M133," SOCIOECONOMIC STATUS CONSISTENCY AND POUTICAL BEHAVIOR-ATTITUDE CONSISTENCY fiissertation for the Degree of Ph. D. MSCHiGAN STATE UNIVERSITY GEORGE FRANKLEN BESHOP 1973 mnnwwmmmmum“ “‘39 HER/1131’ Michigan State m'vcrsity THES'S £ . ,- N \ 35' ‘. f . V, r" ,1 Wait; 5633?}; 9—4421 we 23 3994 ,SEPNE 1995- . $33 ‘ g ’7‘“ MR 3 0 2004 03 3 1 04 ABSTRACT SOCIOECONOMIC STATUS CONSISTENCY AND POLITICAL BEHAVIOR-ATTITUDE CONSISTENCY By George Franklin Bishop This dissertation had three original purposes: (a) to develop a precise conceptual and operational definition of the social-psychological construct of cog- nitive or attitudinal consistency within a field survey context, (b) to integrate this definition with the con— ceptualization and measurement of the sociological construct of status consistency, and (c) to test the social cross-pressures hypothesis that the greater an individual's socioeconomic status inconsistency, the greater will be his cognitive or attitudinal inconsistency with respect to relevant political attitude objects, such as presidential candidates.- Methodologically, the research design involved a secondary analysis of sample survey data from the 1968 American National Election Study conducted originally by the Survey Research Center of the Institute for Social Research at the University of Michigan. The data drawn George Franklin Bishop from this national election study included the traditional ‘measures of an individual's socioeconomic status--leve1 of education, income level, and occupational prestige-- and "feeling thermometer" measures of an individual's affective attitudes toward the three major 1968 presiden- tial candidates: Hubert Humphrey, Richard Nixon, and George Wallace. These measures, in turn, were used to construct the indices of status and attitude consistency, both of which were defined conceptually in terms of deviance and operationally by simple differences between standard scores. The results, in summary, showed that there was a negligible relationship between the variables of socio- economic status consistency and consistency of affective attitudes toward the three 1968 presidential candidates, as well as between status inconsistency and actual voting behavior or preference in 1968, although there were some slight but inconclusive trends in these relationships among those who were highly status inconsistent; that is, the most deviant status types. Additional secondary analyses of the 1968 election data showed, moreover, that there was a statistically sig- nificant positive relationship between voting behavior and candidate attitude consistency, with those who voted in the election being mmre attitudinally consistent than those George Franklin Bishop who didnot vote. It was also found that those who were more behaviorally involved in politics, as measured by participation in the election, were more psychologically involved or interested in politics. And, finally, it was discovered that cognitive or attitudinal consistency was greatest among those individuals who had a high level of interest in government and public affairs and were behaviorally involved in the election through the overt act of voting. Theoretically, it was concluded that the most important reason why the findings on status consistency did not turn out as predicted was an inadequate account of the psychological reality of various socioeconomic status discrepancies to the individual. Thus it was suggested that future research on this subject should focus upon whether or not the individual is aware of being in an incon- sistent status position and on the degree of importance it has for him or her personally. And, finally, the positive evidence for a consistent relationship between voting behavior and affective candidate attitudes was interpreted in terms of the self-perception principle that attitudes- follow-behavior. SOCIOECONOMIC STATUS CONSISTENCY AND POLITICAL BEHAVIOR-ATTITUDE CONSISTENCY By George Franklin Bishop A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1973 I 41 I '3.- To my father and mother, GEORGE AND MARY BISHOP ii ACKNOWLEDGMENTS I would like to express appreciation to my dissertation committee members: Drs. Eugene Jacobson, Philip Marcus, Charles Wrigley, and William Crano, for their helpful guidance and advice. I want to especially thank the chairman of my doctoral, guidance, and dis- sertation committees, Dr. Eugene Jacobson, for his enduring support of my multidisciplinary research efforts in the field of social psychology. I would also like to thank several other indi- viduals for their special assistance in this dissertation project: Elizabeth Powell of the Michigan State Uni- versity Political Science Data Archive, Bill Brown of the Computer Institute for Social Science Research, Roger Halley of the Department of Psychology, Dave Klingman of the Multidisciplinary Social Science Program, and last, but not least, my wife, Lucille. iii LIS’ LIS' CPA TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES CHAPTER I. INTRODUCTION Conceptual and Methodological Problems in the Study of Cognitive Incon- sistency . The Problem of Defining and Measuring Cognitive Inconsistency . . A Conformity and Deviance Model of“ Cognitive Inconsistency . . Conceptual and Methodological Problems in the Study of Status Inconsistency Defining and Measuring Status Incon- sistency in Terms of Deviance or Difference . Hypothesis: Status Inconsistency and Cognitive Inconsistency . Status Inconsistency as Social Cross- Pressure II. METHOD Cases and Dataset Variables and Indices . Socioeconomic Status and Status Inconsistency . A Regression Approach to Measuring Status Inconsistency . A Subgroup Approach to Measuring Status Inconsistency A Difference Score Measure of Statuso Inconsistency . iv Page vi viii 14 15 15 18 19 20 20 21 22 23 CHAPTER Candidate Attitudes and Attitude Inconsistency . . A Difference Score Measure of Attitude Inconsistency Data Transformations Status Consistency- Inconsistency Types . Attitude Consistency- Inconsistency Types . III. RESULTS Status Inconsistency and Humphrey-Wallace Attitude Inconsistency . . . . Additional Analyses of the Data . Status Inconsistency and Voting Behavior . / Voting Behavior and Attitude Incon- sistency . Behavioral- -Psychologica1 Involvement in Politics . Behavioral Involvement Band. Level oof Constraint . Behavior-Attitude Consistency . IV. DISCUSSION Negative Evidence for the Correlates or Effects of Status Inconsistency . Inadequate Data, Methodology, or Theory? Positive Evidence for Behavior- Attitude Consistency . . . Attitudes- Follow- Behavior . . Some Directions for Future Research . V. SUMMARY . REFERENCES APPENDIX A. CODEBOOK FOR THE STUDY VARIABLES B. SUBGROUP CLASSIFICATION APPROACH TO STATUS AND ATTITUDE INCONSISTENCY . . . Page 27 28 33 33 36 38 38 43 43 47 49 53 56 62 62 64 68 73 77 81 87 8O TABLE 1.1 LIST OF TABLES Hypothetical Relationship Between Attitudes Toward George McGovern and Richard Nixon . Pearson Product-Moment Correlations Among Status Variables: Education Level, Income Level, and Occupational Prestige Feeling Thermometer Means and Standard Deviations for 1968 Presidential Candidates: Humphrey, Nixon, and Wallace Pearson Product-Moment and Eta Correlations Among Affective Attitudes Toward Humphrey, Nixon, and Wallace Relationship Between Level and Type of Education-Occupational Prestige Status Inconsistency and Level and Type of Humphrey-Wallace Attitude Inconsistency . Relationship Between Level and Type of Education- Income Status Inconsistency and Level and Type of Humphrey-Wallace Attitude Inconsistency . . . Relationship Between Level and Type of Income-Occupational Prestige Status Inconsistency and Humphrey-Wallace Attitude Inconsistency Relationship Between Level and Type of Education-Occupational Prestige Status Inconsistency and Presidential Voting Behavior or Preference in 1968 vi Page 25 29 . 31 39 . 40 . 41 . 44 TABLE 3.5 Page Relationship Between Level and Type of Education-Income Status Inconsistency and Presidential Voting Behavior or Preference in 1968 . . . . . . . . . . . . . 45 Relationship Between Level and Type of Income-Occupational Prestige Status Inconsistency and Presidential Voting Behavior or Preference in 1968 . . . . . . . 46 Relationship Between Presidential Voting Behavior or Preference in 1968 and Level and Type of Humphrey-Wallace Attitude Inconsistency . . . . . . . . . . 48 Relationship Between Presidential Voting Behavior or Preference in 1968 and Level of Interest in Government and Public Affairs . . . . . . . . . . . . . . . 50 Multivariate Relationships Among Presiden- tial Vote or Preference for Humphrey or Wallace, Level of Interest in Government and Public Affairs, and Level and Type of Humphrey-Wallace Attitude Incon- sistency . . . . . . . . . . . . . . . . . . 52 Pearson Product-Moment Correlations Among Attitudes Toward Humphrey, Nixon, and Wallace--By Presidential Voting Behavior or Preference in 1968 . . . . . . . . . . . . 55 Relationships Among Candidate Attitudes and Presidential Voting Behavior or Prefer- ence in 1968 . . . . . . . . . . . . . . . . 57 Pearson Product-Moment Correlations Among Affective Candidate Attitudes--Control- ling for Presidential Voting or Preference Group in 1968 . . . . . . . . . . 60 Pearson Product-Moment Correlations Between Occupational Prestige and Humphrey Attitude Within Level and Type of Educa- tion X Level and Type of Wallace Attitude . . . . . . . . . . . . 97 vii LIST OF FIGURES Figure 1. Schematic Diagram of Level and Type of Status Consistency-Inconsistency . 2. Schematic Diagram of Level and Type of Attitude Consistency-Inconsistency . viii Page 35 37 CHAPTER I INTRODUCTION Over the past twenty-five years or so American social psychologists interested in the nature of attitudes and attitude change have developed a number of middle- range theories of the structure and dynamics of human social cognition, generally called theories of cognitive consistency (see Abelson et a1., 1968, for an interpreta- tive summary). Although there are some important dif- ferences in basic constructs and propositions among the major consistency models--balance, congruity, and dis- sonance--they commonly assume that a person's cognitions (beliefs, attitudes, values) will tend to be logically or psychologically consistent with one another and with his overt behavior (Jones and Gerard, 1967; Shaw and Costanzo, 1970). And, further, they assume that if a person's cog- nitions are inconsistently related to one another or to his overt behavior, he will experience some form of psychological tension or discomfort which will then moti- vate him to reduce the tension by changing one or more of his cognitions or his overt behavior in the direction of greater cognitive consistency. This, in brief, is the general consistency notion common to most contemporary theories of attitude structure and attitude change. We now turn our attention to some important problems of conceptualization and measurement in the study of cognitive (or attitudinal) consistency. Conceptual and Methodological Problems in the Study ofICognitive Inconsistency1 While the general consistency notion has aided in the discovery of many new empirical facts and relation- ships concerning attitude structure and change, there are some major problems of conceptualization and measurement associated with the various theories of consistency, prob- lems which tend to limit their testability, scope, and general explanatory capacity. As identified and discussed by Pepitone (1966), and reviewed more recently by Shaw and Costanzo (1970), these problems include: (a) defini- tion of the elements or components involved in the cognitive inconsistency, (b) definition and measurement of cognitive inconsistency itself, (c) the problem of mutually relevant or irrelevant relations between two or 1When the term inconsistency is used in this paper, it should be clear that both consistency and in- consistency are intended as component reference terms. more cognitions, (d) the problem of importance or personal significance that inconsistency has for an individual, and (e) the problem of uncertainty; i.e., the degree of cer- tainty an individual attaches to his attitudes or cogni- tions. In this paper, we will focus upon the most central of these problems: the definition and measurement of cognitive inconsistency itself. Further, in discussing this problem, we will follow on Pepitone's (1966) analysis and also focus on Leon Festinger's (1957) theory of cog- nitive dissonance as an example. The Problem of Defining and Measuring_ Cognitive Inconsistency The problem of defining and measuring cognitive inconsistency itself can be seen most clearly in Festinger's (1957) cognitive dissonance theory, which defines inconsistency in terms of "ill-fitting" or "non- fitting" relations among cognitive elements. Specifically, the theory states that there are three basic types of relations that may exist between pairs of cognitive elements: 1. Consonance (consistency)--"Two cognitive elements are in a consonant relation if, considering these two alone, one element follows from the other [Festinger, 1957, p. 260]." 2. Dissonance (inconsistency)--"Two cognitive elements are in a dissonant relation if, con- sidering these two alone, the obverse of one element follows from the other (pp. 260-61]." 3. Irrelevance (nonconsistency)--"Two cognitive elements are in an irrelevant relation if they have nothing to do with one another [p. 261]." But what, we may ask immediately, is the conceptual H H and operational meaning of "follows from, obverse of," or "have nothing to dO‘With," or even "relation"? Nowhere does dissonance theory define these terms operationally or provide adequate guiding conceptual definitions, includ- ing the most recent reformulations of the theory by Brehm and Cohen (1962), Festinger himself (1964), and Aronson (1969). Pepitone's (1966) discussion of these problems is particularly relevant here. On the problem of defining dissonance in terms of "obverseness," for example, he writes: Obviously, this is the most abstract conception of inconsistency encountered in the models discussed; there are no indications as to the empirical mean- ing of "obverseness." Whereas balance specifies some types of unit formations--e.g., ownership, neighboring, belonging, etc.--and ”neobalance" lists some of the positive or negative attitude relations which can exist between two cognitive elements, dissonance theory specifies no rules of correspondence with the empirical plane. In such a circumstance, how does one proceed to define obverseness in the laboratory? How, indeed, can one formulate and test dissonance hypotheses at all, if the model does not say how to define dissonance empirically? The absence of a definition of obverseness can only mean that dissonance experiments are generated by intuition and by extrapolation from other experiments. (PP. 265-66) In lieu of such needed operational definitions, however, dissonance theorists and researchers typically give some examples of what they mean, such as the follow- ing (see Aronson, 1969): l. The cognition that "smoking cigarettes causes cancer" would be consonant with the cognition that a person "does not smoke cigarettes," but dissonant with the cognition that he "does smoke cigarettes,” and irrelevant to the cogni- tion that he "plans to go to Europe next summer." 2. The cognition that a person "has voted for George McGovern" would be consonant with the cognition that he "is a liberal democrat," but dissonant with the cognition that he "likes Richard Nixon (or George Wallace)," and irrelevant to the cognition that he "believes in flying saucers." But these intuitively reasonable examples-- describing cognitive situations which almost everyone would agree are consonant, dissonant, or irrelevant—- still do not provide us with any useful operational clues to the meaning of ”follows from, obverse of," or "have nothing to do with." What, for instance, is the empirical basis for saying that the cognition that a person "likes Richard Nixon" does not follow from the cognition that he "has voted for George McGovern"? Dissonance theory pro- vides none, but it does suggest a simple rule of thumb for saying so: violation of an expectancy (Aronson, 1969). Thus if we know that a person is a liberal democrat, for example, we expect that he will be more likely to vote for or like George McGovern than Richard Nixon. But if he tells us that he intends to vote for or likes Richard Nixon, this (according to the theory) would violate our expectations about the behavior of liberal democrats. Yet, we may ask further of the theory: What is the empiri- cal foundation for these psychological expectations? How can they be defined, operationally? What is the conceptual and operational meaning of the terms "violation of an expectancy"? A Conformityfand Deviance Model of Cognitive InconsiStency The empirical or experiential foundation for such expectations would seem to be that we have observed that certain combinations of cognitions (or attitudes and behavior) tend to be quite common or frequent; that being a liberal democrat, for example, tends to be frequently associated with voting for or liking liberal democratic presidential candidates such as George McGovern; and, conversely, not associated or infrequently associated with voting for or liking conservative republicans such as Richard Nixon.2 Thus a liberal democrat who votes for or likes Nixon violates our expectancy concerning the typical or usual combination of attitudes and behavior. To put it differently, frequently observed combinations of atti- tudes and behavior tend to be seen and assumed as going II together or "consistent, and infrequently observed com- binations, as not going together or "inconsistent" (or even "nonconsistent"). Conceptually speaking, these notions form the basis for taking what can be called a conformity and deviance approach to the analysis of cognitive or 2This frequency of combinations notion is based, in part, on Jackson and Curtis' (1968) conceptualization of status consistency dimensions. attitudinal inconsistency. To illustrate this approach, let us consider the hypothetical relationship presented in Table 1.1. As shown in this illustrative table, persons who like McGovern definitely tend to dislike Nixon and vice versa. That is, in terms of our typical or common frequency approach, the combination of attitudes "like McGovern" and ”dislike Nixon" may be seen as going together or consistent; and, the combination of "like McGovern" ' as not going together or inconsistent. and "like Nixon,‘ When we consider this approach more carefully, it becomes apparent that we are actually describing a certain kind of conformity or deviance (or nonconformity)!3 This assertion is based on an important distinction between two different descriptive criteria of conformity and nonconformity: what Beloff (1958) has called ”conven- tionality" and "acquiescence” and, more recently, what Hollander and Willis (1967) have called "congruence" and "movement" conformity. Operationally, Beloff has defined conventionality as "high agreement between an individual's response and the mean or modal response of his group or c1ass"--and acquiescence as "the amount of shift from 3Conformity here means conformity to the popula- tion of which an individual is a part; that is, conformity to a membership rather than a reference group. This method of conceptualizing and measuring consistency has also been characterized by McGuire (1960) as "the popula- tion parameter method." Table 1.1 Hypothetical Relationship Between Attitudes Toward George McGovern and Richard Nixon —7 1 j I DO YOU LIKE OR DISLIKE DO YOU LIKE GEORGE McGOVERN? OR DISLIKE RICHARD NIXON? Like Dislike George McGovern George McGovern Like Richard Nixon 20% 80% Dislike Richard Nixon 80 20 Total Percent 100% 100% NOTE: The Yule's Q coefficient for this hypothetical relationship would be -.88. 10 private to public opinion [Hollander and Willis, 1967, p. 64]." Similarly, Hollander and Willis have defined congruence conformity (or deviance) "in terms of the extent of agreement between a given response and the normative idea1--and movement conformity "in terms of a change in response resulting in a greater or lesser degree of congruence [p. 64]." It should be clear from these definitions, then, that there is a close correspondence between our frequent combinations conception of cognitive consistency and Hollander's and Willis' concept of congruence conformity (and Beloff's conventionality). In terms of our example in Table 1.1, we would say that individuals who "like McGovern" and "dislike Nixon" (or vice versa) are exhibit- ing congruence conformity, while those who "like McGovern and also "like Nixon" (or dislike both) are exhibiting incongruence or deviance. To further increase the convergence between these conceptions, we can operationally define cognitive incon- sistency as: the degree of agreement (or difference) between an individual's combination of cognitions (beliefs, attitudes, values) and the mean or modal combination of cognitions for his population group or subgroup. Methodo- logically and statistically speaking, this operational definition means that cognitive inconsistency should be 11 measured in terms of some type of deviation or difference score; for example, residual deviation from a linear regression prediction for the relation between two or more cognitions (i.e., the difference between an individual's predicted score and his actual score); or, simply, the difference between an individual's scores on two related cognitive dimensions. Thus, the greater the deviation or difference score, the greater the cognitive inconsistency and vice versa. In summary, we have seen that one possible solu- tion to the problem of defining and measuring cognitive inconsistency is to use a simple conformity and deviance model. In the following two chapters on Method and Results, we will, in fact, apply this approach. But first we need to briefly consider some conceptually and methodologically related problems in the sociological study of status incon- sistency. Conceptual and Methodological Problems in the Study of Status Inconsistency Beginning with Benoit-Smullyan's (1944) article on "Status, Status Types and Status Interrelationships," the general notion of consistency among an individual's ranks on various dimensions of social status (e.g., income, education, occupational prestige) has been of central 12 importance in the analysis of social stratification. Historically, however, Gerhard Lenski (1954) has been credited with the theoretical identification of the status consistency construct--which he formulated as follows: Theoretically it becomes possible to conceive of a nonvertical dimension to individual or family status--that is, a consistency dimension. In this dimension units may be compared with respect to the degree of consistency of their positions in the several vertical hierarchies. In other words, cer- tain units may be consistently high or consistently low, while others may combine high standing with respect to certain status variables with low stand- ing with respect to others [Lenski, 1954, p. 405]. Lenski's original (1954) research on status con- sistency (or status crystallization as he also called it) showed that it accounted for some of the previously unexplained variance in American political behavior. For example, he found that status inconsistency, regardless of the type of inconsistency, tended to be associated with liberal political attitudes and behavior such as support for the democratic party. 3 But while Lenski's formulation and findings have stimulated a large amount of empirical research on the relationship between status consistency and various socio- political attitudes and behavior (see, for example, Broom, 1970; Goffman, 1957; Kelly, 1966; Kenkel, 1956; Laumann and Segal, 1971; Lenski, 1956 and 1967; Olsen and Tully, 1972; Rush, 1967; Segal, 1969, Smith, 1969; and Treiman, 13 1966), this area of research has been plagued concomitantly with many conceptual and methodological problems (see, for example, Blalock, 1966; Jackson and Curtis, 1968; Kasl, 1969; Mitchell, 1964; and Nam and Powers, 1965). As identified and summarized by Kasl (1969), the major prob- lems in this area include: (a) the problem of choice of the status components or dimensions to be used in measur- ing status inconsistency (e.g., whether to include ethnic characteristics such as race or nationality in a status inconsistency index); (b) the problem of association between certain types of status inconsistency and stages of the life cycle (e.g., low income and high education is a common type of status inconsistency for family heads under 35); (c) the problem of the size of the intercorrela- tions between the various dimensions of status (i.e., the correlations may be too high or too low); (d) the problem of control for overall socioeconomic status in relating status inconsistency to selected dependent variables (e.g., individuals who are very high or very low on overall socioeconomic status cannot be very status inconsistent); and, finally, (e) the problem of defining and measuring status inconsistency, given the conceptual and methodologi— cal complexities in this area. Needless to say, we will not attempt to solve all or even most of these problems as they are beyond the 14 scope of the present research. Rather, we will be concerned HOW'With defining and measuring status incon- sistency, and interrelating it—-operationally and empirically--to cognitive or attitude inconsistency. Defining and Measuring Status Inconsistency in Terms of Deviance or Difference Probably the most efficient way of operationally defining the construct of status inconsistency is in terms of the simple conformity and deviance model used to operationally define the construct of cognitive inconsis- tency. That is, we can operationally define status inconsistency as: the degree of agreement (or difference) between an individual's combination of statuses (e.g., education and income) and the mean or modal combination of statuses for his population group or class. Similarly, this operational definition dictates that status incon- sistency should also be measured in terms of some type of deviation or difference score. And, correspondingly, the greater the deviation or difference score, the greater the status inconsistency and vice versa. New or efficient operational definitions alone, however, are not sufficient for the purpose of scientific research. To gain acceptance, they must sooner or later be empirically illustrated and tested as part of a 15 hypothesis or theory. We now turn our attention to this task; that is, to constructing an hypothesis which empirically relates the variables status inconsistency and cognitive or attitude inconsistency within an overall theoretical framework. Hypothesis: Status Inconsistency and Cognitive Inconsistency The correspondence between our operational defi- nitions of status inconsistency and cognitive inconsistency would, itself, suggest a simple hypothesis: The greater an individual's status inconsistency, the greater will be his cognitive or attitudinal inconsistency with respect to relevant cognitive (attitude) objects. Status Inconsistency as Social Cross-Pressure One theoretical foundation for this hypothesis is that status inconsistency is a special case of the more general phenomenon of social cross-pressure. As originally formulated by Lazarsfeld and his co-workers (Berelson et a1., 1954; Lazarsfeld et a1., 1944), the social cross-pressure hypothesis states that individuals whose social character- istics (e.g., occupation, race, religion) predispose them in different directions (e.g., both pro-democratic and pro-republican) will tend to have more conflicting or 16 inconsistent attitudes, for example, than individuals whose sociodemographic characteristics predispose them in the same partisan direction (see Flanigan, 1972; Sperlich, 1971). Thus, status inconsistency can be conceptualized in terms of socioeconomic characteristics (education, income, and occupation) which may predispose an individual in various partisan directions. For example, previous research on American voting behavior (Berelson et a1., 1954; Campbell et a1., 1960; Lazarsfeld et a1., 1944) would suggest that individuals of high socioeconomic status or class are predisposed in a republican or conservative direction, while those of low socioeconomic class status are predisposed in a democratic or liberal direction. Thus an individual who is high on one status dimension (e.g., income) predisposing him in a republican or "right" direction and low an another status dimension (e g., edu- cation level) predisposing him in a democratic or "left" direction could be said to be under social cross-pressure. Such an individual, according to theory, would be expected to have some inconsistent or conflicting attitudes toward each party, their left-right policy positions, or their candidates--e.g., having positive and/or negative attitudes toward the presidential candidates of both parties or other candidates along a general left-right continuum. 17 To summarize, one theoretical foundation for our hypothesis is that status inconsistency is a special type of social cross-pressure which tends to produce conflict- ing or inconsistent attitudes in individuals who are differentially predisposed by such pressures in both left- liberal-democratic and right-conservative-republican partisan directions, for instance, rather than in the same "consistent" partisan direction. In brief, we have formu- lated a more refined operational test of Paul Lazarsfeld's seminal social cross-pressures hypothesis. We will now describe the data, methods of measure- ment, and analyses used to test the status inconsistency- cognitive or attitude inconsistency hypothesis, as well as additional secondary analyses of the present data. CHAPTER II METHOD Methodologically speaking, the present research involved what is called a secondary analysis of sample survey data (see especially Hyman, 1972). More specifi- cally, the data used in the present research were drawn from the 1968 American National Election Study conducted by the Survey Research Center Political Behavior Program of the Institute for Social Research at The University of Michigan.1 For a detailed description of the 1968 election study design and relevant sampling information, the reader should see the Inter-University Consortium for Political Research's codebook for The SRC 1968 Ameri- can National Election Study (197l)--to be designated here- after as simply the: Codebook. And for a background description and discussion of the major substantive find- ings for the 1968 election study, the reader is referred 1These data were made available through the Inter- University Consortium for Political Research and provided locally by the Michigan State University Political Science Data Archive. Needless to say, neither the original collectors of the data nor the consortium bear any respon- gibility for the analyses or interpretations presented ere. 18 19 to Converse et a1. (1969); see also, Bishop et a1. (1972). Cases and Dataset The total number of data cases used in the present study was 1168, reduced from an original total of 1673 cases available in the 1968 election dataset. The raw N of 1673 consisted of: (a) A national cross- section sample of 1557 respondents and (b) A black supple- ment sample of 116 respondents. The reduced working N of 1168 was obtained by excluding those cases containing 'missing data codes for any one of the following six main study variables: education level, income level, occupa- tional prestige, and attitudes toward the three major 1968 presidential candidates Humphrey, Nixon, and Wallace. Thus the data reported in this paper are not technically based on a valid cross-sectional sample of United States citizens. However, since the present research was focused on the relationships among these variables, rather than on precisely estimating their distributions in the Ameri- can adult population, the resultant reduction in external validity (Campbell and Stanley, 1963) or generalizability is of somewhat lesser significance than in a more conven- tional survey research design. 20 Variables and Indices The indices of the main variables used to test the primary hypothesis of the present study--socioeconomic status, status inconsistency, attitudes toward presi- dential candidates Humphrey, Nixon, and Wallace, and attitude or cognitive inconsistency--were constructed in the following manner. Socioeconomic Status and Status Inconsistency First, the three traditional indicators of an individual's general socioeconomic status were used: 1. Level of Education--measured by the total number of grades of school completed plus any non- college vocational or technical training (Code: b295, 1971, pp. 82-84). 2. Income Level--measured by total family income for the current year (Codebook, 1971, pp. 132-33). 3. Occupational Prestige-~measured by Duncan's (1961) index of occupational prestige (Codebook, 1971, pp. 91-92). For a more detailed description of these three variables, including their means and standard deviations, 21 the reader is referred to Appendix A of the present dissertation. A Regression Approach to Measuring Status Incon- sistency.--Originally, the writer had planned to use a simple linear regression approach to the measurement of status inconsistency suggested by Kasl (1969). This approach would have involved obtaining a linear regression equation for the relation between education level (X) and income level (Y), for example, and then deriving a status inconsistency score for each individual by taking the difference between his predicted (expected) and obtained (actual) score--with a status inconsistency score defined A as = Y-Y, where: Y = a+bX = predicted income level Y = obtained income level X = obtained education level a = a coefficient in a linear regression equation b = coefficient for linear regression of Y on X Following Kasl (1969), a positive status incon- sistency score in this example would indicate that an individual's income level is higher than expected on the basis of his education level and vice versa; that is, the higher the score (positive or negative), the greater the status inconsistency. To put it differently, the absolute 22 size of the score would measure the amount of status inconsistency; and the sign of the score, the type of status inconsistency. Finally, some arbitrary low score would have to be defined as the cutoff-point for "status consistency." Although this approach was quite sufficient for the purpose of measuring status inconsistency, it was abandoned because of problems of zero correlation and non-linearity in the presidential candidate attitude data (see the section below on attitude inconsistency). That is, originally the writer had intended to use a regression approach to the measurement of both status inconsistency and cognitive or attitude inconsistency, an approach which assumed that the form of the relationship between variables is linear. This assumption was sufficiently met for the relationships among the three status dimensions, but only partially or hardly at all for the relationships among some of the three candidate attitude variables. Thus the regression approach was discarded and another approach attempted. A Subgroupprproach to Measuring Status Inconsis- tency.--Another approach which was used to measure status inconsistency involved breaking down both the status and attitude dimensions into a number of ordered subgroups and combina examini attituc (for a result Append rathe: some invol techr it d< in r‘ comb case W88 The sta sco inc ind. Stat Pres 23 combinations of subgroups (see Kasl, 1969) and, then, examining the relationships between pairs of status and attitude dimensions within each combination of subgroups (for a more detailed description of this approach and the results that were generated by it, the reader should see Appendix B). While this subgroup approach resulted in some rather interesting findings, it was considered too cumber- some for the purpose of the present research, since it involved a heavy reliance on traditional countersorter techniques of data processing (see Appendix B). However, it does represent a potentially useful technique (albeit in rudimentary form) for studying deviant subgroups and combinations of subgroup interaction effects. In any case, a third approach to measuring status inconsistency was tried and adopted. A Difference ScoreMeasure of Status Inconsistency.-- The third and final approach that was used to measure status inconsistency involved the use of simple difference scores (see Magnusson, 1967). Specifically, status inconsistency was defined as the difference between an individual's scores on each of the following pairs of status dimensions: (a) Education level and occupational prestige, (b) Education level and income level, and 24 (c) Income level and occupational prestige. This approach assumed, of course, that these three status dimensions were significantly related to one another. Table 2.1 shows the Pearson product-moment correlations among the status variables of education level, income level, and occupational prestige. As shown in Table 2.1, there were moderately sub- stantial, statistically significant correlations among the three socioeconomic status variables.2 These results permitted construction of the following difference score index of status inconsistency. First, each individual's score(s) on the variables of education level, income level, and occupational prestige were transformed into comparable units: standard(Z) scores. Next, status inconsistency scores were obtained for each individual 2Some readers may feel that while these correla- tions are statistically significant, they are not particu- larly high in terms of percentage of variance accounted for; that is, they are not substantively significant. It should be noted, however, that a moderate level of corre- lation (.40 to .60) among these variables is desirable for two main reasons. For one, if the correlations are too high, only a very small proportion of the sample will qualify as status inconsistent; on the other hand, if the correlations are too low, consistency between unrelated or negligibly related status dimensions becomes meaning- less. For another, the difference scores indices of inconsistency described below can be assumed to be more reliable, if the separate scores on which they are based are less correlated or even uncorrelated with one another (see, for example, Magnusson, 1967, pp. 93-94). Thus it would seem that these variables should be moderately cor- related with one another. 25 Table 2.1 Pearson Product-Moment Correlations Among Education Level, Income Level, and Occupational Prestige Status Variables: (N = 1168) STATUS Education Income Occupational VARIABLE Level Level Prestige Education Level ---- .52* .55* Income Level ---- .49* Occupational ____ Prestige *P < .001 26 by taking the difference(s) between his Z-scores on each of the three possible pairs of status variables: (a) Edu- cation 1evel-occupational prestige, (b) Education level- income level, and (c) Income level-occupational prestige. For example, a difference score for income level and occu- pational prestige is defined here as = 21-22, where: Z; a standard score for income level Z2 a standard score for occupational prestige Thus, as in the regression approach described previously,3 a positive difference score would indicate that an individual's occupational prestige is higher than expected on the basis of his income level and vice versa for a negative difference score. Or, as expressed simi- larly before, the absolute size of the difference score measures the amount or level of status inconsistency-- i.e., the higher the difference score, the greater the status inconsistency--while the sign or direction of the score (+ or -) indicates the type of status inconsistency. To summarize, three difference score indicators of an individual's status inconsistency were obtained: 3The careful reader will note the essential statistical equivalencies between these two approaches; e.g., index scores from both methods would be very highly intercorrelated. 27 1. Education level Z-score minus occupational prestige Z-score 2. Education level Z-score minus income level Z-score 3. Income level Z-score minus occupational prestige Z-score While these difference score indicators are obviously empirically related to one another, they may represent quite different psychological realities to particular individuals; for example, the high occupational prestige-low income minister or the high income—low edu- cation truck driver. That is to say, status inconsistency, like most things, is relative to the individual perceiver. Candidate Attitudes and Attitude Inconsistency The "cognitions” used in this study were attitudes toward the three major presidential candidates in the 1968 American national election: Hubert Humphrey, Richard Nixon, and George Wallace. These attitudes were measured in a post-election survey by a "feeling thermometer" which involved having respondents indicate their warm or 321$ feelings toward each candidate along a Qlto 129 degrees rating scale, with 50 degrees representing a neutral point or absence of feeling for the candidate; 28 ._1 to 199 degrees, warm feelings; and 9 to 99 degrees, cold feelings (see Appendix A for a detailed description of this instrument). Table 2.2 shows the feeling thermometer means and standard deviations fOr candidates Humphrey, Nixon, and Wallace. As shown in Table 2.2, the respondents in the present study had moderately favorable feelings, on the average, toward both Humphrey and Nixon and fairly cold or unfavorable feelings toward Wallace.“ It is also interesting to note here that Wallace as an attitude object apparently elicited the strongest feeling reactions in either direction, as indicated by a large standard deviation compared to Humphrey and Nixon. A Difference Score Measure of Attitude Incon- sistency.--As mentioned previously, the writer had originally intended to use a regression approach to the measurement of both status and attitude inconsistency. However, as noted, problems of zero correlation and l’The similarity between the overall mean ratings for Humphrey and Nixon raise several interesting points; for example, how much of Nixon's positive rating is due to post-election halo effects with regard to winning presi- dential candidates? Or are the closeness of these two ratings due to an even more general constant error of positive or negative tendencies to evaluate politicians as all good or bad? For another possible interpretation still, see the discussion section of this dissertation, concerning behavior-attitude consistency among voters and non-voters. 29 Table 2.2 Feeling Thermometer Means and Standard Deviations for 1968 Presidential Candidates: Humphrey, Nixon, and Wallace (N = 1168) 1968 FEELING THERMOMETER RATINGS PRESIDENTIAL . Arithmetic Standard CANDIDATE Mean Deviation Hubert Humphrey 60.76 27.95 Richard Nixon 65.41 23.13 George Wallace 31.17 30.57 NOTE: The feeling thermometer means for Humphrey, Nixon, and wallace in the national cross-section sample (Converse et a1., 1969) were 61.7, 66.5, and 31.4, respectively. Standard deviations were not reported by these researchers, however. 3O non-linearity were found for some of the relationships among candidate attitudes. Table 2.3 shows the Pearson product-moment and eta correlations among affective atti- tudes (feelings) toward Humphrey, Nixon, and Wallace. The data presented in Table 2.3 show that the only relationship which was both sufficiently linear in nature and statistically significant was that between attitude toward Humphrey and attitude toward Wallace. And while the eta values shown for the Humphrey-Nixon and Nixon-Wallace pairs were undoubtedly inflated, to some extent, due to the continuous nature of these vari- ables, they indicated a large enough departure from linearity so as to rule out any simple linear regression approach to the measurement of candidate attitude incon- sistency. . For these and other reasons noted earlier, the simple difference score approach was adopted to measure attitude, as well as status, inconsistency. However, this measure was applicable only to the HumphreyeWallace pair since it represented the one significant linear relation- ship between candidate attitudes. In other words, dif- ference scores as indices of inconsistency are meaningful only to the extent that they are based on correlated variables. Thus since the Humphrey-Nixon and Nixon- Wallace pairs were uncorrelated, difference scores for 31 Table 2.3 Pearson Product-Moment and Eta Correlations Among Affective Attitudes Toward Humphrey, Nixon, and Wallace (N = 1168) AFFECTIVE Humphrey Nixon Wallace ATTITUDE VARIABLE R = Eta = R = Eta = R = Eta - Humphrey ---- ---- .00 .45a -.29 .33b Nixon ---- '---- .00 .20c Wallace ---- ---- . P < .001 The eta value shown is for Nixon as independent vari- able and Humphrey as dependent variable. The converse eta is .34. The eta is for Wallace as independent variable and Humphrey as dependent variable. The converse value is The eta is for Wallace as independent variable and Nixon as dependent variable. The converse value is .09. 32 these pairs should represent mostly random errors. In short, the measure of candidate attitude inconsistency in the present research had to be limited to Humphrey- Wallace attitude inconsistency. The difference score measure of Humphrey-Wallace attitude inconsistency was constructed as follows. First, each individual's scores on attitudes toward Humphrey and Wallace were transformed into standard (Z) score form. Then, an attitude inconsistency score was obtained for each respondent from the difference between his Z-score on Humphrey and his Z-score on wallace, with the differ- ence score defined as = 21-22, where: 21 a standard score for Wallace Z2 = a standard score for Humphrey Because the correlation between attitude toward Humphrey and Wallace was in negative direction, however, large difference scores here represent attitude consistency; small difference scores, attitude inconsistency. Thus, a high positive difference score would indicate that an individual has a consistent attitude which is relatively positive toward Wallace and negative toward Humphrey, while a high negative difference score would show that he has a consistent attitude which is relatively positive toward Humphrey and negative toward Wallace. On the other hand, 33 a small positive difference score would show that a person has an inconsistent attitude which is relatively positive toward both Humphrey and Wallace; and a small negative difference score, an inconsistent attitude which is rela- tively negative toward both Humphrey and Wallace. To summarize, the absolute size of the difference score here measures the amount or level of Humphrey- Wallace attitude inconsistency--i.e., the larger the dif- ference score, the greater the attitude consistency--while the sign or direction of the difference score (+ or -) indicates the type of Humphrey—Wallace attitude inconsis- tency. Data Transformations To keep track of the direction and qualitative types of status and attitude inconsistency (as well as the possibility of additional non-linear relationships), the status and attitude difference scores data were trans- formed into a set of categories according to level and type of inconsistency as follows. Status Consistenc - InconsistencyTyEes The status inconsistency difference scores were recoded into six categories according to the following rules: 34 1. If the difference score was less than -1.00 standard deviation below the mean of the dif- ference scores (M = 0.00), it was recoded into the category: status inconsistent-high negative type. 2. If the difference score was greater than -l.00 standard deviation.and less than -.50 standard deviation below the mean, it was recoded into: status inconsistent-medium negative type. 3. If the difference score was greater than -.50 and less than .00 standard deviation(s), it was recoded into: status consistent-low negative type. 4. If the difference score was greater than .00 and less than +.50 standard deviation(s), it was recoded into: status consistent-low positive type. 5. If the difference score was greater than +.50 and less than +1.00 standard deviation(s), it was recoded into: status inconsistent-medium.positive type. 6. If the difference score was greater than +1.00 standard deviation, it was recoded into: status inconsistent-high positive type. Figure 1 shows a schematic diagram of the level and type of status consistency-inconsistency. Finally, these rules were applied to each of the three possible forms of status inconsistency: 3(a) Educa- tion-occupational prestige status inconsistency, (b) Education-income status inconsistency, and (c) Income- occupational prestige status inconsistency. 35 mommamamaoocHnmoaoumflmaoo msumum mo om%H was Ho>og mo amuwmwo oflumaonom H shawflm monHma Qmwuflmom m>HuHmom m>Hufimom o>wumwmz o>aumwmz o>wumwoz ewes asses: 33 33 ashes: swam HzmHmHmzoqu mDHoq mo amuwmfln oaumfimsom N muowwm monHmn nmwufimom swam m>fiuwmom m>fluflmom m>wumwmz. o>wumwmz o>wumwoz asses: son 304 ashes: swam HZMBmHmZOU MQDHHBH< HzmemHmZOUzH MODHHHH< HzmHmHmZOU MQDHHHH< CHAPTER III RESULTS We can best begin this chapter by restating the primary hypothesis tested in terms of the variables operationalized in the previous chapter: The greater an individual's socioeconomic status inconsistency, the greater will be his cognitive or attitudinal inconsistency with respect to the 1968 presidential candidates-- Humphrey, (Nixon), and Wallace. Status Inconsistency and Humphrey-Wallace Atfitude Inconsistency Tables 3.1, 3.2, and 3.3 summarize the relation- ships between the level and type of status inconsistency for the three possible forms of status inconsistency (i.e., education-occupational prestige, education-income, and incOme-occupational prestige) and the level and type of Humphrey-Wallace attitude inconsistency. . Examination of the data presented in these three tables reveals negative or negligible support for the status inconsistency-attitude inconsistency hypothesis. And although one of the chi-square tests (Table 3.1) 38 H . m. OHDWH 39 .ovaufiuum mooaamz m>auowon vow ovsufiuum houngaam 0>Huumom .ovsuauum houngasm o>wumwma was ovauauuu momaams o>fiuamom .owwumoum Hmsoaummaooo amnu umumouw ma Hu>ma aowumoawm .Ho>mH aofiumosvo mono saunas» ma owfiumoum Hoaoauomaooo ”.0013 .OH. u HOV NamHOHNNmou soamNaHuaou .mO. v m .ON a NO .HO.NN u Nx “maoz NOOH NOOH NOOH NOOH NOOH NOOH Hzmummm HHOuNoz aaHOoz OezmamHszO NOOHHNH< ON HN ON OH HN mH opHuuNmz 3OH OH NH ON «H OH OH o>HuHmom sOH ezmHmHszOzH OOOHHNN< N OH OH OH NH ON m>HuHmom asHOuz NmH NOH NOH NOH NOH NOH m>HuHOON ONHO onmHmHszO NOOHHNH< NON u z HHH u z NON u z OmN u z ONH u z NNH n z mom swam mom no: mom Boa woz 304 wmz to: moz swam honoumwmaooaH ovauHuu4 HZMHmHmZOUZH mDHoa somHHmzlhounqasm mo mama vow Ho>oa honoumwmaoosH menuwuu¢ oomHHmzuhmufiqan: mo oth was Hm>mA tam honoumwmaoocH maumum owwumoum HmGOHummnoUOIGowumospm mo oaks tam Ho>oq ammsuom awsmsowumamm H.m mHOmH 40 .ovsuuuum «sedans o>wummos can ovauduum hounmasm o>OuHmom .wvsuwuum woungasw 0>Huowon was ovsuwuum confide: o>wuqmom .HoboH second amnu Houmouw ow nouuwosum QDU'U .Ho>oa coaumoswo saga wouoouw ma Ho>uH «BoosH .nH. u uamHonmaoo hocownfiucoo .mo. A m .mm u up .om.mm u «N "maoz NOOH NOOH NOOH NOOH NOOH NOOH Hzmommm HOHOH HN OH NH NH NH OH msHuaOoz ONHO OH ON OH ON ON OH osHummoz asHOmz OHzONOHOzOO OOOHHHH< ON OH OH OH , OH OH msHOmwmz sOH OH OH OH OH OH OH O>HOHOON sOH HszOHOzOOzH OOOHHHH< O O OH NH OH OH usHuHOom aOHOmz NOH NOH NOH NHH NHH NON o>HOHOom ONHO UHzmHmHmzoo MQDHHHH< «ma u z med u z HON n z NmN n z NmH n z HOH u z mom swam mom to: mom Bog woz 39H woz cox woz swam hoamumumoooaH ovauuuu< nazmamHmzoozH m=Hog hodmumamaoooH maumum maoocHIaoHumoapu mo omha was Hm>og moamuwwmsoocH opsufiuu< oomaamznhousgasm mo maha tam Ho>mq was honoumwmsooaH msumum ofioonHIGoHumoswm mo some use Ho>oq Hmosumm awnmcowumHom N.m manna 41 .ovauwuum oomHHuz 0>Humwoo vow ovsuHuum housgaam o>HuHmom .ovouHuum Ooungaom 0>Humwoo can ovnuHuum oomHHmz 0>HuHmom .owHumoun Hmaofiummaooo swan umumouw mH Ho>oH uaooaH .Hm>mH oaooaH cosy umumouw mH owHumoum HmGOHummsooo ”.006 .OH. u OOOHOHOOOOO OuammaHuaoo .OH. A m .ON u we .O0.00 u Nx “meoz NOOH NQOH NOOH NGOH NOOH NOOH Hzmommm HOHOB OH ON OH HN OH OH O>Humwmz OOHO mm OH mm OH ow OH m>Oumwmz asHvox vazmemHmzoo unbaHahd OH OH NH NH ON NN m>Humwmz 3OH OH OH OH OH NH OH u>HuHmom 36H HZmamHmzoozH MQDBHOH< w on OH OH OH OH m>HuHmom anHvoz NOH NOH NOH NOH NOH NOH m>HuHOOO OOHO oazmamHmzou MQDHHHO< HOH u z NOH n z OON n z OON u z OOH u z OOH u z mom smHm mom we: mom 30H woz 30H woz H62 woz anm OosoumHmoouaH OOONHNHO HZMHmHmZOUzH mDHmH oomHHmzukmunmadm mo same was Hm>0H honoumHmnoosH onSuHuu¢ oomHHmzuhmusgasm mo mmOH was Hm>oH tam OosmumHmcoocH woumum owHuonm HmaoHummnoUOIoaooaH mo mama vcm Ho>oH Goosuom mHnmsoHumHom m.m oHAMH 42 indicated that there was some type of significant relationship between education-occupational prestige status inconsistency and the Humphrey—Wallace attitude inconsistency, the degree of relationship between these variables, as shown by the contingency coefficient, was quite low.' Moreover, when we consider the extremely large size of the present sample (N = 1168) and also the large size of the contingency table here (6 x 6)--two factors which tend to spuriously inflate the value, and thus the significance, of chi-square and related statis- tics such as the contingency coefficient--the statistical significance of this result becomes trifling compared to its substantive significance; that is, percentage of variance accounted for. Finally, it is worth noting that, among the high status inconsistents in Table 3.1 there was a slight tendency for those whose occupational prestige was much lower than their education level to have a "con- sistent" attitude which was positive toward Wallace and negative toward Humphrey--a trend which runs counter to the originally hypothesized direction of relationship. That is, it was found here that extreme status inconsis- tency tended to be associated with attitude consistency rather than inconsistency. In brief, the findings on the main hypothesis were disconfirming. 43 Additional Analyses of the Data In this section, we will do some additional analyses of the data in order to explore some unhypothe- sized relationships of interest. First, we will investi- gate the relationships between the various types of status inconsistency and presidential voting behavior or prefer- ence in 1968; and second, the relationships among voting behavior, attitude inconsistency, and psychological involvement in politics. Status Inconsistency and Votinngehavior Tables 3.4, 3.5, and 3.6 summarize the relation- ships between the various types of status inconsistency and presidential voting behavior or preference for Humphrey, Nixon, and Wallace in 1968.1 As shown in these three tables, the relationships between the various types of status inconsistency and presidential voting behavior or preference in 1968 were not statistically significant. 1This variable was measured in the post-election survey by the following question: "Who did you vote for in the election for president?" (If didn't vote) "Who would you have voted for for resident if you had voted? [Codebook, 1971, pp. 174-75].' For the present analysis, respondents who did not vote, nor indicate a preference, for Humphrey, Nixon, or Wallace (e.g., voted for some other candidates, refused to say for who, etc.) were re- coded into a general residual category: other. 44' .mmwumoum HwooHumnaouo.nwnu usuwouw mH Hu>mH dOHumoovm A .Ho>uH nowuaosuo amsu nonwouw OH owHumouO HwoOHuomaooo s .OH. a OOOHOHOOOOO NoamOaHuaoo .ON. A O .OO n Om..O0.00 u Nx “maoz NOOH NOOH :Mfllu Iflfll. O m N 0 mm on m 0 CH HH NOm NON NOOH NOOH .Jflll .Imll w m N N mm om O n O O Nmm nom NOOH NOOH O O m O N c on mN O OH m CH non Non Hzmummm HdHOH Honuo “Hwowamom oomHHms How wouo> consummmum OUMHHNS "Houo>looz . ootz you wouo> monmuomoum sotz "Hmuo>looz consummoum Oousmasm uu0u0>lsoz Omungasm mom vauo> NON u z HOH u z mom HOE mom 8: NON n z omN u z mom 30H wmz 30H ONH u z ONH u 2 NE 8: wmz OOHO nflzmflmHmZOUZH mDHoH QOOH aH consummoum no uoH>mnmm waHuo> HMHucouHmmum OOOH CH mocmnomoum Ho HoH>m£om waHuo> HMHusmvaoum vsm monoumHmcoocH maumum owHumon HOGOHum93000usoHumootm mo same was Ho>mH Goosuom OHSOGOHumHoM O.O «HOON 45 .Ho>oH anooH away Housman OH Ho>oH aoHumosvm a .Hm>mH dowumoswo ownu Henson» mH Hm>oH maoonH o .OH. - OOOHOHOOOOO NOOOOOHOOOO .OH. A O .OO a OO .OO.NO a NO ”ONOO NOOH NOOH Om on NH NH NON NHm NCCH NOOH O o O O m N Om Om w m w w Nmm Nmm NOOH NOOH .nmlu .IMII OH NH O O NO HN O N O O NON NNO Hzmomum H oudouowmum . oomHst "umuo>laoz notz How wouo> monoumumum notz "Houo>lsoz ooaoummmum Ooungasm "Houo>looz mounmadm How wouo> ONH u z OOH u 2 OOO OOHO mom Om: HON u z OmN u z mom 3OH woz 30H NOH a z HOH n z Ooz Om: Ouz OOHO ABZMHmHmZODzH mDH¢Hm HZMHmHmZOU mDHmH OOOH :H monouomoum no HoH>msom onuo> HmHuomuHmmum OOOH GO oocoumwoum Ho H0H>m£om waHuo> HwHuamemmHm was OoamumHmsooaH maumum maooaHunoHumonvm mo oaks was Hm>mH somsumm OHAOEOHumHom n.m oHan 46 .owHumouanwaoHummsooo saga Houmoum mH Hu>mH oaoosH .mH. u uaoHonwooo monomaHusou .OH. A O .OO 9 .Ho>oH oaounH omnu Hanson» mH owHumoun HaaOHummsooo a NO .ON.OO u NM "maoz NOOH NOOH m n O HH O O HO om NHO Nmm NOOH mm HH Nmm NOOH mm NH NOm NOOH NOOH IHII. .IOII N O O N OO OO O O OH O NON NON Hzmommm H moaouowoum oomHHmz Numuo>looz aoMHz How vmuo> monouommum notz “Houo>ldoz consummmnm Omunaasm "nouo>lsoz O0H£qaam How wouo> HOH u z NOH u z mom OOHO wen we: OON u z OON n z mom 30H woz 30H OOH u z ONH u z 32 O0: 32 OOHO nfizmHmHmzoozH mDHmH NOOH aH monouomoum Ho H0H>mnom waHuo> HmHuoomemum OOOH CH oofimummmnm no HoH>mnmm wnHuo> HmHusmonmHm was OofimumHanoaH msumum owHumon HmGoHummnoUOIoEooaH mo omOH van Ho>oH somSuom mHnchHumHoM 0.0 OHOOH 47 There are, however, some trends in these data worth noting. For example, among the high status incon- sistents in Table 3.5 there was a tendency for those whose education level was much lower than their income level to vote for Humphrey over Nixon, while those whose education level was much higher than their income level. tended to vote for Nixon over Humphrey. And, also, among the high status inconsistents (Table 3.6) there was a tendency for those whose occupational prestige was much lower than their income to vote for Humphrey over Nixon, while those whose occupational prestige was much higher than their income level tended to vote for Nixon over Humphrey. Such "trends" should, of course, be cautiously! regarded given the general pattern of negative results presented previously.. Voting Behavior and Attitude Inconsistency _ Table 3.7 shows the relationship between presi- dential voting behavior or preference in 1968 and the level and type of Humphrey-Wallace attitude inconsistency. When we examine these data, it can be clearly seen that attitudinal consistency was greater among those who voted for either Humphrey or Wallace than those who "would have voted" for or preferred either Humphrey or 48 .OO. a OOOHOHOOOOO NuaomaHuaou .HOO. v O .OO a NO .OO.NNO a NO .033qu oowHHmz 953me can ovauHuuu OPENS—m OPHuHmom A .ovsuHuuu Ouunqaom o>Huawoa vow ovouHuum ouwHsz o>HuHmom m «MHOZ NOOH NOOH NOOH NOOH NOOH NOOH NOOH Hzmommm HOHON NH O O N N NN OO OOHOOOuz OOHO OH O O OH OH ON ON O>Humwoz auHOmz nquHmHmzoo NQOHHHHO ON H O HN NN ON OH u>HuOOaz OOH NH H m ON Hm HH O o>HuHmom 36H HzmHmHmzoozH NODHHHH¢ ON OH ON OH ON O O o>HuHOom asHOoz NOH NOO NON NOH NN NO NH OOHOHOOO OOHO ayzmamHmzoo MQDHHHH< OO u z OO n 2 OO u 2 OOO n 2 HO n z OOH u 2 OOO u 2 HO u OOMHHQ3 Godwhmmmum dONfiZ OUGOHOMOHW UUn—OHOHOHOH houfingm hodmumflmeOflH—H . «:1 Mom Mom oomHHms How :otz Ooungasm How mufiuHuH< .H OH Omuo>. "Hmuo>1ooz .Osuo>. “Houo>laoz “Houo>laoz vouo> oomHHm31Oounnas: NOOH OH monouommum no HOH>mnom wsHuo> HmHuamOHmoum mo mnha was Ho>0H monoumHmcoofiH mvnuHuu< momHHm3:Oouagaam mo OOOH tam Hm>mH flaw NOOH GH oofimuomoum Ho H0H>O£om wsHuo> HOHuaovHOon ammsuom mHnmfiOHuwHom O.m mHan 49 Wallace.2 It can also be seen that attitudinal consistency was higher among Wallace supporters than Humphrey sup- porters. And, finally, it is also interesting to note that Nixon supporters, as might be expected, were highly attitude inconsistent (or, better, "nonconsistent") with respect to Humphrey-Wallace attitudes. Behavioral—Psychological Involvement ifi PolifiEs Previous research (Campbell et a1., 1960; Converse, 1964; Milbrath, 1965) has shown that the greater the behavioral political involvement of an individual--as measured by his vote participation, for example--the greater will be his psychological interest or involvement in politics. The data in Table 3.8, showing the relation- ship between presidential voting behavior or preference in 1968 and level of interest in government and public affairs, provide a further test of this hypothesis.3 2Afurther analysis of these relationships showed that there was a significant positive (product-moment) correlation between voting behavior (or preference) and attitude inconsistency among both Humphrey supporters (r - .17, P < .001) and Wallace supporters (r = .15, P < .10); that is, voters were more attitude consistent than non-voters, controlling for candidate subgroup. 3This variable was measured in the post-election survey by the following question: "Some people seem to follow what's going on in government and public affairs most of the time, whether there's an election going on or not. Others aren't that interested. ‘Would you say you 50 .NN. u OOOHOHONOOO NOOOOOHOOOO .HOO. v O .OH n OO .ON.HO a NO N ”5.02 NOOH NOOH NOOH .NOOH NOOH NOOH NOOH quommm HOHOH :IIII :IIII IIIII IIIII IIII. :IIII IIII: HHO um ON O ON OH ON ON OH OHOHOC eBOHHom "OOH wmu> CmCu OCm 36C mH ON HN OH ON Nm OH OHCo OBOHHom u3OH m u o u 0 NO NN ON Om ON ON Om OOH“ msuHme "SOHOMZ m o o NNN NOO NHN NNO NON NOH NNO OOHHNOOMNHWO "OOHO me n 2 OO u 2 On n 2 OOO u z HO u z OOH n z mam n z .8 ac somHHm3 «UCmummoum Cotz OUCmuomoum OOCOHONOHO Ooh—COHEN PHONE UHHACm .HOCWHOOM Com momHHm3 Com Cotz Omunmaam How OCm . Omuo> "HouoNTCoz Omuo> "Houo>|Coz ”HouoNTCoz Oouo> uCoaCuoNSO CH OOOH CH OUCOCONQHO Ho uoH>mCom wCHuo> HmHuCovaoum UWOHOHQH HO H0>0~H OHHOMH< UHHOCN OCm uCoaCHo>oU CH ummuouCH mo Hm>0H OCO OOOH CH ooCmHomoum Ho HoH>mCom wCHuo> HmHuCOOHmmHm Coosuom OHCOCoHumHmm m.m mHan 51 As shown in Table 3.8, there was a highly significant relationship between these two variables. More specifically, it can be seen that those who voted for Humphrey, Nixon, or Wallace were much more interested ih government and public affairs than those who "pre- ferred" Humphrey, Nixon, or Wallace. More importantly-- for the purpose of the present research-~it can be seen that thewhighest level of interest or involvement in politics was among Wallace voters (49%) who werealso the highest in level of Humphrey-Wallace attitude incon- sistency (see Table 3.7). - Finally, Table 3.9 summarizes the multivariate“ relationships among presidential vote or preference for Humphrey or Wallace, level of interest in government and 5 public affairs, and level and type of Humphrey-wallace attitude inconsistency. As indicated in Table 3.9, there f0110W’What'S going on in government and public affairs most of the time, some of the time, only now and then, or hardly at all? [Codebook, 1971, p. 246]." For the present analysis, missing data respondents were recoded into the "follows some of the time" category (the modal response category). “The analysis-of-variance oriented researcher will recognize here that the elaboration model analysis (see, for example, Rosenberg, 1968) presented in Table 3.9 is essentially an analogue of a three-factor (non- randomized) design. 5High interest was defined in this analysis as follows ”most of the time" and "some of the time," while lOW’WaS defined as follows "only now and then" and "hardly at all" (see Table 3.8 and Appendix A). 52 .ovsuHuum OoHCOECm o>HumOwC OCm,ovsuHuuw oowHHms obHuHOom a .mOCuHuum mowHst m>HumwoC OCO OOCuHuum Omusmaum m>HuHmom w NNHO HNHO ANNO NOOO NOOO AHOO NNHHO NHNNO NanOz NOOH NOOH NOOH NOOH NOOH NOOH NOOH NOOH Hzmomum HOHOO HN ON ON NO O N O H OOHOHOOO HOH: ON ON ON HH O O N O OOHNHOOO aOHOmz AHZNHOHOZOO MOOHHHH< O O O H OH O O O m>HuHmom 30H O O O H NN ON ON OH OOHOOOOz OOH HZMHOHOZOOZH NOOHHHH< O O O O ON ON OO NN O>Hummmz aOHOoz NO NO NO NO NNN NON NOO NOO OOHOOOoz OOHO ONzONOHOzOO NOONHNNO OOH HOHO OOH OOH: OOH OOH: OOH OOHO muHmmm< oHHnCm OCm quaCHo>oO CH umwumuCH mo Hm>mH OoCoumeCooCH OOONHNNO ooCmuomoum muwHHm3 OUCoumwmum Omanmasm mommwmzlmouamasm mouHHmB How OOCCOECO now «o O OH OCm HO>~H "uwuo>ICoz Oouo> ”Hauo>ICoz Omuo> OoCoumHmCooCH OOCuHuu< momHHmznhmHCOEOm mo OOOB OCm Hm>mH OCm .OHHmmw< UHHACN CCO quECHo>oo CH ummuouCH mo Hm>0H .momHHmz Ho Oounqasm Com oonHommHm Ho ouo> HmHquOHmmHm wCoa< OOHCmCoHuOHmM mumHHm>HuHCz O.m mHan 53 were strong interactions among these variables such that, for example, Humphrey-Wallace attitude consistency Was highest among Wallace supporters with a high level of interest in government and public affairs and lowest among Humphrey non-voters with a low level of interest in government and public affairs. The reader is encouraged to make further examinations of the data presented in Table 3.9. Behavioral Involvement and Level of Constraint The relationship between behavioral political involvement (vote participation) and attitudes toward the three 1968 presidential candidates can also be analyzed in terms of Converse's (1964) concept of constraint, a notion which refers to the amount of "functional inter- dependence" or interconnectedness among elements in a belief-attitude system. Operationally, Converse has defined constraint in terms of the absolute degree of intercorrelation between items measuring various beliefs and attitudes. Thus the h1gher the absolute amount of Icorrelation, the greater the level of constraint. Using this operational definition, Converse (1964) found, for example, that the level of constraint between various issue beliefs (e.g., domestic issues such as federal wel- fare spending, school desegregation) was much higher among 54 the politically involved elite (congressional candidates) than among the political mass (national cross-section sample, 1958). From Converse's findings, we would expect that the level of constraint between attitudes toward Humphrey, Nixon, and Wallace--as measured by the sbsolute amount of correlation--would be higher among those who voted for these candidates (i.e., the more politically involved group) than those who did not vote, but "preferred" these candidates. Table 3.10 shows the Pearson product-moment correlations among attitudes toward Humphrey, Nixon, and Wallace--by presidential voting behavior or preference in 1968. Examination of the data in Table 3.10 shows that there was no systematic pattern of differences between voters and non-voters in the level of constraint among attitudes toward Humphrey, Nixon, and Wallace. That is, the absolute amount of correlation for the Humphrey- Wallace pair was higher among voters (.31) than non- voters (.21), while the Humphrey-Nixon correlation was higher among non-voters (.21) than voters (.09);6 and, 6An additional analysis of these data showed that the difference between the correlations for voters and non-voters on the Humphrey-Wallace pair was not statistically significant (Z = 1.46, P > .10, two-tailed test), while the difference on the Humphrey-Nixon pair approached statistical significance (Z = 1.66, P < .10, two-tailed test). 55 Table 3.10 Pearson Product-Moment Correlations Among Attitudes Toward Humphrey, Nixon, and Wallace--By Presidential Voting Behavior or Preference in 1968 PRESIDENTIAL VOTING BEHAVIOR OR PREFERENCE IN 1968 Voted for Humphrey, Nixon, or Wallace N = 873 Humphrey Nixon Wallace Humphrey ------ +.09* -.31** Nixon ------ -.01 Wallace ------ Non-Voter: Preferred Humphrey, Nixon, or Wallace = 9 Humphrey Nixon ‘Wallace Humphrey ------ +.21* -.21* Nixon ------ +.06 wallace ------ Residual: Other N = 66 Humphrey Nixon Wallace Humphrey ------ +.45** -.32* Nixon ------ +.06 Wallace ...... * P < .01 ** P < .001 56 the difference on the Nixon-Wallace pair was negligible. Finally, further examination shows that the level of constraint on the Humphrey-Nixon and Humphrey-Wallace pairs was actually highest among those in the residual group! When we also consider the direction of these cor- relations, it becomes clear that there were more than simple differences (or non-differences) in the level of constraint among these groups. For example, it can be seen that the Humphrey-Nixon pattern of correlation was slightly negative for the voting group (-.09), low posi- tive for the non-voters (+.21),.and moderately positive for the residual group (+.45). That is, there was a slight tendency for those who voted to have a positive attitude toward Humphrey and a negative attitude toward Nixon (and vice versa), while among those who didn't vote and the residual group there was a definite tendency to have either a positive or negative attitude toward both Humphrey and Nixon! Behavior-Attitude Consistency The somewhat surprising finding for the pattern of Humphrey-Nixon correlation suggested a further analysis of the relationships between voting behavior (or prefer- ence) and candidate attitudes. Table 3.11 shows the 7 5 ...... sz unnuuu Zum OOOOO. uuuuuu 3 OONOO. ...... 3 OO. OH. ...... z OO.- OO. uuuuuu z ONOOO.- NOOOO.- sssOm.. ...... m OOOOO.I *NOOO.- NOONO. uuuuuu m OON O 2 O OON O z .m momHHOB so OousmameOMHmwmoum ”Cmuo>ucoz momHHmz Ho wwwnmamm Com Owuo> ...... 2-x ------ 2-x OH. ...... 3 «OH. ...... 3 OOOOO. *NN. ...... z «NOOO. OOH. uuuuuu z OOOOO.- OO.- OH. ...... O OOOOO.- ONH.- OOON.- ...... O Ozé 3 z O sz O 2 O OOH u z OON u z Cotz Co Omunmasm Oouummmum ”Couo>nCoz Cotz No Omunmasm Com Omuo> OOOH ZH mozmmmmmmm mo MOH> HOCOO OCHuo> HmHuCOOHmmHm OCO mowCuHuu< oumOHOCmO wCoa< OOHCOCoHumHmm HH.O mHan 58 .Cotz Com OoCOHOHOHO Ho lOuo> O H HOHECC mnu OCO OUOHHO3 HOH OoCOHOHOHm Ho Ouo> O HOCOHO OCHuCOOOCOOH N HOOECC OCH CuH3 OHAOHHO> OCoaouoCoHO O OH 3 z .qusmasm How ooCOHOHOHO Ho Imuo> O H HOAECC map OCO OOOHHO3 How OoCOHOHOHa no Ouo> O HOCHHO OCHCCOOOHOOH N HOOBCC Osu CuH3 OHOOHCO> OsoaouoCOHO O OH 3um .annmmsm HOH OoCOHOHOHO Mo Ouo> O H HOAECC OCu OCO Cotz How ooCOHOHOHQ Ho Ouo> O HOCNHO OCHHCOOOHOOH N HOOBCC OCH £HH3.OHQOHHO> OCoaONoCOHO O OH Zum HOO. v O HO. v O NO mo. v m uuuuuu 3nz nun--- 3 ***mo . IIIIII 3 §*m© . llllll 3 NOONO.: NOOOO.: uuuuuu z ONOOO.I ONOHO.u uuuuuu z OOH.: NH.u NOONme ...... m *NHN.I OONN.- ONHN. uuuuuu m o3uz 3 z m 0312 3 z m ONH u 2 OOO u z OUOHHO3 Co Cotz OOHHOHOHO ”Hmuo>uCoz OoOHHO3 Co Cotz Com Omuo> 7' .Y t n g. -9 59 relationships among candidate attitudes and voting behavior or preference in 1968 for three possible can- didate combinations: Humphrey-Nixon, Humphrey-Wallace, and Nixon-Wallace. And, finally, Table 3.12 shows the relationships among candidate attitudes by each presi- dential voting behavior or preference group, separately; that is, holding candidate group constant. Inspection of these two tables shows that the tendency to have either a positive or negative attitude toward all three candidates--particular1y toward both Humphrey and Nixon--was much stronger among non-voters than voters, especially among those who preferred either Humphrey or Wallace (see Table 3.12). The latter two groups were also more likely to have either a positive or negative attitude toward both Nixon and Wallace, although this was also the case for Humphrey voters. Finally, it can be seen that voting behavior was more strongly asso- ciated with the appropriate candidate attitudes than was non- voterp'npreferences." For example (Table 3.11), "voting" for Humphrey or Nixon was more highly correlated with affective attitudes toward Humphrey (r = -.65) and Nixon (r the corresponding attitudes toward Humphrey (r = —.45) and .60) than was "preference" for Humphrey or Nixon with k 60 Table 3.12 Pearson Product-Moment Correlations Among Affective Candidate Attitudes-—Controlling for Presidential Voting Behavior or Preference Group in 1968 PRESIDENTIAL VOTING BEHAVIOR OR PREFERENCE GROUP IN 1968 Voted for Humphrey Non-Voter: Preferred Humphrey (N = 383) (N = 104) II N' 00 I1 N' VJ H ------ 49*** 02 H ------ 68*** .05 N ------ 17** N ------ .20* IV ------ VJ ------ Voted for Nixon Non-Voter: Preferred Nixon (N = 397) (N = 91) 11 N' VJ II N' VJ H ------ .09 -.14** H ------ .16 .03 N ------ .06 N ------ .09 I0 ------ VJ ------ Voted for Wallace Non—Voter: Preferred Wallace (N = 93) (N = 34) H N VJ H N VJ H ------ .14 .03 H ------ 51** -.10 N ------ '-.03 N ------ .29 V0 ------ VJ ------ * P < .05 ** P < .01 *** P < .001 NOTE: H, N, and W again represent affective attitudes toward Humphrey, Nixon, and Wallace, reSpectively. 61 Nixon (r = .57). In brief, these data furnished further evidence for behavior-attitude consistency.7 7These data may also help to explain, in part, why the original correlations for the Humphrey-Nixon and Nixon-Wallace affective attitudes were .00 (see Table 2.3). That is, voting and non-voting may have been a suppressor variable acting to partially cancel out the true rela- tionships between these attitude variables (see Table 3.12). For further interpretation of these relations, see the Discussion. ‘ v k a CHAPTER IV DISCUSSION As one writer has recently noted, "negative or inconclusive results are much harder to interpret than positive results [Kerlinger, 1973, p. 154]." Yet, as Kerlinger has also pointed out, "If we can be fairly sure that the methodology, the measurement, and the analysis are adequate, then negative results can be definite con- tributions to scientific advance, since only then can we have some confidence that our hypotheses are not correct [p . 154]. " We begin this chapter with a discussion of some negative evidence for the correlates or effects of status inconsistency in the present study. Negative Evidence for the Correlates or Effects of Status Inconsistency The present research was designed originally to test the hypothesis: the greater an individual's (socio- economic) status inconsistency, the greater will be his cognitive or attitudinal inconsistency with respect to relevant attitude objects (e.g., Humphrey, Nixon, and 62 ..Il..lllllul I 1 [I'll-'1 OJ 63 Wallace in 1968). The data presented here, however, suggest that except at the extremes of status incon- sistency (where the results actually ran slightly counter to the hypothesized direction) there was a negli- gible relationship between status inconsistency and (Humphrey-Wallace) attitude inconsistency. Additional analyses of the 1968 election data also suggested there was a negligible relationship between status inconsistency and presidential voting behavior or preference, except again at the extremes of status incon- sistency, where some small but inconclusive trends did emerge. In brief, the evidence for the correlates or effects of status inconsistency in the present study was trifling at best. Interestingly, these findings and conclusions are congruent with a recently growing body of negative evidence for the effects of status inconsistency and related constructs such as vertical mobility (Eitzen, 1973; Jackson and Curtis, 1972; Olsen and Tully, 1972). For example, Jackson and Curtis (1972) recently found little or no relationship between status inconsistency and forty-three dependent variables (e.g., political liberalism, racial prejudice) suggested as correlates in previous research. And even more relevant to the present research results, Eitzen (1973) found that general 64 socioeconomic status--rather than status inconsistency itself--accounted for consistency among liberal- conservative political attitudes. Thus the status inconsistency construct does not seem particularly use— ful for explaining variations in contemporary American political behavior. Inadequate Data, Methodology, or Theory? Before concluding our discussion of status incon- sistency, we need to consider why the results did not turn out as expected. One possible reason is that there were very low correlations between the various status dimensions and candidate attitudes,1 which precluded the occurrence of cross-pressure effects from inconsistent statuses. That is, for status inconsistency--which is essentially a special case of the more general phenomenon of social cross-pressure--to have an effect, there must be at least a moderate sized relationship between the different dimensions of status from which inconsistency is derived and the predicted or dependent variables of interest (e.g., candidate attitudes). To put it another way, inconsistent status cannot lead to very much conflict 1The average Pearson product-moment correlation between these two clusters was approximately .11, with a range from .05 to .16. 65 or inconsistency in attitudes which are largely unrelated or irrelevant to the separate status characteristics themselves. This, then, was the implicit meaning of ”with respect to relevant cognitive or attitude objects" in the original formulation of the hypothesis. In other words, the writer had originally assumed, on the basis of previous research (Berelson et a1., 1954; Campbell et a1., 1960; Lazarsfeld et a1., 1944), that there was at least a moderate sized relationship between socioeconomic status or social class and political atti- tudes or behavior, and thus that there would be some degree of relationship between socioeconomic status inconsistency and political (candidate) attitude incon- sistency. But, as we have seen from the data, this assumption was not fully justified since the correlations between these two clusters were quite low. Another possible reason for these negative results is that the measure of status inconsistency used in the present study--the difference score technique--was not a valid one. It can be shown, however, that this difference score measure is similarly related to the correlates of status inconsistency established in previous research (Kasl, 1969; Nam.and Powers, 1965). For example, previous research (Kasl, 1969) has shown that there is a signifi- cant relationship between the life cycle and various types 66 of status inconsistency such as the association of high education and low income among individuals under the age of 35. More generally, the data in the present research show that there was a significant relationship between age and education-occupational prestige status inconsistency, for example (r = -.24, P < .001). In short, the difference score measure of status incon- sistency would appear to have some empirical validity, at least. But both of the foregoing reasons-~as well as the original formulation of the status inconsistency hypothesis--are too methodological, too statistical; that is, too atheoretical in the psychological sense. This criticism also applies to the status inconsistency literature in general which has tended to reduce this problem to one of mostly methodological and statistical considerations (see, for example, Jackson and Curtis, 1972). More specifically, previous research (and the present research) on status inconsistency has assumed that discrepancies among broad socioeconomic character- istics such as education, income, and occupational prestige will have certain effects for various individ- uals or subgroups without determining whether such "status inconsistencies" have any particular psychologi- cal reality for the individuals affected. That is, most 67 of the research in this area has involved correlating and regressing selected dependent variables such as political attitudes on pairs of abstract demographic indicators of socioeconomic status. In other words, status inconsistency mayonly have effects on individuals who are psychologically aware of the inconsistencies among their status characteristics and for whom status consistency is personally significant and thus worth achieving. To put it another way, status inconsistency may only have consequences for an individual when it is translated into cognitive inconsistency. Given this assumption and those of cognitive consistency theory in general (Shaw and Costanzo, 1970), we would hypothesize that individuals who are aware of their status inconsis- tency and consider it important will experience some form of psychological tension which will motivate them to reduce the tension through changing one or more of their status characteristics (e.g., earning more money to bring one's income status in line with one's education level) in the direction of greater status consistency. Another hypothesis of relevance here would be that the greater an individual's status inconsistency, the more likely he will be to become psychologically aware of it. Or it may be that there is a threshold at which status inconsistency is sufficiently great to begin to produce I . 68 an awareness effect; i.e., only at the extremes of status inconsistency (e.g., among highly deviant individuals such as self—made millionaires or college-educated garbage collectors). In summary, probably the most important reason why the present results did not turn out as expected, as well as those of many previous studies, is an inadequate theo— retical analysis and accounting of the psychological reality of status inconsistency to the individual. Positive Evidence for Behavior—Attitude»Consistency Beginning with LaPiere's classic (1934) study of "attitudes vs. actions," social psychologists have been concerned with finding evidence for the degree of con- sistency or inconsistency between attitudes and behavior in both laboratory and naturalistic situations. As reviewed by Kiesler et a1. (1969), the empirical studies most frequently cited as evidence for an inconsistent relation between attitudes and behavior are those by LaPiere (1934), Kutner et a1. (1952), and Minardi(l952), while those by Nettler and Golding (1946), DeFleur and Westie (1958), and Sherif and Hovland (1961) are frequently 69 cited as evidence for consistency between attitudes and behavior.2 Accordingly, within this general context, the data presented here can be seen as providing some field survey evidence for a consistent relationship between behavior and attitudes. Specifically, the present research demonstrates that there is a more consistent relationship between behavior and attitudes among those who participate in an election by voting for a presi- dential candidate than among those who do not participate (see Table 3.11). The present data also demonstrate (Tables 3.11 and 3.12) that non-voters are less consis- tently discriminating in their overall attitudes in that they tend to have either a positive or negative attitude toward all presidential candidates, although this might also be due, in part, to a constant response set to evaluate politicians as all positive or all negative (see, for example, Guilford, 1954). But more importantly, other data in the present study (Table 3.7) furnish evidence for a consistent rela- tionship between behavior and attitude inconsistency itself (as measured by the difference score technique). 2For a detailed review of this evidence and the factors affecting the relationship between attitudes and behavior (e.g., situational differences), the reader is referred directly to Kiesler et a1. (1969); see also Campbell (1963) and Rokeach (1968). 70 Specifically, these data show that there was a significant relationship between presidential voting behavior or preference in 1968 and Humphrey-Wallace attitude incon- sistency, with those who voted for either Humphrey or Wallace being more attitude consistent than their non- voting counterparts. To express it another way, whereas previous research has focussed on obtaining evidence for a correlation between attitudes and behavior, the present study has provided evidence for a direct relationship between behavior (voting, non-voting) and attitudinal con- sistency (Humphrey-Wallace) per se. Finally, the data showing a significant relation- ship between behavioral involvement in politics (voting) and psychological involvement or interest in politics (Table 3.8), as well as those demonstrating the multi- variabe relationships among these variables (Table 3.9), provide further evidence for consistency between behavior and psychological states such as attitudes, interests, and motivation. These data also suggest that one reason why voters are more consistent in their candidate attitudes than non-voters is that they are generally more interested and informed about politics which, according to previous research (Converse, 1964), tends to be associated with a higher level of ideological conceptualization; that is, greater political attitude consistency. However, as we H 71 have also seen (Table 3.10), Converse's (1964) conceptualization of this relationship in terms of simple absolute differences in the level of constraint is inadequate. Rather, one needs to take into account the direction and pattern of relations between behavior and attitudes, as well as other psychological states such as interests. Attitudes-Follow-Behavior The behavior and attitude data in the present study can probably best be interpreted in terms of Bem's (1970) self-perception theory which is based on the notion that beliefs, attitudes, values, and other psycho- logical states tend to follow behavior. For example, one study cited by Bem (Lieberman, 1956) showed that a factory worker's attitudes changed in the pro-management direction if he was promoted to foreman, but became more pro-union if he was elected as union steward. But even more important, perhaps, Lieberman's data showed that when those who had been promoted to foremen were forced to resume their previous position of factory workers, their attitudes reverted to their previous pro-union direction. In short, Lieberman's study provided strong evidence for the notion that attitudes or attitude change follow behavior or behavior change. 72 Similarly, in the present study, those who voted for Humphrey, Nixon, or Wallace can be seen as bringing their affective attitudes toward each candidate in line with their own recent behavior in the election--i.e., by expressing more consistently pro or con feelings on the feeling thermometer during the post-election interviewe- while those who didn't vote but rather "preferred" Humphrey, Nixon, or Wallace can be seen as expressing the same attitude in two different ways: effectively on the feeling thermometer and cognitively in a verbal statement of candidate preference--both of which are lacking behavioral foundation; i.e., links with the overt act of voting.3 Finally, the data on behavioral and psychological involvement or interest in politics can also be inter— preted, in part, in terms of Bem's self-perception theory. That is, those who voted for Humphrey, Nixon, or wallace in the election can be viewed as inferring their interest in politics, to some extent, from their own recent behavioral participation,while those who did not vote 3The act or self-report of voting can also be viewed as a third (behavioral) component of the same attitude. However, the present writer feels that the analytical distinction between inner psychological states such as (affective) attitudes and overt non-psychological actions, including verbal statements about such behavior, is a useful one and should be maintained (but see also Kiesler et a1., 1969; and Rokeach, 1968). 73 would tend to see their recent non-participation as evidence (or further evidence) of their general lack of interest in political affairs. In summary, the present research provides some field survey support for the principle of attitudes- follow-behavior.“ Some Directions for Future Research As suggested indirectly before, one possible direction for future research in the area of status inconsistency is to determine its psychological signifi- cance for the individual. This would involve measuring two components: (a) awareness of the status inconsistency itself and (b) the personal importance of the status inconsistency to the individual. Given these measurements and the general assumptions of cognitive consistency theory, we would, again, hypothesize that the greater an individual's awareness of status inconsistency and the greater its personal importance to him, the more likely “The attitudes-follow-behavior interpretation here should be qualified, however, by pointing out that these data are, again, from a post—election survey. That is, they do not provide evidence that attitudes always follow behavior; rather, they illustrate a special case of this general principle. A more rigorous test of this relation would, of course, use a panel study design in order to determine whether attitudes are more consistently related to voting behavior before or after an election, for example. 74 he will be to attempt to change one or more of his status characteristics in the direction of greater status con- sistency. Testing this hypothesis in a field research study would, of course, involve the use of some type of panel or longitudinal design. 0n the other hand, an experimental test of this hypothesis can probably best be achieved through the use of small groups (see Sampson, 1969, for a review of some efforts in this direction). In general, this direction for new research on status inconsistency would seem to be worth exploring. As this writer sees it, the choice in this research area is between reducing a substantative sociological problem to a methodological-statistical one, as many previous researchers in this area have done (e.g., Jackson and Curtis; 1972; Olsen and Tully, 1972), and reducing it, or better, reformulating it into a truly social-psychological one as outlined above. In the area of attitude consistency, there are several potentially fruitful directions for future research. For one, it would seem worthwhile to further explore the degree of consistency among the affective and cognitive components of attitudes and overt behavior, using the difference score measure of consistency developed here. For example, it would be hypothesized that individuals who are behaviorally involved in 75 politics would not only show more consistent affective attitudes toward candidates in an election, as measured by the feeling thermometer, but also more consistent cognitive beliefs about the candidates' policy positions or stands on issues. And the greater the behavioral political involvement (e.g., voting, attending a political meeting, becoming an active member of a political party), the greater would be the affective and cognitive consis- tency. Such an hypothesis could be tested directly in most cross-sectional surveys and secondary analysis designs of the type used here. Another possible direction in this general area would be to study individual differences in affective or cognitive consistency over time. That is, another aspect or meaning of consistency which we have not explored is the degree to which an individual‘s beliefs, attitudes, and values are stable, reliable, dependable, or predictable over a particular time period (e.g., 3 months before to 3 months after an election). For instance, it might be hypothesized that the greater the behavioral involvement of an individual in an election campaign, the more (con- sistent) stable his cognitive and affective attitude components would be over the course of the campaign. This hypothesis could be tested by using a simple test-retest 76 design with Pearson product-moment correlations, for example, as coefficients of stability. Finally, in the area of attitude change, it might be hypothesized that attitude change, in either a field or laboratory situation, would be greatest among those individuals who are, initially, most inconsistent in their beliefs and attitudes as measured by the dif- ference score indicator of cognitive consistency. In other words, it would be predicted that it is easier to change those individuals whose attitudes are less con- sistent or integrated with one another to begin with than those whose attitudes are highly integrated and consistent. That is to say, in summary, that consistency and stability and inconsistency and change are companion pairs of con- StI’UCtS . HJ CHAPTER V SUMMARY In summary, this dissertation had three original purposes: (a) to develop a precise conceptual and opera- tional definition of the social-psychological construct of cognitive or attitudinal consistency within a field survey context, (b) to integrate this definition with the conceptualization and measurement of the sociological construct of status consistency, and (c) to test the social cross-pressures hypothesis that the greater an individual‘s socioeconomic status inconsistency, the greater will be his cognitive or attitudinal inconsistency with respect to relevant political attitude objects, such ‘ as presidential candidates. ~ Methodologically, the research design involved a secondary analysis of sample survey data from the 1968 American National Election Study conducted originally by the Survey Research Center of the Institute for Social Research at the University of Michigan. The data drawn from this national election study included the traditional measures of an individual's socioeconomic status--leve1 77 1 1i] 78 of education, income level, and occupational prestige-- and "feeling thermometer" measures of an individual's affective attitudes toward the three major 1968 presi- dential candidates: Hubert Humphrey, Richard Nixon, and George Wallace. These measures, in turn, were used to construct the indices of status and attitude consistency, both of which were defined conceptually in terms of deviance and operationally by simple differences between standard scores. The results, in summary, showed that there was a negligible relationship between the variables of socio- economic status consistency and consistency of affective attitudes toward the three 1968 presidential candidates, as well as between status inconsistency and actual voting behavior or preference in 1968, although there were some slight but inconclusive trends in these relationships among those who were highly status inconsistent; that is, the most deviant status types. Additional secondary analyses of the 1968 election data showed, moreover, that there was a statistically sig- nificant positive relationship between voting behavior and candidate attitude consistency, with those who voted in the election being more attitudinally consistent than those who did not vote. It was also found that those who were more behaviorally involved in politics, as measured by 79 participation in the election, were more psychologically involved or interested in politics. And, finally, it was discovered that cognitive or attitudinal consistency was greatest among those individuals who had a high level of interest in government and public affairs and were behaviorally involved in the election through the overt act of voting. Theoretically, it was concluded that the most important reason why the findings on status consistency did not turn out as predicted was an inadequate account of the psychological reality of various socioeconomic status discrepancies to the individual. Accordingly, it was suggested that future research on this subject should focus upon whether or not the individual is aware of being in an inconsistent status position and on the degree of importance it has for him or her personally. And last and most important, the positive evidence for a consistent relationship between voting behavior and affective candi— date attitudes was interpreted in terms of the self- perception principle that attitudes-follow-behavior. Sug- gestions for future research in this area included: (a) investigating the degree of consistency between affective and cognitive components of attitudes as a function of behavioral involvement, (b) studying the degree of individual attitude stability or consistency 80 over time, and (c) examining the relationship between attitude inconsistency, as measured by difference score indices, and attitude change in both laboratory and naturalistic settings. Thus, to a great extent, the present study can be seen as an exploratory one, the true purpose of which was to become more familiar with the phenomena of status and cognitive or attitude consistency, such that a future study can be designed with greater theoretical and methodo- logical sophistication. REFERENCES I. 1L ., REFERENCES Abelson, R. P., Aronson, E., McGuire, W. J., Newcomb, T. M., Rosenberg, M. J., and Tannenbaum, P. H. Theories of Co nitive Consistenc : A.Source- 5005. Chicago: Rind McNaIIy, I568. Aronson, E. The Theory of Cognitive Dissonance:' A Current Perspective. In L. Berkowitz (Ed.), Advances in Ex erimental Social Ps cholo . Vol. 4. BW’ ork: Academic Press, 1966, Beloff, H. Two Forms of Social Conformity: Acquiescence ; . and Conventionality. Journal of Abnormal and Social Psychology, 19587'§§, 99-104. Bem, D. J. Beliefs, Attitudes,and Human Affairs. Belmont, GaIif.: Frocks/Cole, 1970. Benoit-Smullyan, E. Status, Status Types, and Status Interrelationships. American Sociological Review, 1944 2, 151-16I’ ' Berelson, B. R., Lazarsfeld, P. F., and McPhee, W. N. Voting. Chicago: University of Chicago Press, Bishop, G. F., Barclay, A" M., and Rokeach, M. Presi- dential Preferences and Freedom-Equality Value Patterns in the 1968 American Campaign. Journal of Social Psychology, 1972, 88, 207-212. Blalock, H. M. The Identification Problem and Theory Building: The Case of Status Inconsistency. Public Opinion Quarteriy, 1966, 39, 130—132. Brehm, J. W., and Cohen, A. R. Explorations in Cognitive Dissonance. New York: Wiley, 1962. Broom, L., and Jones, F. L. Status Consistency and Political Preference: The Australian Case. American Sociological Review, 1970, 32, 989-1001. 81 .. ‘r E. ‘ —~—— . 82 Campbell, A., Converse, P. E., Miller, W. E., and Stokes, D. E. The American Voter. New York: Wiley, 1960. Campbell, D. T. Social Attitudes and Other Acquired Behavioral Dispositions. In S. Koch (Ed.), Ps cholo : A Stud of a Science. New York: Mcéraw-HiII, I963, pp. 94-I72. Campbell, D. T., and Stanley, J. C. Experimental and , Quasi—Experimental Designs for ResearEh. Chicago: Rand McNalIy,I963. Converse, P. E. The Nature of Belief Systems in Mass Publics. In D. E. Apter (Ed.), Ideology and Discontent. New York: Free Press, 1964, pp. 260- 61. Converse, P. E., Miller, W. E., Rusk, J. C., and WOlfe, A. C. Continuity and Change in American Politics: Parties and Issued in the 1968 Election. American Political Science Review, 1969, 63, pp. 10833II057 DeFleur, M. L., and Westie, F. R. Verbal Attitudes and , Overt Acts: An Experiment on the Salience of {- Attitudes. American Sociological Review, 1958, 2_3, 667-73. Duncan, 0. D. A Socioeconomic Index for All Occupations. In A. J. Reiss, Jr., et al. (Eds.), Occupations and Social Status. New York: Free Press, 1961, pp. 109-38. Eitzen, D. S. Status Consistency and Consistency of Political Beliefs. Public Opinion anrterly, 1973, 36, 541-48. - Festinger, L. A Theory_of Cognitive Dissonance. Stanford: University Press, I957. Festinger, L. Conflict, Decision, and Dissonance. Stan— ford: University Press, 1964. Flanigan, W. H. Political Behavior of the American Electorate. BOston: Allyn and Bacon, 1972. Goffman, I. Status Consistency and Preference for Change in Power Distribution. American Sociological Review, 1957, 22, 275-81. '7 83 Guilford, J. P. Psychometric Methods. New York: McGraw- Hill, 1954. Hollander, E. P., and Willis, R. H. .Some Current Issues in the Psychology of Conformity and Nonconformity. Psychological Bulletin, 1967, 68, 62—76. Hyman, H. H. Secondary Analysis of Sample Surveys. New York: Wiley, 1972. Jackson, E. F., and Curtis, R. F. Conceptualization and Measurement in the Study of Social Stratification. In H. M. Blalock and A. B. Blalock (Eds.), Methodology in Social Research. New York: McGraw- Hill, 1968, pp. 112-49. Jackson, E. F., and Curtis, R. F. Effects of Vertical Mobility and Status Inconsistency: A Body of Negative Evidence. Americany§ociological Review, 1972, 31, 701-13. '7 Jones, E. E., and Gerard, H. B. Foundations of Social ngchology. New York: Wiley, 1967. Kasl, S. V. Status Inconsistency: Some Conceptual and Methodological Considerations. In J. P. Robinson, R. Athanasiou, and K. B. Head (Eds.), Measures of Occupational Attitudes and Occupational Character- isfics. Ann Arbor: Institute for SOciaIiResearch, 1969, pp. 377-95. Kelly, D. K., and Chambliss, W. J. Status Consistency and Political Attitudes. American Sociological Review, 1966, 31, 375-82. Kenkel, W. F. The Relationship Between Status Consistency and Political Economic Attitudes. American Sociological Review, 1956, 21, 365-68. Kerlinger, F. N. Foundations of Behavioral Research. New York: Holt, Rinehart, and Winston, 1973. Kiesler, C. A., Collins, B. E., and Miller, N. Attitude Change. New York: Wiley, 1969. Kutner, B., Wilkins, C., and Yarrow, P. R. Verbal Attitudes and Overt Behavior Involving Racial Prejudice. Journal of Abnormal and Social Psychology, 1952, 5:, 649-52. 84 LaPiere, R. T. Attitudes vs. Action. Social Forces, 1934, 33, 230-37. Laumann, E. 0., and Segal, D. R. Status Inconsistency and Ethno-religious Group Membership as Determinants of Social Participation and Politi- cal Attitudes. American Journal of Sociology, 1971, 11, 36-60. Lazarsfeld, P., Berelson, B., and Gaudet, H. The People's Choice. New York: Columbia University Press, Lenski, G. E. Status Crystallization: A Non-Vertical Dimension of Social Status. American Sociological Review, 1954, 33, 405-13 Lenski, G. E. Status Inconsistency and the Vote: A Four Nation Test. American Sociological Review, 1967, 33, 298-301. Lenski, G. E. Social Participation and Status Crystalliza- tion. American Sociological Review, 1956, 33, 458-64. Lieberman, S. The Effects of Changes in Roles on the Attitudes of Role Occupants. Human Relations, 1956, 3, 385-402. Magnusson, D. Test Theory. Reading, Mass.: Addison- Wesley, I967. McGuire, W. J. A Syllogistic Analysis of Cognitive Rela- tionships. In M. J. Rosenberg et al. (Eds.), Attitude Organization and Change. New Haven, Conn.: YaIe University Press, 1960, pp. 65—111. Milbrath, L. W. Political Participation. Chicago: Rand McNally, 1965. Minard, R. D. Race Relationships in the Pocahontas Coal Field. Journal of Social Issues, 1952, 3, 29-44. Mitchell, R. E. Methodological Notes on a Theory of Status Crystallization. Public Opinion Quarterly, 1964, 28, 315-25. Nam, C. B., and Powers, M. G. Variations in Socioeconomic Structure by Race, Residence, and the Life Cycle. American Sociological Review, 1965, 39, 97-103. 85 Nettler, C., and Golding, E. H. The Measurement of Attitudes Toward the Japanese in America. American Journal of Sociology, 1946, 33, 31-39. Nunnaly, J. C. Psychometric Theory. New York: McGraw- Hill, 1967. Olsen, M. E., and Tully, J. C. Socioeconomic-Ethnic Status Inconsistency and Preference for Political Change. American Sociological Review, 1972, 33, 560-74. Pepitone, A. Some Conceptual and Empirical Problems of Consistency Models. In S. Feldman (Ed.), Cogni- tive Consistenc . New York: Academic Press, 1966, pp. 257-297. Rokeach, M. Beliest_Attitudes, and Values. San Francisco: Jossey-Bass, 1968. Rosenberg, M. The Logic of Survey Analysis. New York: Basic Books, 1968. Rush, G. B. Status Consistency and Right-Wing Extremism. American Sociological Review, 1967, 33, 86-92. Sampson, E. E. Studies of Status Congruence. In L. Berkowitz (Ed.), Advances in Experimental Social Ps cholo . Vol. 4. New York: Academic Press, 1969, pp. 225-70. Segal, D. R. Status Inconsistency, Cross Pressures, and American Political Behavior. American Sociological Review, 1969, 33, 352-59. Shaw, M. E., and Costanzo, P. R. Theories of Social Psy- chology. New York: McGraweHi11,'1970. Sherif, M., and Hovland, C. I. Social Judgment: Assimila- tion and Contrast Effects in Communication and Attitude Change. New Haven, Conn.: Yale University "i Press, 1961.~ Smith, T. S. Structural Crystallization, Status Incon- sistency, and Political Partisanship. American Sociological Review, 1969, 33, 907-21. Sperlich, P. W. Conflict and Harmony in Human Affairs. Chicago: RanndNally, 1971. 86 The SRC 1968 American National Election Study. Ann Arbor: Inter-University Consortium for7Political Research, 1971. Treiman, D. J. Status Discrepancy and Prejudice. Ameri— can Journal of Sociology, 1966, 33, 651-64. APPENDICES APPENDIX A CODEBOOK FOR THE STUDY VARIABLES APPENDIX A CODEBOOK FOR THE STUDY VARIABLES In this section, the main variables used in the f“ present study are described and their location by both variable number and page number in the original codebook for The SRC31968rAm§rican National Election Study (1971) Er” «.5 pr .1 provided. The code values shown for the variables below represent those used in the present research. Var. Page Description of Variable No. No. Respondent's Education Level:1 0156 82-84 1. None 2, 1 grade 3. 2 grades 4. 3 grades 5. 4 grades 6. 5 grades 7. 6 grades 8. 7 grades 9. 7 grades plus non-college training w—w- jw . wV—fi—v v v ‘f 1The‘mean of this variable for the present sample (N = 1168) was 16.17 and the standard deviation was 4.80. 87 88 Var. Page Description of Variable No. No. 10. 8 grades 11. 8 grades plus non-college training 12. 9 grades 13. 10 grades 14. ll grades r. 15. 9 grades plus non-college training 5 16. 10 grades plus non-college training € 17. 11 grades plus non—college training 2 18. 12 grades 19. 12 grades plus non-college training 20. Some college 21. Bachelor's degree (4 or 5 years college) 22. Master's degree or equivalent 23. Ph.D. or equivalent 24. J.D. or equivalent 25. M.D. or equivalent 26. J.C.D. or equivalent 27. Honorary degree (LLD, DD, LHD) Head's Occupational Prestige:2 0171 92 00. Lowest (e.g., tobacco manufacturing laborer) Etc. 96. Highest (e.g., dentist, osteopath) “— vrv—rv— ‘V 3‘1 2Themean of this variable for the present sample (N = 1168) was 40.12 and the standard deviation was 25.14. 89 Var. Page Description of Variable No. No. Respondent and Family's Total Annual Income: 10. Under $1000 11. $1,000 - 1,999 12. $2,000 - 2,999 13. $3,000 - 3,999 14. $4,000 - 4,999 15. $5,000 - 5,999 16. $6,000 - 6,999 17. $7,000 - 7,999 18. $8,000 - 8,999 19. $9,000 - 9,999 20. $10,000 - 11,999 21. $12,000 - 14,999 22. $15,000 - 19,999 23. $20,000 - 24,999 24. $25,000 and over Presidential Vote or Preference in 1968: 0316 174-175 1. Voted for Humphrey 2. Non-Voter: Humphrey preference 3. Non-Voter: Nixon preference 4. Voted for Nixon 5. Non-Voter: Wallace preference 6. Voted for Wallace 7. Residual: Other 3Themean of this variable for the present sample (N = 1168) was 17.18 and the standard deviation was 3.62. 90 Var. Page Description of Variable No. No. Interest in Government and Public Affairs: 0430 246 1. Follows most of the time 2. Follows some of the time 3. Follows only now and then 4. Follows hardly at all Feeling Thermometer Rating of Humphrey: 0479 274 00. Zero degrees (very cold feeling) 01. One degree 02. Two degrees 03. Three degrees Etc. 15. Fifteen degrees (quite cold feeling) 30. Thirty degrees 40. Forty degrees (a bit more cold feeling than 50 degrees) 50. Fifty degrees (no feeling at all) 60. Sixty degrees (a bit more warm feeling than 50 degrees) 70. Seventy degrees (fairly warm feeling) 85. Eighty-five degrees (good warm feeling) 97.“ Ninety-seven degrees (very warm feeling) ‘v— v r y fir fi fi “Responses of 97, 98, 99, or 100 degrees on the thermometer were coded as 97 in the SRC Election Study. 91 Var. Page Description offiVariable No. No. Feeling Thermometer Rating of Nixon: 0480 274 00. Zero degrees (very cold feeling) 01. One degree 02. Two degrees 03. Three degrees Etc. 15. Fifteen degrees (quite cold feeling) 30. Thirty degrees (fairly cold feeling) 40. Forty degrees (a bit more cold feeling than 50 degrees) 50. .Fifty degrees (no feeling at all) 60. Sixty degrees (a bit more warm feeling than 50 degrees) 70. Seventy degrees (fairly warm feeling) 85. Eighty-five degrees (good warm feeling) 97. Ninety-seven degrees (very warm feeling) Feeling Thermometer Rating of Wallace: 0478 274 00. Zero degrees (very cold feeling) 01. One degree 02. Two degrees 03. Three degrees Etc. 15. Fifteen degrees (quite cold feeling) 30. Thirty degrees (fairly cold feeling) 92 Var. Page Descriptionwof Variable No. No. 40. Forty degrees (a bit more cold feeling than 50 degrees) 50. Fifty degrees (no feeling at all) 60. Sixty degrees (a bit more warm feeling than 50 degrees 70. Seventy degrees (fairly warm feeling) 85. Eighty-five degrees (good warm feeling) 97. Ninety-seven degrees (very warm feeling) Respondent‘s Age: 0533 302 21. Twenty—one years Etc. 98. Ninety-eight years 99. Not applicable APPENDIX B SUBGROUP CLASSIFICATION APPROACH TO STATUS AND ATTITUDE INCONSISTENCY APPENDIX B SUBGROUP CLASSIFICATION‘ APPROACH TO STATUS AND ATTITUDE INCONSISTENCY As noted in Chapter II on Methods, a second approach to the analysis of both status and candidate attitude inconsistency involved breaking down the status and candidate attitude dimensions into a number of sub- groups and then examining the relationships between pairs of status and attitude variables within various combina- tions of subgroups. This subgroup classification approach will now be illustrated. For example, one analysis involved examining the relationship between occupational prestige and attitude toward Humphrey within subgroup combinations of level and type of education and level and type of Wellace attitude. This analysis was done as follows: 1. First, level of education was broken down into five naturalistic control categories: (a) Grade iv v—vv— 1For a more detailed discussion of the use of sub- group classification techniques in survey research, see Morris Rosenberg, The Logic of Survey Analysis (New York: Basic Books, 1968)T7 93 i. .9 III" ll‘lllclll l 94 school education, (b) Some high school, (c) High school education, (d) Some college, and (e) College education. Then, the average occupational prestige and standard deviation for each of these five categories was computed. Finally, standard (Z) scores on occupational prestige for individuals ‘within each of these categories were obtained. Second, Wallace attitude (as measured by the feeling thermometer) was broken down into six naturalistic control categories: (a) Less than 10 degrees-~very cold feeling, (b) 10 to 29 degrees--quite cold feeling, (c) 30 to 49 degrees--fairly cold feeling. (d) 50 degrees--neutral or no feeling at all, (e) 51 to 74 degrees--moderately warm feeling, and (f) 75 degrees and over--good or very warm.feeling. Then, the mean Humphrey attitude and standard deviation for each of these six categories was computed. Finally, standard (Z) scores on Humphrey attitude for individuals within each of these categories were obtained. Third, each of the five categories of education level was cross-classified (using the counter- sorter) by each of the six categories of wallace h 95 attitude, resulting in 30 separate subgroup combinations of education level X Wallace attitude. 33323. At this stage in the analysis, it was intended to index status and attitude inconsistency as follows (see Kasl, 1969): first, define those cases that are + or -.50 or 1.00 standard deviation above and below the mean for I” each category type as "inconsistent" (e.g., +1.67 2 on occupational prestige for high school education and -1.43 Z on Humphrey attitude for the category of fairly cold feel- ing toward Wellace) and those within these limits as 5_ "consistent"; and, then examine the relation between these status and attitude inconsistency types (occupational prestige, Humphrey) within each subgroup combination (high school education X Wallace attitude). However, it was realized that this additional breakdown procedure would result in too few cases for an appropriate analysis (e.g., chi-square of the relationship between status and attitude inconsistency types within each of the 30 subgroup combina- tions (see Table B.1 below). Thus it was decided to simply explore the relationship between occupational prestige and Humphrey attitude within the subgroup combinations of edu- cation level X wallace attitude, for example. 4. Fourth, the Pearson product-moment correlation(s) between occupational prestige and attitude toward 96 Humphrey was obtained within each of the 30 separate subgroup combinations of education level X wallace attitude.2 Table B.1 shows the Pearson product-moment corre- lations between occupational prestige and attitude toward Humphrey within each of the subgroup combinations of level and type of education X level and type of Wallace attitude. As shown in Table B.1, only one of the corre- lations was statistically significant at the conventional .05 level, a negative correlation for the subgroup of high school education X neutral Wallace attitude. It can also be seen that most of the correlations--like the correlation for the sample as a whole (r = -.15, N - 1168)--are in the negative direction; that is, the greater the occupational prestige, the more negative the attitude toward Humphrey and vice versa. Thus these findings are not particularly surprising or interesting in themselves. Further analysis of these data, however, show that there was a statistically significant inverse relationship between the size of the subgroup combination and the absolute size of the correlation (rho - -.59, Z - 3.18, 2The reader may note that this particular aspect of the analysis is somewhat analogous to the parametric technique of semipartial correlation (see Jum.C. Nunnaly, Psychometric Theory [New York: MbGraw~Hill, 1967], pp.’154-55). 97 no. v m * maHHoom OHHOHo>Om m «0.- «N «H. Hm OH.- om mH.- mm so. so ausz.muHao aH em.- mm OH.- Hm «H.- as mH.- am m~.- “mswwwmmsmwmwmmwmm OH OH.- «H Hn.- as *mm.- mm Ho.- as sH.- ku NM mmwwmwm Nu Ho.- mm AH.-. .mm NH. om HH.- Hm HH.- waHstm wwmwumwwmwm mH mm.- om mH. as mo.- Hu «6.- AN mH.- waHHmmmoswmwwowwmmw um Ho.- on ~H.- an mH.- «a no. no se.- waHHoomanwmmm>Mwwm u z a m u z u a u z n a u z u a u z u m 203823 Hoodoo .883 .888 :on Hosea 6333... 86:63 momoooo exam OOH: mzom moses mo Home was Ho>mo GoHuOosom mo OQNH UGO Ho>oH monuHuu< ooOHHO3_Ho OQNH osO Ho>OH N soHuOosom mo make osO Ho>oH sHsuH3.ovauHuu< Nouneasm UGO owHumoum HOsoHuOmsooo soosuom OsoHuOHouuou usoaoz:uosooum somuoom H.m OHDOH 98 P < .01).3 Although this relationship can be seen as a somewhat spurious one because of the greater sampling error associated with the small N—values, it can also be interpreted in terms of a sort of deviant subgroup inter- action effect. That is, the stronger relationships between occupational prestige and Humphrey attitude in the small subcells are due to the unusual (deviant) combinations of education level and wallace attitude; for example, the deviant combination of college education X quite warm or favorable feeling toward Wallace (r = -.63, N = 5). In brief, this technique would seem to be a potentially useful one (albeit rudimentary) for detecting interaction type effects which would not be picked up by standard semipartial correlation analysis, for example. 3It should be noted here that this pattern of rela- tionship was also found for other subgroup combinations; for example, there was a significant inverse relationship between the size of the subgroup combination and the absolute size of the correlation (rho - -.50, Z - 2.69, P < .01)--for the correlation between income level and Humphrey attitude within the subgroup combinations of level and type of education X level and type of Nixon attitude. 1 L WICHIGAN STQTE UNIV. LIBRQRIES WWIIWWIWWIIWWVIIIHIIHIIIHIUHHI 31293103273045