othercorrelated political predispositions, such asliberalism-conservatism and party identifi-cation, since the candidates also differedquite markedly on those dimensions. Indeed,in both contests, both predispositions wereassociated with candidate preference, withRepublicans supporting Yorty more thanDemocrats did (in 1969, r = .15; in 1973,/• = .12; ps<.10), and conservatives morelikely to prefer him than liberals were (in1969, r = .25; in 1973, r = .34; ps < .01).

But these political predispositions did notaccount for the effects of symbolic racism.When they were added to the 1969 regres-sion analysis summarized in Table 3, vir-tually nothing changed: Candidate prefer-ence did not become more predictable;neither partisanship nor liberalism-conser-vatism contributed independently to thevote, once the effects due to symbolic racismwere taken into account; and finally, noneof the regression coefficients indexing theeffects of racial attitudes changed with theentrance of political dispositions into theregression analysis. To be sure, adding themto the 1973 regression analysis resulted ina modest but clearly significant incrementin R2 (due entirely to ideology). Neverthe-less, symbolic racism continued to have a

significant contribution to voting (thoughdiminished by about 10%), and racial threatsto personal life continued to have no effectat all, when the unstandardized regressioncoefficients were considered. Symbolic rac-ism had substantially more effect than ide-ology, even when added to the equation afterit. Hence there is some evidence of slightlyincreased shared variance from 1969 to1973. But in general, the effects of symbolicracism on voting preferences seem indeedproperly interpreted as mainly reflecting ra-cial prejudice rather than nonracial politicalpredispositions. The racial issue generally,and symbolic racism in particular, was cen-tral to white voters' choices in these twoelections.

The expectancy-value approach to racialthreat. The suprisingly poor showing of theracial threat hypothesis so far suggests thatwe mainly devote our remaining analyses totesting it more precisely, in the hope of find-ing conditions under which it does hold, evenif they are more limited than originally ex-pected. An obvious first move is to test anexpectancy-value approach to racial threat.This approach suggests in general that choicebehavior is the product of both affect towardsome end state and the likelihood of its oc-currence. So the political impact of any givendisliked aspect of racial change ought to begreater among respondents most personallyvulnerable to being affected by it. For exam-ple, opposition to integrated neighborhoodsshould have maximum political impactamong respondents whose own neighbor-hoods are in danger of being desegregated.

To test this expectancy-value hypothesis,we considered each area of racial threat inturn. In each case we determined the impacton candidate preference of the affect aboutthe threat (e. g., liking or disliking deseg-regation of one's own neighborhood) sepa-rately for vulnerable respondents (e.g., thosefeeling their own neighborhood was likely tobe desegregated) and those not vulnerableto it (e.g., those feeling it was unlikely to bedesegregated). In operational terms, regres-sion Equation 2, the results of which areshown in Table 3, was reestimated, this timeseparately for vulnerable and not vulnerablerespondents, taking the threats one at a time.


If the personal manifestation of racial threatwas fueling antiblack voting, these regres-sion coefficients ought to be larger for vul-nerable than for nonvulnerable respondents.

The results provide no support for thisexpectancy-value version of the racial threathypothesis. The data are shown in Table 4.In every single case, affect toward the spec-ified racial threat has more effect on votingbehavior among the least vulnerable respon-dents—exactly contrary to the hypothesis.This difference is significant in only one case(using Multiplicative Affect X Vulnerabilityinteraction terms; see Cohen & Cohen,1975). It would be inappropriate (thoughtempting) to conclude that supposed racialthreats actually have most political impactamong those for whom they are most hy-pothetical, but the data do consistently runcounter to the racial threat hypothesis.

A second and complementary version ofthe expectancy-value racial threat hypoth-esis predicts that symbolic racism will have

important political effects only among whiteswhose own lives are safe from racial change.Ideological or moral thinking may be a lux-ury that can be afforded only when an issuehas no tangible consequence; for example,only people without children vulnerable tobeing bused will decide about busing on thebasis of their racial prejudice or integration-ist beliefs. Vulnerable parents, in this view,think first about the welfare of their ownchildren and second about what is best forsociety as a whole. So, according to this hy-pothesis, only among the nonvulnerable re-spondents should symbolic racism affect vot-ing behavior.

The data do not support this version of theracial threat hypothesis either, as shown inTable 5. Symbolic racism relates to candi-date preference at almost exactly the samelevel no matter how personally vulnerablethe respondent is. This is true for both ver-sions of symbolic racism: expressive racismand opposition to busing. Not one of the in-

Table 4Impact on Candidate Preference of Affects Associated With Potential Racial Threats AmongRespondents High and Low in Vulnerability

Impact of affect among respondentswho are:

Highly vulnerable Not vulnerable

Affect/vulnerability

Threat of neighborhood desegregation and interracial social contact

Oppose neighborhood desegregationLikelihood of neighborhood desegregation (1973)Proximity to Pacioma (1969)Proximity to Pacoima (1973)

Dislike social contact with blacksLikelihood of neighborhood desegregation (1973)

.05,-.05-.12

.10

856248

85

1.51..25.76

.33

398469

39

Threat of economic competition

Economic resentments of blacksLikelihood of being affected by affirmative action (1973)Likelihood of having a black supervisor (1973)

-.05.01

73103

.21

.1881

113

Threat of black violence

Fear of black violenceProximity to Pacoima (1969)Proximity to Pacoima (1973)

.22

.044862

.26

.466984

Note, b is the unstandardized regression coefficient, the appropriate measure for subgroup comparisons. Entrieswith common subscripts differ significantly. For proximity, the comparison is between adjacent and distant; forlikelihood, between very likely and unlikely, ignoring likely.


Table 5Impact of Symbolic Racism on Candidate Preference Among Respondents High and Low inVulnerability

Impact of symbolic racism amongrespondents who are:

Form of symbolic racism/dimensionof vulnerability

Highly vulnerable Not vulnerable

Expressive racismLikelihood of neighborhood desegregation

(1973)Likelihood of being affected by affirmative

action (1973)Likelihood of having a black supervisor (1973)Likelihood of blacks bused in (1973)Likelihood of whites bused out (1973)Proximity to Pacoima (1969)Proximity to Pacoima (1973)

Opposition to busingLikelihood of blacks bused in (1973) plus

parents of school childrenLikelihood of whites bused out (1973) plus

parents of school children

.08

.13

.11

.14

.13

.17

.14

.09

.08

85

73103115984862

53

43

.08

.09

.11

.08

.09

.11

.09

.02

.02

39

8111399

1216984

52

61

Note, b is the unstandardized regression coefficient for the effect of symbolic racism on candidate preference withinthe high or low vulnerability group specified. In the case of racial busing, the highly vulnerable group was definedby those who thought busing likely or very likely and had children in the Los Angeles public schools. The notvulnerable group was made up of respondents who thought busing unlikely or very unlikely and had no childrenin the Los Angeles public schools. For other operational definitions of vulnerability, see note to Table 4.

teraction terms reflecting these differencesis significant.2

In short, the expectancy-value renditionsof the racial threat prediction fail to find anysupport in our data. Affects associated withpotential threats do not become more polit-ically consequential as the threats becomemore real. Nor does symbolic racism dimin-ish in political significance among the mostpersonally vulnerable whites. If anything,the relationships run in the opposite direc-tion in both cases. Affects tied to potentialracial threats affected voting, if at all, onlyas they became less anchored in personalexperience—and therefore less realistic andmore symbolic (as indicated in Table 4).Likewise, the relationship between symbolicracism and voting appears to, if anything,gently increase as vulnerability increases (asshown in Table 5). In our view, these anal-yses represent our most sensitive tests of thepolitical impact of racial threat, and theydetect none. Indeed, the largely nonsignifi-cant results tend mainly to go in the oppositedirection.

Symbolic Racism Determined byVulnerability to Racial Threat?

Symbolic racism had a strong effect onvoting behavior, whereas racial threat hadnone. Yet there is another possible versionof the racial threat hypothesis. Racial threatmay produce symbolic racism and in thatway have indirect political effects. Doessymbolic racism in fact merely representquasi-ideological rationalizations of feelings

2 We have included opposition to busing in Table 5rather than Table 4, which is consistent with treatingit as an aspect of symbolic racism. However, the twoanalyses are conceptually identical, and the same regres-sion coefficients could have been placed in either. Theopposition-to-busing items are not as precise indicatorsof what we mean by "affect toward a racial threat" aswe would have liked; better, perhaps, would have been,"How would you feel about your own children beingbused for racial integration?" Irrespective of whetheropposition to busing is regarded as an aspect of symbolicracism (as we argue here and elsewhere; see Sears etal., 1979) or the affective component of a realistic racialthreat, the finding remains the same: Personal vulner-ability to it does not increase its political impact.


of personal vulnerability to blacks? To testthis, multiple regressions were conducted,using the several indices of vulnerability toracial threat as predictors of the two versionsof symbolic racism (expressive racism andopposition to busing).

Racial threat is not a strong predictor ofsymbolic racism, as shown in Table 6. Re-gressing the expressive racism index on 11indicators of racial vulnerability yields a linkbetween the two, but it is a tenuous one(R2 = .049). When measures of social back-ground (age, education, and sex) are addedto the equation, just one measure of vulner-ability continues to exert a statistically re-liable effect: Suburbanites who thought itlikely that they (or members of their familyor their friends) would be denied opportu-nities for jobs or promotions because of pref-erential treatment to members of minoritygroups were higher in expressive racism thanwere those who anticipated no interferencefrom affirmative action practices.

Much the same holds for busing. Oppo-sition to racial busing was intense, of course,but this was so largely irrespective of per-sonal vulnerability. In particular, attitudeson busing were unrelated to whether or notrespondents thought black children wereabout to be bused into their local schools,whether white children from their neighbor-hoods were soon to be bused into blackschools, or whether they had children in theLos Angeles school system. As shown inTable 6, the only significant effect was dueto an interaction between judgment that itwas likely that whites would be bused outof their neighborhoods and having childrenin the Los Angeles high schools. But it is,once again, an anomalous result for the ra-cial threat hypothesis: suburbanites who be-lieved that their own high school aged chil-dren might become participants in a racialbusing program were actually less intenselyopposed to busing. However this curious re-sult is interpreted, the influence of vulnera-bility on attitudes toward racial busing wasmarginal at best (R2 - .03). Opposition toracial busing did not come from those per-sonally at risk.

The independence of racial threat andsymbolic racism can be illustrated in yetanother way. Living close to blacks was as-

sociated with suburbanites' sense of racialthreat: Fears that blacks might bring vio-lence to their own neighborhood diminishedwith distance from Pacoima, sharply so in1969 (Gamma = -.47, p < .01) and less sobut still significantly in 1973 (Gamma =-.25, p < .05). In both years, however, sym-bolic racism was completely unrelated toproximity to blacks: Opposition to busingand expressive racism were just as intenseamong suburbanites whose personal liveswere centered far from blacks as for thosewho lived adjacent to Pacoima (meanGamma across the four tests = -.03). Soalthough the personal threat component ofsuburban racism was responsive to the fear-provoking "realities" of adult life, symbolicforms of racism were not. Whether subur-banites lived many miles from blacks orwithin a stone's throw had nothing to do withtheir symbolic racism.

Finally, we explored one last route bywhich racial threat might conceivably influ-ence symbolic forms of racism. Rather thanproducing opposition to racial change, vul-nerability to racial threat might polarizewhites' racial attitudes. Threat may provokemore thinking about race, and according toTesser's (1978) research, thinking about anattitude object may lead to more extremeattitudes. We tested this hypothesis by "fold-ing" the two symbolic racism indexes at theirneutral points to form new dependent vari-ables and then repeating the regression anal-yses summarized in Table 6. As it turnedout, extreme attitudes on expressive racismor busing were no more common among thepersonally vulnerable than among those whowere personally removed from racial threats.For expressive racism, of the 11 indicatorsof vulnerability to racial threat included inthe regression equation, none was related toextremity. Collectively, their predictive fail-ure was complete (R2 = .00). Similarly, sub-urbanites who thought that busing was com-ing soon to Los Angeles or who had childrenenrolled in the Los Angeles school systemheld only slightly more extreme views onbusing than did those who were personallyunaffected (R2 = .05, p < .05). So personalvulnerability to racial change did not pro-duce more extremity on our symbolic racismmeasures.


Table 6Effects of Racial Vulnerability on Symbolic Racism (1973 Only)

Type of symbolic racism

VulnerabilityExpressive

racismOppositionto busing

Racial threat to neighborhood

Likelihood of neighborhood desegregation .036

Threat of economic competition

Likelihood of black supervisorLikelihood of being affected by affirmative action

-.121.209**

Threat of busing

Likelihood of busing blacks in (A)Likelihood of busing whites out (B)Have elementary school children (X)Have high school children (Y)( A X X)( B X X)( A X Y )(B X Y)

R2

.271*-.169

.039-.086-.198

.173

.035-.103

.049*

.083

.110-.074-.050-.042-.046

.080-.283**

.030*

Note. The entry is /3, the standardized regression coefficient. A positive beta means that greater vulnerability goeswith greater symbolic racism. N = 238. R2 is adjusted for number of independent variables included in the equation.* p < .05. **p < .01.

Discussion

Our findings consistently support the sym-bolic racism hypothesis, which is associatedwith the sociocultural perspective on preju-dice, at the expense of the racial threat hy-pothesis, which is drawn from realistic groupconflict theory. Racial attitudes were shownto be major determinants of voting behaviorin both mayoral elections we examined; andabstract, moralistic resentments of blacks—what we have called symbolic racism—proved to be the overriding determinant ofvoting, for whites shielded from possible tan-gible threats posed by blacks as well as forthose in more imminent jeopardy.

At the same time, and in direct contra-diction of realistic group conflict theory, di-rect racial threats to whites' private liveswere largely irrelevant, either in accountingfor racially relevant political choices or inproducing symbolic racism. We went to con-siderable pains to measure carefully thethreats blacks might pose to whites in severaldomains: neighborhood desegregation andsocial contact, economic competition, school

desegregation, and crime and violence. Weincluded both subjective measures of threat(e.g., the judgment that racial busing wasabout to be implemented in neighborhoodschools) and objective measures (e.g., havingchildren in the Los Angeles school systemwho could be bused). Nevertheless, we coulduncover little evidence that threat affectedvoting behavior or generated more symbolicracism. Nor did we find support for the pre-dicted expectancy-value interactions: Sub-urbanites' affective reactions to threat didnot produce more antiblack voting when thethreats were more likely to materialize. Nor,finally, did personal vulnerability to threatintensify or polarize symbolic forms of prej-udice. By our analysis, the political conse-quences of racial prejudice are carried bysymbolic resentments, not by tangible threats.

What methodological problems might vi-tiate these conclusions? We assessed sym-bolic racism with multiple-item scales, andracial threat with individual items. It is con-ceivable that symbolic racism is thereforemore powerful than racial threat just be-cause of its greater consequent variability


and perhaps greater reliability as well. How-ever, there are good reasons for adopting thisprocedure. It would be conceptually inap-propriate to operationalize racial threat asa single, internally consistent scale. The var-ious racial threat items represent a collectionof quite separate respondent characteristics.The appropriate way, then, to deal with theracial threat indicators is to treat them sep-arately in estimation but to pool their indi-vidual contributions for comparative pur-poses, as we have done in the bottom ofTable 3. Second, our most pointed analysesof the racial threat hypothesis feature inter-actions between appropriately paired affectand vulnerability measures (see Table 4).Such analysis can only be done by treatingthe threat measures separately rather thangrouping them into a scale.

In any case, we are confident this proce-dure does not account for the main resultsof the study for several reasons. Both sym-bolic racism and racial threat were analyzedas individual items rather than scales in pre-liminary reports of the 1969 data (Sears& Kinder, 1971, Note 1), and the substan-tive results were identical to those reportedhere, though not as economically repre-sented. Then, after finding (to our surprise)that symbolic racism dominated racial threatin the 1969 data, we substantially bolsteredthe measurement of threat in the 1973 sur-vey. Improving the measurement of threatnevertheless leads to essentially the samedisappointing return for the racial threathypothesis. The greater variability of thesymbolic racism scales might indeed givethem more statistical power, but we took theprecaution of analyzing them also as tri-chotomized variables, and the results wereidentical. Finally, the most sensitive analysesof racial threat (Tables 4 and 5) show itcoming out overwhelmingly in the directioncounter to hypothesis; scaling racial threatwould not reverse this situation.

Generality of Results

Our main conclusion is strengthened bythe consistency of our findings, even over avariety of tests and measures, across two in-dependent replications 4 years apart andduring a period of sharp political change.

But to draw implications for general theoriesof racial prejudice—for the failure of real-istic group conflict theory and for the successof the sociocultural perspective—demandsconsideration of external validity. How cau-tious should we be in extrapolating fromthese results to other settings and popula-tions?

We believe first of all that the pattern offindings roughly represents what would befound in American suburbs generally, sincedemographically our respondents resembledsuburbanites throughout the country. Butgeneralizing beyond suburban America ismore risky. According to their retrospectivejudgments, our respondents had behavedpolitically quite like California voters as awhole in recent elections, and their racialattitudes were similar to those held by north-ern whites more generally at the time (Gree-ley & Sheatsley, 1971; Sears & Kinder,1971). Still, in exploring racism in suburbiarather than in an area where tangible racialconflict is more intense, we may have sys-tematically underestimated the effects ofracial threat. Perhaps symbolic racism, withits overtones of disinterested morality, is theluxury of those not personally threatened.Of course large numbers of our suburbanrespondents did perceive themselves as beingdirectly threatened by blacks, and thesethreats had no discernible impact (Tables 3-6). Nor did symbolic racism flourish only inthe absence of threat (Table 5).

Furthermore, racial threat has failed tohave a significant effect in other research weare aware of as well, all of it centering onconflict over racial busing. For example, op-position to racial busing has proved to belargely irrelevant to measures of the personalimpact of busing, such as having publicschool children living in an area where bus-ing had already been implemented or waspromised to be implemented soon, the lengthof the bus ride, or the racial composition ofthe host school (Gatlin, Giles, & Cataldo,1978; Sears, Hensler, & Speer, 1979; Sears,Lau, Tyler, & Allen, 1980; McConahay &Hawley, Note 2). One exception is that par-ents in Los Angeles, just before the an-nouncement of a long-awaited desegregationplan, were somewhat more opposed to busing


than non-parents or those in adjacent un-affected school districts (Allen & Sears,Note 3). Nevertheless, in this and all theseother studies, prejudice was the major de-terminant of antibusing attitudes. These re-sults corroborate our own and extend, ifsuggestively, their generality.

Another objection argues for more pru-dent, cautious generalization. Perhaps raceinvites symbolic thinking among suburban-ites because they have virtually no friendlypersonal contact with blacks. In the subur-ban world, there is simply no personal ex-perience with blacks that might offset out-dated racial and moral socialization. Thereis no question that suburbanites in our sam-ples had little direct contact with blacks, asindicated earlier. The degree of racial iso-lation and lack of direct interracial contactfor white suburbanites is striking. However,racial isolation is hardly limited to the sub-urb. Friendly social encounters betweenraces are rare even in integrated neighbor-hoods. According to Bradburn, Sudman, andGockel's (1971) survey, just 1% of whitesliving in substantially integrated neighbor-hoods reported going out for dinner or to amovie with blacks in the preceding severalmonths; less than a third (32%) said thatanyone in their family had ever stopped totalk with black neighbors when they met onthe street. Racial isolation is one of thedefining characteristics of contemporaryAmerican social life. Whether or not sym-bolic racism would remain the politicallypotent form of racial prejudice with higherlevels of interracial contact cannot be an-swered with our data.

Symbolic Politics

Our findings imply that the white public'spolitical response to racial issues is based onmoral and symbolic challenges to the racialstatus quo in society generally rather thanon any direct, tangible challenge to their ownpersonal lives. The political problems withbusing do not arise merely from whiteswhose children are threatened with busing.Resistance to busing may be as intense asit is in part because it conjures up imagesthroughout the white population of innocentwhite children being sent far from their safe

white neighborhoods into schools jammedwith academically unmotivated, disorderly,dangerous blacks. Similarly, opposition toaffirmative action may not come dispropor-tionately from those whose own careers arein jeopardy. The political problems sur-rounding affirmative action may be tracedin part to popular images of hardworkingwhites being unfairly displaced by unde-serving and underqualified blacks. Thepolitically important component of preju-dice—symbolic racism—is compartmen-talized away from the personal impact ofracial conflicts.

Such "psychic segregation" not only holdsfor symbolic racism, but also shows up as abroader distinction made by these subur-banites between the quality of their personallife on the one hand and the quality of na-tional life on the other. Most of our respon-dents seemed quite satisfied with theirpersonal lives—with their communities,neighborhood schools, economic progress,and safety—yet in each domain, they ex-pressed considerable pessimism about thebroader society. For example, whereas mostin our samples were optimistic about theirown economic futures and quite satisfiedwith their own recent gains, a majorityagreed that "the condition of the averageman is getting worse, not better" (see An-drews & Withey, 1976, for similar findingsbased on national surveys). Moreover, whatpersonal discontent suburbanites did expressseemed not to spill over into their judgmentsabout society. Concern about personal safety,for instance, had virtually no implicationsfor suburbanites' broader views on crime.Those who believed that crime was on therise in their own neighborhood were not, asa consequence, any more enthusiastic abouta number of "law and order" policies suchas the death penalty and restrictions on bailor any more likely to .think that streets inAmerica were no longer safe. Personal fearsand frustrations, however important andpreoccupying they may be, seemed not toaffect their broader social attitudes.

Additional support for this argument isprovided by a number of recent investiga-tions. We have already referred to the re-search on racial busing, which consistentlydemonstrates that opposition to busing stems


largely from longstanding political and ra-cial values and that the personal impact ofbusing is of far less significance. The failureof personal concerns and discontents to takeon political significance has been demon-strated in several other contexts as well: inattitudes toward the Vietnam war (Lau,Brown, & Sears, 1978), in public responseto the energy crisis (Sears, Tyler, Citrin,& Kinder, 1978), and in the political con-sequences of economic discontent (Kinder,in press; Kinder & Kiewiet, 1979, Note 4).In each case, judgments about the nationalcondition were largely autonomous frompersonal experiences and predicaments, andeach time it was the more general form ofdiscontent, not the personal, that influencedpolitical preference.

One recent study provides an interestingcounterexample. Self-interest did seem toinfluence Californians' attitudes towardProposition 13, the property tax referendum,in 1978 (Sears, Note 5). Owning a home andnot having a government job both were re-lated to greater support for Proposition 13;yet even in this case, where self-interestseemed so clearly linked to a political choice,the impact of symbolic predictors on supportfor Proposition 13 was very great. So, al-though it seems unlikely that attitudes neverserve an instrumental function, it is possiblethat they do so more rarely than is usuallyassumed or under only special circum-stances.

In short, unhappiness with American so-ciety appears to be largely unrelated to per-sonal discontents. The stresses and strainsof private life are insulated from politicaland social malaise. Why this is so—why theconcrete realities of people's everyday livesare so compartmentalized away from theirjudgments about the society as a whole—isa complicated but researchable question. Inpursuing it we hope to learn more about theparticular case examined here: why symbolicresentments have a more salient influencethan tangible threats on whites' political re-sponses to racial conflict.

Reference Notes

1. Sears, D. O., & Kinder, D. R. The "good life," whiteracism, and the Los Angeles voter. Paper presented

at the annual meeting of the Western PsychologicalAssociation, Los Angeles, April 1970.

2. McConahay, J. B., & Hawley, W. D. Is it the busesor the blacks'! Paper presented at the 85th annualmeeting of the American Psychological Association,San Francisco, August 1977.

3. Allen, H. M., Jr., & Sears, D. O. White oppositionto busing in Los Angeles: Is self-interest rejuve-nated"! Paper presented at the 86th annual meetingof the American Psychological Association, Toronto,August 1978.

4. Kinder, D. R., & Kiewiet, D. R. Sociotropic politics.Unpublished manuscript, Yale University, 1979.

5. Sears, D. O. The Jarvis Amendment: Self-interest, or symbolic politics. Paper presented at the 86th

annual meeting of the American Psychological As-sociation, Toronto, August 1978.

References

Allport, G. W. The nature of prejudice. New York:Doubleday, 1954.

Andrews, F. M., & Withey, S. B. Social indicators ofwell-being. New York: Plenum, 1976.

Ashmore, R. D. Prejudice: Causes and cures. In B. E.Collins (Ed.), Social Psychology. Reading, Mass.:Addison-Wesley, 1970.

Ashmore, R. D., & DelBoca, F. K. Psychological ap-proaches to understanding intergroup conflict. InP. A. Katz (Ed.), Towards the elimination of racism.New York: Pergamon Press, 1976.

Blalock, H. M., Jr. Toward a theory of minority-grouprelations. New York: Wiley, 1967.

Bradburn, N. M., Sudman, S., & Gockel, G. L. Sideby side: integrated neighborhoods in America. Chi-cago: Quadrangle Books, 1971.

Cohen, J., & Cohen, P. Applied multiple regression/correlation analysis for the behavioral sciences. Hills-dale, N.J.: Erlbaum, 1975.

Edwards, W. The theory of decision-making. Psycho-logical Bulletin, 1954, 51, 380-417.

Ehrlich, H. J. The social psychology of prejudice. NewYork: Wiley, 1973.

Fishbein, M., & Ajzen, I. Belief, attitude, intention, andbehavior. Reading, Mass.: Addison-Wesley, 1975.

Gatlin, D. S., Giles, M. W., & Cataldo, E. F. Policysupport within a target group: The case of school de-segregation. American Political Science Review, 1978,72, 985-995.

Greeley, A. M., & Sheatsley, P. B. Attitudes towardracial integration. Scientific American, 1971, 223,13-19.

Katz, P. A. Racism and social science: Towards a newcommitment. In P. A. Katz (Ed.), Towards the elim-ination of racism. New York: Pergamon Press, 1976.

Key, V. O., Jr. Southern politics. New York: RandomHouse, 1949.

Kinder, D. R. Presidents, prosperity, and public opinion.Public Opinion Quarterly, in press.

Kinder, D. R., & Kiewiet, D. R. Economic grievancesand political behavior: The role of collective discon-tents and symbolic judgments in congressional voting.American Journal of Political Science, 1979, 23,495-527.


Lau, R. R., Brown, T. A., & Sears, D. O. Self-interestand civilians' attitudes toward the Vietnam War. Pub-lic Opinion Quarterly, 1978, 42, 464-483.

LeVine, R. A., & Campbell, D. T. Ethnocentrism. NewYork: Wiley, 1972.

Maykovich, M. K. Correlates of racial prejudice. Jour-nal of Personality and Social Psychology, 1975, 32,1014-1020.

McConahay, J. B., & Hough, J. C., Jr. Symbolic racism.Journal of Social Issues, 1976, 32, 23-45.

Middleton, R. Regional differences in prejudice. Amer-ican Sociological Review, 1976, 41, 94-117.

Pettigrew, T. F., & Cramer, M. R. The demographyof desegregation. Journal of Social Issues, 1959, 15,61-71.

Sears, D. O., Hensler, C. P., & Speer, L. K. Whites'opposition to "busing": Self-interest or symbolic pol-itics? American Political Science Review, 1979, 73,369-384.

Sears, D. O., & Kinder, D. R. Racial tensions and votingin Los Angeles. In W. Z. Hirsch (Ed.), Los Angeles:Viability and prospects for metropolitan leadership.New York: Praeger, 1971.

Sears, D. O., Lau, R. R., Tyler, T., & Allen, H. M.,Jr. Self-interest versus symbolic politics in policy at-

titudes and presidential voting. American PoliticalScience Review, 1980, 74, 670-684.

Sears, D. O., & McConahay, J. B. The politics of vi-olence: The new urban blacks and the Watts Riot.Boston: Houghton-Mifflin, 1973.

Sears, D. O., Tyler, T. R., Citrin, J., & Kinder, D. R.Political system support and public response to theenergy crisis. American Journal of Political Science,1978, 22, 56-82.

Sherif, M., Harvey, O. J., White, B. J., Hood, W. R.,& Sherif, C. W. fntergroup conflict and cooperation:The robber's cave experiment. Norman: Universityof Oklahoma Book Exchange, 1961.

Sudman, S. Applied sampling. Chicago: Aldine, 1976.Taylor, D. G., Sheatsley, P. B., & Greeley, A. M. At-

titudes toward racial integration. Scientific American,1978, 238, 42-49.

Tesser, A. Self-generated attitude change. In L. Ber-kowitz (Ed.), Advances in experimental social psy-chology (Vol. 11). New York: Academic Press, 1978.

Wonnacott, R. J., & Wonnacott, T. H. Econometrics.New York: Wiley, 1970.

Received July 7, 1979Revision received July 24, 1980 •

Documents

Prejudice and Politics: Symbolic Racism Versus Racial ... · Prejudice and Politics: Symbolic Racism Versus Racial Threats to the Good Life Donald R. Kinder Political Science Department,