Hello, I have a paper that is written, and I need you to rewrite my paper so that it is a logical analysis; it needs to be 5 pages, and I will provide all the articles for you to use, and I will provi

786-802 Published by: Midwest Political Science Association Stable URL: https://www.jstor.org/stable/3088434 JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms Midwest Political Science Association is collaborating with JSTOR to digitize, preserve and extend access to American Journal of Political Science This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms Party Polarization and "Conflict Extension" in the American Electorate Geoffrey C. Layifian Vanderbilt University Thomas M. Carsey Florida State University In recent decades, Democratic and Republican party elites have grown increasingly polarized on all three of the major domestic policy agendas: social welfare, racial, and cultural issues. We contend that the mass response has been characterized not by the traditional expectation of "con? flict displacement" or the more recent account of "ideological realignment," but by what we term "conflict exten? sion." Mass attitudes toward the three agendas have remained distinct, but the parties in the electorate have grown more polarized on all three. Conflict extension. rather than conflict displacement or ideological realign? ment, has occurred because there has been a limited mass response to the growth of elite-level party polar? ization. Only party identifiers who are aware of party elite polarization on each of the issue dimensions have brought their social welfare, racial, and cultural issue attitudes toward the consistently liberal or consistently conservative stands of Democratic and Republican elites. Analyses using data from the 1972 through 2000 National Election Studies sup? port both the aggregate- and indi- vidual-level predictions of the conflict extension perspective. 786 An expanding body of research suggests that political party elites in the United States have grown increasingly polarized along a single ideological dimension (Rohde 1991; Aldrich 1995; Poole and Rosenthal 1997; Collie and Mason 2000; Jacobson 2000; Fleisher and Bond 2000; Hetherington 2001yf , Q & R Q J U H V V U D F L D O L V V X H V V X F K D V F L Y L O U L J K W s and racial equality and cultural issues such as abortion, homosexual rights, and school prayer initially divided the parties internally, not externally. However, congressional Republicans grew more conservative than Demo? crats on racial issues in the 1960s (Carmines and Stimson 1989yf D Q G R n cultural issues in the 1980s and 1990s (Adams 1997; Layman 2001yf G U D Z - ing these formerly cross-cutting issue agendas into the dominant liberal- conservative dimension (Poole and Rosenthal 1997yf 0 H D Q Z K L O H W K H H O H F ? tions of Ronald Reagan in 1980 and a Republican congressional majority in 1994 repolarized party debate on the social welfare issues that have structured party conflict since the New Deal (Rohde 1991; Abramowitz and Saunders 1998yf 7 K H U H L V D O V R F R Q V L G H U D E O H H Y L G H Q F H W K D W W K H S D U W L H V ' convention delegates and grassroots-level activists have grown more polar? ized on social welfare, racial, and cultural issues (Stone, Rapoport, and Abramowitz 1990; Layman and Carsey 2000yf . Such definitive changes in party elite ideology generally spark a re? sponse from the mass electorate (Zaller 1992; Nie, Verba, and Petrocik 1976; Carmines and Stimson 1989yf , Q W K L V D U W L F O H Z H F R Q W H Q G W K D W W K H U e should be a limited mass response?confined largely to party identifiers who are aware of party polarization on all three ofthe domestic issue agen? das?which should produce an aggregate result that we call "conflict exten- sion." Citizens' domestic policy attitudes should remain divided into sepa- rate dimensions, but party conflict?in the form of increased party Geoffrey C. Layman is Associate Professor of Political Science, Vanderbilt University, Nashville, TN 37235 ([email protected] 7 K R P D V 0 & D U V H \ L V $ V V R F L D W e Professor of Political Science, Florida State University, Tallahassee, FL 32306-2230 ([email protected] . A previous version of this article was presented at the 2000 Annual Meeting of the American Political Science Association. We wish to thank Ted Carmines, John Geer, Bruce Oppenheimer, Paul Sniderman, and the anonymous reviewers for their helpful comments and suggestions. The data used in this study were obtained from the Inter- University Consortium for Political and Social Research. The Consortium bears no re- sponsibility for their use. American Journal ofPolitical Science, Vol. 46, No. 4, October 2002, Pp. 786-802 ?2002 by the Midwest Political Science Association ISSN 0092-5853 This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms UCONFLICT EXTENSION" IN THE U. S. ELECTORATE 787 polarization?should extend to each of these dimen? sions. After considering the existing perspectives on the mass response to elite-level ideological change, we de? velop our conflict extension argument. We then evaluate it empirically by examining aggregate-level electoral patterns from 1972 through 2000 and by testing our individual-level explanations for these patterns. Existing Views on the Mass Response to Elite Partisan Change Some scholars contend that the growth of elite-level ideological polarization has led to a similar outcome at the mass level: an "ideological realignment" of the par? ties' electoral coalitions characterized by increased lib- eral-conservative polarization (Abramowitz and Saunders 1998, 2000; Levine, Carmines, and Huckfeldt 1997; Putz and Shepherd 2001yf + R Z H Y H U W K H V H U H V H D U F K - ers employ unidimensional indicators of mass ideology, thus assuming (rather than demonstratingyf W K D W P D V V D W ? titudes toward various policy issues are structured along the same single liberal-conservative dimension that de- fines elite-level party conflict. That description runs counter to traditional per? spectives on mass partisan change, which typically argue that party conflict is dominated by?and party polariza? tion takes shape on?a single policy agenda (Key 1955; Schattschneider 1960; Burnham 1970; Riker 1982; Sundquist 1983; Carmines and Stimson 1989yf ' X U L Q g periods of stable party alignment, the parties' mass coali? tions are structured around that agenda, with each coali? tion unified by its attitudes toward the dominant issues but potentially divided on other issues that remain in the political background. Thus, when a new issue agenda that cuts across the existing line of party cleavage emerges, party polarization on the new agenda should result in a decline in party conflict on the old agenda. As Sundquist argues, athe characteristic that identifies a party realignment [is] ... the displacement of one con? flict by another" (1983,13, emphasis in the originalyf . Both viewpoints provide insight into the recent be? havior of the parties in the electorate, but neither ac? count is complete or satisfactory. The conflict displace? ment perspective's focus on the inherent tension between older and newer issue agendas comports well with the substantive differences among the three issue agendas in contemporary domestic politics and the empirical reality that citizens' attitudes toward them are distinct and, to some extent, cross-cutting (Knoke 1979; Abramowitz 1994; Shafer and Claggett 1995; Carmines and Layman 1997yf 5 D F L D O D Q G V R F L D O Z H O I D U H L V V X H V G R S R V H D V L P L O D r philosophical question?should the government take an active role in furthering social and economic equality among its citizens??which helps account for the rela? tionship some researchers find between attitudes on them (Kinder and Sanders 1996yf + R Z H Y H U ' H P R F U D W L c racial liberalism triggered mass defections from the party's traditional base in the white South, and racial concerns continue to divide the Democratic coalition both in and outside ofthe South (Huckfeldt and Kohfeld 1989; Carmines and Layman 1998yf 7 K H S K L O R V R S K L F D l question motivating cultural issues-should the govern? ment take an active role in promoting traditional notions of morality and social order?-differs from that of the other two agendas, and it is cultural conservatives, not liberals, who favor government activism. Consequently, groups that tend to be conservative on social welfare is? sues are often liberal on cultural issues, and vice versa (Ladd and Hadley 1975; Layman 2001yf . Yet, the conflict displacement outlook assumes that "the old cleavage must be played down if the new conflict is to be exploited" (Schattschneider 1960, 63yf 7 K X V L t fails to anticipate party elites polarizing on two or more cross-cutting issue dimensions, as they recently have done. In contrast, the ideological realignment perspective recognizes that party elites have grown increasingly po- larized on multiple issue agendas, but, by modeling the mass response as taking shape along a single liberal- conservative continuum, it ignores the multidimensional structure of mass issue attitudes. More generally, both perspectives offer over- simplified accounts of the mass response to elite-level change. The traditional realignment viewpoint does not anticipate that party elite change on issues may lead indi? vidual voters to alter their views on issues. Instead, it typically assumes that the only individual-level response to party elite polarization is issue-based change in party identification?either party identifiers switching their party allegiances (Erikson and Tedin 1981; Sundquist 1983yf R U L Q G H S H Q G H Q W V F K R R V L Q J D S D U W \ D I I L O L D W L R Q E D V H d on the new source of partisan conflict (Andersen 1979; Clubb, Flanigan, and Zingale 1980yf , I F L W L ] H Q V L V V X H D W W L ? tudes are fixed as such, then issue dimensions such as so? cial welfare and culture that are initially orthogonal to (uncorrelated withyf H D F K R W K H U Z L O O U H P D L Q V R 7 K H Q , even if party elites grow more polarized on both dimen? sions, increased mass party polarization on one issue agenda necessarily results in decreased polarization on the other. We contend that such an outcome is unlikely because the alternative positions championed by party elites struc? ture the political choices offered to the mass public and This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms 788 GEOFFREY C. LAYMAN AND THOMAS M. CARSEY thus play an important role in the development and ex- pression of citizens' views. When Democratic and Repub? lican elites present distinct viewpoints on multiple issues, those issues are, to some extent, packaged together for public consumption. In other words, the policy positions of the two parties help determine "what goes with what" in public policy debates and in the policy attitudes of citi? zens who receive political cues from party elites (Car? mines and Stimson 1989; Sniderman, Brody, and Tetlock 1991; Zaller 1992; Shafer and Claggett 1995yf , Q U H F H Q W G H - cades, the choice increasingly offered to voters is one be? tween a Republican party taking consistently conservative positions on social welfare, racial, and cultural issues and a Democratic party that is consistently liberal on all three agendas. In this context, it is reasonable to expect some citizens to adopt consistently liberal or consistently con? servative positions themselves. The ideological realignment account correctly per- ceives that elite policy positions do structure mass policy attitudes and that the growing ideological consistency of party elite stands on social welfare, racial, and cultural is? sues should lead to increases in mass attitude constraint. However, it envisions a large and uniform response across citizens to elite ideological polarization. In order for mass ideology to become defined by the single lib- eral-conservative dimension assumed by ideological re? alignment, it would be necessary for most citizens to fol? low the lead of party elites and stake out uniformly liberal or uniformly conservative positions on all three domestic issue agendas. Such a widespread increase in attitudinal coherence is unlikely for a number of reasons. First, there remain clear substantive differences between the three domestic agendas, and the major sociodemographic factors associ? ated with attitudes on social welfare, racial, and cultural issues?social class, race, and religion, respectively?are themselves cross-cutting. Second, Converse (1964yf D Q d others show that many, if not most, citizens do not ex- hibit much consistency in their attitudes toward even is? sues within the same policy agenda, much less toward different agendas. Third, the individuals who do have highly constrained attitudes toward the issues in one di? mension are often members of "issue publics" who care deeply about a particular issue or issue agenda, but find other issues to be much less salient and thus lack coher- ent views on them (Converse 1964; Krosnick 1990yf ) L - nally, and most importantly for this article, many, and perhaps most, citizens are unlikely to respond to political cues provided by party elites because they pay little atten? tion to elite-level politics, because they have no ties or only weak ties to a political party, or both. Conflict Extension and the Limited Mass Response to Elite Issue Convergence and Polarization We argue that the growing polarization between and ideological consistency of the social welfare, racial, and cultural stands of Democratic and Republican elites should provoke a response in mass policy attitudes, but it should be a limited one. For a segment of the electorate, we expect an attitudinal convergence similar to that im- plied by the ideological realignment research. However, for much, perhaps most, of the mass electorate, attitudes toward different issue agendas should remain distinct and cross-cutting. The extent to which citizens' issue attitudes reflect the increasingly unidimensional and polarized pattern of party elites should be shaped by two key factors: the strength of individuals' party affiliations and the degree to which they are aware ofthe polarization ofthe Demo? cratic and Republican parties on multiple issue agendas.1 Zaller (1992yf V X J J H V W V W K D W F L W L ] H Q V D U H P R V W O L N H O \ W R U H - ceive and accept political cues from elites who share their own political predispositions. Thus, as party elites grow more polarized on social welfare, racial, and cultural matters, Democratic and Republican identifiers, particu- larly strong identifiers, should be more likely than politi? cal independents to bring their own attitudes on the three agendas toward consistently liberal or consistently conservative positions. Of course, not all party identifiers, even strong party identifiers, are likely to follow the ideological lead of party elites. Many scholars note that politically sophisti- cated citizens are most likely to structure their own pref? erences in a manner consistent with elite-level cues (Converse 1964; Sniderman, Brody, and Tetlock 1991; ^thers have made similar arguments, but we take these ideas in new theoretical and empirical directions. For example, Zaller notes the likely connection between political awareness and attitude constraint (1992, 113-114yf E X W G R H V Q R W H [ D P L Q H D W W L W X G H F R Q - straint empirically nor does he examine the link between aware? ness and partisan change on multiple issue dimensions. Nie, Verba, and Petrocik (1976yf Q R W H W K H U R O H R I S D U W \ G L V W L Q F W L Y H Q H V V R Q L V V X H s in shaping levels of mass attitude constraint, but they do not con? sider the link between either partisanship or political awareness and attitude consistency. Carmines and Stimson (1989, chapter 5yf make an argument similar to ours in terms of the "bundling" of racial and social welfare issues by party elites, but they do not ex? amine empirically how party identification might shape levels of mass attitude constraint. Moreover, they do not consider the pos? sibility of conflict extension across the two issue agendas and in- stead focus on conflict displacement by noting the coincidence of growing party polarization on racial issues and declining polariza? tion on social welfare. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms CONFLICT EXTENSION" IN THE U. S. ELECTORATE 789 Sniderman 2000; Box-Steffensmeier and De Boef 2001yf . Zaller (1992yf V K R Z V W K D W W K H L Q G L Y L G X D O V P R V W O L N H O \ W o receive and accept elite-level cues consistent with their political predispositions are those with high levels of po? litical "awareness." Zaller's definition of awareness en- compasses several indicators of political sophistication and attentiveness, but, for us, the crucial element is rec- ognition of where the parties and their candidates stand relative to each other on political issues (cf. Carmines, Renten, and Stimson 1984yf 7 K H R Q O \ S D U W \ L G H Q W L I L H U s who should receive a clear signal from their party leaders to bring their views on cross-cutting issue agendas such as social welfare and culture together are those who rec- ognize that the Democratic and Republican parties have staked out distinctly liberal and conservative positions, respectively, on each set of issues. In sum, the only citizens we expect to respond to the developments observed among party elites by bringing their own views on different issue dimensions closer to? gether are party identifiers, particularly strong partisans, who are aware of party polarization on all of those di? mensions. The aggregate consequence of this limited in- dividual-level response should be conflict extension: a growth in mass party polarization on multiple distinct issue dimensions. Because many, if not most, citizens are not likely to respond to the changing elite cues, the electorate's attitudes toward the different policy agendas should remain distinct in the aggregate. However, the movement of politically-aware Democrats and Republi- cans toward the uniformly liberal and conservative social welfare, racial, and cultural stands of their parties' elites should lead to a stronger aggregate relationship between the three dimensions of policy attitudes, thereby pushing the aggregate views ofthe parties in the electorate toward more polarized positions on each issue dimension. Mass Awareness, Dimensionality, and Partisan Change: The Aggregate-Level Evidence In this section, we display several aggregate patterns that are more compatible with our conflict extension per? spective than with the conflict displacement or ideologi? cal realignment arguments. The growth in party elite po? larization of course has not been monotonic. It has instead occurred in fits and starts, with the level of polar? ization varying with changes in party leadership, the electoral cycle, and the varied campaign strategies and agendas of particular party candidates. Because citizens respond to both long-term political patterns and short- term fluctuations in elite debate (Nie, Verba, and Petrocik 1976; Zaller 1992yf W K H D J J U H J D W H S D W W H U Q V Z H U H - port exhibit some fluctuation around the general trend. However, that trend is one of growing mass awareness of party elite polarization and growing mass party polariza? tion on multiple, distinct issue dimensions. Awareness of issue differences between the parties depends not just on the attentiveness and cognitive abili- ties of individual citizens, but also on the degree of polar? ization between party elites (Hetherington 2001yf 6 R D s the parties' elites grow more polarized on social welfare, racial, and cultural issues, the proportion of the elector? ate that understands that the Republican party is more conservative than the Democratic party on all three policy dimensions should increase. Table 1 shows that this has occurred on three social welfare issues (govern? ment guarantee of jobs and a good standard of living, government services and spending, and government pro- vision of health insuranceyf R Q H U D F L D O L V V X H J R Y H U Q ? ment responsibility to help African-Americansyf D Q G W Z o cultural issues (women's role in society and abortionyf R n which respondents to the 1972 through 2000 National Election Studies (NESyf Z H U H D V N H G W R S O D F H W K H S R V L W L R Q s of the two parties.3 Despite some short-term fluctuation, a clear trend emerges: on every issue except abortion, the percentage of the electorate-and particularly of party identifiers- recognizing that the GOP is more conservative than the Democratic party was higher in 1996 than at any other point in the time series.4 That was true in 1992 for 2Awareness of party differences is measured as follows. First, the only observations treated as missing are those respondents who were not asked one or both questions (placement of Democrats, placement of Republicansyf R Q W K H S D U W L F X O D U L V V X H 6 H F R Q G W K H U H ? spondents coded as placing the two parties correctly on the issue are those who placed the Republicans as more conservative than the Democrats. All other respondents are coded as not placing the parties correctly on the issue. 3The 2000 NES interviewed roughly half (1,006yf R I L W V U H V S R Q G H Q W s in person and roughly half (801yf R I L W V U H V S R Q G H Q W V R Y H U W K e phone. Since the 1972 study, the NES has measured respondents' attitudes and their perceptions of party and candidate positions on many issues (including all of the issues in Table 1 except abortionyf by asking them to place themselves, the two parties, and candidates on seven-point scales. The face-to-face interviews in 2000 used the traditional seven-point scale format, but the phone interviews used a branching format resulting in issue scales with only five cat? egories. To maintain continuity with the NES surveys from 1972 thrcmgh 1996, our analysis includes only the face-to-face respon? dents in 2000. 4In 1992, 1996, and 2000, there were some instances in which the NES did not ask respondents to place the two parties on the issue scales, but did ask them to place the major-party presidential This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms 790 GEOFFREY C. LAYMAN AND THOMAS M. CARSEY Table 1 Awareness of the Republican Party Being More Conservative than the Democratic Party on Specific Issues, 1972-2000 (in percentagesyf Source: 1972-2000 National Election Studies Note: Entries are the percentage of respondents who placed both parties on the particular issue and who placed the Republicans to the right of the Democrats. Percentages in italics are based on placements of candidates when respondents were not asked to place the parties. aln 1976, respondents were asked to identify the party more likely to support a constitutional amendment outlawing abortion, rather than to place the two parties on the NES abortion scale. abortion, and the percentage of the electorate realizing that the Republicans are more opposed than the Demo? crats to abortion rights was only slightly lower in 1996 than in 1992. These percentages declined somewhat in 2000, perhaps due to George W. Bush's efforts to portray candidates on those scales. When that is the case, we show, in ital- ics, the percentage of respondents placing the Republican candi? date to the right of the Democratic candidate on the issue. Based on an analysis of the twenty-eight instances from 1972 through 2000 in which the NES asked respondents to place both the two parties and their presidential candidates on the same issue, we are confident that the candidate-based measures produce estimates of awareness of party differences that are very similar to those that would be produced by party-based measures. The average correla- tion between the party-based measure and the candidate-based measure of awareness over those twenty-eight cases is .64 and the a more moderate image of the Republican party, or to Al Gore's reputation as a leader of Democratic centrism, or to both. However, the electorate was more aware of party differences and their direction in 2000 than it was in the 1970s on every issue and in the 1980s on most issues. average correlation for 1992 through 2000, the years where we have to turn to candidate-based measures, is .66. The difference between the percentage of respondents placing parties and candidates cor- rectly was small (an average of four percentage pointsyf D Q G Q R W D t all systematic in direction. In the twenty-eight cases, there were thirteen instances in which a higher percentage of respondents provided a correct relative placement of candidates than a correct relative placement of parties, fourteen instances in which more re? spondents correctly placed the parties than correctly placed the candidates, and one instance in which the exact same percentage of respondents correctly placed the parties and candidates. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms CONFLICT EXTENSION IN THE 17. S. ELECTORATE 791 This growth in awareness of party differences on multiple issue agendas should increase the proportion of party identifiers who move their own views toward the uniformly liberal or conservative stands of party elites. However, because the response should be limited to aware party identifiers, we should still observe a multidi- mensional issue space, but with a growing level of mass party polarization on those dimensions.5 To test this hy? pothesis, we first assess whether attitudes on racial, cul? tural, and social welfare issues form separate dimensions or combine to produce a single issue dimension and if the dimensionality of mass issue attitudes has changed over time. To that end, we conduct confirmatory factor analyses of all of the racial, social welfare, and cultural is? sues in all of the presidential-year NES surveys that con- tained questions about all three policy agendas: those from 1972 to 2000.6 We estimated three different measurement models in each year: one with all of the issues loading on a single latent variable, one with all of the social welfare and ra? cial issues loading on a single dimension and the cultural issues forming a separate dimension, and one with racial, social welfare, and cultural issues all loading on separate 5 The degree to which the mass response to elite unidimensionality and polarization is limited depends in part on whether the size of the group most likely to exhibit that response?strong party iden? tifiers who are aware of party polarization on multiple issue di? mensions?has become a larger part of the electorate over time. From 1972 to 2000 the percentage of strong party identifiers in? creased from 25.4 to 30.6; the percentage of weak partisans and in- dependents who lean toward a party declined from 61.3 to 58.6; and the percentage of independents who do not lean toward either party declined very slightly, from 13.3 to 10.8. The percentage of the electorate that identifles with a party and is aware of party dif? ferences on all three issue agendas (measured by awareness on gov? ernment responsibility to provide jobs, government help for blacks, and women's rightsyf H [ K L E L W H G J U H D W H U F K D Q J H L Q F U H D V L Q g from 13.7 in 1972 to 34.4 in 1996, then declining to 22.5 in 2000. The percentage of the electorate made up by strong party identifi? ers who are aware of party differences on all three agendas in? creased from 4.3 in 1972 to 14.7 in 1996, then declined to 9.8 in 2000. So, there has been a noticeable increase in the size of the group most likely to respond to elite-level developments. However, that group remains a small proportion of the electorate, justifying our expectation of a limited mass response. 6 We do not exclude missing values from our analyses. We estimate our models using Amos 4.0, which computes full information maximum likelihood (FIMLyf H V W L P D W H V H Y H Q L Q W K H S U H V H Q F H R f missing data (Andersen 1957yf : R W K N H D Q G $ U E X F N O H \f de- scribe the FIML procedure used by Amos and show that the esti? mates produced by it are more consistent and efficient than those produced by methods using pairwise or listwise deletion of miss? ing observations. The specific issue questions used vary from year to year because NES did not ask the same set of questions in each survey. This makes it more difficult to compare results across years, but was necessary in order to extend the analysis back to 1972. Subsequent analysis avoids this problem by shortening the time frame. latent variables.7 To address the question of dimensional- ity, we turn to the chi-square difference test, which is the difference between the chi-square test of overall fit for a model with more latent factors and that for a model with fewer latent factors. This difference itself follows the chi- square distribution with degrees of freedom equal to the difference in the degrees of freedom between the more and less restricted models and indicates whether the model with more factors explains a significantly larger proportion of the variance in the observed indicators than does the model with fewer factors (Kline 1998yf , n the interests of space, we present, in Table 2, the full re? sults of these analyses only for 1972 and 2000. In both years, a three-factor solution captures do? mestic issue attitudes better than does a one- or two-fac- tor solution. The indicators of attitudes on cultural issues like abortion and women's rights load much more strongly on their own factor than they do on a single fac- tor with racial and social welfare attitudes. In 1972, the standardized loadings for social welfare attitudes are also stronger on a separate factor than in the single-factor model. In 2000, both social welfare and racial attitudes load somewhat more strongly on their own separate fac? tors than they do on a single factor. Most importantly, the chi-square tests of the difference in the goodness of fit between the three-factor solution and both the one- and two-factor solutions reach very high levels of statisti? cal significance (p<.0001yf L Q E R W K \ H D U V 7 K D W L V D O V R W U X e of the chi-square difference tests in every year between 1972 and 2000.8 It is clear that the issue space ofthe mass public has not come to parallel the unidimensional structure displayed by party elites. Mass policy prefer? ences have been and continue to be multidimensional. 7For each model, we take into account measurement error in each observed indicator, and when there is more than one latent vari? able, we assume that each one covaries with every other latent vari? able. To provide a scale for the latent variables, we constrain the factor loading for one observed indicator to be equal to one. For the single-factor model, that indicator is attitude toward govern? ment responsibility to ensure jobs and a good standard of living. For the two-factor model, we set to one the loadings ofthe govern? ment ensuring jobs indicator on the social welfare factor and the abortion indicator on the cultural factor. For the three-factor model, we set to one the loadings of the government help for blacks indicator on the racial factor, the government ensuring jobs indicator on the social welfare factor, and the abortion indicator on the cultural factor. All observed indicators are coded to range from-1 (most liberalyf W R P R V W F R Q V H U Y D W L Y H \f. 8 The chi-square difference test between the three-factor and one- factor solutions is 530.83 in 1976, 359.21 in 1980, 492.68 in 1984, 731.81 in 1988, 3,347.19 in 1992, and 736.64 in 1996. The chi- square difference test between the three-factor and two-factor so? lutions is 126.46 in 1976,105.10 in 1980,203.22 in 1984,424.33 in 1988, 1,010.28 in 1992, and 256.85. The factor loadings for 1976- 1996 are available from the authors upon request. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms 792 GEOFFREY C. LAYMAN AND THOMAS M. CARSEY Table 2 Confirmatory Factor Analyses of Issue Attitudes, 1972 and 2000a Source: 1972 and 2000 National Election Studies Note: Entries are unstandardized factor loadings. The standardized loadings are in parentheses. The number of observations is 2,705 in 1972 and 1,006 in 2000. aAnalyses computed using full information maximum likelihood estimation with missing values *p>.05; all other factor loadings/correlations are significant at p<.05 Since ideological realignment's assumption of mass unidimensionality does not hold, the next question is whether partisan change on these three distinct issue di? mensions has been characterized by conflict displace? ment or conflict extension. To examine the level of party polarization on the latent social welfare, racial, and cul? tural attitude dimensions uncovered by the analyses in Table 2, we inserted dummy variables for Republican and Democratic identifiers (with independent leaners coded as partisansyf L Q W R W K H W K U H H I D F W R U P R G H O I R U H D F K \ H D r and allowed them to affect each of the three latent issue variables.9 The regression coefficients on the dummy 9Because this analysis uses only cross-sectional data with only one indicator of party identification, we assume that party identifica? tion is measured without error. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms CONFLICT EXTENSION IN THE 17. S. ELECTORATE 793 Figure 1 Party Polarization on Three Issue Dimensions, 1972-2000 0.3 0.2 Party Positions on Social Welfare Issues -0.2 -0.3 Republicans 0.3 0.2 CD co c o O 0 CD 13 CO _co CO CD -0.2 1972 1976 1980 1984 1988 1992 1996 2000 Year Party Positions on Cultural Issues -0.3 Republicans Democrats 0.3 Party Positions on Racial Issues Republicans Democrats -0.2 -0.3 L"-?- 0.3 0.2 k 1972 1976 1980 1984 1988 1992 1996 2000 Year Party Difference on Each Dimension* CD o 0.1 h >r 0 co Q_ -0.2 -0.3 1972 1976 1980 1984 1988 1992 1996 2000 Year 1972 1976 1980 1984 1988 1992 1996 2000 Year Source: 1972-2000 National Election Studies Note: Party positions are the estimated means for each party on latent variables (ranging from the most liberal to the most conservative positionyf I U R P F R Q I L U P D W R U \ I D F W R U D Q D O \ V H V . *Party difference is the Republican mean minus the Democratic mean. variables indicate the difference between the mean for Republicans and that of independents and the mean for Democrats and that of independents. Figure 1 presents the estimated means for Democratic and Republican identifiers and the differences between the two party coa? litions from 1972 to 2000.10 10In order to assign a scale to each latent variable, the intercept of the latent variable is set to zero. Thus, the mean value on each la? tent variable for independents is zero and the predicted mean val? ues for Republicans and Democrats in each year are simply the es? timated difference between their mean positions and those of independents (zeroyf 7 K H G L I I H U H Q F H E H W Z H H Q W K H ' H P R F U D W L c mean and the independent mean is statistically significant (p<.05yf in every year on social welfare, from 1976 to 2000 on racial issues, Contrary to the prediction of conflict displacement, increases in party polarization on newer agendas such as racial and cultural issues have not resulted in partisan de- polarization on older social welfare issues. Instead, on all three issue dimensions, the levels of mass party polariza? tion that existed in 1972 were either equally as large or more substantial in 2000. and in 1984 and 1992 to 2000 on cultural issues. The difference be? tween the Republican mean and the independent mean is signifi? cant in every year on social welfare, in 1972 and 1980 to 2000 on racial issues, and 1980 and 1988 to 2000 on cultural issues. The dif? ference between the Democratic mean and the Republican mean is significant in every year on social welfare and racial issues, and from 1980 to 2000 on cultural issues. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms 794 GEOFFREY C. LAYMAN AND THOMAS M. CARSEY Not surprisingly, the largest growth in party polar? ization was on the newest agenda, cultural issues. Much like their party platforms and government officials, Re? publican and Democratic identifiers did not differ much on cultural issues in the 1970s. Partisan differences in? creased sharply in 1980 with the Republican nomination of the culturally-conservative Ronald Reagan. The cul? tural cleavage temporarily narrowed in 1988, but as the Christian Right became increasingly influential within the Republican party and religious liberals and secular individuals responded with a countermobilization into the Democratic party (Layman 2001yf S D U W \ S R O D U L ] D W L R n on cultural issues increased sharply after 1988. Carmines and Stimson (1989yf S R L Q W W R D F U L W L F D l moment" in the racial series between 1960 and 1964. Re? publicans became clearly more conservative than Demo? crats, and those differences increased gradually through the 1960s and early 1970s. Accordingly, we find clear par? tisan differences on racial issues in 1972. But, what is most interesting is that the mid-1970s, the very point at which Carmines and Stimson's analysis ends, marked the beginning of another period of growth in partisan racial polarization. Party differences increased sharply between 1972 and 1984 and then increased gradually over the rest of the 1980s and the 1990s. As we would expect, the parties' coalitions were al- ready divided on social welfare issues in 1972. But, that partisan division did not decrease even as polarization increased on the racial and cultural dimensions. Rather, party polarization on social welfare actually grew through the 1980s and 1990s. In keeping with the elec- tions of 1980 and 1994 signaling rightward turns in the economic and social welfare views of the GOP, most of the increase resulted from the growing conservatism of Republicans. While there is some year-to-year volatility in the level of mass party polarization, it is clear that the level of party polarization on all three issue agendas has increased over the past thirty years. Party polarization was greater in 1996 than it had been at any point in the previous twenty- five years, and even in 2000, party polarization was greater than it was in 1972 on all three agendas.11 nTo test whether the over-time increases in levels of party polar? ization on the three issue dimensions we reported are statistically significant, we constrained the levels of polarization in 1996 and 2000 to equal the level of polarization in 1972 and computed the chi-square difference test for the goodness of fit of constrained and unconstrained models for 1996 and 2000. Specifically, we con? strained the coefficients on the Democratic and Republican dummy variables included in the confirmatory models that pro- duced the estimates of polarization illustrated in Figure 1 to be the same distance apart from each other in 1996 and 2000 as they were found to be in 1972. The differences in the chi-square statistics for the constrained and unconstrained models for both 1996 and 2000 Testing the Explanation for Mass Conflict Extension We contend that conflict extension has occurred instead of conflict displacement because citizens have responded to growing consistency in the social welfare, racial, and cultural stands of party elites by adopting more consis? tent stands themselves. Conflict extension has occurred instead of a more complete ideological realignment be? cause only some citizens, party identifiers who are aware of party differences on the various issue agendas, are likely to have brought their social welfare, racial, and cul? tural attitudes toward the uniformly liberal or conserva? tive stands of party elites. In this section, we evaluate this individual-level explanation for conflict extension. Unfortunately, changes in question wording and, more importantly, in the specific issues about which NES surveys have asked make comparisons ofthe relationship between issue attitudes over the full 1972-2000 time pe- riod very difficult. To ensure that observed changes in at- titude constraint are not due to changes in question wording or the content of particular agendas, it is neces? sary to examine attitudes on the same issues, measured in the same way, over time. So, in Table 3, we examine the relationship between attitudes on the eleven social wel? fare, racial, and cultural issues that were common to all ofthe presidential-year NES surveys from 1988 through 2000 for three categories of citizens: individuals who identify strongly with the Democratic or Republican par? ties, weak identifiers and independents who say they lean toward one of the two parties, and independents who do not lean toward either party ("pure" independentsyf : e show the loadings of all of these issues on a single factor and the average standardized loading of all issues on that were statistically significant, indicating that mass party preferences were significantly more polarized in 1996 and in 2000 than they were in 1972. 12Beginning the analysis in 1988 is appropriate for several reasons. First, Figure 1 indicates that the years from 1988 through 1996 evi- denced increasing party polarization on all three issue dimensions simultaneously, thus providing a clear example of conflict exten? sion. Second, the number of issue questions common to all years becomes quite small if we begin the analysis before 1988. Third, if we begin the analysis before 1988, there is only one racial issue, government help for blacks, that is common to all years, making it impossible to estimate a separate factor for racial attitudes. As a check, we examined the correlations between pairs of issue dimen? sions for the three-factor model applied to the 1972 through 2000 NES studies using the same models which produced the results presented in Table 2, broken down by strength of partisanship. De- spite differences in the questions asked from year to year, in every year the correlations between issue dimensions for strong parti- sans was higher than for pure independents, with the correlations among issue dimensions for weak and leaning partisans almost al- ways falling in between. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms CONFLICT EXTENSION IN THE U. S. ELECTORATE 795 Table 3 Factor Analysis by Strength of Party Identification, 1988-2000: Loadings of the Same 11 Issues on a Single Factor and Factor Correlations from Three-Factor Models of the Same Issues Source: 1988-2000 National Election Studies Note: Entries are unstandardized factor loadings. Standardized loadings are in parentheses. *Not significant at p<.05. All other factor loadings and correlations are significant at p<.05. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms 796 GEOFFREY C. LAYMAN AND THOMAS M. CARSEY factor.13 Because cultural issues tend to load most weakly on a single factor, we also show the average standardized loading of all cultural attitudes. Finally, we show the cor- relations between latent issue attitudes in the three-factor model and the estimated level of party polarization on the three latent factors in each year.14 The results show that attitude constraint on these eleven issues increased between 1988 and 1996 for strong party identifiers. The average loading of all issues on a single factor increased from .43 to .54 and the average loading of cultural issues more than doubled from .14 to .34. The correlation in three-factor models between la? tent cultural attitudes and latent social welfare attitudes nearly quadrupled between 1988 and 1996, and the cor? relation between the racial and cultural factors increased from .22 to .55. An increase in constraint is also evident for weak and leaning partisans, with both the loadings on a single factor and correlations between the cultural fac? tor and the social welfare and racial factors growing stronger. However, these increases are less dramatic than they are for strong partisans. The relationships between the cultural, racial, and social welfare attitudes of party identifiers grew weaker between 1996 and 2000. How? ever, those relationships were generally stronger in 2000 than they were in 1988. Importantly, there is no evidence of growth in attitu- dinal constraint among pure independents over time. The average loading of all issues on a single factor was lower in 2000 than it was in 1988, and the loadings of cultural issues display no pattern of growth. In fact, the correlation between the cultural factor and the social welfare and racial factors in a three-factor solution is sta? tistically insignificant in each year. An alternative hypothesis to ours is that the patterns in Table 3 are due to the mobilization and demobiliza- tion of individuals into and out of the parties in the elec? torate rather than increased attitudinal convergence among individual party identifiers. The individuals most likely to come into and/or stay in parties may be those who share the consistently liberal or conservative views 13 To provide a scale for this single latent variable, we set the factor loading of observed attitudes toward government responsibility to ensure jobs to one in each analysis. 14We also estimated one- and three-factor models for the attitudes of all respondents on these eleven issues, and found a growth in at? titude constraint between 1988 and 1996. The average loading of all issues and of cultural issues on a single factor increased consid- erably, as did the correlations between the cultural factor and the other two factors in the three-factor model. There was also a clear increase in party polarization, estimated in the same way as in Fig- ure 1, on social welfare, racial, and cultural issues. Levels of atti? tude constraint and party polarization were lower in 2000 than in 1996, but remained higher than in 1988. of party elites on all three major issue agendas, while in- dividuals who do not have uniformly liberal or conserva? tive views on the three agendas are more likely to become independents. This is most likely occurring to some de? gree, but the question is whether the growth in attitude constraint among party identifiers is due solely to mobi- lization/demobilization or is at least partly due to attitu- dinal convergence among individual partisans. To answer that, we turn to panel data from the 1992,1994, and 1996 NES surveys and compare the levels of attitude con? straint across the three panel waves of individuals who identified with the same party in each wave to those who were independents in each wave. Table 4 shows the load? ings of attitudes toward the five social welfare and three cultural issues that were in each panel wave for both par? tisan groups.15 It also shows the average standardized loadings of all issues and of cultural issues on that factor, as well as the correlations between the social welfare and cultural factors from two-factor models. Even over a short four-year period, there is clear evi? dence of individual party identifiers merging their atti? tudes on social welfare and cultural issues. The average loading of all issues and of cultural issues on a single fac? tor increased noticeably between 1992 and 1996, particu- larly for attitudes on abortion and women's rights. The correlation between social welfare and cultural attitudes in a two-factor model nearly doubled for partisans over this period. In contrast, there is no evidence of increased attitudinal constraint among independents. The average loadings, either of all issues or of cultural issues, on a single factor did not increase at all, and the correlation between the social welfare and cultural factors was smaller in 1996 than in 1992 for independents. The level of attitu? dinal constraint among independents is also lower than it is for partisans in all three waves ofthe panel. Together, the results in Tables 3 and 4 confirm that the individuals who are most likely to take cues from party leaders?party identifiers, particularly strong par? tisans?have followed their lead by growing more consis? tently conservative or more consistently liberal in their attitudes toward the major domestic issue agendas. As expected, citizens without attachments to the parties have lower levels of attitudinal constraint and have not displayed any increase in attitudinal constraint across distinct issue dimensions. 15We do not include racial issues in this analysis because there were only two racial issues that were in each of the three panel waves and because the very small sample size for independents cre- ated difficulties for estimating a three-factor model. However, in? cluding racial issues in the analysis for partisans, or in the single- factor models for independents, does not alter the results shown here. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms CONFLICT EXTENSION IN THE U. S. ELECTORATE 797 Table 4 Factor Analysis of the Same Eight Social Welfare and Cultural Issues for Partisans and Independents in the 1992-1994-1996 Panel Partisan Group and Issue 1992 1994 1996 Partisans* Government ensure jobs Government services/spending Government health insurance Spending for food stamps Spending for social security Abortion Women's rights School prayer (Nyf Average standardized loading Average loading of cultural issues 2-factor: rS0Cja| we|fare & cultural Independents^ Government ensure jobs Government services/spending Government health insurance Spending for food stamps Spending for social security Abortion Women's rights School prayer (Nyf Average standardized loading Average loading of cultural issues ^-tactor: rS0Cja| Welfare & cultural Source: 1992-1994-1996 National Election Studies Panel Note: Entries are unstandardized factor loadings. Standardized loadings are in parentheses. aPanel respondents who identified themselves as Democrats (strong or weakyf L Q D Q G D Q G U H V S R Q G H Q W V Z K R L G H Q W L I L H G W K H P V H O Y H V D s Republicans (strong or weakyf L Q D Q G 6 bPanel respondents who identified themselves as independents (leaning or non-leaningyf L Q D Q G 6 *Not significant at p<.05. All other factor loadings and correlations are significant at p<.05. We contend that the catalyst for growing attitude constraint among party identifiers is increased awareness of party elite differences on various issue dimensions. If that is true, then the relationship between attitudes to? ward distinct issue dimensions such as social welfare and culture should be stronger among partisans who are aware of party differences on both dimensions than among partisans who are not aware of these differences. To test this, we use data pooled from the 1992,1996, and 2000 NES, and measure awareness of party issue differ? ences using the three issues on which all three studies asked respondents to place the parties and/or the presi? dential candidates.16 We combined the indicators of 16Because this part of the analysis divides the sample by both strength of partisanship and awareness of party differences, the sample size of some categories is very small. We pooled the 1992, 1996, and 2000 NES to increase the number of observations in these categories. We could not include the 1988 NES here because awareness of party/candidate differences on government guarantee of jobs and government services and spending to create a measure of awareness of party differences on social welfare issues.17 Awareness of party differences on cultural issues is defined as identifying the Republican party/candidate as more pro-life on abortion than the Democratic party/candidate. For each partisan group, we conducted factor analyses of the eight social welfare and cultural issues examined in Table 3 for individuals who were aware that the Republicans were more conservative than the Democrats on neither issue agenda, on only one it did not ask about party/candidate issue differences on abortion, which is the only cultural issue on which 1992 respondents were asked to place the parties or candidates. 17Awareness of party differences on social welfare issues is defined simply as being aware that the Republican party/candidate is more conservative than the Democratic party/candidate on both of these issues. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms 798 GEOFFREY C. LAYMAN AND THOMAS M. CARSEY Table 5 Results of Factor Analysis of the Same Eight Social Welfare and Cultural Issues By Strength of Party Identification and Awareness of Partisan Differences on Social Welfare and Cultural Issues Source: 1992, 1996, and 2000 National Election Studies (pooledyf Note: Entries are unstandardized factor loadings. Standardized loadings are in parentheses. In both the one-factor and two-factor models, each ob? served indicator is a function of a latent variable, a random measurement error, and dummy variables for 1996 and 2000 respondents. The effect of the year dummy variables is not shown. *Not significant at p<.05. All other factor loadings and correlations are significant at p<.05. of the agendas, and on both agendas.18 For each level of partisan strength and awareness, Table 5 shows the esti- 18Racial issues are excluded from this analysis because there is no measure of awareness of party differences on racial issues in the 1992 study. mates of a single-factor model and the correlations be? tween latent social welfare and cultural attitudes in a two-factor model.19 19 The chi-square difference tests indicate that the two-factor solu? tion is superior to the one-factor solution for each group. To ac- This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms CONFLICT EXTENSION IN THE U. S. ELECTORATE 799 The impact of awareness is quite marked for strong partisans. As expected, it is only when strong partisans are aware of party polarization on both social welfare and cultural issues that their attitudes toward these di? mensions come closely together. As we move from strong partisans who are not aware of party differences on ei- ther social welfare or cultural issues to strong partisans who are aware of party differences on just one agenda, overall constraint increases slightly (the average loading of all issues on a single factor increases from .22 to .34yf . However, the loadings of cultural attitudes on a single factor and the correlation between the factors in a two- factor model reveal no increase in the relationship be? tween social welfare and cultural attitudes. In contrast, when we move to strong partisans who are aware of party differences on both social welfare and cultural con? cerns, issue attitudes become nearly unidimensional. The average loading of all issues increases markedly, the aver? age loading of cultural issues increases enormously, and the formerly weakly negative correlation between the so? cial welfare and cultural factors is now a positive .77. A similar pattern exists among weak and leaning partisans. The relationship between social welfare and cultural attitudes within the groups that are aware of party differences on neither issue agenda or on only one agenda is very weak. It is only among those weak and leaning partisans who are aware of party differences on both agendas that attitudes toward social welfare and cultural issues really come together. These findings suggest that it is not simply political sophistication or general political awareness that leads party identifiers to respond in kind to the growing unidi- mensionality and polarization of party elites. Individuals who are aware of party differences on one agenda are more aware of the political environment than individuals who are aware of differences on neither agenda, but they are no more likely to merge their social welfare and cul? tural attitudes together. We argued that the most appar- ent reason why partisans would bring their own views on seemingly cross-cutting issue dimensions like culture and social welfare together is an awareness that the par? ties have taken polarized stands on both agendas, and the results show that it is only when that condition holds that attitudes toward the two agendas converge. As we expected, awareness that the parties are polar? ized on social welfare and cultural issues produces no in? crease in attitudinal constraint among pure independents. The average loading of both all issues and cultural issues count for the possibility that the intercept (mean valueyf I R U H D F h observed indicator may vary across the three years, we included dummy variables for 1996 and 2000 respondents in our model. Each observed indicator is a function ofa latent variable, a random measurement error, and the dummies for 1996 and 2000. is no greater for independents who are aware of partisan differences on both social welfare and cultural issues than for independents who are aware of party differences on neither issue agenda. The correlation between social wel? fare and cultural attitudes in the two-factor model is not statistically significant for any group of independents. Conclusion We have argued that neither the traditional "conflict dis? placement" view of mass partisan change nor the more recent work on mass "ideological realignment" provides a satisfactory explanation of how the electorate should react to increasingly unidimensional and ideologically polarized party elites. The alternative we have offered is an account of "conflict extension," in which the mass re? sponse to recent elite-level developments should be lim? ited to party identifiers, particularly strong partisans, who are aware of party polarization on each separate is? sue agenda. In the aggregate, this limited mass response should produce conflict extension, with the parties in the electorate growing more polarized on social welfare, ra? cial, and cultural issues, but with attitudes toward all three agendas remaining distinct. The aggregate- and in- dividual-level evidence we presented supports our view. Our findings challenge the conventional wisdom that mass partisan change is driven by a single issue dimension and characterized by conflict displacement. An important question is if the conflict displacement perspective once accurately described the partisan change process, why does it no longer seem to do so? Our account of contem- porary electoral developments points toward an answer: the extent to which mass partisan change is characterized by conflict extension as opposed to conflict displacement depends heavily on the structure of ideological conflict between party elites. If party elites polarize along one ma? jor issue dimension and take centrist or heterogeneous positions on other dimensions, then the parties in the electorate will not receive cues that their own views on different agendas should be consistent. In such circum- stances, we would not expect mass ideology to be polar? ized on multiple agendas nor would we expect attitudes on multiple issue dimensions to converge toward a single liberal-conservative dimension, even among the most aware strong partisans. However, if Democratic and Re? publican elites take positions on multiple issue dimen? sions that are consistently liberal and consistently conser? vative, respectively, then politically-aware party identifiers will receive cues that their views on different issue agendas should go together and they should move toward polar? ized stands on each of those dimensions. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms 800 GEOFFREY C. LAYMAN AND THOMAS M. CARSEY Part of the reason that the conflict displacement the- sis has prevailed may be the common view in the realign? ment literature that party leaders have incentives to keep the focus of political conflict on one particular issue di? mension (Schattschneider 1960; Sundquist 1983yf ' X U ? ing a period of stable alignment, majority party elites want to focus partisan debate on the issues that created the alignment that made them the majority (Riker 1982yf . Some work suggests that minority party elites have in? centives to introduce cross-cutting issue dimensions in an efifort to disrupt the alignment (Riker 1982; Carmines and Stimson 1989yf E X W P L Q R U L W \ S D U W \ O H D G H U V D O V R D F - crue benefits from the existing alignment?they hold po? litical office and they hold the balance of power within their own party?and they may not wish to upset it by taking the focus off the dominant issues (Carmines 1991yf , Q I D F W 6 X Q G T X L V W \f contends that the initial reaction of both parties' leaders to a new issue agenda is to straddle and suppress it. Of course, during a period of party realignment, such efforts fail and new issues are in? troduced onto the agenda. But, realignments also involve some replacement in the leadership of both parties (Sundquist 1983yf D Q G W K H H O L W H V Z K R Z H U H E U R X J K W W o power by the new issue dimension may wish to focus po? litical conflict only on it and avoid continued party po? larization on the old dimension. If party elites do wish to limit partisan conflict to one issue dimension at a time, and if they are successful in that goal, then periods of change in the partisan issue agenda should be associated with conflict displacement. Of course, party leaders never have been able to exert enough control over the agenda of political conflict to limit it to a single issue dimension indefinitely (Riker 1982yf E X W W K H \ P D \ K D Y H E H H Q P X F K E H W W H U D E O H W R F R Q ? trol the agenda in the past than they are now. Before the Democratic party's reforms in the late 1960s and early 1970s created a participatory nominating process at the national level, party leaders exercised considerable influ? ence over candidate nominations and the drafting of party platforms, and thus over the parties' issue agendas. With a nominating process now dominated by primaries and caucuses, however, party leaders have much less con? trol over the agenda. Groups of political activists, each championing different issue agendas, have greater access to party politics and can exert substantial pressure on party candidates, office-holders, and platforms to take extreme stands on these multiple agendas in order to piece together a winning electoral coalition.20 When 20Much of the evidence for growing party elite polarization on multiple policy agendas comes from Congress, where changes in the nomination process have not been as dramatic as at the presi? dential level. Thus, there are certainly other factors contributing to these elite-level developments. The movement of southern whites party elites are unable to limit partisan conflict to a single, dominant issue dimension and are pressured to take consistently liberal or conservative positions on nu- merous issue agendas, this signals to party identifiers that they should do so as well. Thus, conflict extension, rather than conflict displacement, should occur. The conflict extension perspective on partisan change may be more compatible with contemporary theories of electoral politics and candidate strategy than either the ideological realignment or conflict displacement view- points. Spatial theories that rest on assumptions about the desire and ability of parties to shift the relative salience of different issue dimensions (cf. Riker 1990; Shafer and Claggett 1995; Carsey 2000yf D U H S U H F O X G H G X Q G H U L G H R ? logical realignment's notion that mass issue preferences are unidimensional. But, they are perfectly consistent with our view that although the mass parties have grown more polarized on multiple issue agendas, mass ideology has remained multidimensional. Meanwhile, our evidence that attentive partisans bring their own issue positions to? ward the consistently liberal or consistently conservative stands of Democratic and Republican elites, something not predicted by conflict displacement accounts, suggests that parties may find some success using persuasion as a strategic tool. This is consistent with theories arguing that it is necessary or at least often beneficial to parties to take nonmedian positions on issues (cf. Rabinowitz and Macdonald 1989; Gerber and Jackson 1993; Carsey 2000yf . Lastly, the importance that awareness of elite-level parti? san politics holds in the conflict extension account com- ports with theories of the importance of information in electoral politics (cf. Alvarez 1997yf W K X V S R L Q W L Q J P R U e generally to the importance of awareness in linking mass- and elite-level politics. References Abramowitz, Alan I. 1994. "Issue Evolution Reconsidered: Ra? cial Attitudes and Partisanship in the U.S. Electorate." American Journal of Political Science 38 (1yf . with conservative views on social welfare, racial, and cultural issues out of the Democratic coalition and into the Republican coalition is one of those (Rohde 1991yf + R Z H Y H U F R Q J U H V V L R Q D O Q R P L Q D - tions, like most gubernatorial and state legislative nominations are dominated by primaries and caucuses, where single-issue and ideologically-extreme activists exert considerable influence over candidates' policy stands (Fiorina 1999yf 0 R U H R Y H U F K D Q J H V L n presidential nominations may invite more policy-oriented activists into party politics and encourage their participation in sub- presidential nomination contests. Policy-oriented activists drawn into a party by presidential nomination politics also may decide to run for office themselves and become policy-oriented local, state, or congressional candidates (Fiorina 1999yf . This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms CONFLICT EXTENSION IN THE U. S. ELECTORATE 801 Abramowitz, Alan L, and Kyle L. Saunders. 1998. "Ideological Realignment in the U.S. Electorate." Journal of Politics 60 (3yf . Abramowitz, Alan L, and Kyle L. Saunders. 2000. "Ideological Realignment and U.S. Congressional Elections." Presented at the Annual Meeting of the American Political Science As- sociation, Washington, D.C. Adams, Greg D. 1997. "Abortion: Evidence of Issue Evolution." American Journal of Political Science 41 (3yf . Aldrich, John H. 1995. Why Parties? The Origin and Transfor- mation of Political Parties in America. Chicago: The Univer? sity of Chicago Press. Alvarez, R. Michael. 1997. Information and Elections. Ann Ar- bor: University of Michigan Press. Andersen, Kristi. 1979. The Creation of a Democratic Majority, 1928-1936. Chicago: The University of Chicago Press. Andersen, T.W. 1957. "Maximum Likelihood Estimates for a Multivariate Normal Distribution When Some Observa- tions are Missing." Journal of the American Statistical Asso- ciation 52 (278yf . Box-Steffensmeier, Janet M., and Suzanna De Boef. 2001. "Macropartisanship and Macroideology in the Sophisti- cated Electorate." Journal of Politics 63 (lyf . Burnham, Walter Dean. 1970. Critical Elections and the Main- springs of American Politics. New York: Norton. Carmines, Edward G. 1991. "The Logic of Party Alignments." Journal of Theoretical Politics 3(1yf . Carmines, Edward G., and Geoffrey C. Layman. 1997. "Issue Evolution in Postwar American Politics: Old Certainties and Fresh Tensions." In Present Discontents: American Politics in the Very Late Twentieth Century, ed. Byron E. Shafer. Chatham, NJ: Chatham House, 89-134. Carmines, Edward G., and Geoffrey C. Layman. 1998. "When Prejudice Matters: The Impact of Racial Stereotypes on the Racial Policy Preferences of Democrats and Republicans." In Prejudice and Politics, ed., Jon Hurwitz and Mark Peffley. New Haven, CT: Yale University Press, 100-34. Carmines, Edward G., Steven H. Renten, and James A. Stimson.1984. "Events and Alignments: The Party Image Link." In Controversies in Voting Behavior, 2nd ed., ed. Rich? ard E. Niemi and Herbert F. Weisberg. Washington: Con? gressional Quarterly. Carmines, Edward G., and James A. Stimson. 1989. Issue Evolu? tion: Race and the Transformation of American Politics. Princeton, NJ: Princeton University Press. Carsey, Thomas M. 2000. Campaign Dynamics: The Race for Governor. Ann Arbor: University of Michigan Press. Clubb, Jerome M., William H. Flanigan, and Nancy H. Zingale. 1980. Partisan Realignment. Beverly Hills: Sage. Collie, Melissa R, and John Lyman Mason. 2000. "The Electoral Connection between Party and Constituency Considered: Evidence from the U.S. House of Representatives, 1972- 1994." In Continuity and Change in House Elections, ed. David Brady, John Cogan, and Morris P Fiorina. Stanford: Stanford University Press, 211-34. Converse, Philip E. 1964. "The Nature of Belief Systems in Mass Publics." In Ideology and Discontent, ed. David E. Apter. New York: Free Press. Erikson, Robert S. and Kent L. Tedin. 1981. The 1928-1936 Partisan Realignment: The Case for the Conversion Hy? pothesis. American Political Science Review 75 (4yf . Fiorina, Morris P. 1999. "Whatever Happened to the Median Voter?" Presented at the MIT Conference on Parties and Congress, Cambridge, MA. Fleisher, Richard, and Jon R. Bond. 2000. "Partisanship and the President's Quest for Votes on the Floor of Congress." In Po? larized Politics: Congress and the President in a Partisan Era, ed. Jon R. Bond and Richard Fleisher. Washington: Congres? sional Quarterly Press, 154-85. Gerber, Elisabeth R, and John E. Jackson. 1993. "Endogenous Preferences and the Study of Institutions." American Politi? cal Science Review 87 (3yf . Hetherington, Marc J. 2001. "Resurgent Mass Partisanship: The Role of Elite Polarization." American Political Science Review 95 (3yf . Huckfeldt, Robert, and Carol Weitzel Kohfeld. 1989. Race and the Decline ofClass in American Politics. Urbana: University of Illinois Press. Jacobson, Gary C. 2000. "The Electoral Basis of Partisan Polar? ization in Congress." Presented at the Annual Meeting ofthe American Political Science Association, Washington, D.C. Key, V.0.1955. "A Theory of Critical Elections." Journal of Poli? tics 17 (1yf . Kinder, Donald R., and Lynn M. Sanders. 1996. Divided by Color: Racial Politics and Democratic Ideals. Chicago: Uni? versity of Chicago Press. Kline, Rex B. 1998. Principles and Practice ofStructural Equa? tion Modeling. New York: Guilford. Knoke, David. 1979. "Stratification and the Dimensions of American Political Orientations." American Journal ofPoliti? cal Science 23 (4yf . Krosnick, Jon A. 1990. "Government Policy and Citizen Pas- sion: A Study of Issue Publics in Contemporary America." Political Behavior 12 (1yf . Ladd, Everett Carll, Jr., with Charles D. Hadley. 1975. Transfor- mations of the American Party System. New York: W.W. Norton. Layman, Geoffrey C. 2001. The Great Divide: Religious and Cul? tural Conflict in American Party Politics. New York: Colum- bia University Press. Layman, Geoffrey C, and Thomas M. Carsey. 2000. "Ideologi? cal Realignment in Contemporary American Politics: The Case of Party Activists." Presented at the Annual Meeting of the Midwest Political Science Association, Chicago. Levine, Jeffrey, Edward G. Carmines, and Robert Huckfeldt. 1997. "The Rise of Ideology in the Post-New Deal Party Sys? tem, 1972-1992." American Politics Quarterly 25 (1yf . Nie, Norman H., Sidney Verba, and John R. Petrocik. 1976. The Changing American Voter. Cambridge: Harvard University Press. Poole, Keith T., and Howard Rosenthal. 1997. Congress: A Po? litical-Economic History ofRoll Call Voting. New York: Ox? ford University Press. Putz, David W., and Adrian J. Shepherd. 2001. "The Dynamics of Ideological Realignment Among Elite and Mass Parti? sans." Presented at the Annual Meeting of the Midwest Po? litical Science Association, Chicago. Rabinowitz, George and Stuart Elaine Macdonald. 1989. "A Di- rectional Theory of Issue Voting." American Political Science Review 83 (3yf . Riker, William H. 1982. Liberalism Against Populism. San Fran- cisco: Freeman. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms 802 GEOFFREY C. LAYMAN AND THOMAS M. CARSEY Riker, William H. 1990. "Heresthetic and Rhetoric in the Spa- tial Model." In Advances in the Spatial Theory of Voting, ed. James M. Enelow and Melvin J. Hinich. Cambridge: Cam? bridge University Press. Rohde, David W. 1991. Parties and Leaders in the Postreform House. Chicago: University of Chicago Press. Schattschneider, E.E. 1960. The Semisovereign People: A Realisfs View of Democracy in America. Hinsdale, IL: The Dryden Press. Shafer, Byron E., and William J.M. Claggett. 1995. The Two Ma? jorities: The Issue Context ofModern American Politics. Balti- more: Johns Hopkins University Press. Sniderman, Paul M. 2000. "Taking Sides: A Fixed Choice Theory of Political Reasoning." In Elements of Reason: Cog- nition, Choice, and the Bounds of Rationality, ed. Arthur Lupia, Mathew D. McCubbins, and Samuel L. Popkin. New York: Cambridge University Press, 67-84. Sniderman, Paul M., Richard A. Brody, and Philip E. Tetlock. 1991. Reasoning and Choice: Explorations in Political Psy- chology. Cambridge: Cambridge University Press. Stone, Walter J., Ronald B. Rapoport, and Alan I. Abramowitz. 1990. "The Reagan Revolution and Party Polarization in the 1980s." In The Parties Respond, ed. L. Sandy Maisel. Boulder, CO: Westview Press. Sundquist, James L. 1983. Dynamics ofthe Party System: Align? ment and Realignment ofPolitical Parties in the United States. Washington: The Brookings Institution. Wothke, Werner, and James L. Arbuckle. 1996. "Full-Informa- tion Missing Data Analysis with Amos." SPSS White Paper. Zaller, John. 1992. The Nature and Origins ofMass Opinion. New York: Cambridge University Press. This content downloaded from 130.101.81.186 on Sun, 25 Apr 2021 21:12:23 UTC All use subject to https://about.jstor.org/terms