CIVILRIGHTSANDPOLITICALREALIGNMENT 483 CivilRightsLiberalismandthe SuppressionofaRepublican PoliticalRealignmentinthe UnitedStates,1972to1996

ClemBrooks Indiana University

Arguably the most dramatic trend in U.S. public opinion during the past 30 years has been Americans’ liberalizing attitudes toward the civil rights of African Ameri- cans, women, and more recently, gays and lesbians. Do these changing attitudes have any behavioral or organizational effects, or are they of little consequence for understanding social and political changes since the 1960s? While an influential strain of research has questioned the causal significance of trends in civil rights attitudes, past studies have not developed a systematic test of relevant hypotheses. This study presents such a test, analyzing the effects of trends in civil rights attitudes on an important type of behavior (presidential vote choice) and on a major type of organization (political parties). Guided by theories of issue voting and political realignment, results show that changing attitudes have had significant behavioral and organizational effects on vote choice and election outcomes since 1972. In par- ticular, the margins of Republican presidential victories in the 1980s were sup- pressed by liberal trends in attitudes, while Democratic presidential victories in the 1990s depended upon these trends. These findings lead to a better understanding of the substantial but poorly understood effects of changing attitudes toward civil rights and of the widely debated scenario of a Republican political realignment.

ne of the largest trends in U.S. pub- Schuman et al. 1997; Smith and Sheatsley Olic opinion during the past 30 years 1984), women (Ferree 1974; Mason, Czajka, has been Americans’ changing attitudes to- and Arber 1976; Mason and Lu 1988; ward the civil rights of Thornton and Freedman 1979), and more re- (Farley 1997; Firebaugh and Davis 1988; cently, gays and lesbians (Yang 1997; Zaller 1992, chap. 12). These trends are distin- Direct all correspondence to Clem Brooks, De- guished by three features. First, they involve partment of Sociology, Indiana University, Bloomington, IN, 47405 ([email protected]). beliefs about general, normative principles Data and codebooks from the American National rather than preferences about what govern- Election Studies were provided by the Inter-uni- ment policies should be adopted to imple- versity Consortium for Political and Social Re- ment such principles.1 Second, they repre- search. Data and codebooks from the General Social Surveys were provided by the Roper Cen- Timpone, three ASR reviewers, a deputy editor, ter. Responsibility for the analysis and interpreta- and Editor Glenn Firebaugh for their comments. tion of the data rests solely with the author. A 1 The “principle versus policy” distinction has version of this paper was presented at the 1999 been used extensively to study racial attitudes meeting of the American Sociological Associa- (Schuman et al. 1997, chap. 3; Sniderman, tion in Chicago. I thank Arthur Alderson, David Brody, and Tetlock 1991, chap. 4). It is relevant Brady, Paul Burstein, Thomas DiPrete, Martin to the current study, given that attitudes toward Gilens, Andrew Greeley, Michael Hout, Jeff civil rights principles but not attitudes for or Manza, Whitney Pope, Brian Powell, Kent Red- against government policies have experienced ding, John Skrentny, Marylee Taylor, Richard liberal trends. AmericanSociologicalReview,2000,Vol.65(August:483–505) 483 484 AMERICANSOCIOLOGICALREVIEW sent liberal shifts involving the extension of Taken together, these changes in attitudes rights to members of a particular group who appear to represent a profound transforma- have historically been denied such rights. tion. Surprisingly, however, the behavioral Third, these trends have not given rise to an and organizational consequences of these illiberal reversal: When public support for a trends—indeed, whether there have been any specific group’s civil rights increases, it has such consequences—are poorly understood. tended not to decline, oftentimes continuing This partly reflects growing scholarly inter- to rise or stabilizing at a high level (Smith est in other types of attitudes, especially 1990). among analysts of race.4 But it also stems The magnitude of these trends can be best from a prevailing tendency to view trends in appreciated when it is compared with the civil rights attitudes as superficial in nature high levels of intolerance found by opinion or of little causal consequence. researchers in the 1950s (Farley 1997; Jackman’s (1978, 1994; Jackman and Schuman et al. 1997; Stouffer [1955] 1963). Muha 1984) influential work offers the most For instance, where once whites were un- comprehensive formulations of this thesis. willing to extend the most basic civil rights In an early study, Jackman (1978) presents to African Americans, they now overwhelm- what can be termed the “superficial liberal- ingly endorse racial equality as a matter of ism” version of this thesis. While her ex- principle.2 Likewise, a nearly universal ac- plicit focus is on the relationship between ceptance of highly restrictive gender roles education and support for African Ameri- has given way to much greater willingness cans’ civil rights, her interpretation of this (among both men and women) to grant relationship implies that liberal responses to women equal access to political and eco- survey questions about civil rights are often nomic opportunities (Mason and Lu 1988; superficially held and incapable of dispos- Page and Shapiro 1992). And in recent years, ing individuals to act consistently on the ba- despite little change in beliefs about the ac- sis of their professed beliefs. This thesis ceptability of homosexuality, there has been calls into question the sincerity of liberal re- a significant increase in public support for sponses to civil rights items as well as their the rights of gays and lesbians in employ- significance as a causal factor influencing ment and housing.3 the behavior of individuals. Jackman’s sub- sequent work (Jackman 1994; Jackman and 2 Suggesting widespread public acceptance of Muha 1984) suggests a second version of Jim Crow laws, 60 percent of whites in 1963 this thesis, which hypothesizes that because agreed that “whites have a right to keep Blacks dominant groups’ interests dispose them to out of their neighborhoods.” The percentage adopt liberal ideologies primarily to legiti- agreeing decreased to 40 in 1976, and decreased mate their advantageous positions, changing further to just 13 percent in 1996 (Schuman et al. civil rights attitudes are by themselves gen- 1997:106–107). This liberalizing trend has not erally insufficient to lead to changes in ma- extended to policy attitudes, however; relatively jor organizations and institutions. While this stable majorities have opposed government pro- grams designed to implement racial equality since the 1960s (Schuman et al. 1997; Steeh and 4 These scholars have increasingly turned to Krysan 1996). questions about symbolic racism (Sears et al. 3 While attitudes toward gays and lesbians re- 1997), conflict among racial groups (Bobo and flect high levels of negative affect compared with Hutchings 1996), and attitudes toward race-re- attitudes toward other major social groups lated public policies (Kinder and Sanders 1996; (Sherrill 1996), support for the position that “ho- Sniderman et al. 1991). This important research mosexuals should have equal rights in terms of has begun to develop a systematic portrait of the job opportunities” has nevertheless increased origins and consequences of racism (Bobo 1997), from 56 percent in 1977 to 83 percent in 1996 but it does not answer questions about the conse- (Yang 1997:498). The persistence and degree of quences of growing support for the civil rights negative affect make such increases in support of African Americans (and also for women and for gay and lesbian civil rights all the more note- gays/lesbians) despite the persistence of gener- worthy, providing another instance of the “prin- ally low levels of support for policy implementa- ciple versus policy” phenomenon that character- tion or patterns of negative affect toward specific izes civil rights-related attitudes. groups. CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 485

“insignificant liberalism” thesis makes no consequences. In addition to contributing to assumptions about the sincerity of liberal at- the study of political change, such an analy- titudes, it casts doubt on the latter as a factor sis provides a systematic means of taking up capable of contributing to social change. the challenge presented by the “superficial Although Jackman’s work presents the liberalism” and “insignificant liberalism” most explicit statement, much of contempo- theses. If the superficial liberalism thesis is rary research on racial attitudes has implic- true, we should expect that the political pref- itly discounted the significance of changing erences of “superficial” liberals will differ attitudes toward women’s and minorities’ little from those of their more explicitly il- civil rights since the 1960s. By raising the liberal counterparts, leading to no significant possibility that changes in civil rights atti- association between civil rights attitudes and tudes are either cognitively superficial or too voting behavior. Conversely, finding such a insignificant as a causal factor to influence relationship would indicate that liberal re- social change, both versions of this thesis sponses to survey questions reflect real, un- present a constructive but unmet challenge derlying attitudes that provide people with a to research on civil rights attitudes basis for evaluating and choosing between (Schuman et al. 1997:290). More specifi- political candidates. Furthermore, if trends cally, if either version of the thesis is true, in civil rights attitudes have significantly af- the attitudinal component of the “civil rights fected the relative standing of the two major revolution” (Farley 1996:336–37) is an arti- political parties, this would cast doubt on the fact of scholarly misinterpretation. However, insignificant liberalism thesis by providing if civil rights attitudes have led to significant evidence that such changes in attitudes have behavioral and organizational changes, then contributed to an important type of political- many social scientists have ignored an im- organizational change. portant transformation in public attitudes I emphasize at the outset that this study that is worthy of more systematic study. does not investigate the causal origins of changing attitudes toward civil rights. This CIVILRIGHTSATTITUDES is a separate question and has been partially ANDTHESTUDYOF addressed in a literature on the roles of co- ELECTORALPOLITICS hort and period effects in generating opinion change (Davis 1992; Firebaugh and Davis I use electoral politics as the empirical con- 1988). Instead, my focus is on the behavioral text in which to develop a systematic analy- and organizational consequences of changes sis of the behavioral and organizational con- in civil rights attitudes. sequences of changing attitudes toward civil In the first section of the paper, I use theo- rights.5 Electoral politics involves an impor- ries of issue voting and political realignment tant behavior (voting), type of organization to examine the grounds for expecting chang- (political parties), and a major U.S. institu- ing civil rights attitudes to have influenced tion (elections). If civil rights attitudes have voting behavior and the outcome of elections changed the behavior of individual voters or during the past three decades. The paper’s the outcome of elections, these effects would second section presents the data, measures, increase our understanding of their political and models used in this study; analyses of the political effects of civil rights attitudes 5 No previous study has attempted to analyze are presented in the third section. In conclu- the effects of liberal trends in civil right attitudes on voting behavior and elections. Instead, con- temporary political behavior research has tended in civil rights attitudes as a source of political to focus on economic factors, information-pro- change, focusing instead on racial conflict as a cessing and media effects, social networks, and factor contributing to Democratic electoral losses racial cleavages (Carmines and Huckfeldt 1996; and to the breakup of that party’s coa- Dalton and Wattenberg 1993). While path-break- lition. By contrast, my analyses show that chang- ing research by scholars such as Carmines and ing attitudes toward civil rights were a major fac- Stimson (1989) has established the emergence of tor keeping the Democratic Party competitive in a racial among political elites and ordi- presidential elections during the past three de- nary citizens, such research ignores liberal trends cades. 486 AMERICANSOCIOLOGICALREVIEW sion, I discuss the significance of the study’s titudes, judge which party’s candidate is results for understanding the substantial im- closer to their own position, and vote ac- pact of civil rights attitudes and the widely cordingly. Using this theory, voters who are debated scenario of a Republican political influenced by their beliefs about civil rights realignment. can be expected to choose the candidate per- ceived to be closest to their own views.6 THEORIZINGTHEPOLITICAL To illustrate the logic of civil rights issue EFFECTSOFTRENDSINCIVIL voting, consider the following data from the RIGHTSATTITUDES 1988 National Election Study. The three identically coded seven-point scales in Fig- Any empirical analysis of the political ef- ure 1 measure voters’ attitudes toward fects of changing attitudes toward civil women’s rights, their perceptions of the Re- rights requires that we address two theoreti- publican and Democratic parties’ corre- cal questions. First, why might civil rights sponding issue positions, and respondents’ attitudes affect the behavior of individual perceptions of the issue positions of the Re- voters? Second, why might trends in the ag- publican and Democratic candidates. Com- gregate level of support for civil rights af- paring across the three rows shows whether fect the outcome of elections and thus the voters consistently perceived differences be- relative advantage of political parties? I use tween the major parties and their candidates theories of issue voting and political realign- on the issue of women’s rights in 1988. ment to identify the causal mechanisms that Figure 1 shows that the women’s rights is- link changing civil rights attitudes to the be- sue provided voters with a clear basis for havior of individual voters and also to the evaluating and thus choosing between presi- organizational level of election outcomes. dential candidates in 1988. Differences in perceptions of the two parties’ positions on CivilRightsAttitudesasaSource women’s rights (.72) and also the candidates’ ofVoteChoice corresponding positions (.88) are substan- tial—voters perceived Democrats as more Since the 1964 presidential election pitted supportive of women’s rights than Republi- racial liberal Lyndon Johnson against racial cans. Such findings are consistent with a conservative Barry Goldwater, analysts of body of contemporary research investigating U.S. political parties have identified signifi- how changing attitudes toward various social cant differences in the positions endorsed by issues have affected the political alignments the Democratic versus Republican parties on of specific classes and religious groups civil rights issues. Emerging first with re- (Brooks and Manza 1997; Manza and Brooks spect to African American civil rights (Car- 1997). Although that prior research focused mines and Stimson 1989), these divergent on particular segments of the electorate, it positions subsequently have come to encom- nevertheless provides initial grounds for ex- pass women’s rights, and more recently gay/ pecting that civil rights attitudes will influ- lesbian civil rights (Brennan 1995; Freeman ence the behavior of all voters. My more sys- 1993; Mansbridge 1986). For civil rights is- tematic analysis considers attitudes toward sues to affect the behavior of voters, how- ever, political parties must not only endorse 6 Issue-voting theory assumes that attitudes divergent policy positions, voters must also cause vote choice. While this theory is widely perceive such differences. Perceptions of accepted, Wlezien, Franklin, and Twiggs’ (1997) party/candidate differences are a central ele- study of national economic perceptions offers an ment in theoretical models of issue voting intriguing, alternative interpretation, namely, that (Downs 1957; Fiorina 1981), and while vote choice itself can sometimes affect attitudes. these models have been applied most exten- But the monotonic pattern of change in civil sively to the study of economic issues rights attitudes since the 1960s casts doubt on the relevance of this type of theory to understanding (Alvarez and Nagler 1995), they can readily the civil rights attitudes/vote choice relationship. be applied to other issues. More specifically, This is because the latter theory presupposes that issue voting can be seen as a three-step se- a growing Democratic advantage would have had quence in which voters first consult their at- to precede increases in support for civil rights. CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 487

Recently there has been a ot of talk about women’s rights. Some people feel that women should have an equal role with men in running business, industry, and government. Others feel that women’s place is in the home.

a. Average Self-Placement Self

0123456 Women’s place is Women and men should in the home. have an equal role.

b. Average Placement of the Republican and Democratic Parties

GOP Dem

0123456 Women’s place is Women and men should in the home. have an equal role.

c. Average Placement of George Bush and Michael Dukakis

Bush Dukakis

0123456 Women’s place is Women and men should in the home. have an equal role.

Figure 1. Respondents’ Attitudes toward Women’s Rights and Their Perceptions of the Positions of the Democratic and Republican Candidates and Parties: 1988 Presidential Election

civil rights for multiple groups over a longer legislation at the federal level is strongly in- time period, while also ensuring that the as- fluenced by the political party that controls sociation between civil rights attitudes and the presidency and the Congress. vote choice is not spurious. While the size of these organizational ef- fects may vary considerably, a particularly OrganizationalEffects: interesting outcome involves the widely-de- ElectionOutcomesand bated scenario of political realignment. Key PoliticalRealignment (1955, 1959) defines a realignment as a change in voters’ political allegiances that An association between civil rights attitudes culminates in the creation of a new majority and voting behavior would indicate that such party that subsequently dominates elections. attitudes have behavioral consequences. Whereas “critical” realignments (Key 1955) However, change over time in the aggregate involve abrupt changes that occur during the level of support for civil rights may affects course of a single election, “secular” realign- not only the behavior of individual voters but ments are assumed to evolve more slowly also the outcome of elections. If these effects over a longer period of time (Key 1959). consistently favored a specific political party, Both types of realignments are accompanied this would indicate that changing views of by a new issue that differentiates parties, and civil rights have had significant organiza- unless the party (or parties) whose policy tional effects on the success of that party and position is less popular manages to neutral- its ability to win elections. Organizational ize voters’ perceptions of such differences, effects are doubly significant because the the dominant party has an opportunity to so- passage and implementation of civil rights lidify its electoral advantage (Burnham 488 AMERICANSOCIOLOGICALREVIEW

1970; Clubb, Flanagan, and Zingale 1990; known factors that have disproportionately Seagull 1980; Sundquist 1983; for a review, benefited Republican candidates (e.g., post- see Bass 1991). war increases in economic affluence and de- The precise temporal duration of a major- fection by white southern voters from the ity party’s dominance over elections has Democratic Party [Brooks and Brady 1999; never been fully specified in realignment Carmines and Stimson 1989; Petrocik 1981]) theory. However, taking as a paradigmatic have been neutralized by the poorly under- example Franklin Roosevelt’s dramatic 1932 stood effects of changing civil rights atti- and 1936 presidential victories that ushered tudes, these divergent sources of political in the New Deal, a majority party should win change may have canceled one another out. at least five consecutive elections for a re- Because this scenario involves the simulta- to have occurred (as exemplified neous operation of multiple factors, it is a in the Democratic Party’s control over the somewhat more complex result than the tra- U.S. presidency from 1932 to 1952). Using ditional expectations of realignment theory. this criterion, the Republican advantage in However, it can readily be investigated using presidential politics (five nonconsecutive the research design discussed below. victories in the six presidential elections be- tween 1968 and 1988) falls just short of a DATA,MEASURES,ANDMODELS realignment. Given that there is thus little evidence for an electorate-wide realignment My goal is to analyze the effects of chang- since the 1960s (Ladd 1999; Shafer 1991), ing attitudes toward civil rights on voting changing attitudes toward civil rights cannot behavior and election outcomes. These have produced this dramatic and extremely analyses provide a means of evaluating the rare type of political change. superficiality or causal significance of However, one intriguing scenario, and the changes in civil rights attitudes in the insti- one I consider in this study, is that changing tutional arena of national elections. My views of civil rights could have had impor- analyses also compare the effects of civil tant organizational effects by suppressing rights attitudes with the corresponding ef- what would otherwise have been a secular fects of other, better-known sources of the realignment.7 To express this hypothesis an- vote. This ensures that the effects of civil other way, liberal trends in civil rights atti- rights attitudes are not spurious, while pro- tudes held in check what would have been a viding useful information about their com- complete Republican dominance of the presi- parative magnitude. I focus on presidential dency since the 1960s. If true, this scenario elections to keep the scope of the analysis of suppressed realignment would explain manageable, acknowledging that the effects why researchers have searched unsuccess- of civil rights attitudes may differ in other fully for a Republican realignment during national or in subnational elections. this time.8 More specifically, if the better- Data 7 The steady, approximately linear rate of change in civil rights attitudes implies that such I analyze data from the National Election changes could not have suppressed a critical re- Studies’ (NES) surveys of voting behavior in alignment, given that a critical realignment presidential elections (Center for Political would have had to occur during the course of one Studies 1995, 1997). The NES surveys rep- or two elections. Note that changing civil rights resent the highest-quality available data on attitudes could have produced significant but voters’ attitudes, sociodemographic at- nonrealigning effects on the relative advantage tributes, and voting behavior. These data ful- enjoyed by the two major parties. However, the fill three critical requirements for an analy- magnitude of the effects found in this study is such that the suppressed-realignment character- ization is appropriate. 1983). The suppressed-realignment hypothesis 8 Theories of political realignment generally implies that the causal processes affecting presi- emphasize that a new majority party will subse- dential elections may also have affected House quently dominate both the Congress and the and Senate elections, but my focus here is re- presidency (Burnham 1970; Key 1955; Sundquist stricted to presidential elections. CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 489 sis of the political effects of civil rights atti- I have pooled the seven separate NES tudes: Identically worded survey items9 presidential election year surveys (for 1972 measuring support for civil rights, items through 1996) into a data set in which time measuring individuals’ voting behavior, and is a covariate. I measure time in two ways: additional items measuring other known first as a series of dummy variables in the sources of the vote. These surveys cover a regression models (treating 1972 as the ref- relatively long time period during which erence category), and second as a continu- public support for the civil rights of African ous variable in which election year is coded Americans, women, and gays and lesbians 1 for 1972, 2 for 1976,..., and 7 for 1996. increased dramatically.10 The data thus per- The dummy variables are included in the mit an analysis of the political effects of model in order to control for the main effects these aggregate-level shifts in opinion. of individual elections. I use the continuous The items analyzed in this study refer ex- time covariate to measure interaction effects plicitly to civil rights–related laws or to nor- with civil rights attitudes. These interactions mative principles underlying civil rights for show whether the impact of civil rights atti- a specific group. The single exception is an tudes on vote choice has increased (or de- item that asks respondents their opinion of creased) over time. African American civil rights leaders (rather Because the item measuring attitudes to- than civil rights laws or principles), but it is ward gay and lesbian civil rights has only the best available NES item on African been fielded since 1988, the analysis of this American civil rights that covers the post- issue is presented separately. In the first 1960s period. However, the General Social stage of the analysis, I use the NES data to Survey (Davis and Smith 1997) has consis- estimate the effects of attitudes toward the tently fielded items that ask respondents civil rights of African Americans and women about their attitudes toward blacks’ civil on vote choice between 1972 and 1992 (ex- rights, thus providing a means of corrobo- amining also whether these effects have rating the results of my analysis of the NES changed over time). In the final stage, I con- item. Given that the two surveys present sider the most recent presidential elections, similar findings about the political effects of analyzing the effects of attitudes toward gay/ attitudes toward African Americans’ civil lesbian civil rights on vote choice in 1992 rights, the difference in question wording and 1996. between NES and GSS items is largely irrel- 11 evant to the current study. DependentVariable

9 Items with identical wording are essential to Throughout the analyses, my dependent any analysis of the political effects of changing variable is vote choice in presidential elec- attitudes, given that changes in wording may pro- tions.12 In many elections, presidential vot- duce different estimates of association with other items, thereby introducing significant errors into modest association between measures of civil the measurement of over-time processes. rights support and measures of government 10 I do not analyze the effects of attitudes to- policy implementation as providing evidence in ward abortion in this study. The NES items on support of the thesis. However, this interpretation abortion have undergone significant changes in does not necessarily follow, given that many wording and response categories during this time Americans (including civil rights supporters) fa- period, and the abortion items fielded in the Gen- vor individualistic values and thus do not support eral Social Surveys (GSS) show few attitudinal government activism independently of their atti- trends between 1972 and 1996. Analysis of the tudes toward civil rights (Sniderman et al. 1991, GSS items provides evidence that abortion atti- chap. 4). The prominence of such values in U.S. tudes have an independent effect on major party public opinion is well known (see Feldman and vote choice, but given the absence of trends and Zaller 1992; McClosky and Brill 1983), and us- the internal complexity of the abortion issue, I ing measures of policy implementation to study postpone presentation of these results (for analy- support for civil rights principles is thus prob- sis of the emergence of partisan conflicts over lematic because policy implementation also in- abortion, see Adams 1997). volves attitudes toward government. 11 In her original statement of the superficial 12 I also developed an analysis of party identi- liberalism thesis, Jackman (1978) interpreted the fication as a dependent variable. This analysis 490 AMERICANSOCIOLOGICALREVIEW ing is dominated by the binary choice of women’s rights. The gay/lesbian civil rights Democratic versus Republican candidate. item asks respondents: “Do you favor or However, political independent Ross Perot oppose laws to protect homosexuals against received a nontrivial share of the popular job discrimination?” This item is dichoto- vote in both 1992 (19.3 percent) and 1996 mous, coded 1 if a respondent favors anti- (8.5 percent). Given the possibility that discrimination laws, and 0 if the respondent changing attitudes toward civil rights have opposes such laws. disproportionately benefited both Demo- cratic and third-party candidates relative to OtherIndependentVariables Republican candidates, I measure presiden- tial vote choice as a multi-category variable To ensure that the effects of civil rights atti- in the analyses. The vote choice variable is tudes on vote choice are not spurious (i.e., coded 1 if the respondent voted for the Re- the product of a third factor that influences publican candidate, 2 for the Democratic both variables while also explaining away candidate, 3 for third-party candidates, and their association), I include in the analysis 4 if a respondent reported not voting in a variables measuring well-known sources of particular election. the presidential vote (Abramson, Aldrich, and Rohde 1994; Dalton and Wattenberg MeasuresofCivilRightsAttitudes 1993). The inclusion of these variables in the multivariate analysis also enables me to I analyze two NES civil rights items in the compare their effects on voting behavior first stage of the analysis. The item measur- with the corresponding effects of civil rights ing attitudes toward African Americans’ civil attitudes. rights asks respondents: “Do you think that As summarized in Tables 1a and 1b, these civil rights leaders are trying to push too variables measure voters’ memberships in fast, are going too slowly, or are they mov- four major social groups: region (three ing at about the right speed?” This item is dummy variables for Northeast, South, and analyzed as two dummy variables for the re- West, with Midwest serving as the refer- sponses “moving at about the right speed” ence), race (coded 1 for African Americans and “going too slowly,” with the response and 0 otherwise), gender (coded 1 for “too fast” serving as the reference category women, 0 for men), and class.13 The three in the regression models. remaining sociodemographic variables relate The item measuring attitudes toward to other politically relevant, demographic at- women’s rights asks: “Some people feel that tributes: age and years of education (both women should have an equal role with men measured as continuous variables), and in running business, industry, and govern- household income (measured in constant ment. Others feel that women’s place is in 1992 dollars). the home.” This item is a seven-point Likert Given the powerful effects of economic scale in which higher scores indicate greater calculations on vote choice and political support for women’s rights. change (Downs 1957; Fiorina 1981), I in- In the second stage of the analysis, I use clude in the analysis two economic vari- data from the 1992 and 1996 NES surveys ables. Retrospective evaluations ask respon- to analyze the political effects of attitudes dents if they are “better off or worse off fi- toward gay and lesbian civil rights. The nancially than a year ago.” Prospective women’s rights item is also available in evaluations ask respondents if they expect both 1992 and 1996, permitting me to esti- mate the political effects of attitudes toward 13 I analyze class as seven nominal categories gay/lesbian civil rights net of the corre- for professionals, managers, routine white-collar sponding effects of attitudes toward employees, self-employed nonprofessionals, skilled workers, and nonskilled workers (with non-full-time labor force participants serving as shows that the effects of changing attitudes to- the reference category in the regressions). See ward civil rights on are simi- Hout, Brooks, and Manza (1995) and Brooks and lar to their effects on vote choice. (Results are Manza (1997) for additional details on these class available on available on request). categories. CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 491

Table 1a. Sample Means for Variables in Stage Table 1b. Sample Means for Variables in Stage One of the Analysis Two of the Analysis

Mean Mean Variable 1972 1992 Variable 1992 1996

Sociodemographic Variables Civil Rights Variables Age (in years) 43.26 44.42 Women’s rights 4.84 4.81 Education (in years) 12.25 13.26 (7-point scale score) Household income $36,771 $37,261 Gay/lesbian civil rights (favor .63 .65 (in 1992 U.S. dollars) antidiscrimination laws) Region Dependent Variable Northeast .21 .19 Presidential vote choice: South .34 .33 Democratic candidate .38 .40 West .14 .20 Third-party candidate .15 .06 Race Nonvoter .19 .22 African American .10 .11 Source: National Election Studies. Gender Note: Number of cases equals 2,370. Woman .54 .52 Class to be “better off or worse off financially” a Professional .12 .16 year from now. These two items are mea- Manager .08 .08 sured using dummy variables for the Routine white collar .11 .12 “same” and “worse off” responses, with the reference categories being the response, Self-employed .05 .08 “better off.”14 Skilled worker .08 .08 The final control variable in the analysis Nonskilled worker .15 .13 measures attitudes toward an important Economic Variables public policy domain: the welfare state. Retrospective economic evaluations: This variable is continuous and respon- Same as a year ago .42 .33 dents’ scores are the result of their self- Worse than a year ago .20 .34 placement on a seven-category scale indi- Prospective economic evaluations: cating their degree of support for govern- Same as now .52 .55 ment provision of “a job and a good stan- dard of living.” Although past research has Worse than now .09 .10 found little relationship between such pub- Government Policy Variables lic policy attitudes and views of civil rights, Welfare state attitudes 2.77 2.67 the inclusion of the welfare state item en- (7-point scale score) ables an instructive comparison between the Civil Rights Variables Blacks’ civil rights: 14 While these variables measure economic Moving at about the .43 .56 calculations relating to egocentric concerns, past right speed research has found that sociotropic concerns (re- Going too slowly .09 .17 lating to perceptions of national economic pros- Women’s rights 3.41 4.84 perity) have an independent and larger effect on (7-point scale score) vote choice (Kiewiet 1981). Because the NES items measuring sociotropic calculations have Dependent Variables undergone changes in wording between 1972 Presidential vote choice: through 1992, I include only the egocentric items Democratic candidate .26 .38 in the current analysis. Given that economic Third-party candidate .00 .15 evaluations and attitudes toward noneconomic is- Nonvoter .23 .22 sues tend to be unrelated to one another, the ab- sence of sociotropic items has no impact on the Source: National Election Studies. current analyses (additional analysis of indi- Note: Number of cases equals 5,480. vidual elections corroborated this expectation). 492 AMERICANSOCIOLOGICALREVIEW magnitude of its effect on vote choice and DecompositionoftheEffectsof the corresponding effects of civil rights atti- TrendsinCivilRightsAttitudeson tudes. Tables 1a and 1b present sample ElectionOutcomes means for the variables in the analysis. For the analysis of organizational effects, I StatisticalModels use coefficients from the models of vote choice and the year-specific sample means I use multinomial logistic regression models presented in Tables 1a and 2b to make causal to analyze the four-category vote choice inferences about the contribution of each in- variable.15 Equation 1 summarizes the mul- dependent variable to explaining changes in tinomial logistic regression model, where is vote choice between elections. My regres- the logit transform of the expected probabil- sion decompositions yield estimates of the ity that person i will make vote choice j (1 = impact of changes in each independent vari- Republican; 2 = Democrat; 3 = third-party able (including civil rights attitudes) on the candidate; 4 = nonvoter): predicted change in election outcomes (for a discussion of various types of regression de- L M composition using repeated surveys, see à =+αβ + β yXXij j ∑∑lj il mj im Firebaugh 1997). l==11m N +β ∑njX in. (1) RESULTS n=1 Effectson1972-1992VoteChoice β In this model, the lj are the main effects of specific elections, and Xil are dummy vari- Table 2 examines whether attitudes toward ables for these elections (l = 1 for 1972, 2 civil rights had a significant impact on vot- for 1976,... , 7 for 1996); the 1972 election ing behavior in presidential elections from β serves as the reference category. The mj are 1972 through 1992. The table presents fit sta- coefficients for the effects of civil rights at- tistics—the –2 log-likelihood statistic (–2LL) titudes, and Xim (m = 1,... , M) are the m and Raftery’s (1995) Bayesian Information civil rights variables in the analyses. Finally, Criterion (BIC)—for evaluating competing β the nj are coefficients for the effects of multinomial logistic regression models of other measured sources of vote choice, and vote choice. Model 1’s coefficients measure Xin (n=1,... ,N) are the n control vari- only the main effects of election year, Model ables. 2 also includes the main effects of African Because j has four categories, it is only American’s civil rights and women’s rights. necessary to analyze j – 1 logits to estimate Model 2’s fit is superior to Model 1’s accord- the model. Republican vote choice (j = 1) is ing to both measures of fit, indicating that the reference category, yielding three sepa- civil rights attitudes affected vote choice dur- rate logits: Democrat versus Republican, ing the 1972 through 1992 elections. third-party candidate versus Republican, and Models 3 through 5 examine whether the nonvoter versus Republican. The model’s effects of civil rights attitudes changed since coefficients show the estimated effect of a 1972. Model 3 has an additional coefficient given independent variable on each of these for the interaction between women’s rights three logits. In addition to presenting the co- and election year. Both measures of fit show efficients and fit statistics for these models, that Model 3 is preferred over Model 2, pro- I use graphical displays that summarize the viding evidence that the political effects of effects of particular variables of interest on women’s rights attitudes changed over the vote choice. 1972–1992 period. The continuous measure- ment of the year covariate in Model 3 im- plies that the changing effects of women’s 15 I performed a check on the multinomial lo- gistic regression model’s requirement of inde- rights attitudes follow a linear pattern, but pendence from irrelevant alternatives (Hausman note that this interaction is constrained to and McFadden 1984) and found no violation of apply only to the choice of Democratic ver- this assumption. sus Republican candidates. CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 493

Table 2. Fit Statistics for Multinomial Logistic Regression Models of Vote Choice: Presidential Elections, 1972 to 1992

Fit Statistics –2 Log- Model/Description Likelihood (d.f.) BIC

1 Election years only 13,064.6 (5,462) –33,957 2 Model 1 + Civil rights attitudes 12,759.9 (5,453) –34,184 3 Model 2 + [Women’s rights attitudes × Year (continuous)]a 12,746.0 (5,452) –34,189 4 Model 3 + [Blacks’ civil rights attitudes × Year (continuous)]a 12,745.2 (5,450) –34,173 5 Model 3 + [Blacks’ civil rights attitudes × Year (dummy variables)] 12,743.7 (5,440) –34,089 6 Model 3 + [Sociodemographic variables + Welfare state attitudes + Economic variables + (Economic variables × Year1980)] 11,113.3 (5,389) –35,280 Note: The vote choice variable is coded 1 for the Republican candidate, 2 for the Democratic candidate, 3 for the third-party candidate, and 4 for nonvoters; N = 5,480. a Coefficients for attitudes × year interactions are constrained to 0 in the equations for third-party choice versus Republican candidate and for nonvoter versus Republican candidate.

Using the same type of constraints em- third-party candidate over a Republican ployed in Model 3, Model 4 has two addi- candidate (column 2); and the log-odds of tional coefficients for interactions between not voting versus choosing the Republican African American civil rights attitudes and candidate (column 3). A coefficient thus year. However, both measures of fit favor represents the estimated effect (in logits) of Model 3. Model 5 allows the political effects a variable relative to the constant for a col- of attitudes toward blacks’ civil rights to umn-specific logit. vary during each election, but both fit statis- The coefficients for the sociodemogra- tics again select Model 3, providing evi- phic, economic, and welfare state attitude dence that these effects were stable during variables represent factors whose impacts on the 1972 through 1992 period. vote choice are relatively well understood; In contrast to Model 3, Model 6 also has their estimated effects are consistent with coefficients for the additional items measur- past studies (Abramson, Aldrich, and Rohde ing better-known sources of presidential 1994; Alvarez and Nagler 1995; Brooks and vote choice: social group memberships, atti- Brady 1999; Kiewiet 1983; Manza and tudes toward government, and retrospective Brooks 1999). My primary interest is in the and prospective economic evaluations. coefficients representing the effects of civil Model 6 also allows economic evaluations rights attitudes. In column 1, the two coeffi- to have different effects during Republican cients for attitudes toward African Ameri- and Democratic administrations. This is cans’ civil rights represent sizable effects. measured by an interaction between the eco- For instance, viewing civil rights leaders as nomic items and a covariate for the 1980 moving “too slow” raises the log-odds of election (1980 is the only election between choosing the Democratic over the Republi- 1972 and 1992 in which there was a Demo- can candidate by a substantial 1.14 over the cratic incumbent). Model 6 easily improves reference category. All else being equal, sup- over Model 3’s fit, and thus Model 6 is my porters of African Americans’ civil rights are preferred model of vote choice for the 1972 substantially more likely than are opponents to 1992 period. to favor Democratic candidates. Table 3 presents the coefficients of Model The effects of these attitudes on the 6. Given the four-category dependent vari- choice of third-party candidates are similar: able, there are three separate sets of coeffi- Supporters of blacks’ civil rights are con- cients for the log-odds of choosing the siderably more likely to favor third-party Democratic over the Republican candidate candidates over Republicans. The coeffi- (column 1); the log-odds of choosing a cient for the highest level of civil rights 494 AMERICANSOCIOLOGICALREVIEW

Table 3. Multinomial Logistic Regression Coefficients from Preferred Model: Presidential Vote Choice, 1972 to 1992

(1) (2) (3) Democratic Third-Party Nonvoter vs. GOP Voter vs. GOP Voter vs. GOP Voter Independent Variable Coef. (S.E.) Coef. (S.E.) Coef. (S.E.)

Constant –1.43* (.32) –5.61* (1.14) 3.08* (.33) Election Years 1976 .68* (.14) 2.55* (1.03) .47* (.14) 1980 1.08* (.22 4.56* (1.03) 1.05* (.22) 1984 –.27 (.19) .98 (1.16) .24 (.16) 1988 –.01 (.21) 1.86 (1.17) .60* (.16) 1992 .18 (.25) 4.99* (1.06) .87* (.14) Sociodemographic Variables Age <.01 (<.01) –.02* (.01) –.04* (.01) Education –.07* (.02) –.04 (.03) –.23* (.02) Household income (× 1,000) –.01* (.01) –.01* (.01) –.02* (.01) Region: Northeast .19 (.10) <.01 (.18) .47* (.12) South –.05 (.09) –.56* (.18) .57* (.10) West .09 (.10) .08 (.18) .30* (.12) African American 1.80* (.20) –.43 (.50) 1.10* (.21) Women .07 (.08) –.30* (.14) .08 (.09) Class: Professionals .31* (.12) .02 (.22) –.66* (.16) Managers –.03 (.15) –.08 (.26) –.32 (.17) Routine white-collar .23 (.12) –.09 (.24) –.37* (.13) Self-employed .01 (.16) .47 (.25) –.27 (.18) Skilled workers .27 (.16) .13 (.28) –.04 (.17) Nonskilled workers .27* (.13) .07 (.24) .16 (.13) Economic Variables Retrospective economic evaluations: Same as year ago .27* (.09) .17 (.18) .11 (.10) Worse than year ago .76* (.10) .67* (.19) .31* (.12) Prospective economic evaluations: Same as year ago .28* (.09) .18 (.15) .14 (.09) Worse than year ago .61* (.14) .58* (.26) .38* (.16)

Welfare state attitudes .30* (.02) .13* (.04) .20* (.02) Blacks’ Civil Rights Civil rights leaders moving just about right .45* (.08) .49* (.16) .17 (.09) Civil rights leaders moving too slow 1.14* (.14) .95* (.23) .69* (.16)

Women’s rights attitudes –.02 (.04) .18* (.04) .05* (.02) Interaction Terms Women’s rights × Year .03* (.01) .—a .—a * * Retrospective econ. evaluations × Year1980 –.56 (.16) –.39 (.19) –.15 (.13) × * * Prospective econ. evaluations Year1980 –.55 (.16) –.28 (.24) –.53 (.17) Note: Numbers in parentheses are standard errors. Coefficients are from Model 6 in Table 2; N = 5,480. a Coefficients are constrained to 0 in the model. *p < .05 (two-tailed tests) CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 495 support on the log-odds of not voting versus of each covariate in both 1972 (dashed lines) choosing the Republican candidate is also and 1992 (solid lines). statistically significant. Taken as a whole, The effects on vote choice of attitudes to- these results demonstrate that support for ward African Americans’ civil rights are African Americans’ civil rights makes virtu- generally larger than the effects of prospec- ally any political option attractive relative tive and retrospective economic evaluations to supporting a Republican presidential can- and slightly less than the effect of welfare didate. state attitudes (the effects of race are the Table 2’s results show that the effects of largest of any of the variables). Whereas the attitudes toward African Americans’ civil effects of African American civil rights atti- rights on vote choice were stable between tudes are comparable in 1972 and 1992, the 1972 and 1992 (given the absence of any corresponding effects of women’s rights at- significance interactions with time). How- titudes were negligible in 1972 but substan- ever, the results also show that attitudes to- tial in 1992. Taken as a whole, these results ward women’s rights have followed a dif- establish that civil rights attitudes have sub- ferent pattern. As measured by the women’s stantial political-behavioral effects whose rights × year coefficient (and the main ef- magnitude compares favorably with other, fect of women’s rights attitudes), the politi- better-known sources of voter alignments. cal effects of these attitudes have steadily increased over time, with the estimated ef- Effectson1972–1992Election fect on Democratic versus Republican vote Outcomes choice growing from .01 (in 1972) to. 04 (in 1976), to .16 (in 1992). The effects of Using the same data, I now consider the ef- women’s rights attitudes on third-party vote fects of trends in civil rights attitudes on ex- choice and voter turnout are stable over plaining changes in the outcome of elections time, and reveal the same pattern found ear- between 1972 and 1992. The analyses use lier: Support for women’s rights makes the the coefficients of the vote choice model options of supporting third-party candidates (Table 3) and the sample means (Table 1a) or even not voting attractive in comparison to estimate the impact of each causal factor to choosing the Republican candidate. measured in the model. Taking blacks’ civil How large are the effects of civil rights at- rights as an example, the estimate in the first titudes compared with other, better-known column shows that trends in these attitudes sources of voting behavior? I answer this by themselves raised the log-odds of favor- question using the graphical displays in Fig- ing the Democratic over the Republican can- ure 2. Each panel in the figure summarizes didate by .15; this represents 14 percent of the effects of a key coefficient in the model. the total predicted change in major party Because the predicted probabilities are cal- vote choice between 1972 and 1992. culated across the full range of each Starting with column 1’s results, changes covariate (and heights and lengths are iden- in sociodemographic factors alone would tical across charts), their effects can be com- have slightly increased Republican Party ad- pared.16 Given that the effects of attitudes vantage, lowering the log-odds of favoring toward women’s rights increased between the Democrats by .04. Similarly, a small de- 1972 and 1992, Figure 2 presents the effects cline in support for the welfare state is pre- dicted as leading to a –.03 logit change, 16 For these calculations, the predicted prob- again enhancing Republican Party advan- ability of choosing the Democratic candidate tage. Given that the log-odds of favoring the (among major party voters) is calculated across Democratic over the Republican candidate is the full range of each independent variable, hold- predicted to have increased by 1.10 between ing all other continuous covariates constant at 1972 and 1992, the latter factors cannot ex- their means and all categorical covariates at their reference categories. See Long (1997) or plain these political changes. The estimated Kaufman (1996) for additional discussion of impact of changes relating to economic fac- various types of calculations for comparing and tors has the correct (positive) sign, but ex- presenting the effects of covariates in categorical plains only 9 percent of the net change in the data models. outcome of elections. 496 AMERICANSOCIOLOGICALREVIEW

1.00 1.00 1.00 Blacks’ Civil Rights Women’s Rights Welfare State

.75 .75 .75

1992 .50 .50 .50 1992 1992

1972 .25 .25 .25 1972 1972 Probability of Democratic Vote Choice Probability of Democratic Vote

.00 .00 .00 012 0123456 0123456

1.00 1.00 1.00 Race Retrospective Prospective Economic Evaluations Economic Evaluations

.75 .75 .75

1992

.50 .50 1992 .50 1992

1972

.25 .25 1972 .25 1972 Probability of Democratic Vote Choice Probability of Democratic Vote

.00 .00 .00 01012 012

Item Categories Item Categories Item Categories

Figure 2. Predicted Effects of Civil Rights Attitudes, Welfare State Attitudes versus Other Sources of Major Party Presidential Vote Choice in 1972 and 1992

The key lies with changing civil rights at- elections. Taken together, changing attitudes titudes, with changes relating to women’s toward women’s rights and African Ameri- rights accounting for the lion’s share of this can civil rights17 account for just over 80 transition. Changing women’s rights atti- tudes are predicted as raising the log-odds 17 My separate analysis of two GSS items of choosing the Democratic over the Repub- (Davis and Smith 1997) measuring attitudes to- lican candidate by .74 logits, thereby ex- ward African Americans’ civil rights yield simi- plaining 67 percent of the net increase in lar results. The .11 logit estimate measuring the Democratic Party support between the two combined effect of liberal changes in attitudes to- CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 497

Table 4. Decomposition for Explaining Change in Presidential Vote Choice, 1972 to 1992

(1) (2) (3) Democratic Third-Party Nonvoter vs. GOP Voter vs. GOP Voter vs. GOP Voter Predicted Predicted Predicted Independent Variable Change Prop. Change Prop. Change Prop.

Change in all sociodemographic factors: –.04 (–.04) –.03 (–.01) –.32 (–.44) Age .00 (.00) –.02 (.00) –.05 (–.07) Education –.07 (–.06) –.04 (–.01) –.23 (–.32) Household income .00 (.00) .00 (.00) –.01 (–.01) Region .00 (.00) .01 (.00) .00 (.00) Race .02 (.02) .00 (.00) .01 (.01) Class .01 (.01) .01 (.00) –.04 (–.06) Gender .00 (.00) .01 (.00) .00 (.00)

Change in economic evaluations .10 (.09) .09 (.02) .04 (.06)

Change in welfare state attitudes –.03 (–.03) –.01 (.00) –.02 (–.03)

Change in civil rights attitudes: .89 (.81) .40 (.07) .15 (.21) Blacks’ civil rights .15 (.14) .14 (.03) .08 (.11) Women’s rights .74 (.67) .26 (.05) .07 (.10)

Residual change (not explained by model) .18 (.16) 4.99 (.92) .87 (1.21)

Total change in logit, 1972–1992 1.10 (.99) 5.44 (1.00) .72 (1.00) Note: Entries are the predicted change in a given column’s logit attributable to changes in a row-specific factor between 1972 and 1992. Numbers in parentheses are the proportion of the total predicted change explained by a row-specific factor. percent of the change in election outcomes, The decomposition in Table 4 also sheds leaving only a small portion (16 percent) of light on the sources of voting trends affect- the total unexplained. These results provide ing third-party candidates (column 2). Com- strong evidence that changing attitudes to- pared with 1972, support for third-party can- ward civil rights had substantial organiza- didates is predicted as increasing by 5.44 tional effects on the fortunes of the two ma- logits, reflecting the unusual level of support jor political parties.18 given to Ross Perot in the 1992 election. Predictably, most of this change (92 percent) is not explained by the causal factors in- ward blacks’ civil rights compares favorably with the (.15) estimate from the NES data, providing cluded in the model, attesting to candidate- some corroboration of the results. centered characteristics of the billionaire 18 These results are similar to those found by businessman from Texas. While changes in analyzing major social groups separately. For in- sociodemographic factors, economic evalu- stance, the predicted increase in the log-odds of ations, and welfare state policy attitudes Democratic vote choice due to changing attitudes have virtually no bearing on these voting toward women’s rights is .75 for women, .72 for trends, changes in civil rights attitudes are men, .74 for non–African Americans, and .75 for estimated as raising the log-odds of third- African Americans. The corresponding results party versus Republican vote choice by a for changing attitudes toward blacks’ civil rights is .16 for women, .12 for men, .16 for non–Afri- can Americans, and .08 for African Americans. tudes represent a separate phenomenon from— Given the similarity of political effects across and cannot be explained by—the group-based these social groups, changing civil rights atti- race and gender gaps in U.S. politics. 498 AMERICANSOCIOLOGICALREVIEW

Table 5. Multinomial Logistic Regression Coefficients Predicting Presidential Vote Choice, 1992 to 1996

(1) (2) (3) Democratic Third-Party Nonvoter vs. GOP Voter vs. GOP Voter vs. GOP Voter Independent Variable Coef. (S.E.) Coef. (S.E.) Coef. (S.E.)

Constant –1.41* (.18) –1.64* (.24) 1.15* (.19)

Election year –.12 (.11) –1.01* (.16) –.03 (.12) Women’s rights attitudes .17* (.03) .17* (.05) .07 (.04) Gay/lesbian attitudes 1.50* (.11) .53* (.15) .93* (.13) Note: Numbers in parentheses are standard errors. The reference category for election year is 1992. Add- ing interaction terms for Women’s rights × Year and Gay/lesbian civil rights × Year to the models yields a nonsignificant reduction in –2 log-likelihood (.10; d.f. = 2). N = 2,370. *p < .05 (two-tailed tests) nontrivial .40 logits. Although changing Attitudes toward gay and lesbian civil civil rights attitudes have the greatest posi- rights have had substantial effects on presi- tive effect on Democrats, these results sug- dential vote choice in recent elections. Sup- gest that third-party candidates can also ben- port for the civil rights of gays and lesbians efit by such trends. increased the log-odds of favoring the Finally, the decomposition of voter turn- Democratic over the Republican candidate out shows that increasingly liberal attitudes by a substantial 1.50. As was found for toward civil rights resulted in a slight in- women’s and African Americans’ civil crease in nonvoting versus Republican vote rights, support for gays’ and lesbians’ civil choice. This underscores the earlier findings rights also reduces the attractiveness of Re- that relative to Republican vote choice, sup- publican candidates relative to other politi- porters of African Americans’ and women’s cal options (third-party support or nonvot- civil rights are more likely not to vote. Re- ing). The results provide no evidence that garding other causal factors, the –.32 logit the effects of attitudes toward gay/lesbian estimate for the combined effects of civil rights on vote choice changed between sociodemographic factors illustrates the 1992 and 1996.19 Attitudes toward women’s well-known paradox of over-time decline in rights and also gay and lesbian civil rights U.S. voter turnout: Such changes, especially changed little in the 1990s and are thus pre- the spread of education, should have in- dicted as having a minor impact on Demo- creased voter turnout, even though the ac- cratic versus Republican electoral chances tual trend during this time has been toward from 1992 to 1996. More specifically, declining turnout (Teixeira 1987). changing attitudes toward women’s rights lower by .01 the log-odds of Democratic Effectson1992–1996VoteChoice victory between 1992 and 1996; changing andElectionOutcomes attitudes toward gays and lesbian civil rights raise by .03 the log-odds of Demo- Using data from the 1992 and 1996 NES, I cratic victory. now extend the analysis to the most recent presidential elections. Table 5 presents the 19 Despite limitations in the availability of the same type of estimates as in the preceding NES item measuring gay and lesbian civil rights, analyses, but I now examine the effects of the effects of such attitudes on major party vote attitudes toward gay and lesbian civil rights. choice undoubtedly increased in size over the 1972–1996 period. Clear-cut differences in the The inclusion of the NES women’s rights parties’ policy positions relating to gay/lesbian item ensures that the effects of gay/lesbian civil rights are a relatively new development— civil rights attitudes are estimated net of the for instance, as recently as 1988, the Republican corresponding effects of attitudes toward Party’s platform made no reference to homosexu- women’s rights. ality (Freeman 1996). CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 499

1.0 1.0

.9

.8 Democratic .9 Victory .7

.6

.5 .8 .4 .3 Republican .2 Victory .7 .1 Probability of a Democratic Victory 0 Democratic 1972 1976 1980 1984 1988 1992 1996 Victory .6

.5

.4 Probability of a Democratic Victory .3 Republican Victory

.2 Actual prediction from the model Using 1972 means for civil rights items .1 Using 1992 means for civil rights items

0 1972 1976 1980 1984 1988 1992 1996 Presidential Election Year Figure 3. Predicted Outcome of Presidential Elections from Models Using Actual, 1972, and 1992 Aggregate Levels of Civil Rights Attitudes

SummarizingtheAggregate-Level (dotted line) allow us to compare the pre- Effectson1972-1996Election dicted outcome of each election with the hy- Outcomes pothetical scenario in which civil rights atti- tudes experienced no trend (e.g., remaining Figure 3 summarizes the effects of changing at their 1972 levels throughout the entire attitudes toward civil rights on each of the 1972 through 1996 period). By comparing seven elections from 1972 through 20 1996. The 1972 estimates of the figure models for the 1972–1992 NES data (Table 3) and the 1992–1996 NES data (Table 5). The “ac- 20 The predicted probabilities in Figure 3 are tual” estimates are the probability of a Demo- derived using the multinomial logistic regression cratic versus Republican victory (among major 500 AMERICANSOCIOLOGICALREVIEW the predicted probabilities of a Democratic The estimates using the 1992 means for versus Republican victory according to the the civil rights items (see the inset graph) “actual” versus the 1972 estimates, we can represent a second hypothetical scenario in observe whether Republicans would have which the 1992 levels of civil rights attitudes dominated elections in the absence of liberal are substituted into the calculations for each trends in civil rights attitudes. election. These estimates enable us to an- These estimates have far-reaching impli- swer a different counterfactual question: cations: Without (liberal) changes in civil Namely, would higher levels of support for rights attitudes, Democratic candidates civil rights in earlier elections have im- would have lost every presidential election proved Democratic Party chances? The con- between 1972 and 1996 (as summarized by siderable boost given to Democratic candi- the dotted line). These trends in attitudes dates in this scenario (in elections prior to suppressed what would have been consider- 1992) again attests to the powerful political ably larger Republican victories in the 1980s effects of civil rights attitudes on vote while also enabling Democrat to choice. These estimates also show, however, win the White House in the 1990s. The in- that despite changing the margins of victory, creasing divergence between the actual esti- higher levels of support for civil rights mates and the estimates using 1972 means would nevertheless have been insufficient to for the civil rights items reflects the grow- convert Republican wins into Democratic ing impact of civil rights attitudes on elec- victories (in 1972, 1980, 1984, and 1988). tions. Note, however, that the small changes This implies that although changing views of in civil rights attitudes between 1972 and civil rights suppressed a Republican realign- 1976 appear to have been critical to Demo- ment, such changes—even with the most crat ’s narrow margin of vic- generous of counterfactual assumptions— tory over Republican Gerald Ford.21 could not by themselves have led to a Demo- cratic realignment. In other words, even if American voters had been as supportive of party voters) predicted by the models. The two civil rights in the 1970s and 1980s as they sets of counterfactual estimates are derived by were in the 1990s, the Republican Party substituting the 1972 or 1992 means for the civil would have still controlled the Presidency in rights items into the models and calculating the the 1980s (and also in 1972). subsequent predicted probabilities of a Demo- cratic versus Republican victory (again assuming a two-party contest). Note that for the 1972 esti- DISCUSSION mates, I calculate the predicted probability for the 1996 election by substituting the 1992 mean The results of this study advance our under- for the gay/lesbian civil rights item for the (non- standing of some long-standing puzzles in existent) 1972 mean for this item. This conserva- the study of political behavior and in re- tive calculation probably underestimates the search on opinion trends; it also suggests magnitude of a hypothetical Republican victory some fruitful directions for research in po- in that election. However, this makes the evi- litical sociology. With regard to political be- dence that liberal trends in civil rights attitudes havior, I find evidence that attitudes toward suppressed a Republican realignment all the more convincing. the civil rights of African Americans, 21 Although counterfactual inferences of this women, and more recently gays and lesbi- sort involve inherent uncertainties, the differ- ans provide voters with a clear basis for ences between the vote choice model’s predic- evaluating and choosing among candidates tions and the actual outcome of presidential elec- in presidential elections. The effects of these tions are small enough (less than or equal to 2 attitudes on vote choice are comparable to percentage points for five of the seven elections) other, better-known causal sources of the to lend some support to this inference. In the vote. These effects also reveal highly con- election with the greatest difference (1992), the sistent partisan preferences, with support for statistical model overestimates the Democratic margin of victory, suggesting that the effect of civil rights disposing voters to favor Demo- aggregate-level opinion trends (if estimated ac- cratic over Republican candidates, and also curately) probably resulted in a greater contribu- third-party candidates (and even the option tion to Bill Clinton’s electoral victory. of not voting), over Republican candidates. CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 501

These results provide firm evidence for the sive attention they receive in the research lit- institutionalization of civil rights conflicts erature, the findings of the current study im- within the realm of electoral politics. ply that they are nevertheless compatible Moreover, because attitudes toward the with the phenomenon of civil rights liberal- civil rights of African Americans, women, ism. In other words, the persistence of rac- and gays and lesbians also experienced large ist, sexist, or homophobic attitudes does not changes over time, these attitudes have had mean that Americans have not become, on important organizational effects, shaping the average, more supportive of civil rights. outcome of presidential elections and the These considerations suggest an important relative standing of the two major parties. In lesson: While growing support for civil particular, liberal trends in civil rights atti- rights principles has not transformed all rel- tudes substantially reduced the margin of evant public attitudes, it would be a mistake Republican victories in the 1980s, while en- for scholars to dismiss such aggregate abling Democratic candidates to win the changes in opinion as historically or politi- presidency in 1976, 1992, and 1996. All else cally unimportant. Indeed, this study has being equal, had civil rights attitudes re- demonstrated that scholarly neglect of liber- mained unchanged from their 1972 levels, alizing attitudes toward civil rights has de- Republican candidates would have won ev- prived researchers of an understanding of ery subsequent presidential election through causal factors that have kept Democratic 1996, thereby ushering in a 36-year period candidates competitive in presidential elec- of GOP control of the presidency (a consid- tions since the 1960s. Ironically, it appears erably longer time span than the 20 years of in recent years that it has been a handful of Democratic control following the New Republican Party strategists who best under- Deal). Given that this remarkable scenario stand the partisan importance of civil rights would have resulted from holding constant principles, calling for greater moderation in only civil rights attitudes, I interpret these the party’s platform and focus on issues re- results as providing evidence that changing lating to women’s rights, homosexuality, and attitudes toward civil rights suppressed what also abortion (see Berke 1994) in an attempt otherwise would have been a Republican po- to neutralize cleavages that favor Demo- litical realignment.22 cratic candidates.23 Past research has not discovered the large, electorate-wide effects of changing attitudes CONCLUSION toward civil rights. Complementing this problem, previous theorizing has discounted By advancing our understanding of the po- such factors as significant sources of politi- litical effects of liberalizing trends in civil cal change in the United States. This neglect rights attitudes, the results of the current is reinforced by a tendency among scholars to dismiss liberal trends in attitudes toward 23 The results of the current study also demon- civil rights principles in light of generally strate the importance of conceptualizing and ana- low levels of support for civil rights policy lyzing liberal changes in civil rights attitudes pertaining to women and to gays and lesbians. implementation or negative affect toward Some caution should be exercised in interpret- African Americans and especially gays and ing the comparatively larger political effects of lesbians. While the sociological importance attitudes toward women’s rights given the lim- of the latter phenomena justifies the exten- ited number of civil rights items available over the relatively long period of time examined in 22 Given these findings, a fruitful question for this study. Future research may be able to shed further research is whether liberal trends in civil additional light on the comparative political ef- rights attitudes have had similar effects in con- fects of these three types of attitudes by focus- gressional elections, suppressing Republican ing on recent elections for which additional gains in the House and Senate since the 1960s. If items are available. Analyses of such data may such processes have also occurred in the congres- also yield a better understanding of the political sional arena, it would provide evidence that the consequences of (changing) attitudes toward scope of the suppressed Republican realignment civil rights principles versus the corresponding discovered in this study extends to national elec- effects of (comparatively stable) attitudes toward tions as a whole. civil rights policies. 502 AMERICANSOCIOLOGICALREVIEW study provide a means of evaluating the “su- cratic and Republican parties. Given the perficial” and “insignificant liberalism” the- links between party control over government ses discussed in the introduction. If people’s and civil rights–related legislation, these po- willingness to support women’s and minor- litical-organizational consequences may in ity civil rights conformed to the superficial turn have affected the direction of public liberalism thesis, these attitudes would pro- policy, constraining what might have been vide little basis for value-laden judgments greater retreats from the civil rights agenda and decision-making—precisely the kind of of the mid-1960s and early 1970s. capacities that appear to be central to the Complementing the thrust of recent work “not so simple” act of voting (Carmines and by Burstein (1998), the current study pro- Huckfeldt 1996; Dalton and Wattenberg vides a clear rationale for political sociolo- 1993). Indeed, according to the thesis, the gists to take public opinion seriously as an political preferences of individuals with lib- important causal factor in their investiga- eral civil rights attitudes should differ little tions. The phenomenon of civil rights liber- from their more explicitly illiberal counter- alism, in particular, merits more systematic parts, given that these attitudes are assumed investigation. Electoral politics represents a to be superficial and thus an unreliable guide particularly informative context in which to to actual behavior. study the effects of changing attitudes to- I find, however, precisely the opposite pat- ward civil rights. My results also suggest the tern: In comparison to explicitly illiberal utility of analyzing the effects of these voters, people who indicate support for the trends in other institutional arenas, possibly civil rights of African Americans, women, forging connections with contemporary and gays and lesbians differ substantially in work on the interrelationship of movement their political preferences, being consider- frames, public opinion, and the media ably more likely to favor Democratic over (Gamson 1992; McAdam 1996). A second Republican candidates (and also third-party line of inquiry would be to examine the con- candidates and not voting to voting for Re- ditions under which civil rights principles publican candidates). This strong relation- affect (or do not affect) public policy, build- ship between civil rights attitudes and vote ing from an emerging line of research on po- choice implies that voters are both sincere litical elites’ conflicts over the legitimation in, and willing to act upon, their attitudes of policy alternatives (Kinder and Sanders when evaluating and selecting political can- 1996; Skrentny 1996). Developing such in- didates. In turn, this implies that increas- vestigations would help to advance scholarly ingly liberal responses to questions about knowledge about an important, yet under- civil rights represent genuine changes in at- appreciated, transformation of public opin- titudes over time. ion in the United States. The results also suggest limitations to the insignificant liberalism thesis. If this thesis Clem Brooks is Associate Professor of Sociology were true, liberal trends in civil rights atti- at Indiana University, Bloomington. His research tudes would not by themselves lead to any interests focus on political behavior, values and significant social or political changes, given public opinion, social change, and quantitative the (hypothesized) nature of dominant methods. Current projects include a study of pub- groups’ interests in adopting liberal ideolo- lic perceptions of family decline, a study of the gies primarily as a means of legitimating effects of declining confidence in government on their advantageous positions. I find, how- policy preferences, and a longer manuscript ex- ever, that changing levels of public support panding upon the political processes analyzed in the current study. He is also working with Jeff for the civil rights of African Americans, Manza on studies of gender, religious processes, women, and gays and lesbians have contrib- and electoral conflict since the 1960s. uted to an important type of political change. Far from being inconsequential, these aggre- REFERENCES gate-level shifts in opinion have had a pro- found impact within the institutional arena Abramson, Paul, John Aldrich, and David Rohde. of electoral politics, significantly affecting 1994. Change and Continuity in the 1992 Elec- the relative advantage enjoyed by the Demo- tions. Washington, DC: Congressional Quar- CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 503

terly. and codebooks]. Ann Arbor, MI: National Adams, Craig. 1997. “Abortion: Evidence of an Election Studies and Inter-university Consor- Issue Evolution.” American Journal of Politi- tium for Political and Social Research [pro- cal Science 41:718–37. ducer/distributor] (http://www.umich.edu/ Alvarez, R. Michael and Jonathan Nagler 1995. ~nes). “Economics, Issues and the Perot Candidacy: Clubb, Jerome, William Flanagan, and Nancy Voter Choice in the 1992 Presidential Elec- Zingale. 1990. Partisan Realignment: Voters, tion.” American Journal of Parties, and Government in American History. 39:714–44. Boulder, CO: Westview. Bass, Harold. 1991. “Background to Debate: A Dalton, Russell and Martin Wattenberg. 1993. Reader’s Guide and Bibliography.” Pp. 141– “The Not So Simple Act of Voting.” Pp. 193– 78 in The End of Realignment? Interpreting 218 in Political Science: The State of the Dis- American Electoral Eras, edited by B. Shafer. cipline, edited by A. Finifter. Washington, DC: Madison, WI: University of Press. American Political Science Association. Berke, Richard. 1994. “From the Right, Some Davis, James. 1992. “Changeable Weather in a Words of Restraint.” New York Times (Na- Cooling Climate Atop the Liberal Plateau: tional ed.), September 17, p. 9A. Conversion and Replacement in Forty-Two Bobo, Lawrence. 1997. “Race, Public Opinion, General Social Survey Items, 1972–1989.” and the Social Sphere.” Public Opinion Quar- Public Opinion Quarterly 56:261–306. terly 61:1–15. Davis, James and Tom W. Smith. 1997. General Bobo, Lawrence and Vincent Hutchings. 1996. Social Surveys, 1972–1996. [MRDF and “Perceptions of Racial Group Competition: codebook]. Storrs, CT: Roper Center for Pub- Extending Blumer’s Theory of Group Position lic Opinion Research [producer/distributor]. to a Multiracial Social Context.” American So- Downs, Anthony. 1957. An Economic Theory of ciological Review 61:951–72. . New York: Harper and Row. Brennan, Mary. 1995. Turning Right in the Six- Farley, Reynolds. 1996. The New American Re- ties: the Conservative Capture of the GOP. ality: Who We Are, How We Got Here, Where Chapel Hill, NC: University of North Carolina We Are Going. New York: Russell Sage Foun- Press. dation. Brooks, Clem and David Brady. 1999. “Income, ———. 1997. “Racial Trends and Differences in Economic Voting, and Long-Term Political the United States 30 Years after the Civil Change in the U.S., 1952–1996.” Social Rights Decade.” Social Science Research Forces: 77:1339–74. 26:235–62. Brooks, Clem and Jeff Manza. 1997. “Class Poli- Feldman, Stanley and John Zaller. 1992. “The tics and Political Change in the U.S., 1952– Political Culture of Ambivalence: Ideological 1992.” Social Forces 76:379–408. Responses to the Welfare State.” American Burnham, Walter Dean. 1970. Critical Elections Journal of Political Science 36:268–307. and the Mainsprings of American Politics. Ferree, Myra Marx. 1974. “A Woman for Presi- New York: W. W. Norton. dent? Changing Responses: 1958–1972.” Pub- Burstein, Paul. 1998. “Bringing the Public Back lic Opinion Quarterly 38:390–99. In: Should Sociologists Consider the Impact of Fiorina, Morris. 1981. Retrospective Voting in Public Opinion on Public Policy?” Social American National Elections. New Haven, CT: Forces 77:27–62. Yale University Press. Carmines, Edward and Robert Huckfeldt. 1996. Firebaugh, Glenn. 1997. Analyzing Repeated “Political Behavior: An Overview.” Pp. 223– Surveys. Thousand Oaks, CA: Sage. 54 in A New Handbook of Political Science, Firebaugh, Glenn and Kenneth Davis. 1988. edited by R. Goodin and H.-D. Klingemann. “Trends in Antiblack Prejudice, 1972–1984: New York: Oxford University Press. Region and Cohort Effects.” American Jour- Carmines, Edward and James Stimson. 1989. Is- nal of Sociology 94:251–72. sue Evolution: Race and the Transformation of Freeman, Jo. 1993. “Feminism and Family Val- American Politics. Princeton, NJ: Princeton ues: Women at the 1992 Democratic and Re- University Press. publican Conventions.” PS: Political Science Center for Political Studies. 1995. American Na- and Politics 26:21–28. tional Election Studies 1948–1994. [MRDF Gamson, William. 1992. Talking Politics. New and codebooks]. Ann Arbor, MI: National York: Cambridge University Press. Election Studies, Institute for Social Research, Hausman, Jerry and Daniel McFadden. 1984. and Inter-university Consortium for Political “Specification Tests for the Multinomial Logit and Social Research [producer/distributor]. Model.” Econometrica 52:1219–40. ———. 1997. American National Election Study Hout, Michael, Clem Brooks, and Jeff Manza. 1996: Pre- and Post-Election Surveys. [MRDF 1995. “The Democratic Class Struggle in the 504 AMERICANSOCIOLOGICALREVIEW

United States, 1948–1992.” American Socio- the American Civil Rights Movement.” Pp. logical Review 60:805–28. 338–55 in Comparative Perspectives on Social Jackman, Mary. 1978. “General and Applied Tol- Movements: Political Opportunities, Mobiliz- erance: Does Education Increase Commitment ing Structures, and Cultural Framings, edited to Racial Integration?” American Journal of by D. McAdam, J. McCarthy, and M. Zald. Political Science 22:302–24. New York: Cambridge University Press. ———. 1994. The Velvet Glove: Paternalism and McClosky, Herbert and Alida Brill. 1983. Di- Conflict in Gender, Class, and Race. Berkeley, mensions of Tolerance: What Americans Be- CA: University of Press. lieve About Civil Liberties. New York: Russell Jackman, Mary and Michael Muha. 1984. “Edu- Sage Foundation. cation and Intergroup Attitudes: Moral En- Page, Benjamin and Robert Shapiro. 1992. The lightenment, Superficial Democratic Commit- Rational Public: Fifty Years of Trends in ment, or Ideological Refinement?” American Americans’ Policy Preference. Chicago, IL: Sociological Review 49:751–69. University of Chicago Press. Kaufman, Robert. 1996. “Comparing Effects in Petrocik, John 1981. Party Coalitions. Chicago, Dichotomous Logistic Regression: A Variety IL: University of Chicago Press. of Standardized Coefficients.” Social Science Raftery, Adrian. 1995. “Bayesian Model Selec- Quarterly 77:90–109. tion in Sociology.” Sociological Methodology Key, V. O. 1955. “A Theory of Critical Elec- 25:111–63. tions.” Journal of Politics 17:3–18. Schuman, Howard, Charlotte Steeh, Lawrence ———. 1959. “Secular Realignment and the Bobo, and Maria Krysan. 1997. Racial Atti- .” Journal of Politics 21:198– tudes in America: Trends and Interpretations. 210. Rev. ed. Cambridge, MA: Harvard University Kiewiet, D. Roderick. 1983. Macroeconomics Press. and Micropolitics: The Electoral Effects of Seagull, Louis. 1980. “Secular Realignment: The Economic Issues. Chicago, IL: University of Concept and its Utility.” Pp. 69–81 in Realign- Chicago Press. ment in American Politics: Toward a Theory Kinder, Donald and Lynn Sanders. 1996. Divided edited by Bruce Campbell and Richard J. Trill- By Color. Chicago, IL: University of Chicago ing. Austin, TX: University of Texas Press. Press. Sears, David, Collete van Laar, Mary Carrillo, Ladd, Everett. 1991. “Like Waiting for Godot: and Rick Kosterman. 1997. “Is It Really Rac- The Uselessness of ‘Realignment’ for Under- ism? The Origins of White Americans’ Oppo- standing Change in Contemporary American sition to Race-Targeted Policies.” Public Politics.” Pp. 24–36 in The End of Realign- Opinion Quarterly 61:16–53. ment? Interpreting American Electoral Eras, Shafer, Byron, ed. 1991. The End of Realign- edited by B. Shafer. Madison, WI: University ment? Interpreting American Electoral Eras. of Wisconsin Press. Madison, WI: University of Wisconsin Press. Long, J. Scott. 1997. The Analysis of Categori- Sherrill, Kenneth. 1996. “The Political Power of cal and Limited Dependent Variables. Thou- Lesbians, Gays and Bisexuals.” PS: Political sand Oaks, CA: Sage. Science and Politics 29:469–73. Mansbridge, Jane. 1986. Why We Lost the ERA. Skrentny, John. 1996. The Ironies of Affirmative Chicago, IL: University of Chicago Press. Action: Politics, Culture, and Justice in Manza, Jeff and Clem Brooks. 1997. “The Reli- America. Chicago, IL: University of Chicago gious Factor in U.S. Presidential Elections, Press. 1960–1992.” American Journal of Sociology Smith, Tom W. 1990. “Liberal and Conservative 103:38–81. Trends in the United States since World War ———. 1999. Social Cleavages and Political II.” Public Opinion Quarterly 54:479–507. Change: Voter Alignments and U.S. Party Smith, Tom W. and Paul Sheatsley. 1984. Coalitions. New York: Oxford University “American Attitudes toward Race Relations.” Press. Public Opinion 7(October/November):14–15, Mason, Karen, John Czajka, and Sara Arber. 50–53. 1976. “Change in U.S. Women’s Sex-Role At- Sniderman, Paul, Richard Brody, and Philip titudes, 1964–1974.” American Sociological Tetlock. 1991. Reasoning and Choice: Explo- Review 41:573–96. rations in Political Psychology. New York: Mason, Karen and Yu-Hsia Lu. 1988. “Attitudes Cambridge University Press. toward Women’s Familial Roles: Changes in Steeh, Charlotte and Maria Krysan. 1996. “Affir- the United States, 1977–1985.” Gender and mative Action and the Public, 1970–1995.” Society 2:39–57. Public Opinion Quarterly 60:128–58. McAdam, Doug. 1996. “The Framing Function Stouffer, Samuel. [1955] 1963. Communism, of Movement Tactics: Strategic Dramaturgy in Conformity, and Civil Liberties: A Cross-Sec- CIVILRIGHTSLIBERALISMANDPOLITICALREALIGNMENT 505

tion of the Nation Speaks Its Mind. Reprint, Study.” American Sociological Review 44: Gloucester, England: Peter Smith. 831–42. Sundquist, James 1983. The Dynamics of the Wlezien, Christopher, Mark Franklin, and Daniel Party System: Alignment and Realignment of Twiggs. 1997. “Economic Perceptions and Political Parties in the United States. Rev. ed. Vote Choice: Disentangling the Endogeneity.” Washington, DC: Brookings Institution. Political Behavior 19:7–17. Teixeira, Ruy. 1987. Why Americans Don’t Vote: Yang, Alan. 1997. “The Polls-Trends: Attitudes Turnout Decline in the United States 1960– toward Homosexuality.” Public Opinion Quar- 1984. New York: Greenwood. terly 61:477–507. Thornton, Arland and Deborah Freedman. 1979. Zaller, John. 1992. The Nature and Origins of “Changes in the Sex Role Attitudes of Mass Opinion. New York: Cambridge Univer- Women, 1962–1977: Evidence from a Panel sity Press.