STAWS‘ :NGGGSAGTENGY, 'CRGSSs PRESSURE, AND Gomez: AGE 1968 AAA’EREGAN AGEGIGEAGAL GGGTIGN AS A TEST GAGE “z'hes'is G‘s? €18 Degree «3:? FA. EA WCHQAN STAA‘E GAMERSXW 553% WELUMA 880:6 39:22: . ‘ ..... VERSITY LIB BIRAFI ES IIIIIIIIIIIIIIIIIIIIIIIIIIIII III III 3 1293 00627 928 IIII This is to certify that the thesis entitled STATUS INCONSISTENCY, CROSS-PRESSURES, AND VOTING: THE 1968 AMERICAN PRESIDENTIAL ELECTION AS A TEST CASE presented by has been accepted towards fulfillment of the requirements for Ph.D, degree in Political Science ‘\‘ 1 , r/w" fr. , /. fl/flflKWH ABSTRACT STATUS INCONSISTENCY, CROSS-PRESSURES, AND VOTING: THE 1968 AMERICAN PRESIDENTIAL ELECTION AS A TEST CASE BY John William Books This dissertation had two goals: it examined the theoret— ical underpinnings of the twin concepts of status inconsistency and cross-pressures and suggested improvements in the methods used to examine each. The first task involved a reconceptual- ization of cross-pressures along lines already suggested for status inconsistency. It also involved specification of the status inconsistency concept's applicability, spelling out the potential impact of discrepant statuses. In both cases it was suggested that previous treatments were often inappropriate because they employed each concept as fundamentally additive when the more proper interpretation was non-additive. That is, if these concepts are to be construed as variables they must be representable as separate terms in a regression equation: they must not be explainable in terms of the main effects of the status or pressure variables. The methodology to investigate these phenomena suggests itself: multiple regression. In order to cope with non-interval level measurement, dummy-variable multiple regression was employed, using multiplicative John William Books interaction terms to represent inconsistency and cross- pressure effects. A parallel analysis of the traditional conceptualization was carried out by tabular analysis, and comparisons were drawn between the two approaches for evalu- ative purposes. Departing from Lenski's work, the analysis considered two dependent variables: turnout and Wallace vote in 1968. The Survey Research Center's l968 election study was used. It was proposed that inconsistency or cross—pressure (or the combi- nation of both) should lead to lower levels of voting and higher levels of voting for Wallace than would be expected, given the main effects of status and pressure variables. This expecta- tion was not uniformly confirmed. Even the expectation that status inconsistent and cross-pressured individuals would vote less and vote more often for Wallace than individuals consistent along each dimension failed to find substantial support. Finally, analysis demonstrated that the effects of status incon- sistency and two types of cross-pressure, affiliative and atti- tudinal, were not cumulative. It was concluded that neither the traditional nor the modified concept of cross-pressures was adequate. Status incon- sistency, likewise, explained little variance in voting behavior in this case. The intuitive link between the concepts appar- ently does not exist. Non-voting and Wallace voting were apparently not functional alternatives for the inconsistent or cross-pressured since the two were unrelated. Cross-pressures also appears insignificant in explaining left voting in 1968. John William Books Given these empirical conclusions, it is suggested that both concepts be re-evaluated as well as the methods by which they are investigated. STATUS INCONSISTENCY, CROSS-PRESSURES, AND VOTING: THE 1968 AMERICAN PRESIDENTIAL ELECTION AS A TEST CASE By John William Books A THESIS . Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Political Science 1972 I; ACKNOWLEDGMENTS This thesis is the outgrowth of an idea which itself grew from research begun and sustained by Professor Paul R. Abramson. His enduring interest in social structural problems and effects stimulated me before, during and after a period of joint effort, and led me to ask some of the questions contained here about the role of social marginality on political behavior. It was at this stage that a realization that status characteristics were not all uniform and that they could have a variety of effects was established, leading eventually to the focus on status inconsistency and cross-pressures as potentially explan— atory variables. While this dissertation in no way solves most of the more fundamental problems involved, it does, I h0pe, shed some light on the meaning and utility of these twin concepts. Such success as this research has is due in large part to the initial stimulation and continuing support of my Dissertation Advisor and Committee Chairman, Professor Paul Abramson. His analytic skills and academic judgment are not equalled here, but they served as a model for the analysis presented. Without his example, guidance and tactful but insistent prodding this project would not have been accom- plished. I was constantly buoyed by his confidence and ii and perspective. In a similar vein I owe a great deal to Professor Timothy Hennessey who also tutored me patiently on the intricacies of research design, methodology and the logic of inquiry. His comments and support throughout this endeavor were most useful. To Professor Cleo Cherryholmes I must express appreciation as well. His many insights into methodological problems constantly arising no doubt allowed me to escape from mistakes all Unacommon to the unSOphisticated. His patient and careful reading of several drafts and revisions were a source of enlightenment often sorely needed. Professor Paul Conn deserves a special thanks here, for in a very short period of time he reviewed critically and usefully a major portion of this thesis. He provided perspectives of considerable utility and gave rise to consideration of several problems not carefully considered to that time. Several other individuals and organizations were helpful. Professor Elizabeth Powell and Harriet Dhanak of the Michigan State University Political Science Data Archive provided invaluable assistance in preparing the data for analysis, a task I was alone incapable of performing. Their long hours are, hopefully, rewarded here. Likewise, Steve Minnis and the North Texas State University Computer Center Consulting Service deserve special recognition fOr their contribution of time and skill in the face of an often frustrating "user". At critical points in the analysis Professor Charles Prysby provided a sounding board for ideas and a methodological critic of the first class. I received support for this project iii from both the Department of Political Science at Michigan State University and the Department of Political Science at North Texas State University. I would also be remiss if I did not recognize the support, both financial and intellectual, provided by the Political DevelOpment Studies Group at Michigan State University. The data were made available by the Inter- University Consortium for Political Research and were collected by the Survey Research Center of the University of Michigan from September through December of 1968. None of these individuals or organizations is in any way responsible for errors of design, method or judgment: those I dutifully reserve to myself. Finally, I think I must thank my wife and son who have experienced long periods of absence on my part. Whether they have suffered by this I leave to the more disinterested. iv TABLE OF CONTENTS Page LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . vii LIST OF FIGURES. . . . . . . . . . . . . . . . . . . . . xii CHAPTER I. OVERVIEW. . . . . . . . . . . . . . . . . . . . 1 Status Inconsistency and Voting Behavior. . . . . . . . . . . 4 Cross- Pressures and Voting Behavior. . . . . . . 11 Status Inconsistency, Cross- Pressures and Voting Behavior . . . . . . . . . . 20 The Context. . . . . . . . . . . . . . . . 24 The Data Base. . . . . . . . . . . . . . . 30 II. STATUS INCONSISTENCY AND VOTING BEHAVIOR. . . . 37 Operationalizations and Procedures. . . . . . . . 43 Status Inconsistency and Wallace Vote. . . . . . . 46 Status Inconsistency and Non- -Voting. . . . 66 Summary and Conclusions. . . . . . . . . . 83 III. CROSS-PRESSURES AND POLITICAL BEHAVIOR. . . . . 90 The Concept of Cross-Pressures . . . . . . 90 Operationalizations and Procedures . . . . lOl Affiliative Cross- Pressures and Voting Behavior . . . . . . . 103 Attitudinal Cross- Pressures and Voting Behavior . . . . . . . . . . . . 114 Summary and Conclusions. . . . . . . . . . 136 IV. STATUS INCONSISTENCY AND CROSS-PRESSURES. . . . 143 The Relationship Among Status Inconsistency and Cross-Pressures . . . 144 The Cumulative Effects of Status Inconsistency and Cross- Pressures . Summary and Conclusions. BIBLIOGRAPHY . APPENDIX: SUMMARY OF VARIABLES, VARIABLE LOCATIONS, VARIABLE CONSTRUCTION, AND ANALYSIS TECHNIQUES . . . . . . Analysis Techniques, Assumptions, Restrictions. . vi Page 153 157 169 177 190 LIST OF TABLES Table Title Page 2.1 % Wallace Vote by Level and Type of Inconsistency . . . . . . . . . . . . . 48 2.2 Predicted, Observed and Residual % Wallace Vote for Religious-Ethnic Prestige and Occupation. . . . . . . . . . . . . . . 52 2.3 Predicted, Observed and Residual % Wallace Vote, for Religious-Ethnic Prestige and Income. . . . . . . . . . . . . . . . . 53 2.4 Predicted, Observed and Residual % Wallace Vote for Religious-Ethnic Prestige and Education . . . . . . . . . . . . . . . 54 2.5 Predicted, Observed and Residual % Wallace Vote, for Occupation and Income . . . . . . 55 2.6 Predicted, Observed and Residual % Wallace Vote for Occupation and Education . . . . . 56 2.7 Predicted, Observed and Residual % Wallace Vote, for Income and Education. . . . . . . 57 2.8 Residual PrOportions of Wallace Vote for Religious-Ethnic Prestige, Occupation and Income. . . . . . . . . . . . . . . . . 61 2.9 Residual Proportions of Wallace Vote for Religious-Ethnic Prestige, Occupation and Education . . . . . . . . . . . . . . . 62 2.10 Residual Proportions of Wallace Vote for Religious-Ethnic Prestige, Income, and Education . . . . . . . . . . . . . . . 63 2.11 Residual Proportions of Wallace Vote for Occupation, Income, and Education . . . . . 64 2.12 % Non-Voters by Level and Type of Incon— sistency. . . . . . . . . . . . . . . . . . 67 vii Table 3.1 3.2 Title o\° Predicted, Observed and Residual Voters for Religious-Ethnic Prestige and Occupation Non- Predicted, Observed and Residual % Non- Voters for Religious-Ethnic Prestige and Income . . . . o\° Predicted, Observed and Residual Non- Voters for Religious-Ethnic Prestige and Education. o\° Predicted, Observed and Residual Voters for Occupation and Income. . . . . . . . . Non- Predicted, Observed and Residual Voters for Occupation and Education . . . . . . . o\° Non- Predicted, Observed and Residual % Non- Voters for Income and Education Residual Proportions of Turnout for Religious-Ethnic Prestige, Occupation, and Income; . Residual PrOportions of Turnout for Religious-Ethnic Prestige, Occupation, and Education . Residual PrOportions of Turnout for Religious-Ethnic Prestige, Income, and Education . Residual PrOportions of Turnout for Occupation, Income, and Education F-Tests and Levels of Significance for Reduction of Unexplained Variance Attributable to the Addition of Selected Interaction Terms . . Affiliative Cross-Pressures: Percent Non-Voters, by Spouse and Relative Cues. . . . . . Affiliative Cross-Pressures: Percent Non-Voters, by Spouse and Friend Cues. . . viii Page 69 70 71 72 73 74 77 78 79 80 82 104 105 Table 3. 3 .10 .11 .12 .13 .14 .15 Title Affiliative Cross-Pressures: Percent Non-Voters, by Relative and Friend Cues Affiliative Cross-Pressures: Percent Wallace Voters, by Spouse and Relative Cues . . . . Affiliative Cross-Pressures: Percent Wallace Voters, by Spouse and Friend Cues . . . . . Affiliative Cross-Pressures: Percent Wallace Voters, by Relative and Friend Cues . Affiliative Cross-Pressures, F-Tests Between Additive and Interactive Equations, for Turnout. Affiliative Cross-Pressures, F-Tests Between Additive and Interactive Equations, for Wallace Vote Attitudinal Cross-Pressures: Percent Non-Voters, by Party and Candidate Orientations. . . . . . . . Attitudinal Cross-Pressures: Percent Non-Voters, by Party and Issue Orientations. . . . Attitudinal Cross-Pressures: Percent Non-Voters, by Candidate and Issue Orientations. . Attitudinal Cross-Pressures: Percent Non-Voters, by Party and Candidate and Issue Orientation . Attitudinal Cross-Pressures: Percent Wallace Voters, by Party and Candidate Orientations. Attitudinal Cross-Pressures: Percent Wallace Voters, by Party and Issue Orientations. . . Attitudinal Cross—Pressures: Percent Wallace Voters, by Candidate and Issue Orientation . ix Page 106 108 108 109 111 113 116 116 117 119 120 121 121 Table 3.16 Title Attitudinal Cross-Pressures: Percent Wallace, by Candidate and Issue Orientations. . . . . . . . . . Attitudinal Cross-Pressures: Observed, Expected and Residual Percent Non- Voters, by Party and Issue Orientations. . . Attitudinal Cross-Pressures: Observed, Expected and Residual Percent Non- Voters, by Party and Issue Orientations. . . . . . Attitudinal Cross-Pressures: Observed, Expected and Residual Percent Non- Voters, by Candidate and Issue Orientations. Attitudinal Cross-Pressures: Residual Percent Non—Voters, by Party, Candidate and Issue Orientations. Attitudinal Cross-Pressures: F-Tests Between Additive and Interactive Equations, for Turnout. Attitudinal Cross-Pressures: Percent Republican Voters, by Party and Candidate Orientation . . . Attitudinal Cross-Pressures: Percent Republican Voters, by Party and Issue Orientations. . . Attitudinal Cross-Pressures: Percent Republican Voters, by Candidate and Issue Orientations. . . . . . Attitudinal Cross-Pressures: Observed, Expected and Residual Percent Republican, by Party and Candidate Orientations. . . . . Attitudinal Cross-Pressures: Observed, Expected and Residual Percent Republican, by Party and Issue Orientations. . . . . Page 123 125 125 126 127 128 129 130 130 132 133 Table 3. 4. A1 A2 A3 A4 A5 A6 A7 A8 A9 27 .28 .29 4 Title Attitudinal Cross-Pressures: Observed, Expected and Residual Percent Republican, by Candidate and Issue Orientations. . . Attitudinal Cross-Pressures: Residual Percent Republican, by Party, Candidate and Issue Orientations. Attitudinal Cross-Pressures, F-Tests Between Additive and Interactive Equations, for Percent Republican . Status Inconsistency and Affiliative Cross-Pressure Combinations Affiliative Cross—Pressure Combinations and Attitudinal Cross—Pressures Status Inconsistency and Attitudinal Cross-Pressures Spouse-Relative Affiliative Cross- Pressures and Voting Turnout, by Status Inconsistency . Variable Names, Numbers, and Location. Frequency Distribution for Dependent Variables Frequency Distribution for Status Variable Combinations. Distribution of Inconsistency Types. Attitudinal Cross-Pressure Frequency Distribution. . . . . . . Frequency Distribution of Attitudinal Cross-Pressure Types. Frequency Distribution for Affiliative Cross-Pressure Combinations Frequency Distribution for Affiliative Cross-Pressure Types. Frequency Distribution for Independent Variables in Trichotomous Form. xi Page 133 134 135 146 149 152 155 178 180 181 184 185 186 188 189 191 LIST OF FIGURES Figure Title 1 Expected Relationships Between Status Inconsistency, Affiliative Cross- Pressures, and Attitudinal Cross- Pressures 2 Idealized Cross-Pressure Table 3 Model Matrix for Additive Relationship xii Page 22 100 193 CHAPTER I OVERVIEW The study of voting behavior has made great strides in the last few decades. We now know a good deal about the determinants of turnout and partisan preference. The relationships between these variables and dimensions of social status such as income, occupation, education and race are well documented. The importance of a person's social affiliations and attitudes are similarly known. A large amount of effort has been expended in refining our knowl- edge of these important relationships. And, while most were originally uncovered in American voting studies, subsequent analyses have established the universality of nearly all the findings.1 The degree of cross-national reliability of these findings is remarkable, given the presumed and actual cultural differences.2 When, however, political scientists cease discussing the main effects of these variables and focus on their inter-relationships, the level of agreement on effects diminishes rapidly.3 When the expected results of several variables are not cumulative, the behavior of the individual becomes less easily predicted and understood. The effect of "inconsistency" upon the voter is still relatively uninvestigated and the findings which do exist display a singular lack of accord. They demonstrate that, in this area, much more is unknown than known. This situation apparently has two causes, one theoret- ical and one methodological. First, even a cursory scan of the literature alerts the student of inconsistency to the variety of mechanisms cited as responsible for inconsistent behavior, or at least behavior which is not deemed average or "normal" for similar, but not inconsistent individuals. In general these mechanisms can be reduced to two: a socio-psychological "balancing" approach,4 and an "over- 5 The former asserts that compensation" or "stress" model. individuals finding themselves in inconsistent status, social or attitudinal configurations will adjust to these by rearranging the elements apprOpriately. In the short run their behavior will likely include withdrawal from situations which increase the salience of the inconsistency, or tempo- rary suppression of one cue and exclusion of others. The second approach claims that inconsistency is more psycho- logically distressing than the first allows and that this stress resulting from inconsistency makes it unlikely that a person will balance or merely select cues. Rather, he is expected to adOpt an extreme position on inconsistency- salient behavior and relieve his stress through radical and extreme cathartic action or attitudes. In this case, groups of inconsistent individuals find themselves on the periphery of socially accepted attitudes and behavior. The second approach then sees discharge of tension as the mechanism by which persons overcome inconsistency. In spite of the divergence of these two alternative approaches to inconsistency, the investigators of each have used very similar tools to test their hypotheses, and have been indifferent to testing the two approaches together. This is to some extent a function of the novelty of the concept and the vagueness of its applied meaning. Never- theless, the various measures of inconsistency have rarely allowed for useful comparison of results. While the theo- retical implications of some of the methods used are not totally clear, it is certain that conclusive tests of the predictions of the two approaches have not been achieved. This dissertation has two purposes, stemming from the twin problems of theory and method in the study of incon- sistency. First I shall attempt to test the relative valid- ity of the two approaches outlined above, with an eye toward improving the theoretical coherence of the approaches and their logical results. This is essentially a process of concept clarification. Second, I will utilize overlapping methods used in previous studies. It will also involve testing hypotheses explicit in these studies, or implicit in their logic. The final product of this analysis should be a better understanding of the concept of inconsistency. The dissertation will, then, focus primarily on the relationships between two dependent variables, voter turnout and partisan preference, and three independent variables, status inconsistency, affiliative cross-pressures and atti- tudinal cross-pressures. Each of these independent vari- ables is, of course, a combination of other variables. They are conceptually similar, at differing levels Of analysis. Moreover, they are assumed to be inter-related in a causal sense, to be outlined later. The hypotheses which follow are drawn from the literature on each of the independent variables and from the two general approaches outlined above. Status Inconsistency and Voting Behavior Status inconsistency (or status crystallization, or status congruence) while not an entirely new concept, was given a precise definition and formulation in 1954 by Gerhard Lenski.6 Noting that nearly all previous studies of status relationships had focused only on "vertical" ones, Lenski suggested that there might be a non-vertical dimension as well. He argued that people actually occupy positions in several vertical hierarchies simultaneously, and that these positions need not be consistent, or of the same social rank. Lenski proposed that, "apparently the individual with a poorly crystallized status is a particular type of marginal man, and is subjected to certain pressures by the social order_which are not felt (at least to the same degree) by individuals with more highly crystallized status."7 He suggested, on the basis of his analysis, that people with inconsistent statuses feel less integrated into the commu- nity and are lead to support political movements of social change in the hope that these will lead to a restructuring of the social order and potentially a more felicitous posi- tion for themselves. Specifically, Lenski claimed that this political involvement would (and did) take the form of sup- port for leftist parties. In a later article Lenski favor- ably noted Gordon's suggestion that the effects of incon- sistency may be more diffuse and that we might well expect inconsistent individuals to "adOpt extreme political posi- tions, whether on the extreme left or the extreme right." Lenski's original formulation, then, appears to be ungeneralizable, or perhaps more accurately, not general enough. Rush, for example, finds that "individuals whose statuses are inconsistent are more likely to be right-wing extremists in their political attitudes than are individuals whose statuses are consistent."9 He suggests that "right- wing extremism is a form of political expression offering simplistic, highly structured solutions to the frustrations arising from status inconsistency, and that persons suffer- ing from status inconsistencies will be more likely to main- tain this political ideology than those whose statuses are crystallized."10 In his study he cites a body of literature which adds some support to the thesis that rightism is a possible outcome of status inconsistency. Moreover he pro- poses that extreme rightism and extreme leftism may be very similar events, each easily produces by status inconsistency. Kelly and Chamblis report on research indicating no relation- ship between inconsistency and political liberalism.11 Kenkel, reviewing Lenski's original data, also claims that there is no clear relationship between these two variables.12 Jackson and Burke document a relationship between status inconsistency and symptoms of personal stress.13 The last finding gains centrality from the importance attached to the role of stress in determining political attitudes and actions in most theoretical writings about inconsistency. Treiman states the presumed relationship well: "The status dis- crepancy hypothesis asserts that conflict engendered by status discrepancy is likely to lead to strain and that this strain will manifest itself in pathological behavior. ."14 He comments further that, "according to the status- discrepancy hypothesis, incumbents in discrepant status con- figurations should, on the average, exhibit higher levels of these pathological symptoms than would be expected on the basis of the independent contributions of each of their "15 Herein lies the crux of Lenski's status characteristics. status inconsistency argument. If individuals use common psychological mechanisms, such as exclusion, and act as if only one status were relevant, then, at the aggregate level inconsistent groups will fall between the consistent groups on a given dependent variable. Some individuals will choose each attitudinal or behavioral alternative, resulting in an average choice within the bounds set by the consistent groups. If, on the other hand, the stress model of incon- sistency is valid, we can expect the inconsistencies to fall into some extreme category, which may be situationally defined. This approach assumes that the stress produced by incon- sistency leads to behavior which is non-normal in that few individuals in the society choose it. Therefore, groups of inconsistents should not fall between the relevant consistent groups. Testing whether one of these approaches is more satisfactory is facilitated by framing the question in this fashion, since the statistical concepts of additivity and interaction parallel the logic of each of the arguments, respectively. Thus, deciding whether status inconsistency is a significant independent variable in explaining voting behavior falls on the relatively simple question of whether there is statistical interaction involved in the relationship. That is, is an additive model sufficient to explain the variance, or is there a non-additive relationship which may be attributed to the inconsistency effects? Several studies have been concerned with the possible existence of statistical interaction in the relationship between status inconsistency and various dependent vari- ables. Treiman found, in his study of prejudice, that an additive model was sufficient and that no interaction between status variables was apparent.16 However these results are not completely satisfactory for reasons which illustrate several of the problems encountered in studying status incon- sistency. First, Treiman used only two of the many possible instances of status inconsistency: education-income dis- crepancies and education-spouse's education discrepancies. The latter is not usually included in studies of inconsistency. Secondly, as Treiman himself notes, his dependent variable, prejudice, is probably less sensitive to status inconsistency than other potential variables. Finally, Treiman neglected to use both ascribed and achieved statuses and thus cannot comment on the relationship between them, a relationship which Lenski, for one, considers of prime significance. On the other hand, his method allows him more precision than simple cross-tabulation techniques. This method allows the researcher to pinpoint the location of the interaction, if any. Jackson and Burke, for example, have used a similar approach to find that status inconsis- tency appears interactively related to symptoms of stress.17 As Blalock notes, this technique is perhaps the best method yet devised to overcome the "identification" problem: i.e., the problem caused by overspecification of an equation.18 Jackson and Burke overcome the problem by assuming that certain interaction coefficients equal zero. Thus they need not estimate them. In Jackson and Burke's research, a simple additive model does not explain the relationship between inconsistency and stress as well as one which allows for specific interaction combinations among the independent variables. Especially interesting, and emphasized by the somewhat conservative nature of the "dummy-variable” regression approach, is the finding that status inconsistency produces more stress than would be expected by adding the separate status effects. This means that the suggested link between status inconsistency and extreme behavior is at least partially established empirically: individuals who are inconsistent evidence more symptoms of stress. In his research, Boyd found that there appeared to be little evidence of interaction between status inconsistency and attitudinal cross-pressures, a relationship to be considered here.19 Boyd hypothesized that individuals evidencing status inconsistency would be more cross-pressured than an additive model would suggest. Since there appeared to be sound theoretical reasons for expecting this relationship to exist, Boyd was reluctant to reject his hypothesis even though the data did not support it, preferring instead to suggest alternative explanations for this failure. I am also hesitant to reject this hypothesis and have included one similar to Boyd's in this section. In his study based on the Almond and Verba five nation data, Segal found that there was interaction between status variables and both non- voting and Democratic voting preference in the United States.20 This finding cannot be interpreted unambiguously, however, because his method consisted of cross-tabulation of statuses --a procedure which is incapable of specifying the precise 21 nature of the relationship. Portes' study of left radical- ism in Chile also searched for interaction in the relation- ship between status inconsistency and voting behavior.22 Portes found none, using only education-income and education- occupation inconsistency. Finally, Pool, Abelson and Popkin found, for the most part, little evidence of interaction in their United States voting study.23 They did find, however, 10 a considerable degree of interaction between certain combi- nations of status variables, such as occupation and religion, and partisan preference. In the last instance, an inter- active equation was clearly preferable to an additive one. The conclusion suggested by this brief summary is that the correct conceptualization of status inconsistency is not yet at hand. It may well be that status variables are rele- vant only for a subset of conceivable related dependent vari- ables, as Boyd and Treiman suggest. That is, status incon- sistency may be important in explaining some types of behavior and not others. It may also be that one's method considerably influences, if not determines, his findings. In any case it appears that the existence of status inconsis- tency as an independent variable has not been conclusively demonstrated, and its presumed relationships to significant dependent variables are relatively undocumented. Finally, although there is no way to control for or guage its effect in the present study, the problems of "clustering," first noted by Tingsten, cannot be ignored.24 The distribution of status characteristics is not random across the United States. Clearly this means that incon- sistency does not everywhere have the same meaning. When national samples are employed the likely effect of this phenomenon is to increase error and make hypotheses more easily rejected. However it may be that clustering may distort the impact of inconsistency in some systematic fashion. It is obvious neither why and how such systematic 11 distortion would occur, nor precisely what effect it would have. I shall assume that clustering effects do, in actuality, cancel each other and that their only impact is to weaken the relationships explored here. However, if it is correct to conceive of the effects of status inconsistency in terms of statistical interaction, and if the relationship proposed by Lenski holds, then we may hypothesize that: H. 1: Status inconsistency leads to higher rates of voting for deviant parties than are predicted by an additive model using only the main effects of the status variables. Also, if status inconsistency is Operative, it seems likely that high levels of stress should lead to inordinately high rates of withdrawal among the inconsistent. If status incon- sistency is seen as an inability to decide which cues to accept from one's social environment, then it may be pro- posed that inconsistent individuals might cut psychic costs and avoid decision altogether in greater proportion than would be expected, given their social characteristics. Thus: H. 2: Status inconsistency leads to higher rates of non-voting than are predicted by an additive model using only the main effects of status variables. Cross-Pressures and Voting Behavior Students of voting behavior have long been concerned with the "forces" which determine a voter's choice. While many of the major social determinants have been uncovered, 12 their inter-relationship is still a matter of concern. It has become customary to refer to situations in which major determinants would not lead to the same decision as cross— pressure cases. Lipset, for example, in Political Man discusses a number of studies illustrating the effects of 2 . . . cross-pressures. 5 The d1scuss1on 15 germane here because it demonstrates the lack of precision which often accompanies the use of the concept. Lipset uses cross-pressures to refer to conflict among nearly any and all possible attributes, regardless of their level of analysis. He cites a study by Janowitz and Marvick, for example, which used as pressures, residence, socio-economic status, religion, issue attitudes, and candidate attitudes. Using these variables which obvi- ously represent differing levels of analysis, Janowitz and Marvick found "increasing cross-pressures regularly decreased the relevant party vote. Non-voting also increased regularly . . . H26 With 1ncreas1ng cross-pressures. . . . Other researchers have used cross-pressures in as many ways. This thesis incorporates two of the most common: affiliative and attitudinal cross-pressures. I will follow Pinner's differentiation: Attitudinal conflict may occur when a person is faced with a choice between alternative beliefs or courses of action under conditions which bring into play attitudes motivating different and Opposing choices. Affiliative conflict can result from a person's attachment to several groups which have preferences for different alternatives. . While the mechanics of cross-pressures are apparently very similar to those of status inconsistency, it is clear that 13 the former Operates on different levels of analysis. In fact, each type of cross-pressure represents a different level. I shall discuss affiliative and attitudinal cross- pressures in turn and present hypotheses relating each to turnout and voting preference. In the literature on voting behavior the concept of affiliative cross-pressures represents one of the earliest attempts to explain behavior that was "non-normal." Affili- ative cross-pressures represent, of course, the situation in which an individual is presented with different cues by various relevant others, usually his family, friends, and work-place associates. Lazarsfeld documented the impact of inconsistencies in preferences among a voter's relevant Others on both the voting decision and partisan preference in The Pe0ple's Choice.28 While he used several variables which are here labelled status variables and not affiliative ones, Lazarsfeld did find that peOple who were cross- pressured were less likely to vote, more likely to change vote direction during a campaign, and more likely to post- pone decision on both turnout and party preference. Berelson, Lazarsfeld and McPhee report similar findings in Voting.29 In this study the authors found that "nominal conflicts between social categories like religion and class are associated less with delay (of vote decision) than primary group discords, and the more so the more intimate the conflict."30 This actually reflects a comparison between the impacts of status inconsistency and affiliative 14 cross-pressures and amounts to a claim that the more likely an inconsistency will be salient, the more potent it will be. The authors also found clear relationships between voting change and the preferences of one's family, friends and co-workers. Unfortunately, neither the Voting study nor its predecessor offers substantial assistance in explaining these findings. Campbell and his cohorts are also mute here. After reviewing their findings on primary group influences on voting, they conclude that "the mechanisms by which the group influences the individual are exceedingly complex."31 Berelson, Lazarsfeld, and McPhee note that "while theory regarding these cross-pressures is not yet particularly advanced, an impressive series of empirical results has accumulated over the past fifteen years of research."32 Yet we cannot adequately comprehend the concept Of cross- pressures without such theory. The relevant question here is: Why should individuals who exhibit cross-pressures be less likely to vote, make up their minds later in a campaign and vacillate in the choice of party and candidate? Is it due to indecision over which relevant other should be given priority? Is it because individuals weight cues from each relevant other in making a choice? Or is it because the cross-pressures arising from inconsistent cues lead to stress which makes consistent behavior unlikely or impossible? These questions imply somewhat different behavior patterns and are thus susceptible to test. If we assume that cross-pressured individuals solve 15 their dilemma by selecting one relevant other or cue as their basis Of choice, then the cross-pressured, as a group, should not exhibit lower turnout than consistent individuals, but they should be less likely to vote for the candidate of their choice. Moreover, and most important here, there should be no significant differences among the cross-pressured sub- groups in terms of their turnout or deviation from their own partisan preference. That is, the existence of more cross—pressures should not lead to different kinds of behavior. On the other hand, if cross-pressures reduce con- sistency by increasing psychological stress, we would expect to find that increasing cross-pressures lead to increased stress and, consequently, to decreased turnout and increased deviation from own partisan preference. This means that as groups become more cross-pressured, they should exhibit less consistency. This line of argument also implies that high levels of stress produced by high cross-pressures should lead to high levels of voting for non-system parties or parties of social change, since they represent an outlet for pressures and provide a potential source of relief in a long- term sense. Perhaps cross-pressured individuals are, in some sense, released from normal restraints by the pressures on them and will be more easily recruited by radical parties than will the more socially integrated non-cross-pressured voter. The similarity of this conceptualization of cross- pressures to the preceding conceptualization of status 16 inconsistency is obvious and natural. If cross-pressures have any meaning, they must have an effect independent of the individual "pressures." Thus, we may argue that the stress produced by cross-pressures does not accumulate in an additive fashion, but rather that cross-pressures combine interactively to produce higher symptom levels than might be expected under an additive assumption. This approach suggests the following hypotheses: H. 3: Individuals characterized by affiliative cross-pressures will be less likely to vote than those without cross—pressures, and the more so the greater the cross-pressures. H. 4: Individuals characterized by affiliative cross-pressures will be more likely to vote for marginal parties than those without cross-pressures, and the more so the greater the cross-pressures. H. 5: Individuals characterized by affiliative cross-pressures will be less likely to vote than is suggested by an additive model utilizing only the main effects of the affiliative cues. H. 6: Individuals characterized by affiliative cross-pressures will be more likely to vote for marginal parties than is suggested by an additive model utilizing only the main effects of the affiliative cues. Like affiliative cross—pressures, the concept of atti- tudinal cross-pressures is one with a long history in the literature on voting behavior. Whereas the Columbia researchers were most concerned with affiliative cross— pressures, the Survey Research Center's group focused primarily on attitudinal ones, reflecting the basic differ- ences in approach used by each. This is not to say that 17 either ignored affiliative or attitudinal cross-pressures. The Lazarsfeld group reported on them indirectly, noting that attitudes have a "strain toward consistency" during 33 campaigns." In The Voter Decides, Campbell, Gurin and Miller document the effects of family, friends and work- place peers on an individual's partisan preference.34 In The American Voter, however, the Campbell group relies heavily on analysis of attitudes to explain voting behavior.35 The authors of The American Voter emphasized the crit- ical influence of party identification on the vote of any individual. Compared to party identification, both candi- date orientation and issue orientation had little predict- ive power. Campbell and his colleagues argue, consequently, that party identification is the variable of most concern to students of voting behavior, the one which explains the most variance, and the one which is most enduring. They present evidence that individuals rarely change party identification and that this identification is usually transmitted inter- generationally with little question, prebably because it is of such low salience. This finding is of considerable relevance to this dissertation and will be reviewed below. On the other hand, some have quarreled with the emphasis placed on party identification by the Michigan researchers and/or their interpreters. Key, for instance, in The Responsible Electorate, suggests that voters are not nearly so irrational as this blind reliance on party identification suggests.36 He notes that earlier studies of voting behavior 18 were principally focused on the effect of each of these attitudinal variables, and their combinations, on voters' changes of preferences as an election drew near. Key, however, was more concerned with switches of vote between elections. His analysis supports the view that switchers are not necessarily the most incompetent voters and that they do pay considerable attention to political affairs, even basing decisions upon them. A viable theory of voting would certainly explain inter—election switches and campaign deviations as well as non-voting, a tOpic which Key does not investigate. In analyzing data from the 1968 Presidential election, researchers from the University of North Carolina also accented the importance of a person's issue position for his voting decision.37 This analysis is of particular importance here because 1968 was the first recent election featuring a significant vote for a third-part candidate, allowing the voter a choice other than Democrat and Republican. This factor probably increased the role of issues in the outcome of the election. Moreover, the work of Kovenock and his colleagues raises interesting questions about what, precisely, constitutes an issue. This question is obviously central to the study of attitudinal cross-pressures and has been too little discussed. The way one defines an "issue" may pre- determine his results, as Kovenock argues. While this problem will deserve more comment later, the point noted here is that both Key and Kovenock find issue orientation to be 19 more potent than previously believed, and that conflicts between party identification and issue position can cause defection from previous voting patterns. When Campbell et al. attempt to explain individual deviations from voting preference, cross-pressures is one of their key variables. In their scheme only other attitudes mediate between an individual's partisan preference and his vote. The precise form of this mediation is not specified in The American Voter, but it does not seem unfair to infer that the relationship was assumed to be linear additive. In the case of attitudinal cross-pressures, however, there is evidence that an additive assumption is unwarranted. In his study of electoral behavior of voters in the 1956, 1960, and 1964 Presidential elections, Richard Boyd found considerable evidence of statistical interaction among the factors involved in attitudinal cross-pressures: an individual's own partisan preference, his evaluation of the candidates, and his issue position relative to the parties.38 In short, it appears that attitudinal cross-pressures do not conform to an addi- tive model but rather that they combine interactively to produce levels of voting defection (to use Boyd's term) above that predicted by summing the individual effects. Of course this finding is important here since it lends support to the preceding hypotheses concerning interaction in the cases of status inconsistency and especially affiliative cross-pressures. Thus, we can test the effect of attitudinal cross-pressures on turnout with the expectation that interaction 20 is present. From the standpoint of the theoretical outlook adopted here, this is crucial. Boyd's analysis indicates support for the stress approach: it appears that individuals defect more than would be the case if they selected one paramount value as their decision-criterion. This is the outcome suggested by the stress approach. In the light of this research, and in keeping with the arguments formulated previously, the following hypotheses are suggested: H. 7: Individuals characterized by attitudinal cross-pressures will be less likely to vote than those without cross-pressures, and the more so the greater the cross-pressures. H. 8: Individuals characterized by attitudinal ' cross-pressures will be more likely to vote for marginal parties than those without cross- pressures, and the more so the greater the cross-pressures. H. 9: Individuals characterized by attitudinal cross—pressures will be less likely to vote than is suggested by an additive model utilizing only the main effects of the attitudinal pressures. H. 10: Individuals characterized by attitudinal cross-pressures will be more likely to vote for their preferred party than is suggested by an additive model utilizing only the main effects of the attitudinal pressures. Status Inconsistency, Cross-Pressures and Voting Behavior As noted earlier, the three variables considered independent in this study, status inconsistency, affiliative cross-pressures and attitudinal cross-pressures, are pre- sumably inter-related. Since they actually represent con- / ceptually congruent phenomena at different levels of analysis, we should not be surprised at this relationship. 21 Nevertheless, the specification of the inter-relationship is relatively complex. Status inconsistency denotes a situation in which an individual's status groups are not completely congruent. This means that status inconsistency describes a condition of holding relatively highly divergent statuses in various hierarchies. Individuals who are inconsistent may, then, be exposed to a wide variety of people, a larger range than that to which consistent individuals are ordinarily exposed. This is the traditional notion of the marginal man. Such exposure may easily lead to affiliative cross-pressures, since holding very different positions in these hierarchies is likely to result in one's family, friends and work- place cohorts being from different status levels. It is not unreasonable to expect political views encountered in this context to diverge relatively more than those nurtured in a homogeneous milieu, subjecting the individual to quite different cues. Affiliative relationships may, in turn, affect a person's attitudes. Socialization studies tell us that attitudes are learned from diverse agencies, including the family, the school, the peer group, voluntary associa- tions, the mass media, and even broad social class.39 If this is the case, individuals with affiliative cross-pressures may be internalizing attitudes which conflict, and/or may be more conscious of attitude conflicts already present but not necessarily salient. Thus status inconsistency should be associated with affiliative cross-pressures, which, in 22 turn, should be associated with attitudinal cross-pressures. There may also be a direct link between status inconsistency and attitudinal cross-pressures, although this relationship would probably be weaker than the other two and affected by them both. The only evidence we have bearing on these relationships is presented by Boyd. He examined the relationship between status inconsistency and attitudinal cross-pressures and found that no interaction existed in the relationship.40 Moreover, Boyd found no evidence that status inconsistency was related to attitudinal cross-pressures at all. However, he did not reject his hypothesis relating the two variables since he still felt it had considerable intuitive validity and there were several plausible alternative explanations.41 An explanation of his failure to confirm not considered by Boyd is suggested by the above discussion: the relationship between status inconsistency and attitudinal cross-pressures, if any, may be Operative only through the intervening vari- able of affiliative cross-pressures. That is, status inconsistency may affect attitudinal cross-pressures only by first affecting affiliative cross-pressures. We may represent the prOposed relationship diagramatically in this way: Affiliative ///,//3 Cross-PressuresIT“~\\\§§ Status Attitudinal Inconsistency *7 Cross-Pressures Figure 1: Expected Relationships Between Status Inconsistency, Affiliative Cross-Pressures and Attitudinal Cross-Pressures 23 It seems essential, therefore, to test the entire set of postulated relationships rather than one alone. The fore- going discussion suggests the following hypotheses: H. 11: Individuals characterized by inconsistent statuses are more likely to have affiliative cross-pressures than those with consistent statuses. H. 12: Individuals characterized by affiliative cross-pressures are more likely to have attitudinal cross-pressures than those not affiliatively cross-pressured. H. 13: Individuals characterized by inconsistent statuses are more likely to have attitudinal cross-pressures than those with consistent statuses. Although Boyd proposed that status inconsistency should inter- act statistically with attitudinal cross-pressures, I find no theoretical justification for this expectation and shall refrain from proposing any interactive relationships between these variables. It may be, finally, that the effects of status incon- sistency, affiliative cross-pressures, and attitudinal cross- pressures are cumulative. That is, we might expect that the more inconsistent and cross-pressured an individual becomes, the more likely he is to exhibit deviant behavior. Thus, peOple high on status inconsistency, and high on affiliative and attitudinal cross-pressures may refrain from voting and vote for radical parties more than people not so character- ized. For simplicity's sake, and for lack of any evidence to the contrary, we can here assume an additive relationship. Thus we have two final hypotheses: 24 H. 14: Status inconsistency, affiliative cross- pressures and attitudinal cross-pressures have a cumulative effect on voting turn- out. Individuals high on three Will turn out less than individuals high on two and individuals high on two will turn out less than individuals high on one. H. 15: Status inconsistency, affiliative cross- pressures and attitudinal cross-pressures have a cumulative effect on partisan pre- ference. Individuals high on three will be more likely to vote for a radical party than individuals high on two, and indiv- iduals high on two will be more likely to vote for a radical party than individuals high on only one. The Context The 1968 American Presidential Election may have been unique: it was certainly unusual. Both major parties experienced peculiarities in nominating candidates. Although the Republicans conformed to expectations in nominating Richard Nixon, many within that party had serious doubts concerning his ability to win and others were concerned about his strategy--the now-famous Southern strategy. The Democrats found themselves faced with a potentially fatal internal dissensus caused by the Vietnam war. Eugene McCarthy's challenge to the incumbent President over his conduct of the war was an early indication. Following the New Hampshire primary, Lyndon Johnson withdrew as a candidate for a second term with an abruptness that startled the nation. Coming late to the race, Robert Kennedy, until his assassination (itself one of the most bizarre aspects of the election) provided an attractive alternative to Hubert Humphrey. Finally, in a convention marred by ugly street riots and charges of 25 police brutality and convention rigging, Humphrey was annointed. He was never completely accepted by all Democrats, however, and his association with the policies of the Johnson admin— istration and the Chicago Spectacle diminished his potential attractiveness. The most unusual aspect of this already incredible election, however, was the emergence of a third party of respectable proportion: The American Independent Party led by George Wallace, erstwhile Governor of Alabama.42 Labelling himself an alternative to Tweedlededum and Tweedlededee, Wallace had apparently tapped a responsive chord in the electorate and his presence threatened to make the Presidential election inconclusive, forcing a resort to the archaic process of selection by the House of Representatives. The major parties reacted to this challenge by attempting to neutral- ize the major Wallace Issue, law and order, and by pointing out that a vote for Wallace was a vote wasted since he could not be elected. In a society feeling itself afflicted with anarchy and a breakdown of Values, the impact of the law and order issue could not be underestimated. To some extent the salience of this issue can be measured by the fact that Wallace's candidacy was the first significant minor party effort in twenty years and attracted a larger proportion of the popular vote than any third party since 1924, and a greater electoral vote total than any since 1860.43 The Wallace effort thus makes the 1968 election different from the normal pattern of United States Presidential elections. 26 Students Of voting behavior must, of course, investigate such discontinuities. Certainly a central question for analy— sis is what contributed to this unusual election. However, the topic under consideration here is a good deal more cir- cumscribed. I am concerned primarily with the effects of status inconsistency and cross-pressures on voting behavior in the American public. In this context the anomalous nature of the 1968 election and the Wallace candidacy is an asset. Wallace offered voters an alternative clearly distinct from the two major parties. Thus, although the election was undeniably non-normal, it offers the student of inconsistency the Opportunity to examine an instance in which inconsistency can well be expected to have had a visible impact. If incon- sistency Operates as we expect it does, then the presence of the Wallace alternative should allow inconsistent and cross- pressured individuals an outlet for their tensions. Of course we cannot expect that all such individuals will utilize this outlet. There may be reasons totally unconnected with incon- sistency or cross-pressures which will lead a voter to select Wallace. It may be argued that some voters, particularly those in Alabama, do not perceive Wallace as the candidate of the third party at all, but rather view him as the legiti- mate Democratic party candidate.44 Of course it may also be that some voters do not consider voting for Wallace to be a rejection of the two major parties in general, but that they see voting for Wallace as a means of expressing themselves on the issue of law and order. Presumably voters still highly 27 committed to either major party could choose this alternative. Other voters might consciously be using a Wallace vote to object to the tendency of the major parties to assume similar stances on most of the issues. Finally, certain inconsistent or cross-pressured individuals might be ideologically Opposed to voting for Wallace in spite of a preference for changes and a rejection of the two major parties. These individuals might well be expected to obstain.4S While any of these explanations might aid our under- standing of Wallace support, it may also be that Wallace votes are correlated with our measures Of inconsistency and cross- pressures. Certainly if the theories of inconsistency and cross-pressures have any validity, we can expect them to hold in this instance. In Wallace we have a candidate who was not commonly admired, seen as a bigot by the media, presumably appealing to the lesser instincts of man. His party was portrayed in the press as a radical fringe group intent upon setting back the clock and promoting inequality. In a national sense George Wallace and his party constituted a negative valence figure. Those individuals voting for Wallace thus had to cast a ballot they knew to be outside the "mainstream” of American politics. If inconsistent and cross-pressured individuals are seen as unsatisfied with their present positions and if we assume that they view political action as a means of ameliOrating these circumstances, then the Wallace candidacy should be particularly attractive since it expresses great dissatisfaction with the present state of affairs. 28 So the 1968 election holds advantages as well as perils for the study of inconsistency and cross-pressures. The possible pitfalls Of this case may not be overlooked, however. First, in this election Wallace support was not proportional across the entire nation. Rather, it was clustered in the South, among working class Voters, among protestants and among whites. It appears that Wallace's support cut into the traditional demographic strongholds of both major parties and the question of which major candidate was most adversely affected by Wallace is still being debated.46 The uneven distribution of Wallace's support means that we must be cautious about generalizing our findings, however, since any given area or region may diverge rather greatly from such generalizations. Certainly controls for race and region seem particularly appropriate. A control for race, however, is meaningless, since, among those sampled no blacks voted for Wallace (none even supported him). Region may be a powerful factor, but so few respondents in our sample are Northern Wallace voters that comparison by region also becomes impos- sible. Thus, regrettably, these differences must be assumed away. That is, I shall assume that there are no differences in the way inconsistency and cross-pressures affect peOple's behavior in the North and South. Blacks are eliminated from consideration. Admittedly the problem of the Southern voter regarding Wallace as the legitimate candidate of the Democratic Party may still affect the results, but it should bias them in a negative direction: if it has an effect, that effect 29 should be to lessen the impact of inconsistency and cross- pressures on voting behavior. It will increase error and decrease the likelihood of discovering significant relation- ships. A second, and potentially more damaging, difficulty is related to the problem of social "clustering" originally suggested by Tingsten.47 If inconsistency, in particular, is to have any effect, there must be some general agreement on rankings of the status variables used here: occupation, education, income and religious—ethnic prestige. At a national level, there is such a relationship. Individuals, however, do not live at the national level. They instead tend to live in clusters of socially similar people. The effect of this clustering may be to dilute the meaning and impact of inconsistency because individuals whose objective characteristics indicate inconsistency in a national sense may feel completely consistent in their micro-environment. In other words, the measure of inconsistency may be too crude. We should prefer some measure which takes into account these probable community differences in rank order of statuses. Unfortunately, there is no such measure available in our national sample. There is, in fact, a tension between the desire for accurate measurement and the desire to use a national sample. Once again, however, errors caused by this difficulty will prejudice our results against the acceptance of the inconsistency hypothesis by introducing more error. There is no reason to suspect, finally, that these errors are in any way systematic. 30 While the 1968 election is not flawless for the study of inconsistency, it appears that the virtue of having an alter- native Opposed to those usually offered presents an Opportunity which ought not be passed. Voting for Wallace may be construed as a likely response to inconsistency and cross-pressures and the question of whether it really is an empirical one. The Data Base The data used in this thesis were gathered by the Survey Research Center of the Institute for Social Research of the University of Michigan immediately preceding and following the election of 1968.48 The Survey Research Center uses a multi-stage probability sample of the dwellings in the United States.49 This procedure is commonly known as ”cluster" sampling and greatly reduces the cost of the survey. At the same time, however, the method increases sampling error and violates the assumptions underlying several of the statistics employed later. Particularly the F-tests in the regression analysis must therefore be viewed with caution. It is not clear how serious this defect is, however, and the decision to utilize statistics for which the assumptions were not perfectly met was made on the conscious calculation that the power of the tests made them desirable at the level of risk implied by using a cluster instead of a simple random sample. This is not an unusual judgment, and given the fact that it is probably impossible to conduct a simple random sample of the American public, the trade-off seems altogether reasonable. 31 The sample has several other limitations which should be noted. First, because the sampling unit is the dwelling unit, the sample systematically under-represents the young, the poor, and the infirm (probably resulting in an under- representation of the old as well). This deficiency is of little consequence here, because none of these characteristics is particularly important to the study. At worst it impairs the generalizability of the findings, but this thesis is not particularly concerned with describing the American population. A second, and more important defect is the non-response rate. The survey has a non-response rate of 15 percent, most of which is accounted for by individuals who were not at home, even for as many as three call-backs, and those who refuse to answer. The sampling procedure does not allow for replace- ment, so these reSpondents are lost. More importantly, they show consistent biases in terms of their presumed social and economic situations. They are disproportionately poorer, probably less educated and of lower occupational status. These variables are relevant in this study and thus these groups are probably under-represented, and our results will not accurately describe the pOpulation. More important here, since our purpose is not to describe the national pOpulation, is the fact that this under—representation may systematically distort the results of our analysis, particularly that of status inconsistency. It may be that we are tapping only a certain type of inconsistent individual. Put another way, it is not inconceivable that some kinds of inconsistency lead to 32 refusal to respond to an interview. Although this danger is present and perhaps plausible (especially given the arguments made later concerning the effects of inconsistency), there is no way to control for this contingency and its effect, if any, must be assumed away. Of course, if such a pattern is present it will prejudice the resulting analysis against accepting the relevant hypotheses. In spite of these likely difficulties, the 1968 Social Research Center Election Study has much to recommend it. Traditionally SRC surveys have been excellent in terms of sampling design, item construction, pre-test and validity check procedures. We have every reason to believe that this survey is as good as possible in these regards. Secondly, a national sample of this type provides a large N for analysis (total N = 1673) and hopefully is not unduly affected by regional variations. Thirdly, the SRC has refined its item sufficiently over the years so that the questionnaire used contained all the information needed for this thesis. And finally, because pre- and post-election interviews were con- ducted, we have the opportunity to compare attitudes with outcomes, offering a better insight into the workings of the independent variables, status inconsistency and cross- pressures. FOOTNOTES 1The major studies in the United States have been reported in: Berelson, Bernard, Paul F. Lazarsfeld, and William N. McPhee, Votin , (Chicago: The University of Chicago Press, 1966); LazarsIeld, Paul F., Bernard Berelson, and Hazel Gaudet, The Pe0p1e's Choice, (New York: Columbia University Press, 1968); Campbell, Angus, Gerald Gurin, and Warren E. Miller, The Voter Decides, (Evanston, Illinois: Row, Peterson and Company, 1954); Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes, The American Voter, (New York: John Wiley and Sons, 1964); Key, V. 0., Public Opinion and American Democracy, (New York: KnOpf, 1967); Pool, Ithiel de Sola, Robert P. Abelson, and Samuel POpkin, Candidates, Issues and Strategies, (Cambridge, Massachusetts: The M. I. T. Press, I964); Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes, Elections and the Political Order, (New York: John Wiley and Sons, 1965). 2The number of studies in other nations is too large to be reviewed comprehensively here. Some representative examples are: Blondel, Jean, Voters, Parties and Leaders, (Baltimore: Penguin Books, 1963i; Converse, PHIIip B., and Georges Depeux, "Politicization of the Electorate in France and the United States," in Campbell et al., Elections and the Political Order; and Janowitz, Morris, "Social StratIfication EEO Mobility in West Germany," The AmeriCan Journal 9f_Socio- logy, 64 (July, 1958). 3This phenomenon is particularly evident in the literature on status inconsistency. See, Lenski, Gerhard, "Status Crystallization: A Non-Vertical Dimension of Social Status," American Sociological Review, 19 (August, 1954), 405-413; KeIIy, K. Dennis, and WlIIlam J. Chambliss, "Status Consis- tency and Political Attitudes," American Sociological Review, 31 (June, 1966), 375-382; Segal,7David R., "Status Incon- sistency, Cross Pressures, and American Political Behavior," American SoCiological ReView, 34 (June, 1969), 352-358; Rush, Gary B., "Status Consistency and Right-Wing Extremism," American SociologiCal Review, 32 (February, 1967), 86-92. 4Generally, this approach was used in the voting studies executed in the 1950's. For example, see Berelson et al., Votin , and Campbell et al., The AmeriCan VOter. Berelson et aI. clearly adopt the view that the resolution of inconsistency is simply a matter of choosing which cue to accept. See 33 34 pp. 129-131 and pp. 200-206. More specifically, Segal, in ”Status Inconsistency," bases his argument explicitly on the "balance" models of Heider and Festinger. 5Of course Lenski's original article implied that incon- sistency effects are non-additive. Several later researchers have made this assumption explicit. See, for example, Treiman, Donald J., "Status Discrepancy and Prejudice," The American Journal 9f SOciology, 71 (May, 1966), 651-664; Boyd, Richard W., ”Presidential Elections: An Explanation of Voting Defections," American POlitical Science Review, 63 (June, 1969), 498-5143 and, for critical comment, Blalock, Hubert M., "Status Inconsistency and Interaction: Some Alternative Models," The American Journal 9f Sociology, 73 (November, 1967), 305-315. 6Lenski, "Status Crystallization." 71bid., 412. 8Gerhard Lenski, "Social Participation and Status Crystallization," American Sociological Review, 21 (August, 1956), footnote 3, 459. WhIle Lenski's original reasoning on the causes of extreme left political partisanship may be generalized to include both left and right because rightist parties as well as leftist do usually prOpose some kind of social change (perhaps regressive), it appears more plausible that a less clearly calculating mechanism is employed in partisan choice by inconsistents. This line of argument prOposes that extremist votes result primarily from stress, at least among inconsistents, and that these votes may be either right or left depending on availability and accept- ability of alternatives. 9Rush, ”Status Consistency," 91. 101bid., 88. 11 Kelly and Chamblis, "Status Consistency." 12William F. Kenkel, "The Relationship between Status Consistency and Political-Economic Attitudes," American Sociological Review, 21 (June, 1956), 365-368. 13Elton F. Jackson and Peter J. Burke, ”Status and Symptoms of Stress," American Soc101ogical Review, 30 (August, 1965), 556-563. 14 Treiman, "Status Discrepancy,” 654. 151bid., 654. 16Ibid. 35 17Jackson and Burke, "Status and Symptoms of Stress." 18Hubert-M. Blalock, "The Identification Problem and Theory Building: .The Case of Status Inconsistency,"'American SOciOlogical ReView, 31 (February, 1966), 52-61. 19 Boyd, ”Presidential Elections." 20Segal, "Status Inconsistency." 21For a discussion of this difficulty, see Blalock, Hubert M., "Tests of Status Inconsistency Theory: A Note of Caution," Pacific Sociological Review, 10 (Fall, 1967), 69-74. 22Alejandro Portes, "Leftist Radicalism in Chile," Comparative Politics, 2 (January, 1970), 251-274. 23Pool, Abelson and Popkin, Candidates, Issues, and Strategies. 24 Herbert Tingsten, Political Behavior, (London, 1939). 5Seymour Martin Lipset, Political Man, (New York: Doubleday, 1960). 26 Ibid., 223. 27Frank A. Pinner, "Cross-Pressures," International Encyclopedia 9f Social Science, 3 (1968), 519. 28Lazarsfeld, et al., The People's Choice. 29Berelson, et al., Voting. 301bid., 131. 31 Campbell, et al., The American Voter, 204. 32Berelson, et al., Voting, 284. 33Lazarsfeld, et al., The PeOple's Choice. 34Campbell, et al., The Voter Decides, 202-204. 35Campbell, et al., The AmeriCan Voter. 36V. 0. Key, The Responsible Electorate, (Cambridge, Massachusetts: The Belknap Press,I1966). 37David M. Kovenock, Philip L. Beardsley, and James W. Prothro, "Status, Party, Ideology, Issues, and Candidate Choice: A Preliminary Theory-Relevant Analysis of the 1968 36 American Presidential Election," Paper presented at the Eighth World Congress of the International Political Science Association, Munich, West Germany, August 31-September 5, 1970. 38Boyd, ”Presidential Elections." 39See, for example, Hyman, Herbert,'POlitiCal SOcial- ization, (New York: The Free Press, 1959IT 40Boyd, "Presidential Elections," 506. 411bid., 506-507. 42 The election has been analyzed several places. See Converse, Philip B., Warren E. Miller, Jerrold G. Rusk, and Arthur C. WOlfe, "Continuity and Change in American Politics: Parties and Issues in the 1968 Election," American Political Science Review, 63 (December, 1969), 1083-1105; Scammon, RicHard M., and Ben J. Wattenberg, The Real Majority (New York: Coward, McCann, and GeoghegafiT—1970I; an Kovenock et al., "Status,Party, Ideology, Issues and Candidate Choice." 43Converse et al., "Continuity and Change," p. 1083. 44Kovenock, et al., "Status, Party, Ideology, Issues and Candidate Choice," p. 13. 45See Lenski, "Status Crystallization," p. 412, for an explanation of this phenomenon in terms of the inconsistency hypothesis. 46Converse, et al.,"Continuity and Change." See also Crespi, Irving, "Structural Sources of the George Wallace Constituency," Social Science anrterly, 52 (June, 1971), 115- 132. 7Tingsten, Political Behavior. 48This is the Survey Research Center 1968 Election Study (SRC 8523). The Intermediate Codebook, January, 1971, was used. 49This procedure is described in Kish, Leslie, and Irene Hess, "The Survey Research Center's National Sample of Dwellings,” Institute for Social Research, The University of Michigan, ISR NO. 2315, (June, 1965), Mimeo. CHAPTER II STATUS INCONSISTENCY AND VOTING BEHAVIOR While the concept of status inconsistency as used here 1 is a direct outgrowth of the work of Gerhard Lenski, the underlying notion can be traced back at least to Weber's 2 writings on the multiplicity of status hierarchies. In 4 and Adams5 had been con- fact, Bernoit-Smullyan,3 Hughes, cerned earlier with the same problems: the behavior of individuals holding incongruent statuses in different hier- archies. Lenski's 1954 article, however, unleashed a torrent of studies of inconsistency, presumably because he specified a method for analyzing the concept.6 Since then disputes have raged over the content of the concept itself and especially over the various methods utilized to measure it. A brief review of each controversy will aid in relating the present study to previous research. At a superficial level, of course, the concept of status inconsistency is clear: it is, as Lenski says, "a consistency dimension. In this dimension units may be compared with respect to the degree of consistency of the positions in the several vertical hierarchies."7 Thus, Lenski labels incon- sistency a "non-vertical dimension of social status." It is 37 38 clear also that it is completely defined by an individual's various positions. In his original article, Lenski hypo- thesized that inconsistents would differ in certain political attitudes from consistents, "when status differences in the vertical dimensions are controlled."8 This type of control as Blalock notes, is not possible.9 Realizing this, Lenski, in a later article, suggested that what is really expected of status inconsistents is a behavior not predictable from the main effects of the relevant status dimensions.10 That is, his proposal is not that inconsistent individuals deviate because of the canceling effects of the status dimensions when taken together, but rather that they deviate more than would be expected by an examination of the additive effects of each status. Moreover, Lenski proposes that the results of status inconsistency are the same regardless Of the type of inconsistency. At this point, status inconsistency detaches itself most clearly from traditional status analysis. Whether, for instance, a person is high on occupation and low on income or vice versa, his response ought to be the same. Lenski's rationale implies an intermediate psychological step between status inconsistency and behavior, and that is internal pressure or stress. Regardless of the type of inconsistency, an individual should feel some psychological stress and the stress will determine his behavior. Lenski also suggested that frequent types of political response would be left voting, and withdrawal, or non-voting. 39 Some have quarreled with Lenski's articulation of status inconsistency. Still others have suggested modifications while endorsing the fundamental thrust of the inconsistency thesis. Kenkel, for instance, reanalyzed Lenski's own data and found no support for the inconsistency argument.11 He pointed out that Lenski's relationships could have been the result of only one status variable--race. Indeed, inspection of Lenski's own tables leads the reader to this conclusion. Kenkel, then, rejected the validity of the concept. His position is supported by Portes, in his analysis of left voting in Chile.12 Portes used multiple regression methods to examine the hypothesis that inconsistent Chileans vote more left than was suggested by an additive model. The hypothesis failed. Portes noted that, in his sample, the relationship might actually be negative. Lenski has also asserted that status inconsistency may lead to right-wing as well as left-wing voting, claiming the stress produced by inconsistency leads to anti-system or anti-status quo reactions.13 This has not been the only suggested modification of the inconsistency hypothesis. After analyzing the Almond and Verba American Data, Segal concluded that the status incon- sistency thesis was contingent upon the saliency of the statuses to the individual.14 Only cases in which the dis- crepancy was clearly apparent to the individual were "true" cases of inconsistency for which the hypothesis should hold. Segal also supported Lenski's argument that the most pronounced 40 inconsistency effects should be found in comparisons of ascribed-achieved status conflicts. This view is also adOpted 15 To a certain extent this by Smith and Broom and Jones. argument parallels the saliency prOposition in that, as Segal notes, ascribed-achieved discrepancies are more likely to be salient for the individual, and as Smith suggests, these discrepancies allow for less individual control. That is, ascribed-achieved status conflicts are more likely to be perceived and less likely to be alterable than achieved- achieved conflicts. Of course, ascribed-ascribed conflicts may be salient as well, but they are often perceived as unalterable. It is not the intent of this discussion to reiterate all the modifications which have been proposed for status incon- sistency theory. Rather this abbreviated summary has been an attempt to illustrate the lack of specification surrounding the concept. While each modification may have merit, I intend to incorporate only one of them in this study. I will avoid most of Lenski's suggested procedures and depart from his early theory in examining voting behavior. I shall adopt Gordon's suggestion that status inconsistency leads to the adoption of extreme positions, whether left or right.16 More specifically, I shall propose that inconsistency leads to higher levels of voting for the candidate of the "third" party in the 1968 election--George Wallace. There is already some empirical evidence that inconsistency is related positively to radical rightist attitudes, and the extension does no harm 41 to the essence of Lenski's original thesis. In fact, Lenski himself has allowed that rightism could be an "expected" reaction to inconsistency.17 In this case, however, it should be clear that the prediction is based on the usual assumption that inconsistency-produced stress will lead to non-normal responses. I am defining Wallace votes as such a response: that is, a vote for Wallace is being viewed as a rejection of the traditional party system and a signal of a desire for some change, either progressive or regressive.18 The debate over the formulation and applicability of the status inconsistency hypothesis pales compared to scholarly exchanges regarding the appropriate methodology with which to investigate the phenomenon. Originally, Lenski constructed an ”inconsistency index" which ranked individuals according to their relative distances from the means of four statuses.19 This method has the apparent charm of straightforward application. The charm is only apparent, however, as Treiman has shown.20 The method is incapable of identifying inter- action which Lenski allows is the crux of the consistency problem. Like simple cross-breaks, while they may be sugges- tive, distributions of scores cannot pinpoint inconsistency effects. This is due, of course, to the fact that cross- tabulations cannot sort out the effects of the main variables: that is, both status variables cannot be controlled simul- taneously.21 Unfortunately all of Lenski's work, and much of that done by others, has used some form of this tabular approach. 42 Several researchers, however, have not. They have employed regression techniques to determine directly whether there is interaction among status variables. Jackson and Burke were the first to use this method.22 By means of dummy-variable regression they demonstrated that an additive model was not fully adequate to explain variance in individual stress levels. They suggested several interactive relationships which better fit the data. This article was a break-through both theoret- ically and methodologically because it used a new and proper method to support a relationship fundamental to the analysis of inconsistency--the relationship between inconsistency and stress. Following Jackson'and Burke the same method has been 23 used (in order) by Treiman to reject hypotheses relating inconsistency and prejudice, by Boyd24 to reject hypotheses relating inconsistency and voting defection, by PortesZS to reject hypotheses relating inconsistency and left voting in Chile, and by Broom and Jones to test hypotheses relating 26 In each case inconsistency and left voting in Australia. the method consists of matching mean scores or cell frequencies predicted by the regression equation to the actual scores or frequencies. Treiman, Boyd, and Portes all found an additive model sufficient. I will employ this technique to examine the relationship between status inconsistency and both turnout and vote. Both additive and interactive models will be examined; and comparisons between the two will be presented. I shall further make use 43 Of a cross-tabulation technique suggested by Jackson in which consistent individuals are.compared to progressively less consistent ones.27 This technique suffers from some of the same limitations as those employed by Lenski. However, it is suggestive of interaction, should such exist, and it allows for a useful comparison with the more sophisticated regression method. Operationalizations and Procedures In this chapter, percent voting for Wallace and percent not voting in the 1968 United States Presidential election will be dependent variables. They are relatively straight- forward. Percent voting for Wallace is merely the proportion of the voters in my sample who claim to have voted for George Wallace. Percent non-voters is the proportion of respondents who answered no to a question inquiring if they had voted, excluding "don't knows," "no answer," and "no answer avail- able" codes. The independent variable is status inconsistency. Status inconsistency is presented in several forms by simultaneously considering sets of two status variables. Four variables are employed: religious-ethnic prestige, occupation, income, and education. Following Jackson28 and Boyd29 religious-ethnic prestige was measured by assigning individuals to categories based on their religious affiliation and ethnic background. Thus Rank I, the highest, consisted of Old English or Old American Protestants; Rank 11 consisted of Northwest European, 44 Protestant or Catholic; Rank III, the lowest,.consisted of Eastern and Southern European and Jews. Everyone else was excluded. Individuals were first classified on their own ethnic status. If they had none, the classification pro- gressed to father's origin and grandfather's origin. Failing this, perceived origin of family was employed. Since no Blacks voted for Wallace, they were all excluded. The measure of occupation used was adapted from the Survey Research Center's Political Behavior Code and corres- ponds roughly to that of the NORC study of occupational prestige.30 Rank I, the highest, includes professional and business occupations; Rank II consists of clerical, white- collar workers and skilled laborers; and Rank III includes semi-skilled, unskilled and service occupations. Doubtful occupations were either excluded or coded in the middle category. Income groups were divided roughly into thirds by cutting at the following points: Rank I (high), over $10,000 per year; Rank II, $6,000 to $9,999; and Rank III, $5,999 and under. On the assumption that family income most fairly represents the level of affluence assigned to an individual by his peers, total family income was employed. Finally, education was measured, following Lenski's suggestion, by establishing different criteria for those over and under forty-five years of age.31 For those over forty-five the categories are: High equals a college education or attended college; medium equals high school graduate and attended high 45 school; low means eight years of schooling. For younger respondents, those who.attended high school and had no non- college training were also assigned to the low education category. In order to test the hypotheses presented in this chapter, cross-tabulations of the statuses and a dummy- variable regression were used. The cross-tabulations were straightforward and no tests of significance were attached, since some of the sub-samples were not independent. The regression was obtained by the use of a computer routine.32 The dummy variable regression procedure is identical to that used by Jackson and Burke, Treiman, and Portes. Dummy variables were established, treating each category of a status variable as a distinct variable with a vaer of zero or one. The additive equation follows the form W = a1 + blxl + bzxz +c1y1 + czyz, where a is a constant, g1 is the coefficient of the first category of 5 (x1), and so forth. This equation is used to generate a set of expected cell proportions which is then compared to the observed proportions to check for inconsistency effects. It should be noted that this equation is only additive and not necessarily linear. Its shape is contingent upon the distribution of the status variables. This technique avoids the complications of identification present in the cross-tabulation. Its inter- pretation is also more straightforward. Finally, although the independent status variables are correlated, the strength 46 of the correlations is quite low (under 0.50) which effectively dismisses the potential difficulty of multi-collinearity.33 In the comparisons of the models examined, F-tests and significance levels are presented to assess the explanatory power of the interactive terms added. The size and signifi- cance of individual coefficients are not reported, since there is no way to report the impact of the middle category. It is always summed into the constant term and thus, depending upon the choice of categories used in the regression, it would be possible to alter these coefficients. In all cases the middle category is assimilated into the constant so that inconsistencies between polar groups could be examined. Status Inconsistency and Wallace Vote Of the several cross-tabulation methods for investigating status inconsistency relationships, Elton Jackson's suggested approach is probably the clearest and most intuitively satis- fying.34 In this procedure individuals are grouped according to whether they are consistent, whether they deviate from consistency by one rank, or whether they deviate by two ranks. Thus persons who are high and high, medium and medium, or low and low on two given dimensions of social status are consistent. The combinations high-medium, and low-medium are labeled one rank deviates, and the extreme permutations high-low and low- high are two rank deviates. These groups are then compared with each other on some dependent variable. In this case, the variable is vote for Wallace. 47 Hypothesis 1, from Chapter I states: Status inconsistency leads to higher rates Of voting for deviant parties than are predicted by an additive model using only the main effects of the status variables. Jackson's argument suggests that, if the proportion of those voting for Wallace increases as the inconsistency increases, inconsistency effects may be present. More importantly, according to Jackson, if the incon- sistent vote for Wallace in proportions exceeding the range established by the consistent, we can infer the presence of inconsistency effects. Table 2.1 presents this relationship. PrOportions in the body of the table are percent Wallace vote. Inconsistency level is as previously described. Type of inconsistency refers to the two possible situations for each diadic status relationship: for example, R/O means cases in which religious- ethnic prestige is higher than occupation, and O/R refers to the inverse. Underlined values are those which fall outside the range bounded by the most extreme consistent groups. Inspection of the table leads to the impression that there is no relationship between inconsistency and Wallace vote. Most Of the two rank deviates fall within the bounds of the consistent groups. Of those which don't, more are bglgg_the range (3) than are above (2). In the case of one rank deviates, the predicted relationship occurs only twice out of twelve Opportunities, although the R/E relationship is particularly deviant. The overall impression gleaned from the table, then, is of little or no relationship. 48 TABLE 2.1: % WALLACE VOTE BY LEVEL AND TYPE OF INCONSISTENCY Inconsistency Two Rank One Rank . Type .Deviates.. . Deviates.. 'Cons1stents $ R/O 20.7 11.2* HH 7.3 '———' ————' MM 9.5 O/R 5.1 8.7 LL 3.1 R/I 6.9 11.0 HH 8.4 MM 9.1 I/R '5.3 9.6 LL 11.5 R/E 7.5 16.9 HH 5.1 ““ MM 8.2 E/R 2.5 8.1 LL 4.8 O/I 10.3 10.6 HH 8.8 MM 13.5 I/O 20.5 13.0 LL 19.3 O/E 13.0 15.2 HH 6.6 MM 13.6 E/O 11.5 14.2 LL 11.5 I/E 15.4 15.1 HH 6.5 MM 14.5 E/I 2.0 11.7 LL 21.9 *Underlined prOportions indicate outside bounds established by consistent groups. $R/O refers to cases in which Religious Ethnic Prestige is higher than Occupation. is employed for all combinations. R Ethnic Prestige, 0 = Occupation, I = Income, and E = Education. This code Religious 49 However, even if the data in the table had appeared to indicate interaction, we would be cOmpelled to use extreme caution in interpreting the results for several reasons. First, the method represents only a rough test of whether relatively large inconsistency effects are present. Values close to the boundaries are not considered significant, although they might well be so. By corollary, the table does not allow unambiguous simultaneous interpretation of scores which are either "too high" or "too low." That is, we cannot say that very low scores by inconsistents represent a "liberal" or "pro-status-quo" or "pro-system" response, even though the table suggests that some inconsistents vote less for Wallace than their social characteristics might warrant. Clearly, not voting for Wallace is not the conceptual complement to voting for him. Secondly, the table has the common cross- tabular difficulty of grouping inconsistents who are sub— stantially different. Within the one-rank deviate category are contained bOth high-medium and medium-low inconsistents. There is no theoretical reason to expect these two groups to act identically. Actually, we should expect them to act somewhat differently if only because the marginal distribution of respondents do not show equal numbers of Wallace voters in each status subgroup. Thus the marginal distributions cannot be ignored in interpreting the table. A third and important qualification of the results of the table also concerns the marginal distributions. If the marginals are not both roughly linear, interpretation of the table becomes even more complex. 50 In this case it is not necessarily proper to expect the consistent cells to set the extreme bounds for the table. It may well be that some inconsistent configuration would be expected to be the most (or least) extreme group, merely on the basis of the marginals. In this case the table distorts the data in the worst possible manner: toward confirmation of an hypothesis which may not be correct. In fact, the 35 Only marginal relationships are not all linear here. occupation and education are. Thus both of the high values displayed by the two rank deviates are probably artifacts of their distributions, although even this knowledge of the marginals does not entirely eliminate ambiguous interpretation. It is worth noting, in this context, that the only cross- tabulation of variables with linear additive marginals (occupation vs. education) shows no evidence of extraordinary interaction effects. Finally, if it is clear that this cross- tabulation cannot confirm the inconsistency hypothesis, it is also clear that it cannot disconfirm it. The table, under the best circumstances, can only suggest the presence of inter- action and inconsistency effects: it can never dgpy their presence. For this task a more sophisticated technique is required. Multiple regression is a statistical approach which allows us to sort out the effects of status inconsistency more adequately. By examining the added explanatory power of inter- action terms in additive equations, we can test the hypothesis more precisely. If it is correct to conceive of status 51 inconsistency as interaction of status variables over ranges of a dependent variable, then multiple regression with the addition of interaction terms is a prOper technique for its analysis. Tables 2.2 through 2.7 present the results Of the dummy- variable multiple regression analysis for sets of two inde- pendent status variables. Reading from the left, the tables display the observed proportions of Wallace vote for each combination of statuses, the expected proportion given an additive model, and the differences between the expected and Observed proportions. Positive deviations indicate larger proportions than would be expected by a consideration only of the individual status variable main effects. If the status inconsistency argument holds, we should expect to find such positive deviations from the expected prOportions in the inconsistent cells. These have been marked with an asterisk. In Table 2.2 we do indeed find such a relationship. Both the high R/O and the high O/R inconsistents vote for Wallace in larger numbers than predicted. This can be taken as tentative evidence that interaction is significant here. It is especially impressive to note that the relationship holds for both types of inconsistency, since Lenski's original thesis disregarded type. In Table 2.3 results are mixed, and both differences are small. Although in Table 2.4 results are again mixed, the differences are quite large. The largest difference in any of our tables occurs here for the medium religious-ethnic prestige--low education group. This is 52 H «co 2 Hmpoe mHHOU pampmfimqouaH Saamwme Nowfl.e- omm.o-Nm¢N.u+HmeH.n+mN.w n 3 ”:Ofiumzco menu kn poumpoqow who: mcofiuuomoum wouuficOHm» fifimv mNeO 55mg. H.m- o.~- «u.m m.w m.mH e.mH N.m m.NH H.HN 304 Home ”Nev fiva n.o m.o o.o- H.N e.m m.m w.~ n.m 5.x. suave: Anny aheav heady «m.m m.H m.~- m.H o.m m.m m.m m.oa «.0 cm“: 303 eez emu: 304 e82, ewe: :03 em: emu: mcofiuhomonm Mama wouufipoum-vo>nomno amnofipuomoum wmuufiwowm m:0fluHOQOHm wo>uomno Hmcowpmmsuuo xcem ewfimepm. 35pm 3.0.1.”me mHO> mummmmo .QmHUHQmmm ZOHHHOmno mm:OfiHHOQOHm pouuapogm mcofiuHOQOHm pm>aomno anunH Mama.emapmoem.beenpm.m:OHwHHem. .mHO> mummmmo .QmHUHQmmm MZOUZH mz< moHHmmmm qumHmImDOHUHAmm mom “m.N mamHomno . meOHpHOmOHm wouuflwmhm . mQOHpHOQOHm wo>HOmno Hanowwmoswm xemm.omfipmmce.Oficnum,m:onHHom. one mummmmo .mmeonmme ”4.N mqmhomno amq044409044 wouufiwOHQ mGOepHOQOHA po>HOmno anunH Mcwm.am:owpmmauuo mZOUzH mz< ZOHH mu<44mmmmo .mmeu4nm44 ”4.4 m4m<4 56 coo n z HmHOH 44460 acoumflmnouaH zanmwmg 4 4 4444.4+4444.4- 004.4- Omm.4-4v.44 u 3 ”qoflpmsco 44;“ kg woumhodom who: mcofluhomoum wouufiwoumm 4444 4444 4444. 4.4 4.4- «4.4 . 4.44 4.4 4.4 4.44 4.4 4.44” 304 4444 44444 4444M 4.4- 4.4 m.o- 4.44 4.4 4.44 4.44 4.44 4.04. 554402 4444 _4444 44444+ 44.4 4.4 4.4- 4.4 4.4 4.4 4.44 4.4 4.4 . 444m :04 we: 444m w46.4 442. 4444 . 344 we: 444: 4:044404044 _ ... .... _ . _ _ _ ... . _. popuprHm-wo>nomno 44:044404044 wouofiwoum meowphomoum po>HOmno mama Hmcomumuscm 4:4m4meo4pmmwouo. ZOHH mu<44<3 w Ammmmo .mmBUHQme no.m mqm40mno 40:044404044 404044044 4:044404044 w0>40mno 44:0440034m 0444.34. 94400444. . ZOHH mu<44<3 4 4mmmmo “QMHUHDmmm ”4.N mqm mu mU mu muHomno mmzowuHOQOHQ wopofiwoum mnofiuHOQOHm ©o>momno x:mm,owflpmonm oflnnum-mzowmwaom Hm:0fipmmsooo ZOHBuzoz w Ammmmo .mmHuHQmmm ”mH.N mqm<9 70 005 z Hmuoe mafiou ucmwmflmaoocH 3HAMfi:« N H NHw¢.¢H-HHow.NN- mmo.o- mmm.a+0a.mm u H "cowpmsco wasp kn wopmhocom 0903 mcoflphomohm wouofiuohme ( .xsns . me0._ Away H.H m.m «0.m- N.~m H.Nm H.vm m.mm 0.0m 0.0m 304 momv mAHHV mNHHv 0.0- 0.H- 0.m n.5H 0.5a 0.0a H.HH N.0H N.mm 5:000: ,..~4030 . 33°30 Ammfiv «H.m 0.0- 0.0- v.m m.m m.HH m.NH v.0 0.0H swam :04 em: 0000. 304. . .,0oz . ‘ .0w0:. :00 1.4 002.. swam. mnemonomoam. .... . _. 6.. .. . .., ...... ..._.__. ..,. ‘ xqmm 0000a00nm-00>a0m00 amqofiuhomohm wouowwoum maofiphomoum 00>H0mno , oaoqu gnaw omflpmmhm Ufleepm-msoflmfifiom mZOozH Qz< mOHHmmmm Uszem-mDoHqumm mom mmth>-Zoz w Ammmmo .QmHUHQme "QH.N mqm00000 , 000000000000 000000000 00000009000 00>00mno . Hmcofiumuswm 0000.00000000.000000-0000w00om ZOHH-zoz w Ammmmo .mmeuHmmmm ”mH.N m0m<9 72 000 u 2 00000 00000 000000000000 00000m0 0 0 N 0 Hmo.m0- 00m.ma- 000.m- ooo.m-0o.om u a ”nowpmscm 00:0 0n woumaonow 0003 0:00000m00m umuuwwohmm < . 0000 . 0000 0000 m.0 0.0 «0.0- o.om 0.0m 0.00 m.0m m.mm 0.00 3o0 ..,.0000.4. 0000 . 0000 m.0- «.0 0.0- o.mN o.m0 0.m0 m.0~ m.om m.~0 £50002 ;....,.A0000 .0000 00000 «0.0- 0.0- 0.0 0.00 0.00 0.0 0.00 0.00 0.0 000: 300 002 000:. 0 :00 . 0 002, ,. 000:. ....,:o0.,,4 002. , 000m WHHOfin—JHOQOHQ . ,.. ., ......... ,. . .......... . . . . . v—Gmm wm000wmpm-wo>pmmno amflofiu00900m wouufiwmhm macauhom00m wo>Hmmno oEOUGH 0c0m 000000000000 mZOozH nz< ZOHH-zoz w Ammmmo .DmHUHQmmm ”0H.N mqmHomno amCOHpHOQOHm umuuflwmhm macauhomoum wm>Hmmpo Hanoflumuswm Mama Ha:OHpmmdooo mmmho>uzoz w ZOHHmmmmo .mmeonmmm ”NH.N mqmHmmno wmcofiwHOQOHm copufiwmgm m:0flpH090Hm wm>Homnow chofipmoswm Mama mEoonH ZOHHuzoz w Ammmmo .mmBUHQmmm umH.m mqm<9 75 we recall that this high-occupation--low-education cell was more pro-Wallace than expected. These inconsistents both vote more often and vote more for Wallace than is expected. Earlier we suggested that this would not be the case, that non-voting and Wallace support would coincide. Actually, this relation- ship does not exist in the data. A slightly negative tendency is observed, seven of twelve comparisons being inverted. This finding will be discussed more thoroughly later. Considered as a set, Tables 2.13 through 2.18, then, do not provide support for the inconsistency hypothesis. For the inconsistents, the average deviation is actually negative, both for moderate (-O.53) and extreme groups (-0.73). Con- versely the consistents average less turnout than is expected (+0.2). This is, of course, precisely the Opposite pattern from that which is predicted. It appears then that the hypothesis cannot be accepted, and should, in fact, be inverted, as inconsistency seems to produce marginally higher turnout rates than expected. Perhaps this is a more logical exten- sion of the original Lenski formulation as well. If incon— sistency leads to stress, presumably voting is one way of relieving that stress. If Lenski's argument concerning desire to change the social order has any merit, voting might appear a logical result of inconsistency.37 It should also be noted that inconsistency effects are more likely to occur among the cells generated by ascriptive- achievement status comparisons. This fits well with the findings on Wallace voting, but is, of course, overturned if 76 the hypothesis is reformulated. The results can be viewed either as evidence of the need for reformulation or as mixed results indicating the importance of achieved-ascribed status differences. No final determination can be made here. As with Wallace vote, we can examine the residual pro- portions of observed minus expected voters for combinations of three independent variables. This relationship is presented in Tables 2.19 through 2.22. The results of this analysis for turnout parallel those for Wallace vote. Again the hypothesis predicts positive residuals in the highly incon- sistent cells. However, this expectation is not consistently confirmed in the tables. In fact only half the residuals are in the predicted direction, not an inspiring result. It seems, then, that inconsistency is no more strongly related to turnout than it is to Wallace vote, and in neither case is the relation- ship significantly other than random. Of course, these tables often suffer from small cell sizes, but even where cell frequencies are sizable, the predictions are not consistently supported. There is also no discernable pattern to the positive or negative residuals, i.e., no particular sets of statuses seem to operate in the expected fashion more frequently than others. Earlier we suggested that there might be an inverse relationship between Wallace vote and turnout and that for inconsistency types which do not fit our expectations on vote we could expect to find that they would fit on turnout, since Wallace vote and non-voting may be functionally alternative 77 mHHou ucmomfimcoucH sfinmfimg so“ mo 2 Hmpou a no woman mu mflnmu whey“ o.OH+ m.~- «m.H- m.m+ w.w+ m.m- «m.o + H.o- «m.¢H- 304 o.v - m.m- m.N+ o.m- ¢.o- 0.0- o.m + v.m+ H.o - Esfivoz «v.om+ m.n+ *N.H- o.H- w.m+ m.m- «m.NH- H.m- w.H - swam :04 em: awn: so; em: awn: :04 em: swam Mama coaummsooo :owummsooo :ofiummzooo 304 anacoz gm“: oEoocH omwumopm owcsum-msowwfiHom mmZOUzH Qz< .ZOHH 0: nmme 30H OCH WOHOGQMU mwcoaovcfl sumo op mofifimmm ohzwooohm mafiaaonma 0500 one 20 m m .omaumohm oficapm-m:oww040m mo zuomoumo cmwn may mouoaow mmn Amy 030 Eowoonm mo moohwoc mo .mfimn opv moohwov genes: map woonwoh magma coauomhopaH 0:0 mommo 440 :40 040.0u0 m040.0u0 000.000 0450.0u0 00044o+4m440o+ 000.0u0 0040.0u0 000.000 0000.0n0 004400+4m0440+ 400.000 0000.000 000.0»0 0040.000 40040o+0m444o+ 4m + 04 + 44 + 0o + 40 000.000 "000.0u0 000.0u0 0045.000 000440+4m4400+ 000.0u0 ”000.0u0 000.0u0 0005.0u0 004400+4m0440+ 050.000 0000.000 000.000 0004.0u0 400400+0m4440+ 4m + 04 + 44 + 00 + 40 400.000 0050.000 054.000 0505.400 000040+4m4000+ 400.000 0040.000 400.000 0000.000 004000+4m0040+ 000.0u0 0004.0u0 000.000 0550.0u0 400000+0m4040+ 4m + 00 + 40 + 00 + 40 405.000 0400.0u0 040.0«0 0040.000 040040+444000+ 000.0u0 m000.034 040.0u0 0000.4u0 044000+440040+ 000.0u0 0000.0u0 000.000 0005.0+0 440000+044o40+ .44 + 0o + 4o + 00 + 40 ouo> oumHHmz 0:0:059 mahoe :ofluomhoch ncowpwscm o>0uwww< mmdeH ZOHHU QmZHm4 Qz< mhmmb+m umN.N mqm<9 83 Summary and ConcluSions The most obvious comment on these status inconsistency data is that very little evidence of the expected effect exists. None of the multiple regressions, including that with all four independent variables, was able to explain more than 8.7% of the variance in Wallace vote. There was no expectation that a great deal of variance would be explained, but 8.7% is still very little. More important, of course, is the fact that interaction terms, when added to the additive equations, failed to increase the explained variance measur- ably. Although the process of dummy variable regression makes the interpretation of significance tests for beta-coefficients a very delicate procedure, very few of these tests attained significance. Only rarely did an interaction term matter, and there were sufficient numbers of them so that this result could easily have happened by chance. From a methodological vieWpoint, the analysis demon- strated that tabular presentation of these data is usually uninformative, can be misleading and even when essentially correct cannot be relied upon. This method is crude, imprecise and throws out useful information. Obviously the multiple-regression procedure is more accurate and reliable. Although a four-independent variable multiple regression equation was run, it was not reported, principally because the number of cells required made cell frequencies low or zero in too many cases (the table has 162 cells and the N was only 760). Its results paralleled those already reported. 84 Moreover, while the three-independent variable multiple regression runs are reported, the problem of small cell Sizes may have been significant here as well. There is no way to be sure of this, of course, but patterns found in the two- variable regression runs cease to exist in the three- variable case. This may be a result of small cell frequencies in the latter. Needless to say, these comments do not apply to the F-tests in Table 2.23 which are not related to cell sizes. Nevertheless, the two-variable regression runs should not be viewed as inferior to the three-variable case and may, in fact, be more appropriate tests of the hypothesis. The theoretical significance of the findings is clear: they lend little support to the inconsistency hypothesis. For the most part, there is no evidence of inconsistency effects. A possible exception to this generalization is the case of ascribed vs. achieved status. As Segal and others note, conflicts between ascribed and achieved statuses might be 38 Both expected to show inconsistency effects more clearly. the two-variable and three-variable regressions indicate that this sub-hypothesis may have merit. The two—variable finding was previously discussed. In the three-variable case, that is where religious-ethnic status was either higher or lower than any two other status variables, there is a consistent relationship. When REP is higher, Wallace vote is lower. For combinations of occupation, education and income, higher than religious-ethnic prestige individualsvote for Wallace more than they ought, assuming these characteristics have an 85 additive effect. This finding lends some support to Segal's argument. It is somewhat puzzling, however, if one recalls that this group consists mainly of Eastern European Catholics and Jews. The conventional wisdom does not include this group among Wallace stalwarts, but they vote more for Wallace than they ought. The other group, with achieved statuses above REP would be more conventionally thought of as heavy Wallace supporters, but actually they provide less than they ought. This is not to say that the first group has a larger prOportion of Wallace voters than the second: it does not. It does, however, support Wallace more than it ought, given its characteristics. The ability of this approach to discover such anomalies is another of its virtues. Finally, we must note that although there is little evidence of interaction effects here, the hypothesis may still be correct. This is possible, of course, because of the indirect test being attempted here. Supposedly the relation- ship between inconsistency and vote is not a direct one. If it is not, there is no a priori reason to expect that a strong relationship will exist between these two variables. A definitive answer to this problem awaits the analysis of data containing stress and voting information. For the most part the findings on turnout parallel those on Wallace voting. Very little variance is explained by any regression. No remarkable patterns of inconsistency appear in the residual tables. The ascribed-achieved relationship found for Wallace vote is not reiterated in the turnout data. 86 Although it was originally prOposed that non-voting might be a functional alternative to Wallace voting among groups of inconsistents, this hypothesis had to be discarded. There was no relationship of any kind between non-voting and Wallace vote. Groups which supported Wallace more than they should have were neither more nor less likely to vote. Hypothesis II, then, has to be rejected in a less qualified way than Hypothesis I. There is no support in the data for it, and only little for Hypothesis I. FOOTNOTES 1Gerhard E. Lenski, "Status Crystallization: A Non- Vertical Dimension of Social Status," American Sociological Review, 19 (August, 1954), 405-413. 2H. H. Gerth and C. Wright Mills, From Max Weber, (New York: Oxford University Press, 1958), sep. chapter 7. 3Benoit-Smullyan, "Status, Status Types and Status Inter- relationships," American Sociological Review, 9 (1944), 151-61. 4Everett C. Hughes, "Social Change and Status Protest: An Essay on the Marginal Man," Phylon, X (First Quarter, 1949), 58-65. 5Stuart Adams, "Status Congruency as a Variable in Small Group Performance," Social Forces, 32 (1953), 16-22. 6 Lenski, "Status Crystallization." 71bid., 405. 81bid., 406. 9Hubert M. Blalock, Tests of Status Inconsistency Theory: A Note of Caution, Pacific Sociological Review, 10 (Fall: 1967), 69-70. 1OGerhard E. Lenski, "Social Participation and Status Crystallization," American Sociological Review, 21 (August, 1956), 459. 11William F. Kenkel, "The Relationship between Status Consistency and Politico-Economic Attitudes," American ' SOCiologiCal‘ReView, 21 (June, 1956), 365-368. See also Lenski's reply immediately following in which he suggests the interactive approach to inconsistency. 12Alejandro Portes, "Leftist Radicalism in Chile,” Comparative Pelitics, 2 (January, 1970), 251-274. 13 Lenski, ”Social Participation." 87 88 14David R. Segal, "Status Inconsistency, Cross-Pressures, and American Political Behavior," American SoCiological Review, 34 (June, 1969), 352-358. 15Thomas Smith, "Structural Crystallization, Status Inconsistency and Political Partisanship," American Sociological Review, 34 (December, 1969), 907-921. Broom, Leonard andvF. Lancaster Jones, "Status Consistency and Political Preference: The Australian Case," American Sociological Review, 35 (December, 1970), 989-1001. Broom and Jones suggest that status inconsistency may occur only in certain societies, and only when structural characteristics permit. Australia does not seem to be such a society. 16Milton Gordon, in a conversation with Lenski, reported in Lenski, "Social Participation." 17 Lenski, "Social Participation." 18For a similar interpretation, see Crespi, Irving, "Structural Sources of the Geroge Wallace Constituency,” Social Science Quarterly, 52 (June, 1971), 114-132. 19 Lenski, "Status Crystallization.” 20Donald J. Treiman, "Status Discrepancy and Prejudice," American Journal gf_Sociology, 71 (May, 1966), 652. 21Blalock, "Tests of Status Inconsistency Theory." See also Blalock, Hubert M., "Status Inconsistency, Social Mobility, Status Integration, and Structural Effects," American Socio- logical Review, 32 (October, 1967), 790-801 especially. Also see Hyman, Martin D., "Determining the Effects of Status Inconsistency," Public Opinion Quarterly, 30 (Spring, 1966), 120-129, for a discussion of some flaws in cross-tabulation analysis of inconsistency. 22Elton Jackson and Peter J. Burke, "Status and Symptoms of Stress," American Sociological Review, 30 (August, 1965), 556-563. 23 Treiman, "Status Discrepancy," 651-664. 24Richard W. Boyd, "Presidential Elections: An Explanation of Voting Defection," American Political Science Review, 58 (June, 1969). 25 Portes, "Leftist Radicalism." 26Broom and Jones, "Status Consistency." 27Elton Jackson, "Status Consistency and Symptoms of Stress," American Sociological Review, 27 (August, 1962), 469- 480. 89 281bid. 29Boyd, "Presidential Elections." 30See Survey Research Center 1968 Election Study, Intermediate Codebook, July, 1969. 31Lenski, "Status Crystallization," 407. 32See North Texas State University Statistical User's Guide, Version II, September 1, 1971. 33Hubert M. Blalock, "Correlated Independent Variables: The Problem of Multicollinearity," Social Forces, 42 (December, 1963), 233-237. 34Jackson, ”Status Consistency." 35For Religious-Ethnic Prestige the rank order of frequencies is Middle, High and Low for the High, Medium and Low categories. For occupation the ranks are identical. For income the ranks are arranged as they are for Religious-Ethnic Prestige. And for Education the rank order of frequencies is Low, Medium and High for the High, Medium and Low categories. 36See, for example, Jackson and Burke, "Status and Symptoms of Stress." 37Lenski, "Status Crystallization," 412. For empirical comment of preference for social among inconsistents, see Goffman, Irwin W., "Status Consistency and Preference for Change in Power Distribution," American Sociological Review, 22 (June, 1957), 275-281. 38Segal, "Status Inconsistency." See also Smith, "Structural Crystallization." CHAPTER III CROSS-PRESSURES AND POLITICAL BEHAVIOR The Concept of Cross-Pressures Previously I argued that status inconsistency and cross- pressures were conceptually similar. According to this thesis they differ only in the level of analysis involved: status inconsistency refers to demographic characteristics of individuals while cross-pressures has been used alternately to signify non-cumulative associational cues or non-cumulative attitudinal orientations. While the level of analysis clearly differs, the underlying dynamics of inconsistency resolution appear to be the same. That is, for each variable, the inconsistent are expected to resolve their non-cumulative "pressures" in one of two ways: by tension release, "cathartic” behavior, or by a decision which "balances" the forces and decides on the strongest. This approach was tested for status inconsistency in the last chapter. In this chapter I turn to the analysis of affiliative and attitudinal cross-pressures. Before presenting the analysis, however, an extended digression is necessary. The conceptualization of status inconsistency presented in Chapter II differed from previous studies primarily in the method used to investigate it. This is not the case with cross-pressures. There is a logical 90 91 disagreement with previous research at the core of the recon- ceptualization of cross-pressures to be presented. This difference can best be seen if we examine the assumptions underlying previous uses of the cross-pressure variable. A brief review of these studies reveals, first, that cross- pressures has been applied in various ways. Usually it con- cerns the "factors" influencing an individual's vote. These may all dispose a person to vote but one way, or they may not. -If not all the factors disposing a person to vote lead to the same conclusion, he is Said to be cross-pressured. Ih§_ PeOple's Choice was the first study to use the cross-pressure notion.1 In the investigation of the 1940 election in Erie County, Ohio, Lazarsfeld and his colleagues defined cross- pressures as, "the conflicts and inconsistencies among the 2 The factors which factors which influence vote decision." constituted the stuff of cross-pressures in this study were: religion, socio-economic status, occupation, community group identification, 1936 vote, 1940 vote, family political orientation, friend's political orientation, attitudes toward business and government and party preference. This list includes variables whichare here labelled status: occupation and socio-economic status; others which are labelled affiliative: family political orientation, friend's political orientation; and others still which are (or would be) labelled attitudinal: community group identification, attitudes toward business and government, and party preference. The tendency in cross- pressure analysis to lump very different characteristics 92 together is both frequent and unfortunate. While The People's Choice generally confines its cross-pressure analysis to cases which operate at the same level of analysis this is not always the case and the failure to separate levels is no doubt attributable to the inadequate conceptualization of cross- pressures.3 The reader gains the impression that these factors have equal potency and may be paired regardless of level of analysis differences, which are never explicitly noted. In fact, the authors of The Pe0ple's Choice leap levels when contrasting associational cues and party preference. Nowhere are the relationships between levels discussed. If a person is cross-pressured at one level, it might be reason- able to expect cross-pressures at other levels. This expec- tation draws no attention. Of course The People's Choice was a pioneering effort and it is no doubt too much to ask that new findings such as those produced by this study be inte— grated into a coherent theory. We must not, however, over- look the implicit assumption embodied in this conception of cross-pressure and its effect on later studies. Cross- pressures here, as later in Voting, referred to structures of contradictory cues, presenting an individual with a "choice" (presumably non-cross-pressured voters had not this choice --their vote was effectively determined by the cue they had). This view assumed the adoption of a linear additive model of cross-pressure effects--one which approximates vector analysis of voting ”pressures." This assumption was not clearly articulated, nor were its implications. Neither was it 93 recognized, apparently, that cross-pressure behavior could have more than one explanation or that it might have needed no Special treatment. The sequel to The People's Choice, Voting, is intentionally 4 more theoretically oriented. This study of the 1948 election in Elmira, New York, led the research team (essentially unchanged from the Erie County group) to the same general findings about cross-pressures. People experiencing cross- pressures once again had difficulty deciding for whom to vote and more often decided not to vote (or at least did not vote as often--these may not be equivalent statements).5 In this case, the level of analysis question is further muddied by the conceptual disregard of these levels and the equation 6 The of within—level effects with between-levels ones. authors note that men appear to have a built-in "strain toward consistency which accounts for the difficulties found by the cross-pressured."7 When faced with inconsistent cues, the argument apparently is, voters are deprived of this consis- tency and this results in tension they wish to avoid. This argument bears a resemblance to that proposed for status inconsistency effects, although it is not so explicit and, perhaps, not so justified, as we shall see. The concept of priority variables introduced by Berelson et al. are also suggestive, since they represent the first attempt to differ- entiate between more and less important factors.8 It might appear that this concept is crucial to an understanding of cross-pressures, and recently Sperlich has argued precisely 94 that.9 Later I will take the position that if a linear additive assumption is maintained, this distinction has little meaning. In Voting, the assumption is implicitly made. A final point must be noted in discussing The People's Choice and Voting. Neither study has any measure of the magnitude of the cross-pressures involved. The only way magnitude can be assessed is by examining the change in the dependent variables. This method is, of course, tautological and therefore untenable. Magnitude, however, may be thought significant, especially if a vector model is conceived as appropriate, or even if linear additive assumptions are made. Cross-pressure predictions come to hinge on the magnitudes of the forces involved. The failure to note or attempt to measure this potential factor would appear important if either model mentioned applies. If either model is applied, however, the magnitude of the cross-pressures may be measured by their effects, as will be shown later, and the problem is hidden. Later it will be shown that magnitude is a concept which is automatically measured by the approach prOposed here. The bulk of the Columbia Group's energy was devoted to assessing the impact of avowedly "social" factors--group memberships, family ties, occupational status--on the political behavior of the voter. A different tack was taken by the Survey Research Center's group. In two volumes and several articles they emphasized the importance of intra-psychic space --the voter's attitudes. In The Voter Decides, Campbell, Gurin and Miller report findings which parallel those of the 95 Columbia group: "When the three motivating factors were not congruent peOple were less clear in their choice of candidates, they were not so likely to vote, and those who did showed the conflict in their motivations by a greater incidence of vacillation, postponement, and divided votes."10 Campbell, et al., assume the same conflict model apparently accepted in Voting.11 Although the authors of The Voter Decides recognize that reactions to cross-pressure may be quite varied, they nonetheless expect them to conform to this basic model of tension and conflict avoidance. In The American 12 Voter, the SRC group did not expand this conceptualization. Their findings were by then routine and the explanation has become widely accepted. The level of analysis is much more explicit here: Campbell et al. are concerned almost solely with attitudinal variables (what they call "proximate" causes). They, moreover, attempt to illuminate the links between affiliative and status cross-pressures and thesaattitudinal ones, although their treatment here is mainly illustrative.13 Finally, they add a dimension of magnitude to their analysis and attempt to 14 assess its effect. Thus the analysis in The American Voter represented the most advanced yet in the study of cross- pressures. Since 1960 several attempts have been made to refine further the concept of cross-pressures. Three of these deserve special consideration. The first is that of Boyd.15 In his study of voting defection (whether or not a person 96 votes according to his partisan preference) Boyd, in effect, reconceptualized attitudinal cross-pressures. He first tested cross-pressures in the standard fashion, by assessing the impact of more and more pressures. His findings here parallel those of the American Voter (some of the data overlap those used in that study) and are not surprising. He notes that this test assumes that cross-pressures operate in an additive fashion. His second test drops the additive assumption and argues that certain cross-pressure combinations lead to non- additive results. This hypothesis is not completely confirmed. Thus Boyd's analysis supports the original conception of cross— pressures, at least to the extent that cross-pressures do not always lead to effects greater than an additive model suggests. Boyd does not, however, completely recognize the damage done to the implicit meaning of cross-pressures by this recon- ceptualization. Sperlich, in another attempt at reconceptualization, argues that cross-pressures as traditionally defined and measured do not produce the results usually claimed.16 His analysis of the 1960 election in the traditional manner showed no evidence of cross-pressure effects. The expected, and by now common, relationships between delay, non-voting and split-ticket voting did not materialize. The reason for this, Sperlich argues, is imprOper and vague conceptualization of cross-pressures. Particularly, the concept was based upon faulty assumptions about the "natural" tension and conflict avoidance mechanism in man and it did not sufficiently consider 97 either the "centrality" (salience) or the magnitude of the cross-pressures. Moreover, says Sperlich, cross—pressure analysis never really demonstrated that cross-pressures lead to any more tension than normal, or that presumed cross- pressures were actually felt. All of these cavils are apparently worthwhile. Certainly the psychological mechanisms presumed operative in cross- pressure situations were simplistic and crude. However, Sperlich himself does not investigate whether these are in fact operative, or whether some non-tension reducing behavior might be important. While he notes that some reactive behavior might be merely "task solving" (choosing between cues), he fails to test whether this alternative or another (like tension reduction) is actually more common. Further, he never questions the additive assumptions commonly made, and incorporates them in his own analysis. Of course this makes his failure to find cross-pressure effects all the more interesting, but it means that his reconceptualization is very limited indeed, amounting to the specification of intervening variables (third factors, as he calls them). In this sense a more promising approach is that made by Horan.17 In his article on cross-pressures he rejects the utilization of an additive model as having no theoretical value. In an argument very similar to that made here in Chapter I, he proposes two models for cross-pressures, one additive, the other interactive. He claims that if the additive model holds, then cross-pressures is a meaningless 98 concept since it has no existence outside of the main effects 18 He finds that an interactive model of the "pressures". explains no more than an additive one. This finding is, however, more relevant for Chapter 11 than here, since Horan's cross-pressures refer to variables I label status ones (again demonstrating the 510ppy usage invited by the term cross- pressures). On the other hand, his effort represents a bridge between the congruent notions of status inconsistency and cross-pressures and further sensitizes us to their common- alities. Another interesting aspect of his study is that it involves reanalysis of the Elmira study and refutes the findings presented there. The major contribution made by Horan, however, is his recognition that the conceptualization made originally in The People's Choice was ambiguous. It allowed for strikingly different interpretations, in fact, assuming one model of behavior, while apparently utilizing one quite different. This ambiguity has, for the most part, been accepted and maintained by other researchers. Hopefully this review has sufficiently illustrated the major previous conceptualizations of cross-pressures. Earlier I noted that the difference in the conceptualization to be presented here and that provided originally by Lazarsfeld, Berelson and Gaudet is primarily a logical one. Basically the argument is that if the cross-pressure variable is used as Lazarsfeld et al. suggest, then it is not a variable. If by cross-pressures they mean merely that the additive effects of the relevant pressures should be considered, then they 99 have invented a term essentially without meaning. Lipset, Lazarsfeld, Barton and Linz, for example, point to four mechanisms which "explain" political participation. They are: l) the degree to which interests are involved in government policies, 2) access to information about the relevance of government policies, 3) group pressures to vote, and 4) cross-pressures.19 If we accept the Lazarsfeld conceptualization there is no need for the cross-pressure variable: it is already contained in group pressures to vote. On the other hand, if we follow Horan and Boyd, cross- pressures take on new meaning, since we now must expect non- additive behavior from individuals experiencing them. Now we are forced to decide whether cross-pressures exist or not, in a fashion very similar to that of the status inconsistency analysis. The key question becomes: "will an additive model suffice?" If so we need only know an individual's character- istics to predict his behavior-~50 special configurations will lead to aberrant behavior. If not, our universe is more complex than previously suspected. If so, cross- pressures do not exist: if not, they do. This distinction may be illustrated graphically by the use of a typical tabular presentation. In this idealized case we have a table with three dichotomous variables which may represent any of the factors mentioned above. 100 Factor C D R Factor B D R D R D I II | II lIII Factor A R III II I II I I Figure 2: Idealized Cross—Pressure Table In this table cell entries refer to diffent degrees of cross- pressure. I means no cross-pressures, 11 means moderate cross-pressures and III means severe cross-pressures, assuming, as is customary, that Factor A is the most "central" variable (e.g. party orientation). The traditional cross-pressure hypothesis regarding turnout, for example, would propose that turnout declines from Type I to Type III, probably monotonically although this stipulation may be ignored. Thus Type I has high turnout, Type II less, and Type III the least. The posture adopted here suggests, on the other hand, that relative turnout is of little consequence. What matters is whether an additive model will explain such a table. In this conception of cross-pressures we would only claim to have found cross- pressures effects if, for instance, Type III individuals were less likely to vote than expected, given their characteristics (Factors A through C). This formulation rescues the concept from the conceptual muddle currently surrounding it. Perhaps this is the usage originally intended. Whether it is or not is of little consequence, since, as has been demonstrated, it is not the one usually applied. The argument here is that the 101 normal usage presents cross-pressures as a descriptive term: i.e. one which refers only to certain configurations of factors. In no way is this usage consonant with the present one which suggests that cross-pressures be used as a variable. A final disclaimer is necessary here. None of this, of course, represents an effort to measure cross-pressures. An inferential leap is necessary to accept the existence of cross-pressures (or deny it) in any case, since we have no direct measure of them. Cross-pressures in this conception becomes a state of mind, akin to attitudes, but perhaps even less susceptible to measurement. Operationalizations and Procedures This chapter presents an analysis of three dependent variables: voting turnout, Wallace vote, and major party vote. The first two have already been described. The last is included here to allow for a better comparison of the results of the reconceptualization of cross—pressures with previous studies. It is simply the percentage of the sample voting Republican for President (it could as easily have been percent Democrat). This variable excludes all voters choosing another party or candidate, as well as those who didn't vote or refused to answer. There are two independent variables, or sets of them. The first section deals with affiliative cross-pressures. How a person was told to vote by his spouse, his relative(s) or his friend(s) are the core variables. All permutations 102 of these variables are employed, and they are presented in two distinct fashions, one strictly tabular and one based on a dummy-variable multiple-regression analysis. These variables were measured directly by means of an item which asked, ”Who (sic) did your spouse (relative, friend) tell 20 The variables were trichotomized to you to vote for." indicate Democratic—-Humphrey, American Independent--Wallace, or Republican--Nixon cues. Wallace suggestions were assumed not to mean Democrat as well, since the respondent had that choice independently presented and could not choose two alternatives simultaneously. Attitudinal cross-pressures were measured by means of a series of ten questions asked on party and candidate preferences and by selection of ten important issues in 1968 and the individual's perceived partisan stands on these. This procedure parallels that of Boyd, so results should be comparable.21 Scores ranged from -10 to +10 on each variable and were then trichotomized for use in the analysis. The regression analysis presented in this chapter is identical with that used in Chapter II, although it is simpler since fewer variables are involved. The tabular presentations are different, reflecting the different approach taken by students of cross-pressures. Tests of hypotheses employ both methods and an attempt is made to distinguish standard from non-standard hypotheses and tests. Finally, in the case of affiliative cross-pressures, the lack of a sufficient number of respondents necessitated 103 dropping three way cross-tabulations. This does not affect the analysis significantly, since the first two hypotheses are testable in two-way cross-breaks and the second two rely on a dummy-variable multiple-regression analysis for which the N is ample. Affiliative Cross-Pressures and Voting Behavior The concept of affiliative or associational cross- pressures is not a new one. As conventionally employed it asks whether the cross-pressured are more likely not to vote, delay voting decisions, etc. The effect of cross-pressures is usually established by comparing individuals or groups of individuals with inconsistent characteristics With those having consistent ones. The first hypothesis in this chapter is representative. H. 3: Individuals characterized by affiliative cross-pressures will be less likely to vote than those without cross-pressures, and the more so the greater the cross—pressures. Evidence to test this hypothesis is then presented in tabular form. Table 3.1 is such a table. The entries in Table 3.1 are percentages of non-voters in 1968. The N's on which these percentages are based are given in parentheses. The hypothesis expects that individuals with consistent cues will rarely fail to vote, individuals with slightly inconsistent cues will vote less, and individuals with strongly incon- sistent cues will most often be non-voters. Table 3.1 lends some support to this hypothesis. Individuals with two Democratic cues are relatively unlikely to vote, as are slightly 104 TABLE 3.1: AFFILIATIVE CROSS-PRESSURES: PERCENT NON-VOTERS, BY SPOUSE AND RELATIVE CUES* Relative Spouse Democrat Wallace Republican Democrat 23.1 25.0 25.0 (13) ( 4) ( 4) Wallace 0.0 0.0 0.0 ( 5) ( 4) ( 2) Republican 10.0 0.0 0.0 (10) ( 2) (17) *Numbersjllparentheses indicate N on which the percent- ages are based. cross-pressured individuals who were urged to vote for Wallace by either their spouse or relative. Strongly cross-pressured individuals, those urged to vote for both major parties differ markedly in their turnout. Here, as in all the affiliative cross-pressure tables, we must exercise caution in inter- pretation because of the small cell sizes. In any event an explanation of these differences is not essential. The most important aspect of the table is the apparent proclivity for the cross-pressured, on the average, to vote less. This can be measured by borrowing a procedure from the analysis of status inconsistency. If we subtract the percent of non- voters in the non—cross-pressured cells from the percent of non-voters in the cross-pressured cells we have a crude . Ill III’I’I lull] .«llll'lill llll. 105 measurement of the effect of cross-pressures, traditionally conceived. In this case this index is 11.9. Performing a similar analysis on Tables 3.2 and 3.3 yields essentially the same result: for spouse-friend cross- pressures the index is 14.2, and for relative-friend cross- pressures, the index is 44.7. The last value is particularly high, presumably indicating a strong cross-pressure effect. TABLE 3.2: AFFILIATIVE CROSS-PRESSURES: PERCENT NON-VOTERS, BY SPOUSE AND FRIEND CUES* Friend Spouse Democrat Wallace Republican Democrat 15.0 0.0 37.5 (20) ( 1) ( 8) Wallace 0.0 25.0 0.0 ( 3) ( 4) ( 1) Republican 0.0 0.0 8.3 ( 9) ( 4) (12) *Numbers in parentheses indicate N on which the percent- ages are based. Inspection of Table 3.3 reveals that it is somewhat unusual. Although it has no zero-N cells, only the consistent and highly cross-pressured cells contain percentage entries. However, it fits the cross-pressure hypothesis very well, since the cross-pressured cells have high prOportions of non-voters relative to the consistent cells. Tables 3.1 106 through 3.3, then, appear to support H.3. Cross-pressured cells have, on the average, proportionately more non-voters. TABLE 3.3: AFFILIATIVE CROSS-PRESSURES: PERCENT NON-VOTERS, BY RELATIVE AND FRIEND CUES Friend Relative Democrat Wallace Republican Democrat 8.3 0.0 27.3 (24) ( 2) (11) Wallace 0.0 0.0 0.0 ( 4) (10) ( 4) Republican 22.2 0.0 6.3 (18) ( 5) (16) *Numbers in parentheses indicate N on which percentages are based. An interesting sidelight of these tables is the apparent effect of Wallace cues. A Wallace cue was conceptualized as a stimulus intermediate between Republican and Democratic cues. Although these tables do not display the direction of the vote, it appears that a Wallace cue was particularly potent in sending an individual to the polls. This phenomenon is most striking in Table 3.3, in which no one with even one Wallace cue failed to vote (N=25). Perhaps this is an instance in which the magnitude of a pressure is significant. I shall return to this possibility in the later regression analysis. Conceivably, a Wallace cue is perceived differently than a 107 Democratic or Republican suggestion. However, the hypothesis still seems to hold, especially if we center attention only on major party cues. A second interesting aspect of these tables is the emergence of an expected relationship. Consistent Democratic cues are associated more strongly with higher levels of non- voting than consistent Republican cues, in all cases. Since we know Democrats are less likely to vote than Republicans, this finding reassures us as to the validity of the tables. The second hypothesis must be approached even more carefully. A straight-forward interpretation is impossible because of the confounding of independent variables to a necessary degree. The hypothesis states: H. 4: Individuals characterized by affiliative cross-pressures will be more likely to vote for marginal parties than those without cross-pressures, and the more so the greater the cross-pressures. One level of cross-pressures, however, suggests precisely that a person votes for the candidate of a marginal party. We would expect that, in cells cued for Wallace, Wallace vote would rise. Tables 3.4 through 3.6 confirm this expectation. Thus the analysis of cross-pressures centers here on Republican-Democrat and Democrat-Republican cross- pressures only. Only one person in all six cross-pressured cells across the three tables voted for Wallace (N=48). In the consistent cells all of six peOple selected Wallace in 1968 (N=92). This indicates that fewer peOple, proportionally, voted for 108 TABLE 3.4: AFFILIATIVE CROSS-PRESSURES: PERCENT WALLACE VOTERS, BY SPOUSE AND RELATIVE CUES* ‘ Relative_ Spouse . Democrat Wallace . Republican Democrat 0.0 0.0 0.0 (10) ( 3) ( 3) Wallace 33.3 100.0 50.0 ( 3) ( 4) ( 2) Republican 0.0 0.0 5.9 4 ( 9) ( 2) (17) *Numbers in parentheses indicate N on which percentages are based. TABLE 3.5: AFFILIATIVE CROSS-PRESSURES: PERCENT WALLACE VOTERS, BY SPOUSE AND FRIEND CUES* Friend Spouse Democrat Wallace Republican Democrat 0.0 0.0 (17) ( 1) ( 5) Wallace 33.3 100.0 100.0 ( 3) ( 3) ( 1) Republican 0.0 25.0 18.2 *Numbers in parentheses indicate N on which percentages are based. III III. I): II II. III] II) Ill 109 TABLE 3.6: AFFILIATIVE CROSS-PRESSURES: PERCENT WALLACE VOTERS, BY RELATIVE AND FRIEND CUES* Friend Relative Democrat Wallace Republican Democrat 4.5 50.0 0.0 (22.) ( 2) ( 8) Wallace 0.0 40.0 25.0 ( 4) (10) ( 4) Republican 7.1 20.0 13.3 (14) ( 5) (15) *Numbers in parentheses indicate N on which percentages are based. Wallace among the cross-pressured than among those whose cues were consistent Democrat or Republican. The hypothesis suggests exactly the opposite outcome. Again the small cell sizes dictate prudence, but the evidence strongly suggests that people having inconsistent affiliative cues for major parties did not increase support for the marginal Wallace effort in 1968. On the other hand, analysis of data presented later shows that individuals cross-pressured among major parties follow the pattern expected with regard to vote for either major party. That is, when the same analysis uses percent Democrat or percent Republican as the dependent variable the traditional cross-pressure hypothesis is largely confirmed. This would seem to mean that the Wallace vote hypothesis is a deviant case. ‘. Ag. 110 The Wallace vote case has another anomaly associated with it: individuals with consistent Republican cues are more often willing to vote for Wallace than are those with con- sistent Democrat cues. This is surprising when we recall that Wallace was occasionally, especially in some Southern states, viewed as the "real" Democratic candidate. We would then expect that more Democrats than Republicans would select Wallace. Further, other analyses of the 1968 election have demonstrated Wallace's appeal to the least educated, least affluent segments of the society--the working class. These people again are most likely to be Democrats. It would seem that a vote for Wallace would be a greater effort for an individual with consistent Republican cues. This anomaly occurs again the case of attitudinal cross-pressures and will be discussed further there. Apparently hypothesis four, then, must be rejected. It is contradicted by the data. Cross-pressures do not appear to increase the vote for marginal parties. The preceding analysis has followed the pattern tradition- ally applied--investigation of differences between cross- pressured and consistent cells. Now the analysis shifts to an investigation of whether cross-pressure effects amount to a non-additive phenomenon. In this analysis for affiliative cross-pressures, I am handicapped by small cell sizes and the possibility of distortion in comparisons between observed and expected frequencies generated by an additive dummy-variable regression equation. Therefore this section deletes these 111 comparisons, although they have been made. Where their inclusion seems useful they will be mentioned. The first hypothesis to suggest other than additive effects is: H. 5: Individuals characterized by affiliative cross-pressures will be less likely to vote than is suggested by an additive model utilizing only the main effects of the affiliative cues. Since we must avoid direct comparison of additive results with observed ones, this analysis focuses on two elements: F-tests between additive and interactive models, and beta-weights of additive and interactive dummy—variables. Table 3.7 presents the F-tests for interactive versus additive models of the TABLE 3.7: AFFILIATIVE CROSS-PRESSURES, F-TESTS BETWEEN ADDITIVE AND INTERACTIVE EQUATIONS, FOR TURNOUT Affiliative Combination R-Squared F-Value P Spouse- Add: 0.0966 0.1350 0.8740 Relative Int: 0.1006 Spouse- Add: 0.0935 0.3916 0.6776 Friend Int: 0.1048 Relative- Add: 0.0370 3.0801 0.0506 Friend Int: 0.0962 affiliative cross-pressures utilized. For the spouse-relative and spouse-friend combination the F-values are not large and the probabilities are no where near significant, in any sense. 112 In these cases the R-squared for the interactive model is higher than that of the additive one, but the added explanatory power is offset by the loss of degrees of freedom. A different picture emerges in the case of relative- friend cross-pressures. The F-value is large, and the R— squared generated by the interactive equation is significantly higher than the additive R-squared. Apparently in this case the additive equation can be improved by considering inter- action--this relationship supports the hypothesis. The results of inspection of beta-weights for additive and interactive terms in the dummy-variable regression analysis are similar to the F-test results. None of the interactive terms approaches significance. However, only rarely do the additive terms achieve statistical significance. Usually these are on the same order of magnitude as the interactive terms. This suggests that affiliative cross-pressures have little explanatory meaning. Also it appears that affiliative cues in general have little effect on turnout. Again the case of relative-friend is deviant but even here beta-weights for cross-pressure terms are not significant. Certainly analysis of the results of the regression analysis provide no consis- tent support for H. 5: there is little evidence that cross- pressures lead to lower levels of turnout than an additive model would predict. Neither is there any evidence that an additive model is particularly useful. The case of relative- friend cross-pressures does not fit the pattern, and suggests that in some circumstances affiliative cross-pressures may 113 exist. It must be noted that the relative-friend regression had more degrees of freedom than the others, and this may affect the outcome. The analysis for H. 6 follows the same order. Comparisons of tables observed and expected may be misleading (in fact, they suggest interaction) because of the small N. The analysis is more apprOpriately based on the regression results alone. H. 6 stipulates: H. 6: Individuals characterized by affiliative cross-pressures will be more likely to vote for marginal parties than is suggested by an additive model utilizing only the main effects of the affiliative cues. Table 3.8 presents the results of the F-tests for each inter- active versus additive model combination. Clearly adding TABLE 3.8: AFFILIATIVE CROSS-PRESSURES, F-TESTS BETWEEN ADDITIVE AND INTERACTIVE EQUATIONS, FOR WALLACE VOTE Affiliative Combinations R-Squared F-Value P Spouse- Add: 0.5580 0.1521 0.8593 Relative Int: 0.5605 Spouse- Add: 0.4829 1.5311 0.2255 Friend Int: 0.5107 Relative- , Add: 0.1234 1.5405 0.2202 Friend Int: 0.1544 interaction terms has little effect on the amount of variance explained in the case of the spouse-relative models. For the 114 other two combinations the relationship is also not signifi- cant. However it is much closer to significance. The beta weights for spouse-friend and relative-friend show interactive terms at significance levels similar to the additive ones. While not statistically significant they indicate that inter- action effects may be at least as important as the additive ones. On the other hand it cannot be said that the gain in R-squared provided by the addition of interaction terms adequately compensates for the loss of degrees of freedom: We cannot accept the hypothesis on the basis of these data. While they do not completely contradict the hypothesis they provide only the weakest support: the gains in explanatory power could have happened by chance. An additive model apparently is sufficient. In the first two cases the R-square is respectable, suggesting that the dimensions included are important predictors. Wallace vote is, thus, little affected by affiliative cross-pressures, as a variable. Attitudinal Cross-Pressures and Voting Behavior Presumably because data on them are more easily gathered attitudinal cross-pressures have been studied in considerably more depth than affiliative or associational ones. Again, however, the mode of analysis most often applied involves comparisons of cross-pressured individuals with consistent ones. In the case of affiliative pressures this amounted to assumption of an additive relationship and likewise in the case of attitudinal cross-pressures an additive assumption is 115 normally implicit. In this case the tests are considerably more reliable, due to a much larger N and perhaps more adequate measurement. The investigation of attitudinal cross-pressures proceeds in essentially the same fashion as that of affiliative cross- pressures. First, standard hypotheses are examined. Then interactive characteristics are investigated by means of comparisons of observed and expected values, F-tests between models and beta-weight significance tests. The larger N here, it should be noted, makes it considerably easier for an F-value to be significant. Nevertheless, the results may legitimately be compared with those of affiliative cross- pressures. It must be remembered also, that confirmation of traditionally stated hypotheses in no way affects the potential confirmation of the reconceptualized hypotheses. Finally, the measurement of attitudinal cross-pressures does not involve Wallace preferences, simplifying interpretation. The first attitudinal cross-pressure hypothesis parallels that in the affiliative section: H. 7: Individuals characterized by attitudinal cross-pressures will be less likely to vote than those without cross-pressures, and the more so the greater the cross-pressures. Data relating to this hypothesis are presented in tables 3.9 through 3.11. Interpretation of these tables is, presumably, straightforward. The hypothesis predicts that as we move from totally consistent cells to totally cross-pressured ones, cross-pressure effects become more visible, and cross-pressured TABLE 3.9: 116 ATTITUDINAL CROSS-PRESSURES: PERCENT NON-VOTERS, BY PARTY AND CANDIDATE ORIENTATIONS* Candidate Orientation Party Orientation Democrat No Preference Republican Democrat 20.1 27.8 25.0 (309) (144) ( 44) Weak Identifier 36.8 38.3 25.0 or No Preference ( 87) (227) (120) Republican 18.8 25.8 17.3 ( 32) (132) (393) *Numbers in parentheses indicate N on which percentages are based. TABLE 3.10: ATTITUDINAL CROSS-PRESSURES: PERCENT NON-VOTERS, BY PARTY AND ISSUE ORIENTATIONS* Issue Orientation Party Orientation Democrat No Preference Republican Democrat 20.5 30.3 17.7 (303) (132) ( 62) Weak Identifier 38.0 37.7 24.3 or No Preference (100) (223) (111) Republican 22.2 20.9 17.5 ‘( 81) . (191) (285) *Numbers in parentheses indicate N on which percentages are based. 117 TABLE 3.11: ATTITUDINAL CROSS-PRESSURES: PERCENT NON-VOTERS, BY CANDIDATE AND ISSUE ORIENTATIONS*- Issue Orientation Candidate Orientation Democrat, No Preference .Republican Democrat 22.2 26.8 21.9 (252) (112) ( 76) Weak Identifier 30.1 37.7 22.4 or No Preference (143) (244) (132) Republican 21.3 22.1 17.3 ( 89) (190) (310) *Numbers in parentheses indicate N on which percentages are based. have higher prOportions of non-voters than consistent ones. Although differences are small, the cross-pressure index constructed in the previous section and adapted from status inconsistency analysis shows consistently higher prOportions of non-voters in the cross-pressured cells. The net advantages of the cross-pressured cells are, for Tables 3.9 to 3.11 respectively, 6.4, 1.9, and 3.7. These are not large differ- ences, but they are consistent. They support the first part of the hypothesis. The expected effect of the progression from zero to very strong cross-pressures does not materialize, however. Moderate cross-pressures appear to lead to higher levels of non-voting than strong cross-pressures. At this point the 118 possible effect of intensity of attitude must be recalled. Attitudes toward each of our elements, party, candidate and issue were measured in such a way that medium responses almost necessarily include an element of apathy. Thus people with medium cross-pressures are more often little involved, so one might well expect that these cross-pressures lead to high levels of non-voting. There was no way, in these data, to measure intensity adequately, however, so this interpretation must remain tentative. So, while the second half of the hypothesis cannot be either confirmed or disconfirmed definitely, the first has at least consistent support. The cross-pressured vote less often. Note also that the data in Tables 3.9 to 3.11 show a familiar pattern: consistent Democrats are more likely not to vote than consistent Republicans. This is expected and lends validity to the cross-pressure findings. As a secondary check on the hypothesis, the three attitudes were merged in a three-way table. (See Table 3.12) The pattern noted in the two-way cross-breaks re-emerges here. On the average, those in the highly cross-pressured cells are likely to vote, but the highest non-voting rates occur among those in the medium cross-pressured cells. In some highly cross- pressured cells, however, non-voting is quite low. The pattern here is not so consistent as in the two-way tables. This may be due to the fact that a three-way cross-break means that strong cross-pressures include two pressures in one direction and one in another, whereas the two-way tables 119 TABLE 3.12: ATTITUDINAL CROSS-PRESSURES: PERCENT NON-VOTERS BY PARTY AND CANDIDATE AND ISSUE ORIENTATION* Issue Orientation Party Democrat Republican Orientation Candidate Candidate ~ Candidate Dem N P Rep Dem N P Rep Dem N P Rep Democrat 19.8 18.7 37.5 23.4 40.7 21.4 15.2 26.7 14.3 (212) (75) (16) (64) ( 54) ( 14) (33) (15) ( 14) Weak Identif- 39.4 43.8 21.1 34.2 43.1 27.3 37.5 20.4 23.9 ier or No Preference ( 33) (48) (19) (38) (130) ( 55) (16) (49) ( 46) 4.3 40.0 16.2 20.0 23.2 19.8 20.0 23.1 16.1 ( 7) (20) (54) (10) (60) (121) (15) (52) (218) H Republican *Numbers in parentheses indicate N on which percentages are based. have a more stark contradiction. The important fact of Table 3.12 is, of course, that it largely reiterates the findings of Tables 3.9 to 3.11. None of this, of course, asks whether cross-pressures lead to non-voting at rates significantly higher than are to be expected, given the marginal distributions of attitudes. These findings may, therefore, be artifacts of the marginals. We know, for example, that Republican orientations toward any of the objects result in lower non-voting rates. Investigation of this possibility occurs in H. 9. First we focus on the presumed effect of attitude cross- pressures on Wallace vote, in conventional terms. Although 120 no previous research has investigated this variable, the expected relationship is stated in H. 8: H. 8: Individuals characterized by attitudinal cross-pressures will be more likely to vote for marginal parties than those without cross- pressures, and the more so the greater the cross-pressures. Tables 3.13 through 3.15 present data bearing on this hypo- thesis. Table 3.13 is particularly edifying. It shows the [1 classic cross-pressure pattern: both high cross-pressure :- cells contain greater proportions of Wallace voters than either consistent cell. Clearly the difference between the cross-pressured and consistent cells is large and in the predicted direction (18.6). The other tables are less ideal, TABLE 3.13: ATTITUDINAL CROSS-PRESSURES: PERCENT WALLACE VOTERS, BY PARTY AND CANDIDATE ORIENTATIONS* Candidate Orientation Party Orientation Democrat No Preference Republican Democrat 2.9 2.0 16.1 (245) (102) ( 31) Weak Identi- fier or No 11.5 22.2 27.9 Preference ( 52) (126) ( 86) Republican 15.4 8.2 10.0 ( 26) ( 97) (321) *Numbers in parentheses indicate N on which percentages are based. TABLE 3.14: 121 ATTITUDINAL CROSS-PRESSURES: PERCENT WALLACE VOTERS, BY PARTY AND ISSUE ORIENTATION* Issue Orientation .Party. Orientation Democrat No Preference Republican Democrat 0.8 4.4 16.0 (237) ( 91) ( 50) W93k Identi’ 5.5 26.9 25.3 fier or No ( 55) (130) ( 79) Preference Republican 9.5 7.4 11.6 ( 63) (149) (232) *Numbers in parentheses indicate N on which percentages are based. TABLE 3.15: ATTITUDINAL CROSS-PRESSURES: PERCENT WALLACE VOTERS, BY CANDIDATE AND ISSUE ORIENTATIONS* Issue Orientation Candidate Orientation Democrat No Preference Republican Democrat 1.0 11.2 12.0 (193) ( 80) ( 50) ”fak Identi' 5.3 13.8 15.1 fier or No ( 94) (145) ( 86) Preference Republican 5.9 14.5 16.0 ( 68) (145) (225) *Numbers in parentheses indicate N on which percentages are based. 122 perhaps, but the cross-pressure index is positive in each case, 13.1 and 0.9 respectively. These data support the first half of H. 8. The findings on the second part are more complicated. The pattern found among non-voters is not so clearly evident: medium cross-pressures are not associated with consistently higher Wallace voting than strong cross-pressures. Only in Table 3.15 is the relationship found in the case of non- voting reiterated for Wallace vote. It would be illegitimate to conclude from this that the hypothesis be confirmed, but equally overzealous to reject it. The data give it some support. Table 3.16 combines the effects of the three attitudinal variables. The pattern of this table is similar to that found in the two-way Splits. The average Wallace vote among highly cross-pressured cells is larger than among the consiStent cells as before. Medium cross-pressures here are character- ized by the highest rates of Wallace voting. Perhaps once again the effect of apathy is present, but this interpretation is not unambiguous. Two further findings merit comment. Both the two- and three-way cross-breaks indicate higher Wallace voting among consistent Republicans than among consistent Democrats. This finding duplicates that discovered in the analysis of affiliative cross-pressures. It also runs counter to what we might expect, given the regional and social class basis of Wallace support as reported elsewhere. The source of this —'. 123 TABLE 3.16: ATTITUDINAL CROSS-PRESSURES: PERCENT WALLACE, BY PARTY, CANDIDATE AND ISSUE ORIENTATIONS* Issue Orientation Party . . Democrat Republican Orientation ..Candidate,.. Candidate ..Candidate Dem N P Rep Dem N P Rep Dem N P. Rep Democrat 0.6 0.0 11.1 4.2 3.1 9.1 14.3 9.1 27.3 — —_< (169) (59) 9)(48) (32) (11) (28) (11) ( 11) Weak Identi- fier or No 0.0 13.0 0.0 20.8 22.1 39.5 10.0 28.6 26.5 Preference ( 18) (23) (14) (24) (62) (38) (10) (35) ( 34) Republican 16.7 16.7 6;Z_25.2 8.9 5.2 ii; 5.0 13.3 ( 6) (12) (45) ( 8) (45) (96) (1 ) (40) (180) *Numbers in parentheses indicate N on which percentages are based. phenomenon cannot be isolated here. Perhaps Wallace supporters, while nominal Democrats, found their candidate and issue preferences more Republican than usual in 1968. To some extent this interpretation is supported by Table 3.13 to 3.15 in which the candidate-issue cross-break produces the highest Wallace vote among Republicans. This explanation cannot be more than speculative, however, as a complete investigation would take us too far afield. A second observation, relevant later, is that cross- pressure effects are more visible when partisan orientation is involved. Both the two and three-way cross-tabulations evidence this. Of course, this is not unexpected, given the 124 findings of The American Voter. In this case it may mean that when individuals experience attitudes contradicting their party preference they are freed to vote for third parties more than if only candidate and issue preferences conflict. Again this interpretation is, at best, only tentative, but the finding is interesting. So far the analysis has proceeded in traditional terms. I have argued, however, that accepting hypotheses such as H. 7 and H. 8 does not establish the existence of attitudinal cross-pressures. To do this requires propositions concerning cross-pressures as a separate variable. The ninth hypothesis is such a proposition: H. 9: Individuals characterized by attitudinal cross-pressures will be less likely to vote than is suggested by an additive model utilizing only the main effects of the attitudinal pressures. Tables 3.17 through 3.19 present data on this hypothesis. These tables contain the observed frequencies, the expected frequencies assuming an additive model, and the differences between the two. The hypothesis predicts that the differ- ences in the highly cross-pressured cells will be positive. That is, observed frequencies will exceed expected frequencies if cross-pressure effects are present. In five of six cases, this expectation is fulfilled, the exception occurring in the case of Democratic party orientation and Republican candidate orientation. Only in the opposite cross-pressure instance is the difference large, however. For the most part, then, the data fit the expectation. It is difficult to set limits TABLE 3.17: 125 ATTITUDINAL CROSS-PRESSURES: OBSERVED, EXPECTED AND RESIDUAL PERCENT NON-VOTERS, BY PARTY AND CANDIDATE ORIENTATIONS Party 'Candidate'orientation Orientation Observed Expected Observed- ., Expected IDem NEP Rep Dem N P Rep. Dem N P Rep Democrat 20.1 27.8 25.0 21.2 27.4 17.9 -l.l 0.4 7.1 Weak Identi- fier or No 36.8 38.3 25.0 32.0 38.2 28.7 4.8 0.1 -3.7 Preference Republican 18.8 25.8 17.3 20.3 26.4 27.0 -l.5 -0.6 -9 7 TABLE 3.18: ATTITUDINAL CROSS-PRESSURES: OBSERVED, EXPECTED AND RESIDUAL PERCENT NON-VOTERS, BY PARTY AND ISSUE ORIENTATIONS Issue'Orientation Party Observed Expectedf ggsgzggg- Orientation p Dem N P Rep Dem N P Rep Dem N P Rep Democrat 20.5 30.3 17.7 22.4 26.0 17.6 -l.9 4.3 0.1 Weak Identi- _ fier or No 38.0 37.7 24.3 33.6 37.4 26.9 4.4 0.4 2.6 Preference Republican 22.2 20.9 17.5 20.6 24.3 15.8 1.6 -3.4 1 7 126 TABLE 3.19: ATTITUDINAL CROSS-PRESSURES: OBSERVED, EXPECTED AND RESIDUAL PERCENT NON-VOTERS, BY CANDIDATE AND ISSUE ORIENTATIONS 'Issue Orientation Candidate Observed Expected Observed- Orientation . _ __ Expected Dem N P Rep Dem N P Rep Dem N P Rep Democrat 22.2 26.8 21.9 22.6 27.6 19.0 -0.4 -0.8 2 KO ”- Weak Identié~ fier or No 30.0 37.7 22.4 30.4 35.4 26.8 -0.3 2.3 -4.4 Preference Republican 21.3 22.1 17.3 19.6 24.5 16.0 ' 1 7 -2.4 1.3 on what prOportion of "correct" differences constitutes the support for the hypothesis. At the least the data do not suggest rejection. In Table 3.20 differences between observed and expected values are presented for an expanded analysis using all three variables. In this case three of the six highly cross- pressured cells are negative. This result could, of course, easily have happened by chance no matter what criteria are applied. However the expanded table also defines cross- pressures somewhat differently than the two-way cross-break and thus may not be strictly comparable. This phenomenon was noted above. While these results must make us even more cautious, they do not allow rejection yet. A more precise analysis is necessary. 127 TABLE 3.20: ATTITUDINAL CROSS-PRESSURES: RESIDUAL PERCENT NON-VOTERS, BY PARTY, CANDIDATE AND ISSUE ORIENTATIONS Issue Orientation Party Democrat . ..., Republican Orientation -.-Cand1date.. ...Candidatek.. Candidate Dem N P Rep Dem N P Rep. .Dem .N P Rep Democrat -l.2 -6.0 l9.2-0.8 10.9 0.0 41.7 4.2 0.1 :2: Weak Identi- fier or No 7.8 6.6 -7.8 0.5 2.7 -4.7 10.5-12.7 0.8 Preference Republican -6.7 13.3 -l.6 -4.1 6.6 1.6 3.1 0.6 1.9 Table 3.21 represents such an analysis: it presents F-tests for the differences in explanatory power of several additive and associated interactive models. If the hypo- thesis is correct we would expect that adding interaction terms would increase the explanatory power of an additive model significantly. In two of the four comparisons, the test is significant. However, in the party/issue comparison, the R-squared of the additive model is higher than that of the interactive one. The same occurs for party/candidate, although the difference is not significant. The hypothesis finds no support here. Nor does an inspection of the beta- weights for individual interaction terms yield support for H. 9. The beta-weights in almost all cases could have occurred by chance. The ones which achieve significance are 128 TABLE 3.21: ATTITUDINAL CROSS-PRESSURES: F-TESTS BETWEEN ADDITIVE AND INTERACTIVE EQUATIONS, FOR TURNOUT Attitudinal Combinations . R-Squared . . . F-Value.. P Party- Add: 0.0285 1.1282 0.3224 Candidate Int: 0.0270 Party- Add: 0.0268 3.7102 0.0246 Issue Int: 0.0220 Candidate- Add: 0.0216 8.6022 0.0002 Issue Int: 0.0328 Party' Add' 0 0297 1 4725 0 1835 Candidate- Int: 0'0355 ' ' Issue ' ° so few that their occurrence is not unsurprising: an occasional significant beta-weight is also to be expected under random conditions. This hypothesis, then, must be rejected. It appears that an additive equation explains the relation- ship between attitudinal cross-pressures and non-voting as well as or better than an interactive one. There is no con- sistent evidence that interaction exists. Of course, the amount of variation explained is miniscule in either case. Thus cross-pressures remains as it was, a descriptive term and not a variable, although it may be that the cross- pressure relationship may take some other form not specified. What shape that might be is not easily imagined, however, especially given the results of this analysis. 129 Perhaps, as others have suggested, cross-pressures are relevant only for some dependent variables. The pro- cedure just utilized for the analysis of turnout will now be applied to vote for preferred party in 1968--voting defection. In this instance our concern is with whether cross-pressures result in higher rates of defection from party preference than an additive model suggests. First, for the sake of completeness, we analyze the relationship in the traditional manner. Tables 3.22 through 3.24 display the two-way cross- tabulations for percent Republican for each combination of cross-pressures. We expect to find deviation from TABLE 3.22: ATTITUDINAL CROSS-PRESSURES: PERCENT REPUBLICAN VOTERS, BY PARTY AND CANDIDATE ORIENTATIONS* Party Candidate OrientationS' Orientations Democrat No Preference ’Republican Democrat - 2.1 12.0 23.1 (238) (100) ( 26) gig: giefigi' 19.6 42.9 75.8 Preference ( 46) ( 98) ( 62) Republican 72.7 91.0 94.8 (22), (89) (289) *Numbers in parentheses indicate N on which percentages are based. TABLE 3.23: ATTITUDINAL CROSS-PRESSURES: 130 PERCENT REPUBLICAN VOTERS, BY PARTY AND ISSUE ORIENTATIONS* Issue Orientations Party Or1entat1ons Democrat No Preference . Republican Democrat 3.8 9.2 14.3 (235) ( 87) ( 42) WEak Identl' 28.8 48.4 62.7 fier or No 2 59 Preference ( S ) ( 95) ( ) Republican 82.5 92.8 95.6 ( 57) (138) (205) *Numbers in parentheses indicat N on which percentages are based. TABLE 3.24: ATTITUDINAL CROSS:PRESSURES: PERCENT REPUBLICAN VOTERS, BY CANDIDATE AND ISSUE ORIENTATIONS* Issue Orientations Candidate Orientation Democrat No Preference Republican Democrat 4.7 11.3 29.5 (191) ( 71) ( 44) gig: igefigi‘ 21.3 52.8 68.5 Preference ( 89) (125) ( 73) Republican 67.2 87.1 93.1 ( 64) (124) (189) *Numbers in parentheses indicate N on which percentages are based. 131 Republican vote as cross-pressures increase, and as we move from consistent Republican to consistent Democrat. These expectations are confirmed. The lowest percent Republican in all cases occurs among consistent Democrats, the highest among consistent Republicans. The relationship appears basically to be additive. The index of cross-pressures is negative in all three cases, indicating that attitudinal cross-pressures lead to a surplus of Democratic voting. In the three-way cross-break essentially the same results emerge. This resembles Lenski's results for status inconsistency. These differences are small, however, and unconvincing. They gain in interest when it is recalled that the Wallace analysis suggested a tendency for attitudinal cross-pressures to lead to right-wing responses. This is puzzling, but it is not inconceivable that both suggestions are correct (they may both be incorrect as well). The type of analysis employed may generate these results consistently and it would be easy to exaggerate their significance. Our major concern here, however, lies with testing the presumed and apparent additive nature of this relationship. The hypothesis in this case encourages us to look for evidence of non-additivity in this relationship. If cross- pressures are a factor, in themselves, in increasing voting defection, then the relationship ought to be interactive. A voter could not have preferred the American Independent Party in this case, so analysis is restricted to Democratic and Republican defection. Hypothesis ten specifies the relationship: 132 H. 10: Individuals characterized by attitudinal cross-pressures will be more likely to vote for their preferred party than is suggested by an additive model utilizing only the main effects of the attitudinal pressures. Data bearing on H. 10 are presented in Tables 3.25 through 3.27. These tables again display observed, expected and residual proportions of Republican voters. If interaction is present and Operates in the predicted direction differences in the cross-pressured cells should be positive. Half of them are. The tables indicate that party orientation is the most significant variable affecting vote, and candidate TABLE 3.25: ATTITUDINAL CROSS-PRESSURES: OBSERVED, EXPECTED AND RESIDUAL PERCENT REPUBLICAN, BY PARTY AND CANDIDATE ORIENTATIONS Candidate Orientation Party Observed Expected Observed- Orientation Expected Dem N P Rep Dem N P Rep Dem N P Rep Democrat 2.1 12.0 23.1 0.1 15.1 30.0 2.0 -3.1 - 6.9 Weak Identi- fier or No Preference 19.6 42.9 75.8 31.7 46.6 60.7 -12.1 -3.7 15.1 Republican 72.7 91.0 94.8 68.4 83.4 97.5 4.3 7.6 - 2.7 133 TABLE 3.26: ATTITUDINAL CROSS-PRESSURES: OBSERVED, EXPECTED AND RESIDUAL PERCENT REPUBLICAN, BY PARTY AND ISSUE ORIENTATIONS Issue Orientation Party Observed Expected Observed- Orientation Expected Dem N P Rep Dem N P Rep Dem N P Rep Democrat 3.8 9.2 14.3 2.1 12.0 18.2 1.7 -2.8 -3 9 ll- Weak Identi- fier or No 28.8 48.4 62.7 38.4 48.3 54.5 -9.6 0.1’ 8.2 Preference Republican 82.5 92.8 95.6 81.0 91.0 97.2 1 5 1.8 -1.6 TABLE 3.27: ATTITUDINAL CROSS:PRESSURES: OBSERVED, EXPECTED AND RESIDUAL PERCENT REPUBLICAN, BY CANDIDATE AND ISSUE ORIENTATIONS Issue Orientation Candidate Observed Expected 3:5::¥:3- Orientation p Dem N P Rep Dem N P Rep Dem N P Rep Democrat 4.7 11.3 29.5 1.0 20.3 31.3 3.7 -9.0 ~l.8 Weak Identi- _ fier or No 21.3 52.8 68.5 30.9 50.2 61.2 9.6 0.6 7.3 Preference Republican 67.2 87.1 93.1 65.2 84.5 95.5 2.0 2.6 -2.4 134 orientation is second. They do not indicate support for the cross-pressure hypothesis. Likewise, the expanded table (3.28) presents results which fail to support the hypothesis. In both the two- and three-way cross-breaks the relationship appears roughly additive. But, of course, appearances can be deceiving. The results of the regression analysis allow a less biased judgment. TABLE 3.28: ATTITUDINAL CROSS-PRESSURES: RESIDUAL PERCENT REPUBLICAN, BY PARTY, CANDIDATE AND ISSUE ORIENTATION Issue Orientation Party Democrat _ RepUblican Orientation Candidate Candldate Candidate (Dem N P Rep ”Dem N P Rep Dem N P Rep Democrat ' 5.1 - 4.1-11.7— 3.0 -2.7 -12.2 -10.0 6.3 0.5 Weak Identi- fier or No Preference -14.5 - 9.2- 2.6-17.8 0.0 17.7 6.9 -7.1 23.6 Republican - 0.9 15.5 3.9- 2.2 5.3 0.8 00 O 7.3 3.9 Table 3.29 presents the results of F-tests for differences in explanatory (R-squared) between associated additive and interactive models. All are significant. This, however, fails to confirm the hypothesis since in the first two cases (Party/candidate, and Party/issue) the additive model has a higher R-squared than the interactive one. While in every case the differences are significant, only in two are they 135 TABLE 3.29: ATTITUDINAL CROSS-PRESSURES, F-TESTS BETWEEN ADDITIVE AND INTERACTIVE EQUATIONS, FOR PERCENT REPUBLICAN Attitudinal Combinations R-Squared F-Value P Party- Add: 0.6213 35.2679 0.0000 Candidate Int: 0.6023 Party- Add: 0.6017 186.6176 0.0000 Issue Int: 0.4675 Candidate- Add: 0.4671 324.8076 0.0000 Issue Int: 0.6296 Party‘ Add- 0 6222 6 5736 0 0000 candldate‘ Int: 0.6320 Issue in the predicted direction. Both the additive and inter- active equations employing all three variables explain about 63 percent of the variance. By the standards of status incon- sistency and affiliative cross-pressure analysis this is high, but the important fact here is that an additive model is as useful as an interactive one. While the difference between these last two models is statistically significant, this is doubtless due more to the c0pious degrees of freedom than to the magnitude of the absolute difference. The individual beta-weight F—tests present the same profile. The additive terms are all significant in every model, while no interactive term in any model approaches 136 significance. Obviously the addition of these terms has little effect on the ability of any model to predict the actual values of percent Republican. These results reinforce the inter- pretation of the model vs. model F-tests. There are no important interaction effects. Thus H. 10 must be rejected. Summary and Conclusions In general, then, this chapter has provided support for the concept of cross-pressures as a descriptive term, but not as a variable, at least under the conditions specified. In the case of affiliative cross-pressures, H. 3, relating cross- pressures to non-voting, was largely confirmed. Individuals who were cross-pressured voted less than those who were not. In the case of H. 4, relating affiliative cross-pressures to Wallace support, the hypothesis was not supported. In fact the data suggested that non-cross-pressured individuals were more likely to vote for Wallace than cross-pressured ones. Of course Wallace cues uniformly increased the likelihood of a vote for Wallace. Medium cross-pressures do not fit the expected pattern. In the case of turnout medium cross- pressures were associated with very low levels of non—voting. This was not expected, but perhaps it is due to the salience of a Wallace cue. Certainly being urged to vote for a small, ostensibly radical party may indicate higher politicization and it may be that the magnitude (force) of such a cue is greater than one for a major party. None of the findings relating to H. 3 and H. 4 prove or disprove the existence and/or effect of cross-pressures as 3—_*“TH 137 a variance. For this task F-tests and beta-weights for the dummy-variable multiple-regression were analyzed. For both dependent variables, turnout and Wallace vote (H. 5 and H. 6) the results were identical: there was little support for the non-additive hypothesis. The only exception to the general pattern occurred in the impact of relative-friend cross- pressures on turnout. Here an interactive model was clearly more potent than an additive one. While interesting, this finding alone could not save the hypothesis. All of the results of the affiliative cross-pressure analysis must, of course, remain tentative since the N's involved were small, varying from 51 to 94. A more convincing test of these hypotheses awaits better data. The traditionally formulated hypotheses regarding atti- tudinal cross-pressures were largely supported. Tests of H. 7 and H. 8 indicate that the cross—pressured are less likely to vote and more likely to vote for Wallace than their consistent cohorts. In the turnout case medium cross- pressures again failed to fit the expected pattern: those characterized by medium cross-pressures were most likely not to vote. This deviation from the prediction was attributed to the method used in constructing cross-pressures, which probably includes apathy in the middle categories. The analysis of attitudinal cross-pressures as a vari- able again evidenced little support for H. 9 and H. 10. Apparently an additive model is sufficient to explain the relationship between attitudinal cross—pressures, and turnout 138 and vote for preferred party: in some cases the additive model is significantly mgrg_potent. In no case were the inter- action terms significant. They contribute nothing to reducing the unexplained variance. Cross-pressures as a variable is apparently a chimera. What does this mean for cross-pressure theory? At least two points are clear. First, we cannot accept completely even the use of cross-pressures as a descriptive term. The evidence only partially supports this view. Cross-pressures do not always have the prescribed effects in the limited sense implied by traditional analysis. Secondly, the non- additive conception of cross-pressures is unwarranted. Conceived as as an interactive variable, cross-pressures apparently does not exist. There are alternative ways of conceptualizing cross—pressures as a non-additive function, but an additive model seems to be the most appropriate. This means that only the main effects of the pressures need to be considered in explaining and predicting behavior. For the present, at least, we can avoid complicating research in this area. From a conceptual standpoint this analysis has, hope- fully, demonstrated the need for clearly articulated concepts, ones which are susceptible to unambiguous interpretation and validation. The original cross-pressures formulation was not such a concept. A related question concerns the measurement of cross-pressures. A prOper conceptualization implies a means of measurement and test. Regarding measurement, it is 139 fair to say that cross-pressures have only been measured indirectly and that any real test of their effects would require a more direct method of measurement. Such precision seems impossible at this time. The method of dummy-variable multiple-regression employed here seems particularly suited for the study of cross- pressures, and that of voting behavior in general. Other similar methods, such as multiple classification analysis (tree analysis) or probit analysis may prove even more useful under specific conditions. These methods help greatly in the specification of concepts and in the analysis of the main effects of variables. They may, thus, be very useful in the development of theory. Several incidental points deserve comment here. The conceptual problem of studying cross;pressures' effect on voting when there are more than two parties has not received sufficient attention. The analysis of affiliative cross- pressures above was no doubt affected by this. Assumptions have to be made about the position of third or minor parties, assumptions which may distort results. Cross-pressures may have a different meaning in these circumstances. Likewise decisions must be made concerning the possibility of partisan identification toward new parties. In other polities this could be a critical factor, and it may have influenced the analysis of attitudinal cross-pressures. A last point concerns the difficulties involved in an analysis using dependent and independent varibles as closely related as 140 vote and party orientation. The problem of multi-collinearity inevitable in this analysis is compounded here by use of a dependent variable not totally distinct from an independent one. We may, in some instances, be predicting a variable from itself in another form. Finally, this analysis has, I think, demonstrated that theory and method go hand in hand. Students of voting behavior must become more rigorous in the deve10pment of their conceptual tools and pay more attention to selecting the most apprOpriate technique for testing their hypotheses. Concepts such as cross-pressures, hand-me-downs from another era, largely untested and occasionally untestable, abound in the literature. Systematic evaluation of these notions is essential to the deve10pment of a sound approach to voting behavior. Hopefully this analysis has been a step, however small, in that direction. FOOTNOTES 1Paul F. Lazarsfeld, Bernard Berelson, and Hazel Gaudet, The PeOple's Choice, (New York: Columbia University Press, 1968). 21bid., p. 53. 3Ibid., p. 61—64. The analysis presented here uses cross- pressures without regard to levels of analysis. 4Bernard Berelson, Paul F. Lazarsfeld, and William McPhee, Voting, (Chicago: The University of Chicago Press, 1966). See especially Part IV, the conclusion, in which theory is stressed. 51bid., p. 284. 6Ibid., especially pp. 130-131. 71bid., p. 285. 8Ibid., p. 286-7. 9Peter Werner Sperlich, Cross-Pressure and Conflict in Political Behavior, (Doctoral Dissertatibn, University of_— Michigan, 1966), pp. 135-148. For an expanded version of this thesis see Sperlich, Peter Werner, Conflict and Harmony 13 Human Affairs: A Stud of Cross-Pressures and PoIitical Behavior, (Chicago: Rand—McNally 8 Co., 197T). 10Angus Campbell, Gerald Gurin, and Warren E. Miller, The Voter Decides, (Evanston, Illinois: Row, Peterson and Company, 1954), p. 183. 11Berelson et al., p. 183. 12Angus Campbell, Philip E. Converse, Warren E. Miller, Donald F. Stokes, The American Voter, (New York: John Wiley and Sons, 1964). lsIbid., pp. 46-48. 14Ibid., pp. 72-79. 15Richard W. Boyd, "Presidential Elections: An Explanation of Voting Defection," American PolitiCal Science Review, 63, no. 2 (June, 1969), pp. 498-514. 141 142 16Sperlich, op, cit., pp. 1-2 and chapter 3. 17Patrick M. Horan, "Social Positions and Political Cross- Pressures: A Re-Examination,” American Sociological Review, 36 (August, 1971), pp. 650-660. 18On this important point see, Blalock, Hubert M., Jr., "The Identification Problem and Theory Building: The Case of Status Inconsistency," American Sociological Review, 30 (June, 1965), pp. 374-380. 19Seymour Martin Lipset, Paul F. Lazarsfeld, Allen H. Barton, and Juan Linz, ”The Psychology of Voting: An Analysis of Political Behavior," in Gardner Lindzey, ed., Handbook 9f Social Psychology, (Cambridge, Mass.: Addison- Wesley Publishing Co., 1954), pp. 1124-1175; see especially pp. 1126-1134. 20Inter-University Consortium for Political Research, The SRC 1968 American National Election Study (August, 1971 Codebook), p. 235. 21 Boyd, "Presidential Elections," pp. 501-502. CHAPTER IV STATUS INCONSISTENCY AND CROSS-PRESSURES The previous two chapters have analyzed, in turn, the effects of status inconsistency and two variations of cross- pressures on voting behavior. In Chapter I, I suggested that these are parallel concepts and that they might well be expected to be inter-related. This suggestion is not entirely new. On the other hand, the possible relationships between these variables have never been studied systematically. This chapter presents such an analysis, subject to the limitations of the data. We suspect, on the basis of previous research and to a lesser extent in view of the data previously presented here, that status inconsistency, affiliative cross-pressured and attitudinal cross-pressures are related in the fashion suggested in the first chapter. This logic concerns levels of analysis and social interaction patterns. Reasoning that social categories imply social milieus, that inconsistent categories imply inconsistent milieus and that inconsistent milieus imply inconsistent social cues, I prOposed that status inconsistency and affiliative cross-pressures ought to be related. More precisely highly inconsistent statuses should breed affiliative cross-pressures. Similarly I reasoned that inconsistent social 143 144 cues might either lead directly to contradictory attitudes or make them more salient to the individual. Affiliative cross- pressures should lead to attitudinal ones. Finally I argued that inconsistent social categories might well lead to atti- tudinal inconsistency directly through intra-psychic compari- sons of relative status positions and associated attitudinal dimensions. Presumably this last relationship ought to be less pronounced than the others. Not necessarily implicit in this argument was the suggestion that, for both turnout and Wallace vote, the effects of these three independent variables would be cumulative. That is, in the same fashion that descriptive cross-pressures predict lower turnout among the cross-pressured, we might expect that those who are status inconsistents and cross-pressured both affilia- tively and attitudinally would vote less and for Wallace more often than those who are conSistent and without cross-pressures. This is not an interactive prOposition; there is no reason, theoretical or empirical, to expect interaction. The Relationships Among Status Inconsistency and Cross-Pressures These hypotheses are not startling, especially in the light of previous research. I have earlier noted that The Pe0ple's Choice and VOting both used "factors" influencing voters which overlap the levels employed here.1 Each of these studies reported cross-pressure effects. Lipset, in Political Man, reports similar instances.2 In The American 145 V3333, the SRC group found both affiliative and attitudinal cross-pressures operating.3 Pool, Abelson and Popkin, again mixing levels found only marginal evidence of cross-pressures.4 Clearly the most germane research here is that conducted by Boyd.5 He explicitly analyzed the relationship between status inconsistency and attitudinal cross-pressures. Boyd was interested in whether inconsistency increased cross-pressures above the level expected by an additive model. He found that they did not. He further discovered that even an additive model did not fit the data well. He concluded, "there is no additive relationship between the dimensions of status and political cross-pressures." His measures were nearly identical to those employed here. His finding is not particularly alarming given the relationship hypothesized here. Boyd himself suggested several reasons for his failure to discover this expected relationship.6 He did not, however, consider the possibility that affiliative cross-pressures may act as an intervening variable here. I shall test this potentiality. The first relationship to be examined here is formal- ized as: H. 11: Individuals characterized by inconsistent statuses are more likely to have affiliative cross-pressures than those with consistent statuses. This relationship is easily tested by means of a simple cross-tabulation. Table 4.1 presents these results. In this table status inconsistency has been trichotomized. Individuals whose statuses are completely congruent are highly consistent. 146 .momoaunopmm :M one 50:0pmwmcoucw maumpm mo xwomoueu some pom mrzm .Hohho m:a@:§0p on one wooa ou 53m Ho: mmom+ Amy Ameqm. flame c.OOH o.OOH o.ooa Happy m.um w.wm.. m.wa mono pcmumwmcoucH m.mm v.- ._n.mN oumwwofihmuqH -o>WMMMMm o.m~ m-wm 4.Hm mono Dappmflmaou same .. ANmV,, .fimmv‘ I ,_ o.ooa +H.O¢H o.oo~ H8669 o.ov m.Hm o.o~ mono unoumwmcoqu o.o m.H~ c.4N pppflppsppqu -wmmwmm o.oo m.o¢ o.om mono “nowmflmqoo Ame flame «Ammv._ o.oo~ o.ooa o.ooa H8669 o.o~ w.mm ,c.oN. mono unoumfimnounH o.oo o.aN o.o~ oumH66596ch pawwmwmm o.o~ N.me o.op mono unppmfimqou p:meMmmm6:H papuppEppucH Demwmwmmpo. .xocmumflmqoueH.msompm mmhzmmmhmlmmOHU ®>HHNwHHWM< mZOHHHHwumfiaflmmm mo kpomoumo some how m.ze flame flva nose ANU_. ...mwv. .54HV‘1.. fine,. .0 And .3, Emmy o.ooH o.ooa o.ooH o.OOH o.ooa o.ooH o.ooH o.ooH o.ooa o.o o.o 6.m~ o.w~ o.o o.m~ m.¢H m.¢H 4.44 DepumflmcoueH sfigmfim o.oo~ o.ooa 4.Ha e.Ha o.ooa 4.Hn a.mm A.mm o.mm pumfleoepppeH o.o o.o o.o o.o 0.0 o.o o.o o.o o.o peppmfimcou saga“: mmdu mu.“ . WOQU W050 . $96 mOSU . , . mono. . . . _ mum . mODU pdmumfim ....Eumfi wcmu Hcoumwm nwmva “Emu Hfimwmwm ...mwoe HGOH uCOUGH ...HOHGH rmfimfiou uGOUQH uhmucH umflmfiou uQOUQH IHOHQH nmwmfiou mmhdmmmhm _. ,..... _ , . -mmOho w:0fihm - o>wumHom vnowpm - omzomm o>fiumfiem - omsomm HmawwSHHuu< moasmmppmamBOHU 6>fippflaflwm< mmmammmmm-mmomu 4HH o.OOH o.o o.o m.NH H.HH m.sH o.o o.o~ n.p 062 ppm o.o o.ooa o.ooa m.nm m.wm “.mw o.ooa o.om n.nm pppo> UOHDmmmhm Ohflmmmhm wmhdmmmhm QHDmmOHm VOHSmmmhm OHSmmmHm -mmOhu mumfiwoz -mmouu -mmOho Vuwflwoz -mmonu -mmoyu oumfipoz -mmonu Aflammm -pppea oz Raemmm -pppeH oz ReamHm -ppch 62 moudmmOHm-mmouu 0>Hpmfiafimm< o>wumfiom-omsomm fineness HGOHmHmGOUCH mumflwOEHOHCH “floumwmdou >0:oumflm:ou:H mspmum .HDOZMDH UZHHO> Qz< mmMDmmmmmummomu m>HBHHUzmHmHmZOUZH mDHm ”v.v mam