The Social and Ideological Bases of Middle-Class Political Realignment

Home , Political realignment

The Social and Ideological Bases of Middle-Class Political Realignment in the United States, 1972 to 1992 Author(s): Clem Brooks and Jeff Manza Source: American Sociological Review, Vol. 62, No. 2 (Apr., 1997), pp. 191-208 Published by: American Sociological Association Stable URL: http://www.jstor.org/stable/2657299 Accessed: 05/01/2010 13:17

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=asa.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

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].

American Sociological Association is collaborating with JSTOR to digitize, preserve and extend access to American Sociological Review.

http://www.jstor.org THE SOCIAL AND IDEOLOGICAL BASES OF MIDDLE-CLASS POLITICAL REALIGNMENT IN THE UNITED STATES, 1972 TO 1992*

Clem Brooks Jeff Manza Indiana University, Bloomington Pennsylvania State University

Research on the nature and bases of the political alignments of the middle class in the United States has produced extensive disagreements but little resolution of fundamental controversies. We address unresolved questions about the political alignments of the middle class through an investigation of change in voting behavior among two of its principal segments: managers and professionals. Among professionals, but not among managers, there has been a long-term shift away from support for Republican presidential candidates to support for Democratic candidates. Competing hypotheses about the sources of these trends are tested using data from the National Election Studies. Increasingly liberal attitudes toward social issues, not changes in economic evaluations or sociodemographic composition, explain the growing tendency of professionals to vote Democratic and their increasing divergence from managers. Party identification and partisan affect substantially mediate the effects of social group membership, views of the welfare state, and attitudes towards social issues. The relevance of these findings to understanding the Democratic realignment of professionals and the stable Repub- lican alignment of managers is discussed.

he political alignments of the middle [1960] 1981). Classical sociological theories class have long fascinated and perplexed of middle-class political alignments empha- social scientists and analysts of U.S. politics size their sober conservatism and stable (Tocqueville [1835] 1969; Mills 1951; Lipset democratic values (Parsons 1954:39-54; Lip- set 1960). Since the late 1960s, however, the * Direct correspondence to Clem Brooks, De- emergence of pressures for social change partment of Sociology, Indiana University, from "new" social movements drawing sup- Bloomington, IN, 47405 (cbrooks @indiana.edu). port primarily from members of the educated Authors names are listed in alphabetical order; middle class has brought forth a series of both authors contributed equally. An earlier ver- theoretical argumentsabout the receptiveness sion of this paper was presented at the 1994 an- of various segments of the middle class to nual meeting of the American Sociological Asso- liberal political appeals. Some of these argu- ciation in Los Angeles. Data and codebooks from the American National Election Studies were pro- ments emphasize the role of material inter- vided by the Inter-university Consortium for Po- ests in an expanded public sector in dispos- litical and Social Research. The authors are solely ing them to support or participate in strug- responsibility for tabulations, analyses, and inter- gles for social change (Bruce-Briggs 1979; pretations of these data. Michael Hout provided Gouldner 1979), while others point to chang- us with inspiration and incisive comments. We ing values (Inglehart 1990; Clark 1994). The also thank Steven Brint, Terry Clark, the late resurgence of conservative political and eco- Clifford Clogg, John Goldthorpe, David Halle, nomic trends in the Anglo-American democ- Kevin Leicht, Scott Long, Michael Rempel, racies during the 1980s, however, has led to Adrian Raftery, Paul Sniderman, David Weakliem, the ASR Editors (Paula England and a revival of the classical view (Goldthorpe Glenn Firebaugh), one Deputy Editor, and three 1982, 1995; Derber, Schwartz, and Magrass anonymous reviewers for their comments and 1990; Reich 1991; Brint 1994). Analyzing suggestions. the ways in which the economic dependence

American Sociological Review, 1997, Vol. 62 (April:191-208) 191 192 AMERICAN SOCIOLOGICAL REVIEW of middle-class professionals and managers Democratic increased by nearly a full logit gives them a material stake in the status quo, (an increase from a probability of .35 to .60). these analysts suggest that only a relatively In sum, professionals have moved from be- small fraction of the middle class is suscep- ing the most Republican class in the 1950s tible to liberal or left political alignments. to the second most Democratic of the six In our previous analysis of class voting in classes we analyzed by 1992 (see Hout et al. the United States since World War II (Hout, 1995). Brooks, and Manza 1995), we found evi- These findings have significant implica- dence of a significant Democratic voting tions for the debates about the political align- trend since the 1960s among one core ments of the middle class. Managers' voting middle-class segment-professionals-but behavior situates them in a stable, conserva- not others. Those results raise importantbut tive alignment with the Republican Party, as as yet unanswered questions about the thesis the classical view would suggest. Profession- of conservative retrenchment among the als, however, have undergone a realignment middle class. The main panel of Figure 1 with the Democratic Party since the 1960s. summarizes these earlier findings. The trend The latter development has been missed in lines presented in this figure are the pre- scholarly debates over middle-class politics dicted logits of choosing the Democratic in the United States because class has largely over the Republican presidential candidate, disappeared from analyses of voting, with and they are derived from a model of vote research instead focusing instead on policy choice that has been fitted to data from the attitudes or participation in social move- National Election Studies. Insofar as this ments. model documents but does not explain those In this paper, we develop a systematic trends, we term it the "initial" model. The analysis of the voting patternssummarized in solid line in the main panel shows the Demo- Figure 1. We address two questions: (1) What cratic turn among professionals; the dashed factor(s) account for political realignment line illustrates the absence of any such trend among professionals after 1968? (2) What among managers during the same period.' factor(s) account for the electoral differences Whereas managers remain a dependable Re- between professionals and managers that publican constituency in presidential elec- have emerged during this period? The tions, professionals increasingly support smaller, embedded chart in Figure 1 summa- Democratic candidates. rizes the direction in which our analyses pro- Taking into consideration that the y-axis in ceed. The angled trend line in this chart Figure 1 represents the log-odds of choosing shows the Democratic voting trend among the Democratic candidate, the trend line for professionals for the critical 1972 to 1992 professionals crosses the zero-point on the x- period. As before, this trend line has been axis (which translates into a probability of derived from our initial model of middle- .5) in 1972, showing that this election year class voting (which includes no explanatory marks the point after which professionals are variables). The two remaining bold lines, by predicted as moving decisively away from contrast, are derived from the preferred their historic pattern of support for Republi- model which includes explanatory variables can candidates. The magnitude of the Demo- (discussed later in this paper). The bold line cratic voting trend among professionals can for professionals' vote choice is now flat and be gauged by examining the difference be- congruent with the bold, dashed line for tween the logits for the first and last elections managers, showing that the additional effects in the series: Between 1948 and 1992, pro- parameterized in our preferred model have fessionals' (predicted) tendency to vote explained away both professionals' voting trend and their electoral differences with I Figure 1 presents the aspect ratio (the height managers. of the rectangle enclosing the data divided by its In the remainder of the paper, we present width) that banks to 450 the lines representing professionals' voting trend. Banking to 45? en- the steps by which we arrive at the preferred ables researchers to judge rates of change accord- model's estimates and discuss the answers ing to a non-arbitrarystandard (Cleveland 1994, they provide to our two research questions. section 2.4). In the first part of the paper, we review re- MIDDLE-CLASS POLITICAL REALIGNMENT 193

1.0 1.0

CURRENT .8 ANALYSIS .8 .6 PREVIOUSANALYSIS

C .4 Professionals O ~ .4 (initalmodel)

.2 o .4 Professionals (preferredmodel) Professionals ______(inital m odel) (U0 ?Managers.2 E (preferredmodel) 4) -.2 I I I I I I 1972 1976 1980 1984 1988 1992

0X 0 ______

*0 W ------0

..2

-4 Managers (initial model)

-6 1948 1952 1956 1960 1964 1968 1972 1976 1980 1984 1988 1992 Election Year

Figure 1. Maximum-Likelihood Estimates of Vote Choice in U.S. Presidential Elections, by Election Year: Professionals and Managers, 1948 to 1992 cent debates about middle-class political THEORIES OF MIDDLE-CLASS alignments in the United States, focusing on POLITICAL ALIGNMENTS four accounts that yield competing hypotheses about the causal mechanisms explaining Explanations of political behavior among so- these outcomes. In the second section of the cial groups or classes typically identify (1) a paper, we use data from the National Elec- particular demographic or social structural tion Studies to evaluate a series of logistic attribute, (2) an economic attribute relating regression models embodying these compet- to calculations of self- or societal-interest, or ing hypotheses. Once our preferred model (3) an ideological attributerelating to politi- has been chosen, we use a regression decom- cal attitudes or policy preferences to account position to make inferences about the rela- for phenomena observed by analysts (see tive contribution of each factor to explaining Dalton and Wattenberg 1993; Manza, Hout, professionals' voting trend and their differ- and Brooks 1995). While these explanations ences in vote choice relative to managers. In are not mutually exclusive, most accounts of the concluding section of the paper, we dis- middle-class politics do in fact tend to make cuss the relevance of our findings to under- claims about the causal priority of a particu- standing the nature and evolution of political lar type of mechanism. Our analysis evalu- alignments among the middle class. ates four theoretical approaches that yield 194 AMERICAN SOCIOLOGICAL REVIEW competing hypotheses about middle-class ing professional occupations (Brint 1984, political alignments. 1994). Since the late 1970s, however, Brint (1994:115ff.) argues that these same "multiple cumulative trends" worked in reverse, Thesis The "New Class" leading to a renewal of conservatism among The most controversial interpretation of most professionals as well as the rest of the middle-class politics, and hence a key stand- population. While noting the persistence of point in recent scholarly debates, is the "new "social issue" liberalism among profession- class" thesis (for historical overviews, see als-both in comparison to managers and to Szelenyi and Martin 1988; Gouldner 1979: the rest of the population-Brint argues that 94-100). This theory hypothesizes that grow- the rising salience of economic issues in the ing segments within the middle class are re- recent period has made professionals more ceptive to liberal or left-wing ideas and are politically conservative (1994:14, 106-107). disposed to participate in struggles for social The picture of middle-class voting implied change on the basis of their material interests by the cumulative trends thesis is thus one in in, or ideological commitment to, protecting which an initial affinity with Republican can- and expanding the public and nonprofit sec- didates is (temporarily) displaced during the tors of the economy. The new class thesis sug- 1970s, only to return as a result of composi- gests two complementary explanations of tional shifts and the emergence of conserva- middle-class political behavior: a material tive economic self-interest during the 1980s. interest-based and a symbolic interest-based interpretation. The materialist interpretation The Economic Voting Thesis implies that support for state-building outcomes, and hence the most consistent support A third account of middle-class politics ex- for Democratic candidates, comes from those emplifies the economic approach to under- segments of the middle class that would ben- standing political behavior. Economic mod- efit most from these types of developments: els of voting focus on the expected costs and people employed in the public or nonprofit benefits attached to the outcomes of electoral sectors (Bruce-Briggs 1979; Lamont 1987). choice and voters' economic assessments of The symbolic interpretationimplies that in- candidates and parties (Downs 1957; Kiewiet sofar as Democratic voting among middle- 1983; Hibbs 1987). Some theorists focus on class voters will also be driven by pro-state voters' personal economic situations ("ego- attitudes, such attitudes might explain recent centric voting") or the political consequences voting trends among professionals (an argu- of perceptions of the economy as a whole ment broadly consistent with Gouldner ("sociotropic voting"). Other analysts point 1979). to voters' evaluations of past economic performance. Finally, some analysts emphasize prospective evaluations of the consequences Trends Thesis The Cumulative of parties' and candidates' beliefs about fu- In a series of studies, Brint (1984, 1985, ture economic performance (Kinder and 1994) examined the political beliefs and be- Kiewiet 1981; Haynes and Jacobs 1994). The havior of U.S. professionals, comparing partisan effects of economic well-being, es- them with managers and the rest of the popu- pecially frustration with the status quo lation in the post-World War II period. He caused by declining personal or occupational argues that liberal political attitudes among group fortunes among segments of the professionals in the 1960s and early 1970s middle class (see Ehrenreich 1989; Phillips were the result of changes in their demo- 1993; Greenberg 1995) can be expected to graphic composition during that period. hinge on which party is in control of govern- These changes included: (1) growing oppor- ment. Insofar as Republicans controlled the tunities for employment in the public and presidency during five of the six election nonprofit sectors; (2) diversification of the years considered in the currentanalyses, eco- professions with respect to race and gender; nomic frustrations may explain increases in and (3) an increased proportion of younger support for Democratic candidates during people without family responsibilities enter- these elections. MIDDLE-CLASS POLITICAL REALIGNMENT 195

The New Political Culture Thesis quential, and we do not consider them in the current analyses.2 The final approach to middle-class political In line with our own and others' recent alignments we consider focuses on changes work on class voting (Heath, Jowell, and in attitudes toward "social issues" among the Curtice 1985; Heath et al. 1991; Hout et al. middle class as a source of political cleav- 1995; Brooks and Manza forthcoming) we age. Social issues relate to civil liberties, tol- distinguish six class categories:3 (1) profes- erance of cultural, racial, and ethnic diver- sionals (salaried and self-employed); (2) sity, and gender equality. The increasing managers and administrators; (3) owners, prominence of such issues has led a number proprietors, and other nonprofessional self- of analysts to hypothesize the emergence of employed persons (including farm owners); a "second" left, whose liberal views on so- (4) routine white-collar workers (retail sales, cial issues draw them to support liberal or clerical, and white-collar service workers); left political parties (Lipset [1960] 1981: (5) craft workers and foremen in all indus- 503-23; Ladd 1976-77; Edsall 1984:49-64; tries; and (6) semiskilled and unskilled work- Zipp 1986; Weakliem 1991). Following ers in all industries (including farming and Clark and Inglehart (1988) and Clark (1994), services). Our analyses focus on two key seg- we label this interpretationthe "new politi- ments of the contemporaryU.S. middle class: cal culture" thesis. working professionals (salaried and self-em- In his elaboration of this argument, Clark ployed) and managers (salaried only).4 (Clark and Inglehart 1988; Clark 1994) hypothesizes that a growing concern with social issues in the United States and other Data Western polities-especially among seg- We use National Election Study (NES) data ments of the middle class-has led to a new for presidential elections from the 1972, political culture. This culture is distinctive 1976, 1980, 1984, 1988, and 1992 pre- and in (1) emphasizing individualism; (2) favor- post-election surveys (Center for Political ing the relative priority of social over eco- Studies 1995). The NES is the premier source nomic issues; and (3) conjoining fiscal mod- of U.S. voting data, and its rich battery of at- eration with liberalism on social issues titudinal, occupational, and demographic (Clark and Ferguson 1983, chap. 7; Inglehart 1990: 302). Increasing support for 2 A separate analysis of the bases of the sig- the Democratic candidate among profes- nificant (19 percent) Perot vote in 1992 revealed sionals may be linked to their growing sup- no differences in support for Perot between pro- port for social issue liberalism. fessionals and managers (these analyses are available from the authors upon request). 3 These categories are a modified version of DATA, VARIABLES, AND MODELS Erikson and Goldthorpe's (1992) five-class The new class, cumulative trends, economic scheme. The principal occupational titles coded voting, and new political culture theses iden- as "professional" include accountants, architects, computer programmers, engineers, statisticians, tify a host of causal mechanisms that may physical scientists, physicians, dentists, teachers, potentially explain the Democratic voting lawyers and judges, librarians, reporters, writers trend among professionals and their increas- and editors. Additional details about coding ing electoral divergence from managers. Our schemes are available from the authors on re- empirical analyses incorporate competing quest. claims about these mechanisms into statisti- 4 We limit our analysis to professionals and cal models of voting behavior. Throughout managers for several reasons. First, although these analyses, our dependent variable is some employment conditions of routine white- vote choice in presidential elections, coded collar workers are similar to those of profession- or there are also "1" als managers, many important if a respondent votes for the Democratic differences that place them, at best, in a "contra- candidate, and "0" if (s)he votes for the Re- dictory class location" (Wright 1978, 1985). Rou- publican candidate. Votes for independent or tine white-collar workers typically do not possess third-party candidates during the 1972 to the same degree of autonomy and discretion in 1988 period have been relatively inconse- their work as do professionals and managers 196 AMERICAN SOCIOLOGICAL REVIEW items makes it an ideal data set for our pur- reference category of these dummy variables poses. The items requiredfor our analysis are is the assessment "better than a year ago." available beginning with the 1972 election Negative economic assessments can provide study, and thus our analysis is limited to the individuals with reason to vote against the period including the 1972 through the 1992 incumbent president and to test whether presidential elections. However, because those occurring during Republican presiden- 1972 marks the point at which professionals' cies explain the Democratic voting trend voting trend matures (with professionals' ex- among professionals, we coded the item pected probabilityof favoring the Democratic measuring dissatisfaction to apply to the five Party candidate crossing the .5 threshold), the election years in which the incumbent presi- available data provide a sufficient basis for dent was a Republican (1972, 1976, 1984, making inferences about the phenomena un- 1988, and 1992). der investigation. The next block of variables relates to the ideological bases of politics, and specifically to attitudes towards two types of issues: Variables views of the state and views of social issues. We have pooled the NES items in our analy- The first state item refers to an ideal-typical ses into a single data set in which election attribute of the modern welfare state-the year is itself a variable, allowing us to di- provision of jobs and a guaranteed standard rectly analyze trends in vote choice. Our of living. This item allows us to evaluate the middle-class segment variable is coded "1" "symbolic" version of the new class thesis. for professionals and "2" for managers. The If this version of the thesis is supported, the sociodemographic covariates in the analysis six dummy variables for the 7-point welfare are gender (female = 1), race (African state item should help explain the differences American = 1), birth cohort (1960s = 1), edu- in vote choice between professionals and cation (BA or more = 1), region (South = 1), managers as well as the Democratic realign- and employment sector (public/nonprofit = ment among professionals. A second state 1). These variables are all coded as dummy item asks whether the federal government variables. Means on several variables (e.g., is-as predicted by the "new political cul- gender, cohort, and region) show large ture" thesis-viewed by citizens as too pow- changes over this period, underscoring the erful. We use this dichotomous item to evalu- possibility that changes in the composition of ate a claim advanced by the new political the two segments may explain trends and dif- culture thesis, that the relationship between ferences in the voting behavior of profession- social liberalism and views of governmental als relative to managers after 1972. power is a negative one. We analyze two variables as indicators of The next block of variables measures atti- the economic bases of voting behavior.These tudes towardthree social issues: abortion, the two variables are dichotomous; they are ex- civil rights movement, and gender equality. tracted from a trichotomous item asking re- The abortion item is coded "1" for the pro- spondents to assess their current economic choice response and "O" otherwise.5 The situation in comparison to the past year. The 5 The abortion item is the one item in our analyses whose question wording varied between (Goldthorpe 1982). Second, although the self-em- the 1972 to 1992 surveys. Between 1980 and ployed are often considered part of the "middle 1984, the NES changed the wording of the item's class" in American political discourse (e.g., the four response categories. To minimize any mea- "old middle class" in Mills's [1951] scheme), we surement biases introduced by this change we prefer to maintain a clear distinction with self- recoded the item as a dichotomous contrast be- employed nonprofessionals who are directly ex- tween the most pro-choice position versus all posed to market pressures in their work. Results other positions. While the results of the analyses from our earlier application of Goodman's (1981) are congruent with results obtained by omitting collapsibility test supports this distinction among the abortion item from the analyses, we retained nonprofessionals, but indicates that self-em- the item because of the substantive relevance of ployed and salaried professionals can be com- reproductive rights to the idea of a new political bined into a single class category (Hout et al. culture concerned with individual rights and au- 1995). tonomy. MIDDLE-CLASS POLITICAL REALIGNMENT 197 civil rights item is a three-category item ask- choice are mediated by a change in partisan- ing whether the civil rights movement is ship, it suggests a more lasting political re- moving too fast, about right, or too slow, alignment. while the gender equality item is a seven- To this end, we first fit a preferred model point scale asking whether women and men of middle-class vote choice that omits the should have equal roles in the family and two partisanship items in order to estimate workplace. We analyze these respectively as the total (direct + indirect) effects of our ex- a pair and as a set of six dummy variable planatory factors. We then refit the preferred contrasts. Change in the marginal distribu- model with the partisanship items, deriving tions of responses to these social issue items estimates of the (direct) contribution of our is large and moves in a liberal direction, earlier factors and of the partisanship factors highlighting the possibility raised by the new to explaining trends and differences in vote political culture thesis, that attitudes toward choice among the middle class. This proce- social issues explain trends in, and differ- dure allows us to make inferences about the ences between, the voting behavior of pro- extent to which the effect of the causal fessionals and managers. mechanisms of middle-class voting are me- Two final items in our analyses relate to diated by (changes or differences in) parti- two separate dimensions of partisanship:par- sanship among professionals and managers. tisan affect (the tendency to positively or negatively evaluate each major party), and Logistic Regression Models of Vote Choice party identification (the tendency to identify with one [or neither] of the two major par- We choose a logistic specification because of ties). Our measure of partisan affect is di- our dichotomous dependent variable. By chotomous, coded "1" if respondents give comparing models that incorporate various any positive evaluations of the Democratic combinations of independent variables and Party in the NES's open-ended party evalua- different parameterizations of their relation- tion items and "0" if they do not (see Watten- ship to the dependent variable, we test whe- berg 1994 for related work). Our second ther and in what manner the sociodemo- measure is the classic Michigan seven-point graphic, economic, and ideological attributes scale that taps the strength and direction of of professionals and managers affect their party identification. As with our other poly- voting behavior. chotomous items, we treat the party identifi- The dependent variable is the log of the cation item as a set of six non-redundant odds of choosing the Democratic over the dummy variable contrasts for the seven cat- Republican presidential candidate, which we egories (these range from strong Republican designate by Oij for vote choice j (j = 1 for to strong Democrat). the Democratic candidate and 0 for the Re- The two partisanship items provide addi- publican) for person i. The first model in our tional information about the nature and sta- analyses is a simple, one-parameter model bility of middle-class political (re)align- that hypothesizes a voting trend among pro- ments. If the Democratic voting trend among fessionals but not among managers. Model 1 professionals has led to a stable partisan re- is summarized in equation 1: alignment, the inclusion of the two items in = a+ the model should explain a sizable portion of oi] OkjCik Ii (1) their voting trend (and their growing differ- where Okj is the linear trend parameter, Cik is ences with managers). This is a critical issue a dummy variable for the two middle-class in light of the centrality of partisanship to segments (k = 1 for professionals; 2 for man- voting behavior (Campbell et al. [1960] agers) and Yi1is a fixed score ranging from 1 1980; Miller and Shanks 1996), for if pro- to 6 for the six elections years (1 = 1 for fessionals' voting trend has occurred inde- 1972, 2 for 1976, . . . 6 for 1992). These pendently of their partisan attitudes, it sug- election year scores constrain the interaction gests a far less durable attachment to (and between middle-class segment and election thus realignment with) the Democratic Party. years to be linear (in their logits). In Model By contrast, if we find that the effect of the 1, we set k equal to 1 to restrict the voting causal factors explaining middle-class vote trend parameterto professionals only; the Oj 198 AMERICAN SOCIOLOGICAL REVIEW parameter thus represents the assumption tributes on political behavior explains the that the effect of being a professional (k = 1) trend among professionals and their overall on choosing the Democratic candidate (j = differences with managers in Model 4. The 1) increases at a constant rate (in logits) over comparison of Models 3 and 4 tests whether the six election years. the three-way interaction terms (between Our second model is a nested extension of segment, election year, and the sociodemo- the first, and it hypothesizes the existence of graphic variables) contained in the latter an additional voting trend pertaining to the model result in an improvement in fit, indi- managerial segment of the middle class. This cating that the political salience of social at- model includes a new parameterfor a linear tributes has increased (or decreased) among trend in managers' vote choice over the 1972 professionals (relative to managers). Our re- to 1992 period. The comparison of Models 1 maining models (5 through 8) add terms rep- and 2 allows us to corroborate our earlier resenting compositional and salience effects finding (Hout et al. 1995) that managers' for economic, ideological, and partisanship vote choice reflects a stable Republican attributes. alignment (and hence that 02j = 0). The preceding models provide a compre- Using Model 3 we evaluate whether hensive basis for evaluating hypotheses changes in the sociodemographic composi- about the mechanisms of middle-class vot- tion of the middle class explain the Demo- ing. In our analysis, we choose a preferred cratic voting trend among professionals and model using the -2 log likelihood (-2LL) their overall differences with managers. Note statistic and Raftery's (1995) Bayesian Infor- that because Model 1 turns out to be supe- mation Criterion (BIC). For a pair of nested rior to Model 2, we use its 1 degree of free- models, we interpreta complex model result- dom parameterizationof a linear Democratic ing in a significant improvement in fit over voting trend among professionals. Model 3 the simpler model (using the partitioned is summarized in equation 2: -2LL statistic) as a necessary but insufficient M condition on accepting that model. To accept 'D= Ca+ OkjCik Y1 +X Tt mWitl, t (2) the more complex model, it must also show 1n2=1 significant improvement in fit according to where LDijis the logit for Democratic vote BIC, which is calculated for logistic regres- choice, the Wi,,,(m = 1, . . ., ) are the six sion models as D - (d.f.)ln(N), where D is dichotomous sociodemographic variables the residual deviance (i.e., -2LL) for the and the T,,j are the six new parametersto be model under consideration, d.f. is its degrees estimated. of freedom, In is the natural logarithm, and In addition to changes in composition, a N is the sample size. Negative values of BIC second way in which social structural, eco- indicate a potentially acceptable fit, with nomic, or ideological forces can affect vot- models with larger negative values being ing is by becoming increasingly salient dur- preferable to those with BIC values closer to ing the time period of interest. Changes in 0 (or positive BIC values). Because BIC the political salience of an attribute implies takes sample size and a model's degrees of that the association between a particularitem freedom into consideration, it provides the and vote choice increases (or decreases) over necessary supplementto the practice of mini- time. Note that this type of change in sa- mizing the residual deviance (i.e., sole reli- lience can occur even if the marginal distri- ance on the -2LL test), which can result in bution of a particular item does not change the choice of unparsimonious models and during the time period in question (e.g., the hence Type I errors for large samples (Wong association between being female and voting 1994; Raftery 1995). Democratic can increase over time even if the proportion of women is unchanged). We Regression Decomposition parameterize change in political salience as an interaction effect between election year After choosing a preferredmodel, we use the and the attributein question. model's coefficients to make inferences We investigate the possibility that salience about the relative causal importance of the changes in the effect of demographic at- various factors under examination. With re- MIDDLE-CLASS POLITICAL REALIGNMENT 199

Table 1. Fit Statistics for Logistic Regression Models of Change in Middle-Class Voting

Fit Statistics -2 Log- Degrees of Model Descriptiona Likelihood Freedom BIC 1 Linear voting trend among professionals 1,246.44 939 -5,183 (professionals *vote choice*election year) 2 Linear voting trend among managers 1,246.13 938 -5,176 (manager*vote choice*election year) 3 Sociodemographic variables (gender, race, cohort, 1,181.72 933 -5,206 education, region, and employment sector) 4 Change in salience of sociodemographic variables 1,160.24 918 -5,125 (professional x election year x race, gender, and public sector) 5 Ideological and economic attributes (views of the state 968.52 915 -5,296 and social issue items, economic dissatisfaction) 6 Restricted change in salience of ideology (professionals* 957.59 914 -5,301 gender equality item [7]*election year [1992]) 7 No linear voting trend among professionals 957.89 915 -5,307 (Model 6 minus professionals*vote choice*election year) 8 Partisanship (party identification and partisan affect) 587.96 908 -5,629

Note: BIC for the null model (constant only) is -5,166; N = 941. a Linearly constrained interaction effects are designated by an asterisk (*); unconstrained interaction effects are designated by a times sign (x). gard to our first question, we want to know RESULTS to what extent changes in the marginal distribution of a particular variable-or change Table 1 presents fit statistics for the eight in its interrelationship with the dependent models. Model 1 posits a linear voting trend variable-can explain the 1972 to 1992 for professionals. The comparison with the Democratic voting trend among profession- null model (including only a constant) shows als. With regard to our second question, we decisive improvement (BIC difference = want to know to what extent differences in -17), providing conclusive evidence for a the distribution of the independent variables voting trend among professionals. Model 2, among professionals versus managers can however, is not even remotely close to im- explain their overall voting gap during the proving on Model l's fit, demonstrating the 1972 to 1992 period. absence of any trend in managers' vote To answer these questions, we use two re- choice during the 1972 to 1992 period. These gression decompositions (Teixeira 1987; results corroborateour earlier findings (Hout Firebaugh 1997). In the first of these, we de- et al. 1995) for managers and professionals.6 compose the 1972 to 1992 voting trend We compare Models 3 and 4 with Model 1 among professionals into its components. to evaluate the evidence for changes in the The second decomposes the differences be- effect of sociodemographic factors on tween professionals' and managers' vote middle-class vote choice. BIC handily se- to 1992 choice during the entire 1972 period. 6 For both decompositions, we multiply the We also evaluated a second model of voting that imposed no con- mean for the relevant group in the trends among professionals sample straints on the segment by year interaction. The analysis by the appropriateregression coef- comparison with Model 1 (BIC difference = 25) ficient. By converting this product into a per- decisively favored Model 1, implying that the centage of the total predicted logit, we can evolution of professionals' voting during the directly compare the respective contribution 1972 to 1992 period is best understood as a linear of each explanatory factor. trend. 200 AMERICAN SOCIOLOGICAL REVIEW

lects Model 3 over both Models 1 and 4, pro- sanship-based natureof professionals' politi- viding decisive evidence for the main but not cal realignment in the 1972 to 1992 period. the interaction effects of the sociodemographic variables. Model 5's improvement Decomposing the Sources of Trends and over Model 3 (BIC difference = -90) dem- in onstrates that economic dissatisfaction and Differences Middle-Class Vote Choice attitudes towards state and social issues have The factors parameterized in our preferred a significant impact on middle-class vote models fully explain away the trend in pro- choice. Model 6, which includes an addi- fessionals' vote choice and their differences tional parameter for salience change among from managers. We now enlarge upon these professionals in the gender item's most lib- analyses by decomposing the two sets of re- eral response category, in turn improves the gression results. We consider first the decom- fit of Model 5.7 position of the regression results for explain- Model 7 is derived from Model 6 by delet- ing professionals' Democratic voting trend, ing the interaction term for the linear voting using the gender item's coefficient for Model trend among professionals. The strong im- 7 (.168) to illustrate this procedure.9 The provement in fit (BIC difference = -6) shows sample mean for the gender item in 1992 for that the voting trend among professionals professionals is .377. Multiplying the coeffi- (and also their differences from managers) is cient by this mean (.168 x .377) yields a explained away by the independent variables product of .063. We then multiply the coeffi- in the model.8 Model 7 is thus our preferred cient by the professionals' mean in 1972 model that does not control for the effects of (.168 x .415) to obtain the second product, partisan attitudes and we use its coefficients .070. Subtracting the second product from to estimate the (total) contribution of each the first, we obtain a rounded estimate of sociodemographic, economic, and ideologi- -.01, which is the estimated effect (in logits) cal variable to explaining trends and differ- that change in the gender composition of ences in middle-class vote choice. However, professionals has on their likelihood of the decisive improvement in fit of Model 8 choosing the Democratic candidate. We ap- over Model 7 demonstrates that the partisan- ply this type of calculation for each of the ship variables also have a significant impact independent variables to obtain the results on vote choice. Model 8 is thus our preferred presented in column 1 of Table 2. model that controls for partisanship, and we The total predicted logit change in profes- use its coefficients to reestimate the (direct) sionals' vote choice from 1972 to 1992 is sub- contribution of each variable to explaining stantial-.51 (in logits). Of this total, (liberal) trends and differences in middle-class vote change in professionals' gender attitudes ex- choice. By comparing the two sets of esti- plains by far the largest share of any variable mates obtained under Models 7 and 8, we (84 percent). The effects of changes in abor- can answer the causal question of whether tion attitudes (.15) and views of the civil the political effects of sociodemographic, rights movement (.19) explain respectively 29 economic and ideological factors are medi- percent and 37 percent of the total change in ated by corresponding changes in partisanat- vote choice. Takenas a whole, liberal changes titudes. This enables us to derive an informa- in professionals' attitudes toward social is- tive picture of the issue-based versus parti- sues thus explain 151 percent of the total logit change. This estimate exceeds 100 percent for 7 Allowing all state and social issue items to a simple reason: The combined effect of the interact with the election years yielded an infe- sociodemographic variables (and also the rior model relative to Model 6 (BIC difference = welfare state variables) predict a substantial 449). turn away from Democratic candidates dur- 8 This outcome is summarized in the embed- ing the 1972 to 1992 period. On the basis of ded chart in Figure 1, which shows the sharp contrast between the Democratically inclined trend line for professionals in the initial model versus 9 Due to space considerations, we do not the flat trend lines for both for professionals and present the coefficients or sample means for pro- managers under the preferred models (Models 7 fessionals and managers. These are available and 8). from authors on request. MIDDLE-CLASS POLITICAL REALIGNMENT 201

Table 2. Logistic Regression Decomposition for Explaining Trend in Vote Choice among Profession- als, 1972 to 1992

Percent of Change ( 1) Explaineda Logit Change in (2) (3) Independent Variable Vote Choice Model 7 Model 8

Sociodemographic Variables Change in percent female -.01 -2 0 Change in percent African American -.05 -10 -1 Change in percent in 1960s' generation .00 0 -2 Change in percent college graduate -.03 -6 -1 Change in percent living in South .00 0 -I Change in percent employed in public/nonprofit sector -.10 -20 -5 Total (all sociodemographic variables) -.19 -37 -10

Economic Evaluations Change in level of economic satisfaction .03 6 1

View of the State Change in attitudes toward government power .05 10 I Change in attitudes toward welfare state -.15 -29 -4 Total (all state variables) -.10 -19 -3

Social Issues Change in abortion attitudes .15 29 5 Change in civil rights attitudes .19 37 12 Change in gender attitudes .43 84 34 Total (all social issue variables) .77 151 51

Partisanship Change in Democratic Party affect - 8 Change in party identification 55

Total Change in Vote Choice .51 99 101

aPercentages do not sum precisely to 100 due to rounding. sociodemographic changes among profes- changes in professionals' attitudes on social sionals alone, we would predict a -.19 logit issues were more than sufficient to nullify any drop in levels of Democratic Party support such reversal. Indeed, had only these liberal (representing -37 percent of the total logit ideological changes occurred, professionals' change). Likewise, with regard to their Democratic voting trend would have been changing views of the welfare state, we substantially larger (.77). would predict a -. 15 logit drop in profession- Of the remaining independent variables, als' supportfor Democratic candidates.These the economic variables have a modest, posi- results establish that if only changes in tive impact, explaining 6 percent of the over- sociodemographic composition and welfare all trend. The corresponding estimate for state attitudes occurred, the Democratic vot- change in professionals' views of the power ing trend up to 1972 would have reversed it- of the federal government is 10 percent. This self, leading to a reconstitutionof profession- result establishes that increasingly favorable als as a solidly Republican constituency (as views of state power led professionals to sup- Brint's cumulative trends argumentsuggests). port the more liberal of the two major par- However, the combined effects of liberal ties. By contrast, the -15 logit/-29 percent 202 AMERICAN SOCIOLOGICAL REVIEW estimates for the welfare state item reveal a cratic presidential candidates, in part because trend away from support for the welfare greater social liberalism has disposed profes- state. During the same period in which pro- sionals to identify with, and endorse more fessionals have become more liberal on so- positive views of the Democratic Party. So- cial issues and less hostile to government in cial issue liberalism has thus exerted a pow- general, they have also become more conser- erful effect on vote choice, part of which is vative with respect to social provision. direct and part indirect (mediated by changes To this point, our results do not analyze the in partisanship).Although it is logically pos- role that partisanship may play in explaining sible to reverse the alternative causal infer- professionals' vote choice. In column 3 of ence (i.e., changing partisanship caused so- Table 2, we use the coefficients from Model cial liberalism to increase among profession- 8 to reestimate our decomposition to take into als), this would raise the additional question account the effects of professionals' party of what force was responsible for profession- identification and their partisanaffect toward als' shift in partisanship in the first place. the Democratic Party. Column 3 presents the Notice that this additional factor would have percentage of the total logit change that each to explain not only the partisan shift, but also independent variable explains net of change the relationship between social liberalism in partisan attitudes. By comparing columns and vote choice among professionals. In light 2 and 3, we thus measure the extent to which of professionals' conservative shift in their the sources of professionals' voting trend ac- views of the welfare state (and the modest cording to Model 7's results are mediated by relationship between their changing voting changes in partisanship attitudes. behavior and personal economic evalua- Most of the effects of economic and tions), it is impossible to construe these fac- sociodemographic change (as well as chang- tors as capable of explaining any Democratic ing views of the welfare state) are mediated political trends among professionals. Our in- by partisan change. Whereas the sociodemo- ference, by contrast, provides a succinct and graphic variables explain -37 percent of the compelling answer to the causal question: total predicted change in vote choice not Professionals' change in voting behavior is controlling for partisanship, they explain the result of their increasingly liberal views only -10 percent when partisanshipis taken of social issues.10 into account in Model 8. Likewise, the -29 Turning now to our second research ques- percent estimate for the welfare state vari- tion, we present in Table 3 the results of our able declines to -4 percent when the parti- decomposition of the average differences in san variables are included in Model 8. Over- the voting behavior of professionals versus all, change in partisan attitudes, due in large managers during the entire 1972 to 1992 pe- part to the increased proportion of profes- riod. Differences in demographic composi- sionals who identify themselves as "strong tion explain 17 percent of their overall vot- Democrats," explains 55 percent of the total ing differences in Model 7 (again not con- change in vote choice. Because changing at- trolling for partisanattributes). The main fac- titudes toward social issues account for just tors behind professionals' greatertendency to over 50 percent of the total predicted logit vote Democratic over the 1972 to 1992 pe- change in Model 8, a substantial portion of riod are their higher percentages of African their effect (in Model 7) remains unex- Americans and public/nonprofit sector em- plained, having occurred entirely outside the ployees (partially counterbalanced by their realm of partisanship. These results suggest higher levels of education which lead to sup- that the incorporation of social issues into port for the GOP). Although sociodemo- political partisanship has been incomplete, to the especially when compared congruent 10 patterns of change in welfare state and parti- Space considerations do not permit us to present or fully discuss our analysis of a related san attitudes. question, the origins of social issue liberalism The preceding results imply the following among professionals. This research shows that causal picture of professionals' realignment. sociodemographic and economic factors do not Increasingly liberal views on social issues explain professionals' increasingly liberal atti- have led to increasing support for Demo- tudes in the 1972 to 1992 period. MIDDLE-CLASS POLITICAL REALIGNMENT 203

Table 3. Logistic Regression Decomposition for Explaining Differences in Vote Choice among Pro- fessionals and Managers, 1972 to 1992

Percent of Difference ( 1) Explaineda Logit Difference in (2) (3) Independent Variable Vote Choice Model 7 Model 8

Sociodemographic Variable Difference in percent female .02 3 1 Difference in percent African American .05 7 2 Difference in percent in 1960s' generation .00 0 -1 Difference in percent college graduate -.11 -16 -7 Difference in percent living in South .00 0 -1 Difference in percent employed in public/nonprofit sector .16 23 1 1 Total (all sociodemographic variables) .12 17 5

Economic Evaluations Difference in level of economic satisfaction .03 4 2

State Issues Difference in attitudes toward government power .01 1 0 Difference in attitudes toward welfare state .26 38 19 Total (all state variables) .27 39 19

Social Issues Difference in abortion attitudes .04 6 2 Difference in civil rights attitudes .12 17 11 Difference in gender attitudes .11 16 11 Total (all social issue variables) .27 39 23

Partisanship Difference in Democratic Party affect 9 Difference in party identification 43

Total Difference in Vote Choice .69 99 101

a Percentages do not sum precisely to 100 due to rounding.

graphic changes among professionals per se predicted difference in vote choice relative to predict a conservative shift in vote choice, managers. The reason for this is managers changes in managers' sociodemographic have also become less supportive of the wel- composition (especially the increase in the fare state over time. As a result, the relatively percentage of managers with college educa- higher level of welfare state support among tions since 1972) has occurred at a higher professionals has been left intact. While wel- rate, disposing professionals to be less Re- fare state attitudes are thus of importance for publican in their vote choices in comparison understandingpolitical differences within the to managers. middle class, economic assessments and at- Differences in attitudes toward the welfare titudes toward the power of the federal gov- state explain a sizable portion of the gap in ernment have little relevance to understand- vote choice. While professionals are far from ing the electoral differences between profes- favoring an egalitarian view of social policy sionals and managers. -their attitudes have, in fact, become less The greater prevalence of socially liberal egalitarian over time-their comparatively attitudes among professionals explains the more liberal views explain 38 percent of the largest portion of their voting differences 204 AMERICAN SOCIOLOGICAL REVIEW from managers (39 percent). In similar fash- fessionals) is emerging as an importantmain- ion to the earlier analyses, the gender equal- spring of support for the Democratic Party. ity and civil rights movement items are of We have systematically explored a number greater consequence than the abortion item. of competing hypotheses concerning the The similarity to the decomposition of vot- bases of middle-class political alignments, ing trends becomes even stronger when par- finding that liberal views on social issues tisan attitudes are taken into account in have to this point driven professionals toward Model 8: Attitudes toward the welfare state the Democratic Party.This novel result is dif- and sociodemographic attributes are again ficult to integrate within existing theories of disproportionately associated with partisan- middle-class politics, for previous studies ship factors. The percentage explained by the have largely ignored or downplayed the sig- sociodemographic variables and the welfare nificance of social issue liberalism.12For ex- state variables decline respectively from 17 ample, although Brint (1984, 1994) found percent to 5 percent, and from 38 percent to greaterlevels of social liberalism among pro- 19 percent. The direct effect of attitudes to- fessionals, he placed much greater weight on ward social issues, by contrast, still accounts conservative counter-trendsduring the 1980s for 23 percent of the total difference in vote in describing their political behavior. Jack- choice, with partisanship factors now ex- man and Muha (1984) go even further, argu- plaining the largest share (52 percent). ing not only that social liberalism among the middle class is of little political consequence, but also that liberal attitudes toward social DISCUSSION issues are only superficially held. In an analy- How the middle class votes is of increasing sis of the future of the Democratic Party, consequence for understanding key sources Kuttner(1987) asserts that "to the extent that of change in American politics. Post-indus- liberal positions on social issues have any ef- trial trends have enhanced the electoral sig- fect, they push white working-class voters to nificance of the professional and managerial the Republican column" (p. 112). segments of the electorate, as they have In contrast to such accounts, our analyses grown significantly as a proportion of the demonstrate that socially liberal attitudes population over the course of the past sev- have critical explanatory value in under- eral decades (Hout 1988; Reich 1991). Fur- standing important trends in middle-class thermore, they turn out to vote at a higher political behavior. Not only does social issue rate than the rest of the population (Wolfing- liberalism explain the recent trends and dif- er and Rosenstone 1980; Piven and Cloward ferences in the voting behavior of profession- 1988; Hout et al. 1995). Whether the middle als and managers, it has increased among class is a force for progressive change or a professionals in both magnitude and to a bastion of the status quo is thus a matter of more limited extent in political salience. We considerable importance for political parties can think of no firmer demonstration of the as. well as for the future of the "democratic specifically political importance of socially class struggle."1IYet far from being a homo- liberal attitudes than a strong association geneous constituency, the contemporary with vote choice in presidential elections. middle class is politically heterogeneous. Moreover, liberal views on social issues such One major segment (managers) remains a as abortion, women's roles, and civil rights solidly Republican bloc, while another (pro- are the principal reasons behind professionals' realignment with Democratic candidates I In related work (see Brooks and Manza and their widening differences vis-a-vis man- 1996), we have investigated further the signifi- agers. Without this specific ideological base cance of these developments for understanding of support, our analyses show that Demo- change in the nature of class politics. These analyses show that changing views of social issues affect most classes in the same way (with 12 In related work on religion and politics since the partial exception of professionals). As a re- the 1950s (see Manza and Brooks forthcoming), sult, the increasingly important political cleavage we have found that increases in liberal views of defined by social issues has not decreased the social issues have also been of consequence for overall magnitude of the class cleavage. some religious groups. MIDDLE-CLASS POLITICAL REALIGNMENT 205 cratic presidential candidates would have lost Table 4. Correlation Matrix of State and Social considerable ground among both segments of Issue Items for Professionals and Man- the middle class. Indeed, the net effect of agers marginal changes on the other independent Attitude Item variables in our analyses would have actually Class and Civil Gender Welfare led to a reversal of professionals' Democratic Attitude Item Abortion Rights Equality State voting trend in the post-1972 period. Of the four theories of middle-class poli- Professionals tics we have examined, only the new politi- Government .09* .15* .14* .13* cal culture thesis provides grounds for antici- power pating that attitudestoward social issues mat- Abortion .17* .33* .05 ter for understanding trends in middle-class Civil rights .24* .25* political alignments. Our analyses provide Gender equality - .13* general support for this claim. However, an important limitation of this thesis relates to Managers its claims about the relationship between at- Government -.01 .13* .12* .07 titudes toward social issues and attitudes to- power ward the state. Inglehart (1990) and Clark Abortion -.01 .26* .08 (1994) argue not only that Western publics Civil rights .21* .22* have become more liberal on social issues, Gender equality - .05 but also that the mechanisms producing this shift have led to a growing skepticism about Note: Based on pooled data from all six election the use of the federal government to solve surveys. social problems. In part, what makes the idea p < .05 (two-tailed tests) of a new political culture theoretically distinctive is the assumption that this "new" kind of liberalism is structurally incompat- suits for managers show slightly fewer sta- ible and at odds with the "old" sort typified tistically significant correlations, they too cut by the governmental activism and modest against the new political culture thesis, for egalitarianism associated with the New Deal. each of the latter is again positively signed. If this aspect of the new political culture This is not to deny that the middle class thesis were true, we would expect the rela- (or professionals in particular) are economi- tionship between liberal positions on social cally conservative, for inspection of changes issues and left positions on state/economic on the relevant item shows that since 1972 issues to be zero-sum. In other words, a so- they have become increasingly likely to op- cially liberal outlook should be associated pose the welfare state. Nevertheless profes- with a fiscally conservative/individualistic sionals' growing social liberalism and their view of the state/economy nexus. We exam- ebbing support for the welfare state do not ine whether this assumption is tenable in appear to be aspects of the same ideological Table 4, which presents pairwise correlations trend. Moreover, the positive sum results of among the five attitudinalitems in the analy- our correlational analysis suggest that the ses for managers and professionals. These "old" left and the "new" liberalism are com- correlations indicate that responses to the plementary (or at least not contradictory) in governmental power items are positively re- their relationship to one another (Brooks and lated to responses to the social issue items Manza 1994). among professionals: Endorsing the liberal response on a given social issue item is cor- and Partisan related with rejecting the claim that the state Social Issues Realignment has become too powerful. The same is true The concept of a political realignment (Key of the interrelationshipbetween responses to 1959; Nardulli 1995) provides a useful the welfare state item and responses to the means of understanding the long-term shift social issue items: Holding a liberal view on in voting behavior among professionals, a social issues is associated with greater sup- shift which is particularly remarkable when port for the welfare state. Although the re- viewed from the perspective of their solidly 206 AMERICAN SOCIOLOGICAL REVIEW

Republican voting pattern in the 1950s (see Jeff Manza is Assistant Professor of Sociology at Hout et al. 1995). Had professionals' voting Pennsylvania State University. His research in- trend occurred independently of change in terests are political sociology, class analysis, and partisan attitudes, we could view it as a social policy. In addition to his collaborative purely cultural phenomenon, perhaps inca- work with Brooks, he is conducting research (with Fred Block) on the feasibility and conse- realign- pable of supporting a more lasting quences of negative income tax regimes, and a ment. However, our analysis of the mediat- historical study of policy experts and political ing role of partisanship suggests that chang- change in the United States between 1932 and ing views of social issues have contributed 1950, an earlier version of which won the 1996 to change in both party identification and American Sociological Association Dissertation partisan affect. The mediation of the impact Award. of social liberalism on professionals' vote choice implies that attitudes on social issues REFERENCES are distinctive not only for their explanatory impact, but also for the way in which they Brint, Steven. 1984. "'New Class' and Cumula- have been partially absorbed by-and thus tive Trend Explanations of the Liberal Politi- have contributed to-new partisan align- cal Attitudes of Professionals." American Jour- ments. nal of Sociology 90:30-7 1. Our analyses also demonstratethat profes- 1985. "The Political Attitudes of Profes- sionals are an economically conservative sionals." Annual Review of Sociology 11:389- segment of the electorate whose increasingly 414. inegalitarianviews of the welfare state would 1994. In an Age of Experts: The Chang- otherwise have pushed them toward the Re- ing Role of Professionals in Politics and Pub- lic Life. Princeton, NJ: Princeton University publican rather than the Democratic Party. Press. Changes in their sociodemographic compo- Brooks, Clem and Jeff Manza. 1994. "Do Chang- sition would also have moved them back to a ing Values Explain the New Politics? A Criti- conservative alignment. These results high- cal Assessment of the Postmaterialist Thesis." light the importance of social issue liberal- Sociological Quarterly 35:541-70. ism, showing it to be the single mechanism . 1996. "Class Politics and Political that can effectively counteract the conserva- Change in the United States Since World War tive thrust of these previous factors. While II." Paper presented at the annual meetings of professionals report increases in personal the American Sociological Association, New York. reveal economic dissatisfaction, our analyses . Forthcoming. "Social Cleavages and Po- that their dissatisfaction has had a minimal litical Alignments: U.S. Presidential Elections, payoff for Democratic candidates. To date, 1960-1992." American Sociological Review. increased liberalism on social issues has Bruce-Briggs, B., ed. 1979. The New Class? San been sufficient to drive a wedge through the Diego, CA: HarcourtBrace Jovanovich. middle class and move professionals out of Campbell, Angus, Philip E. Converse, Warren E. their earlier Republican alignment. However, Miller, and Donald E. Stokes. [1960] 1980. The the political impact of social issues is an on- American Voter. Chicago, IL: The University going process, and one whose institutional- of Chicago Press. Center for Political Studies. 1995. American Na- within the system is as yet in- ization party tional Election Studies 1948-1994. [MRDF complete. and Codebooks]. Ann Arbor, MI: Inter-university Consortium for Political and Social Re- Clem Brooks is Assistant Professor of Sociology search, Institute for Social Research [distribu- at Indiana University, Bloomnington.His research tor]. interests are political behavior, values and pub- Clark, Terry N. 1994. "Race and Class Versus the lic opinion, and quantitative methods. He is work- New Political Culture."Pp. 21-78 in Urban In- ing with Jeff Manza on a series of studies of novation: Creative Strategies for Turbulent change in the social bases of U.S. parties in the Times, edited by T. Clark. Thousands Oaks, postwar era. He is also working on a study of CA: Sage. changing public views of liberalism since the Clark, Terry N. and Lorna C. Ferguson. 1983. 1960s, and is beginning work on a study of the City Money: Political Processes, Fiscal Strain, changing relationship between social problems and Retrenchment. New York: Columbia Uni- and partisan alignments in the United States. versity Press. MIDDLE-CLASS POLITICAL REALIGNMENT 207

Clark, Terry N. and Ronald Inglehart. 1988. "The The British Voter, 1964-87. London, England: New Political Culture: Changing Dynamics of Pergammon. Support for the Welfare State and Other Poli- Hibbs, Douglas. 1987. The American Political cies in Post-Industrial Societies." Paper pre- Economy: Macroeconomics and Electoral Pol- sented at the annual meetings of the American itics. Cambridge, MA: Harvard University Political Science Association, Washington, Press. DC. Hout, Michael. 1988. "More Universalism, Less Cleveland, William S. 1994. The Elements of StructuralMobility." American Journal of So- Graphing Data. Rev. ed. Summit, NJ: Hobart ciology 93:1358-400. Press. Hout, Michael, Clem Brooks, and Jeff Manza. Dalton, Russell J. and Martin P. Wattenberg. 1995. "The Democratic Class Struggle in the 1993. "The Not So Simple Act of Voting." Pp. United States, 1948-1992." American Socio- 193-218 in Political Science: The State of the logical Review 60:805-28. Discipline, edited by A. Finifter. Washington Inglehart, Ronald. 1990. Culture Shift in Ad- DC: American Political Science Association. vanced Industrial Society. Princeton, NJ: Derber, Charles, William Schwartz, and Yale Princeton University Press. Magrass. 1990. Power in the Highest Degree: Jackman, Mary R. and Michael J. Muha. 1984. Professionals and the Rise of a New Mandarin "Education and Intergroup Attitudes: Moral Order. New York: Oxford University Press. Enlightenment, Superficial Democratic Com- Downs, Anthony. 1957. An Economic Theory of mitment, or Ideological Refinement?" Ameri- Democracy. New York: Harper and Row. can Sociological Review 49:751-69. Edsall, Thomas. 1984. The New Politics of In- Key, V. O., Jr. 1959. "Secular Realignment and equality. New York: Norton. the Party System." Journal of Politics 17:3-18. Ehrenreich, Barbara. 1989. Fear of Falling: The Kiewiet, D. Roderick. 1983. Macroeconomics Inner Life of the Middle Class. New York: and Micropolitics. Chicago, IL: University of HarperCollins. Chicago Press. Erikson, Robert and John Goldthorpe. 1992. The Kinder, Donald R. and D. Roderick Kiewiet. Constant Flux: A Study of Class Mobility in In- 1981. "Sociotropic Politics: The American dustrial Societies. New York: Oxford Univer- Case." British Journal of Political Science sity Press. 11:129-61. Firebaugh, Glenn. 1997. Analyzing Repeated Sur- Kuttner, Robert. 1987. The Life of the Party. New veys. Thousand Oaks, CA: Sage Publications. York: Penguin. Goldthorpe, John. 1982. "On the Service Class, Ladd, Everett C. 1976-77. "Liberalism Upside Its Formation and Future." Pp. 162-85 in The Down: The Inversion of the New Deal Order." New Working Class, edited by R. Hyman and Political Science Quarterly 91:577-600. R. Price. London, England: MacMillan. Lamont, Michele. 1987. "CulturalCapital and the .1995. "The Service Class Revisited." Pp. Liberal Political Attitudes of Professionals: 313-29 in Social Change and the Middle Class, Comment on Brint." American Journal of So- edited by M. Savage and T. Butler. London, ciology 92:1501-06. England: UCL Press. Lipset, Seymour Martin. 1960. Political Man. Goodman, Leo. 1981. "Criteria for Determining New York: Doubleday. Whether Certain Categories in a Cross-Classi- . [1960] 1981. Political Man. Expanded fication Should Be Combined." American ed. Baltimore, MD: Johns Hopkins University Journal of Sociology 87:612-50. Press. Gouldner, Alvin. 1979. The Future of Intellectu- Manza, Jeff and Clem Brooks. Forthcoming. als and the Rise of the New Class. New York: "The Religious Factor in U.S. Presidential Oxford University Press. Elections, 1960-1992." American Journal of Greenberg, Stanley. 1995. Middle Class Dreams. Sociology. New York: Times Books. Manza, Jeff, Michael Hout, and Clem Brooks. Haynes, Stephen E. and David Jacobs. 1994. 1995. "Class Voting in Capitalist Democracies "Macroeconomics, Economic Stratification, Since World War II: Dealignment, Realign- and Partisanship: A Longitudinal Analysis of ment, or Trendless Fluctuation?" Annual Re- Contingent Shifts in Political Identification." view of Sociology 21:137-63. American Journal of Sociology 100:70-103. Miller, Warren E. and J. Merrill Shanks. 1996. Heath, Anthony, Roger Jowell, and John Curtice. The New American Voter. Cambridge, MA: 1985. How Britain Votes. London, England: HarvardUniversity Press. Pergammon. Mills, C. Wright. 1951. White Collar. New York: Heath, Anthony, John Curtice, Roger Jowell, Oxford University Press. Geoff Evans, Julia Field, and Sharon Wither- Nardulli, Peter F. 1995. "The Concept of a Criti- spoon. 1991. Understanding Political Change: cal Realignment, Electoral Behavior, and Po- 208 AMERICAN SOCIOLOGICAL REVIEW

litical Change." American Political Science Toqueville, Alexis de. [1835] 1969. Democracy Review 89:10-22. in America. Edited by J.P. Mayer. Garden City, Parsons, Talcott. 1954. "The Professions and the NJ: Doubleday and Company, Inc. Social Structure." Pp. 34-49 in Essays in So- Wattenberg, Martin. 1994. The Decline of Ameri- ciological Theory, edited by T. Parsons. New can Political Parties, 1952-1992. Cambridge, York: Free Press. MA: HarvardUniversity Press. Phillips, Kevin. 1993. Boiling Point: Democrats, Weakliem, David L. 1991. "The Two Lefts? Oc- Republicans, and the Decline of Middle Class cupation and Party Choice in France, Italy, and Prosperity. New York: Random House. the Netherlands."American Journal of Sociol- Piven, Frances F. and Richard A. Cloward. 1988. ogy 96:1327-61. Why Americans Don't Vote. New York: Pan- Wolfinger, Raymond and Stephen Rosenstone. theon. 1980. Who Votes? New Haven, CT: Yale Uni- Raftery, Adrian E. 1995. "Bayesian Model Selec- versity Press. tion in Sociology." Sociological Methodology Wong, Raymond Sin-Kwok. 1994. "Model Selec- 25:111-63. tion Strategies and the Use of Association Reich, Robert. 1991. The Work of Nations. New Models to Detect Group Differences." Socio- York: Knopf. logical Methods and Research 22:460-91. Szelenyi, Ivan and Bill Martin. 1988. "The Three Wright, Erik 0. 1978. Class, Crisis, and the Waves of New Class Theories." Theory and State. London, England: New Left Books/ Society 17:645-67. Verso. Teixeira, Ruy. 1987. WhyAmericans Don't Vote: . 1985. Classes. London, England: Verso. Turnout Decline in the United States 1960- Zipp, John. 1986. "Social Class and Social Liber- 1984. New York: Greenwood Press. alism." Sociological Forum 1:301-29.