STATUS ENCGNSISTENCY AND DEMOCRATEC PARTY ‘ ‘3?fo 'EREHCE Thesis fart-13 Degree of M. A. M ANN STA EUAEVERSHY PHiLUP AE‘éTHOW SALGPEK 1973 ML 1,152.43: Michigan State ..« Unix crsity ABSTRACT STATUS INCONSISTENCY AND DEMOCRATIC PARTY PREFERENCE By Phillip Anthony Salopek The purpose of this research endeavor is to test the theory of status inconsistency. The analysis is restricted to an investigation of the relationship between status inconsistency and Democratic party preference. The theory of status inconsistency can be formulated in the following manner. There are a number of statuses in any social system on which an individual is ranked. These status ranks are weighted according to their importance in society and one's overall status is a combination of the relevant rankings and their weights. A person's evaluations on the status indicators may be inconsistent. That is, an individual may hold a high ranking on some of the hierarchies but a low rank on the others. When this is the case, the actor has an inconsistent status configuration. A person whose ranks on the statuses are similar, however, possesses a consistent status configuration. Inconsistent status ranks produce a state of tension within the individual. It is assumed that a person will attempt to reduce this strain by bringing his statuses into line. Possible responses to the stress created by status inconsistency are: revolution, mobility, isolation, or political Phillip Anthony Salopek liberalism. We are concerned with the relationship between status incon- sistency and the latter response mentioned above. The problem of this research is to determine if status inconsistent individuals are more likely to voice a preference for the Democratic party than consistent persons. We utilize multiple regression analysis with dummy variables to test the hypothesis that status inconsistent persons are more likely than consistents to prefer the Democratic party. The relevant status variables for our study are education, occupation, income, and religious preference. The measure of status inconsistency is computed in the following fashion. Each respondent's score on the education, occupation, and income hier- archies is converted into a standard score. The standard deviation among the three standard scores is then figured for each person. The standard deviation, scaled by a factor of 10, is our measure of status inconsistency. The inconsistency score has a range of from zero to twenty-fbur. A low score signifies little variation among the hier— archies and thus, a status consistent individual. A high score indicates large discrepancies among the statuses and a status inconsistent person. The data analyzed in this research represent a random sample of the rural population of Michigan. The data was collected during the summer of 1970 by the Gallup organization through the use of an interview schedule constructed by members of the Sociology Department at Michigan State University. There are 3&3 respondents in the data set. The computing programs which were used to analyze the data could not handle missing data. Therefbre, the N for the regressions is 231. The major finding of this research is that status inconsistency is not related to Democratic party preference. The regression equation including terms for education, occupation, income, and status Phillip Anthony Salopek inconsistency does not significantly reduce the amount of unexplained variance in Democratic preference beyond that explained by the status variables alone. We also test the effects of achieved—ascribed and investment-reward inconsistencies. Neither of these specific types of status inconsistency are related to a preference for the Democratic party. We conclude that the primary factor responsible for our own and others' inability to substantiate the theory of status inconsistency is the incomplete and vague fermulation of its basic assumptions. The indeterminacy of the theory is a result of a lack of conceptualization and understanding of the processes of comparison, balance, reference, response, and activation. Until these processes are more completely comprehended, a specification of the theory of status inconsistency is not possible. Further research is unwarranted until the specification of the theory is achieved. STATUS INCONSISTENCY AND DEMOCRATIC PARTY PREFERENCE By Phillip Anthony Salopek A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Sociology 1973 ACKNOWLEDGMENTS I would like to sincerely thank Dr. Christopher Vanderpool, my committee chairman, for his valuable direction and assistance in this study. In addition, I am very grateful to Dr. J. Allan Beegle for making the data for this study available to me. Lastly, I wish to acknowledge the contribution of my wife Kathy, whose patience and encouragement has been a major factor in the completion of this project. TABLE OF CONTENTS LIST OF TABLES . . . . . . . . . . . . CHAPTER I. II. III. IV. INTRODUCTION . . . . . . . . . The Origins of Status Inconsistency . The Theory of Status Inconsistency . . A General Review of Status Inconsistency Theory and Research . . . Statement of the Problem . . REVIEW OF THE SELECTED RELEVANT LITERATURE Section I: Status Inconsistency and Political Party Preference . . . . Section II: The Measurement and Analysis of Status Inconsistency. . . . . . . . . . . . HYPOTHBSES, DATA, AND METHODS . Hypotheses . . . . . . . . . Data . . . . . . . . . . . . Method . . . . . . . . . . . Preliminary Analysis . . . . RESULTS . . . . . . . . . . . . SUMMARY 0 O O O O O O O O O O 0 iii Page iv 1% 1n 25 an an 37 so 5a 61 72 TABLE 10. 11. 12. 13. 1M. 15. LIST OF TABLES Distribution of Sample by Age . . . . . . . . . . . Distribution of Sample by Sex . Distribution of Sample by Social Class . . . . . . Distribution of Sample by Religion . . . . . . . . Distribution of Sample by Occupation . . . . . . . Distribution of Sample by Education . . . . . . . . Distribution of Sample by Party Preference . . . . Distribution of Sample by Income . . . . . . . . . Matrix of Correlation Coefficients Among the Independent Variables . . . . . . . . . . . . . Distribution of Sample by Party Preference . . . Distribution of Sample by Occupation . . . . . . . Distribution of Sample by Income . . . . . . . . . Correlation Coefficients Between Party Preference and Age, Sex, Social Class, Education, Religion, Occupation, and Income . . . . . . . . . . . . . Multiple Regression Solutions for Estimating the Probability of Claiming a Preference for the Democratic Party . . . . . . . . . . . . . . . . Multiple Regression Solutions from Broom and Jones ' Study 0 O O O O O O O O O O C O O O O 0 iv Page '39 39 HO #2 ”3 an HS us 53 56 57 58 59 63 65 CHAPTER I INTRODUCTION This section traces the theoretical development of the concept of status inconsistency by examining some of the research and theor- izing which preceeded its conceptualization. In light of these origins, alterations in the original theory are discussed and a de— finition of the concept of status inconsistency, as used in socio— logy today, is formulated. This introduction concludes with a state- ment of the research problem. I The Origins of Status Inconsistency During the 19u0's American sociologists rediscovered the works of the German theorist, Max Weber. His writings have played an integral part in the development of sociological theory since that time. Specifically, his "Class, Status, Party" (Weber, 19MB) is of particular importance in the development of the theory of status inconsistency. Following Weber's conceptualization of a multidimen- sional status system, American sociologists began to conceptualize social status in terms of its non-vertical aspects. Prior to the resurgence of Weber's works, the dominant theories of social status in this country were unidimensional (e.g. Warner 8 Lunt, 19M1). Individuals were seen as being ranked on a single scale. However, Weber's paper pointed to a number of components which deter- mined one's status. Weber's pioneering effort led sociologists to hypothesize that each component of status formed its own hierarchy on which an individual was ranked. A person holds different rankings on the various scales and one's overall status is a composite of the relevant hierarchies. These coexisting hierarchies form a multi- dimensional status system and offer a better explanation of how status is determined than the unidimensional theories. In viewing status as a multidimensional system, it became obvi- ous that some individuals would have contrasting ranks on the hier- archies which comprised their status. Given the possibility of incongruent ranks, the affect of being high on some of the hier- archies, but low on others, needed to be formulated and researched. In an early article, Hughes (194R) outlined some of the pos- sible consequences of such contradictions on political behavior. Other theorists postulated an equilibrium effect which operated to keep one's rankings at a comparable level (Benoit-Smullyan, 19nu; Fenchel, Monderer, 5 Hartley, 1951). The status equilibration hypothesis designates those statuses which are most important in society and postulates a tendency for a person's ranks on these scales to reach and maintain a common level. If one's ranking on the economic hierarchy, for instance, rises above the rankings on the political and prestige scales, this state of disequilibrium is cor- rected by a subsequent increase on the latter two hierarchies (Benoit- Smullyan, 19AM). Thus, a common level of these rankings presents a unified picture of one's status, allowing more predictable, orderly interaction with others. Although this theory was not based on empirical evidence, it was theoretically valuable. The equilibra- tion hypothesis supported Hughes' contention that status discrepancies might be related to political behavior by asserting that when the tendency toward a common level is hindered, social tensions of a revolutionary magnitude often result. The Theory of Status Inconsistency Having established a theoretical base for the concept of a multi- dimensional status system, sociologists began to investigate the relationship between such a system and a number of dependent variables. Studies linking status discrepancies with performance in small groups (Adams, 1953), political-economic liberalism (Lenski, 195k), social participation (Lenski, 1956b), preference for change in power distribution (Goffman, 1957), and symptoms of stress (Jackson, 1962) were conducted. The theory of status inconsistency, as proposed by these re- searchers, can be formulated in the following manner. There are a number of statuses in any social system on which an individual is ranked. These status ranks are weighted according to their impor- tance in society and one's overall status in a combination of the rel- evant rankings and their weights. A person's evaluations on the status indicators may be incongruent. That is, an individual may hold a high ranking on some of the hierarchies but a low rank on the others. When this is the case, the actor has an inconsistent status configuration. A person whose ranks on the statuses are similar, however, possesses a consistent status configuration. The basic assumptions of the theory state that while consistent ranks are stable, inconsistent ranks are not. It is assumed that imbalanced ranks will tend to change until they become consistent (Zelditch 8 Anderson, 1966). In this respect, we see that the status equilibration hypothesis is incorporated into the theory of status inconsistency. In addition, incongruent ranks produce a state of tension within the individual. It is assumed that a person will attempt to reduce this strain by bringing his statuses into line. If the attempt to change his statuses is blocked, even greater tension is created. The actual form the tension may take varies a great deal. The stress may result in revolution, mobility, isolation, politcal liberalism or right-wing extremism. Which of these responses will occur in any particular situation has not been determined by the theory. Nor does the theory specify exactly how balance among the ranks is accomplished. It only states that inconsistencies are un- stable, that incongruent ranks will tend to change until they become balanced, that inconsistent ranks generate tension, and that individ- uals will try to reduce this tension by bringing their statuses into line. A General Review of Status Inconsistency Theory and Research One of the foremost figures in status inconsistency theory and research is Gerhard Lenski. Most of the work done in this field since 1954 is based on his two pioneering studies (Lenski, 195a, 1956b). Lenski's reports contain the first formal definition of the concept of status inconsistency, or as he dubbed it, "status crystallization". He was also the first to operationalize the idea of a non-vertical dimension of status for research purposes. He accomplished this operationalization by measuring his subjects on four different vertical hierarchies: education, occupation; income, and ethnicity. To derive a measure of the comparability of one's rankings across these hierarchies, a standard score for each scale was computed and the crystallization value determined "by taking the square root of the sum of the squared deviations from the mean of the four hierarchy scores of the individual and subtracting the resulting figure from one hundred (Lenski, 195“ pp ”07-408)". The crystallization score measures the degree to which a person's rankings on the status hierarchies are at a comparable level. In Lenski's formulation, the consistent individual possesses congruent expectations of how he should behave and how others should interact with him. The inconsistent individual, however, may have conflicting expectations due to his differing status ranks on the hierarchies in question. More specifically, Lenski states that the tendency of the actor is to evaluate himself in terms of his highest single rank. He there- fbre expects others to accord him the prestige associated with this rank. The "other" who is interacting with the actor, however, responds to the actor's lowest rank, trying to increase his own advantage in the situation. According to Lenski, this social ambiguity experienced by the inconsistent individual results in tension, which in turn leads to behavior which is significantly different from that of an inconsistent person. Although the theory of status inconsistency as outlined by Lenski seems entirely plausible and is supported by a number of empirical studies, it has also been criticized by many sociologiSts. The primary reason for this criticism is the difficulty other researchers have encountered in trying to substantiate Lenski's results. A vast number of studies have attempted to test the theory of status inconsistency, but as yet the value of the concept has not been proven. There is no conclusive evidence supporting the theory. Status inconsistency has been measured in a number of different ways and linked to a host of dependent variables, but no single pattern of findings has consistently emerged. For example, one of the major dependent variables that is often linked with status inconsistency is political liberalism (Lenski, 1954). From the very beginning, however, Lenski's findings that inconsistents are more liberal have been difficult to replicate (Kenkel, 1956). There are methodological differences between the first Lenski article and Kenkel's research which may account for their differing results (Lenski, 1956a), but subsequent research fails to settle the issue. A number of researchers claim that inconsistents are more liberal (Geschwender, 1970; Lenski, 1956b, 1967; Segal 8 Knoke, 1968; Segal, 1969), while others find no such relationship (Brandmeyer, 1965; Broom 8 Jones, 1970; Kelly 8 Chambliss, 1966; Runciman 8 Bagley, 1969). Some sociologists who fail to substantiate Lenski's initial findings suggest that social class is a better predictor of various dependent variables than incon- sistency, but even this issue has been disputed (Bauman, 1968; Fauman, 1968). In his two early reports, Lenski also found that certain patterns or types of inconsistency are more closely associated with liberalism than others. For instance, persons having a low ethnic status in conjunction with high rankings on one of the other scales are the most liberal sub- groups in his sample. As a result, research into the significance of various patterns of inconsistency has become one of the major thrusts of inconsistency studies. Some researchers have found that different patterns of inconsistency are indeed, important (Bauman, 1968; Broom 8 Jones, 1970; Fauman, 1968; Geschwender, 1968; Jackson, 1962; Lenski, 1956b, 1967). Others, however, find that status indicators such as occupation (Hyman, 1967; Segal, Segal, 8 Knoke, 1970) or ethnicity (Schweiker, 1968; Treiman, 1966) account for much of the variance in a person's behavior and that there is no inconsistency effect over and above that effect exerted by the status indicators themselves. Perhaps the most profitable outgrowth of this emphasis on specifying which patterns of inconsistency are most important is the research dealing with discrepancies between achieved and ascribed status characteristics. Many researchers recognize this form of discrepancy as the most promising emphasis in inconsistency re— search. Numerous findings lend credence to this belief (Broom 8 Jones, 1970; Chambliss 8 Steele, 1966; Geschwender, 1967; Jackson, 1962; Jackson 8 Burke, 1965; Leavy, 1970; Lenski, 1967; Schweiker, 1968; Segal, 1969). However, there are no conclusive studies testi- fying to the fact that whenever achieved and ascribed statuses are incongruent, significantly different behavior results. Discrepancies which occur between achieved and ascribed status variables as a result of mobility have also been investigated fairly widely, but these studies fail to produce any conclusive findings (Bloombaum, 196k; Geschwender, 1967; Jackson, 1962; Leavy, 1968; Segal 8 Knoke, 1968; Simpson, 1968). Many sociologists accept the basic theoretical underpinnings of status inconsistency theory such as expectancy congruence (Exline 8 Ziller, 1959; Geschwender, 1967; Jackson, 1962; Landecker, 1970; Leavy, 1970; Sampson, 1963; Treiman, 1966) and increased stress on the part of the inconsistents (Adams, 1953; Geschwender, 1968; Hyman, 1967; Jackson 8 Burke, 1965; Lenski, 1956b, 1967; Runciman 8 Bagley, 1969; Rush, 1967; Schweiker, 1968; Segal, 1969). However, these researchers are still unable to Specify exactly how status incon- sistency operates. Various attempts have been made to link status inconsistency to other existing theories but none of the attempts have proven to be of much value (Geschwender, 1962, 1967; Runciman 8 Bagley, 1969; Sampson, 1963; Schweiker, 1968; Segal, Segal, 8 Knoke, 1970). Other miscellaneous dependent variables that have been used in inconsistency research are the visibility of the inconsistency (Hyman, 1967; Schweiker, 1968; Segal, 1969), suicide (Chambliss 8 Steele, 1966; Gibbs 8 Martin, 1958), prejudice (Penman, 1968; Geschwender, 1967; Treiman, 1966), right-wing extremism (Brandmeyer, 1965; Ringer 8 Sills, 1952; Rush, 1967), and social isolation and individual unrest (Bauman, 1968; Geschwender, 1967, 1968; Hyman, 1967). Hence at the present time, the literature on status inconsistency can be viewed as a wide array of related research reports, mainly in agreement concerning the basic tenets of the theory, but patently unable to specify exactly how inconsistency affects behavior. One of the basic problems with this literature is the reluctance of researchers to pattern their studies on previously published reports. In an effort to fhrther the generalizability of the theory, researchers have failed to adequately replicate and substantiate the basic findings on which the theory is based. Most of the research which has been done in this field is sufficiently dissimilar as to -preclude.comparability. As a result, we do not know why those studies which have supported the theory have obtained such results. There is no set of factors common to these reports which leads us to a specification of the theory of status inconsistency. Likewise, there is not enough common ground among those studies which have failed to support the theory to enable us to determine the types of situations in which status inconsistency does not operate. In effect, we do not know why or how status inconsistency effects behavior. Statement of the Problem The purpose of this research endeavor is to test the theory of status inconsistency. We want to determine if status inconsistency is a useful concept for sociologists. Previous research is inconclusive; some researchers claim that status inconsistency does effect behavior, while others fail to discover any inconsistency effects. The question to be answered is which group is correct. Does the theory warrant the further attention of sociologists, or can it be proven empirically that the theory is incorrect? Status inconsistency has been linked to a wide variety of depen- dent variables. It would be impossible in a thesis like this to consider all of them. Therefore, this research endeavor is restricted to an investigation of the relationship between status inconsistency and a preference for the Democratic Party. We have chosen this dependent variable because it is the one most often employed in inconsistency studies. Many researchers have used this dependent variable but the 1O findings are contradictory. Status inconsistency theory states that persons with incon— sistent ranks may opt for programs of social change in an attempt to restore balance to their rankings. If the social order is altered, evaluations on the status hierarchies may change or other statuses may become relevant. Inconsistents support social change, therefore, with the hope that the changes will remove their inconsistencies. In this study a preference for the Democratic party is interpreted as support for programs of social change. In a similar manner, Lenski has interpreted a vote for the Democratic party as a liberal political response. Given the indeterminacy of the theory as outlined previously, we are not able to specify which particular response will occur. Thus, theoretically we may consider political liberalism, a Democratic vote, preference for the Democratic party, and support for programs of social change as equivalent responses to a state of status inconsistency. Whether or not these responses are actually equivalent is a question which will be left to other researchers to answer. Past practice has been to assume that they are equivalent, however. It should be noted again that this research endeavor is a test of the theory of status inconsistency. We are not primarily concerned with why persons claim a preference for the Democratic party. A group of entirely different variables would be appropriate if that were the purpose of this study. Political scientists and political sociologists have found that factors such as parents' political affiliation, past voting performance, the voting behavior 11 of friends, religious preference, and SBS level offer the best prediction of party preference (Berelson, Lazarsfeld,8 McPhee, 195M; Coser, 1966)., This study then, seeks to determine whether or not persons who are inconsistent on the education, occupation, income, and religion hierarchies are significantly more likely to voice a preference for the Democratic party than consistent persons. Although many researchers have included a measure of ethnicity in their studies, this variable is absent in our research because it is not relevant for our sample. We include religion, an ascribed status variable, in the analysis in order to test the effect of achieved-ascribed status inconsistencies. We have noted above that in order to substantiate the theory of status inconsistency, replications of previous studies are needed. Therefore, this study is a replication of one of the most recent research endeavors in this field, the study conducted by Broom 8 Jones (1970). In Finifter's terms, our study is a "virtual replication" of the work by Broom 8 Jones (Pinifter, 1972). The sample employed is not the same but the measurement and analysis are practically identical, the only difference being a slight variation in the computation of the inconsistency score. Broom 8 Jones ranked each respondent on the status variables in question and then took the standard deviation among these rankings as an indication of status inconsistency. In this study we first convert each respondent's score on the status variable into a standard score and then take the standard deviation among the standard scores as an indication of status inconsistency. The two procedures are essentially identical but we feel that our method is 12 an improvement on Broom 8 Jones' since it controls for the fact that the status indicators do not have the same number of categories. As in the Broom 8 Jones article one of the most crucial questions in determining the success of our research effort is how to adequately differentiate between the effect of the inconsistency and the direct effects of the status variables themselves. It must be shown, if the hypotheses are to be supported, that a regression equation including terms for status inconsistency explains a significantly greater amount of the total variance in political behavior than one using terms only for the status variables. Since the primary purpose of the study is to test the theory of status inconsistency, not to explain Democratic party preference, the critical question is not how much of the total variance in party preference is accounted for, but rather, does the inconsistency term significantly reduce the amount of residual variance over that ex- plained by the status indicators. Toward this end, regression equa- tions with dummy variables are used in this study. This method has proven most effective in distinguishing between the main effects of the statuses and the inconsistency effect (Broom 8 Jones, 1970; Jackson 8 Burke, 1965; Treiman, 1966). To summarize, the main purpose of this study is to test the theory of status inconsistency. The analysis is restricted to an investigation of the relationship between status inconsistency and Democratic party preference. This research effort is a virtual replication of the study by Broom 8 Jones and the critical question to be answered is whether or not the inclusion of the status 13 inconsistency term significantly reduces the amount of unexplained variance in the dependent variable, Over and above that accounted for by the status indicators. The research reports which deal specifically with the relation— ship between status inconsistency and political behavior are detailed in the second chapter of this thesis. The third section explains the hypotheses, data, and methodology upon which the research is based. The fourth chapter contains the results of the analysis of the data and the fifth section contains the discussion or summary. We now turn to a review of the selected relevant literature. CHAPTER II REVIEW OF THE SELECTED RELEVANT LITERATURE In this chapter a detailed account of those studies which deal specifically with the relationship between status inconsistency and political party preference is presented. The chapter is divided into two sections. Section I deals with the theoretical formulation of the relationship between inconsistency and Democratic party preference. Section II is concerned with the changes that have occurred in the measurement and analysis of status inconsistency as it relates to political behavior. Section I: Status Inconsistency and Political Party Preference The theoretical formulation of the relationship between inconsistency and party preference, of course, closely reflects the development of the general theory of status inconsistency as described in the preceeding chapter. Initial research (Lenski, 195k; Kenkel, 1956) was concerned with determining the effect of a general measure of status inconsistency on political behavior. The theory ‘states that inconsistent individuals, those whose ranks are not at a comparable level, will be more politically liberal than consistent subjects, those persons having nearly the same rankings on the status variables. There are a number of studies which test this theoretical relationship. Lenski IAnalyzing voting behavior in three different elections, Lenski found that in two of them a significantly (.05) greater proportion of inconsistents than consistents supported the Democratic party 19 15 (Lenski, 1959). He interpreted this result as confirmation of his hypothesis. This finding encouraged him to further test this relationship on a 50% subsample of his respondents using a different set of variables as indicators of liberalism. Lenski asked his subjects their views on government-sponsored health insurance programs, price controls, and general extension of governmental powers. In all three cases, inconsistents reported more liberal views than consistents and on two of the questions the differences were significant (.02). Further analysis revealed that the differences between the two crystallization categories were not a function of certain patterns or combinations of the status variables found more predominantly in one of the categories. In other words, liberal political tendencies were associated with status inconsistency regardless of any specific configuration of the status variables. Kenkel Kenkel based his study of status inconsistency on the same general proposition, i.e. that inconsistencies among any of the status variables would produce liberal political behavior. In this respect, these two reports are characterized by the same theoretical assumptions. However, Kenkel's work is in no way a replication of Lenski's paper, as there are numerous methodological differences between the two studies. (These methodological differences will be outlined in the second part of this chapter). On the basis of Lenski's earlier work, Kenkel did not expect these variations to influence his findings. Given the general character of Lenski's theory, Kenkel believed a change in the operationalization of the variables would not effect the results. 16 However, Kenkel's data showedxthat consistents and inconsistents did not differ with regard to the attitudes measured. This result cast doubt upon the universal applicability of the theory proposed by Lenski. In the original study Lenski found that certain types of inconsistencies were more closely related to liberalism than others. For instance, he found that individuals with low ethnic status and high status on any of the other variables were the most liberal group in his sample. Lenski reported this finding in his initial work but felt that his most important result was that all types of inconsistencies resulted in political liberalism. He thus emphasized the latter in his write-up. However, Kenkel's inability to substantiate this general hypothesis of status inconsistency changed Lenski's viewpoint. In his response to Kenkel's paper, Lenski called for a new emphasis in status inconsistency research (Lenski, 1956a). He stated that researchers should study the consequences of particular patterns or types of inconsistency and that not all types of dis- crepancies had equally potent effects. The original theoretical assumption, which stated that all types of inconsistencies caused liberalism, was therefore replaced by one which claims that only inconsistencies among certain types of statuses will result in liberal tendencies. The trick, of course, is to find out which types of inconsistencies are salient and result in political liberalism. Brandmeyer One of the first studies to specifically look at different 17 patterns of inconsistency is the research by Brandmeyer (1965). This paper is much more similar to Lenski's first article than the Kenkel study was. The comparability of the two reports facilitates a comparison of the results. In his analysis, Brandmeyer first looked for a relationship between each individual's overall inconsistency score and political liberalism. The results show no association between the degree of status inconsistency and political liberalism. This finding concurs with Kenkel's results and casts further doubt on the usefulness of Lensk's general definition of the concept. This is especially true when we consider that Brandmeyer's work is method- ologically very similar to Lenski's. The evidence is mounting, then, that not all types of inconsistency result in political liberalism. Brandmeyer conducted further analysis to determine the effects of certain patterns or types of inconsistency. None of the specific patterns of inconsistency was related to political liberalism. Brandmeyer could not substantiate Lenski's finding that low ethnic, high occupation or education respondents were the most liberal group in the sample. Brandmeyer was forced to conclude that status incon- sistency has no effect on political liberalism. According to his data, differences in political attitudes are the result of differential status rankings and are not caused by any inconsistencies among these status ranks. "Social status remains a far more powerful predictor of political party preference and political attitudes than is the nonvertical dimension of status inconsistency (Brandmeyer, 1965, p. 252)." Needless to say, Brandmeyer's study illuminates the need for further revisions in the theory of status inconsistency, or at the very least for more studies to substantiate his results. 18 Kelly and Chambliss Shortly after the Brandmeyer article, Kelly 8 Chambliss (1966) cc- authored an article which contained a new theoretical thrust for status inconsistency. One of their major objections to the previous literature is the conception of liberalism as a unidimensional phenomenon. Kelly 8 Chambliss do not think that an individual who is liberal on one item should be assumed to be liberal on a whole host of items. It is entirely possible for a person to be liberal on some items, yet conservative on others. Following Lipset (1960), they suggest that the lower classes are more liberal on economic items while the upper classes are more liberal on noneconomic items like civil rights, civil liberties, and internationalism. Their research tests this relation- ship as well as the more traditional hypotheses found in the preceeding studies of inconsistency and liberalism. In this respect, they proposed a new theoretical direction for the field, while still providing a type of replication by testing the traditional hypotheses that had been used in the past. Such a procedure insures comparability with past research as the hypotheses used by other researchers are tested in addition to those not previously formulated. Another innovation offered by Kelly 8 Chambliss is the collection of data on perceived status inconsistency. They claim that it is important to get the respondent's opinion of where he stands in comparison to the rest of the population on the status variables in question. Therefore, their subjects ranked themselves on the status indicators. To test the multidimensionality of liberalism, Kelly 8 Chambliss determine political attitudes from responses to questions 19 comprising four distinct scales. The four dimensions of liberalism are: welfare, civil liberties, internationalism; and civil rights. Kelly 8 Chambliss' results do not support the hypothesis that persons who are status inconsistent are more liberal than individuals with consistent status configurations. Even when the analysis is restricted solely to the economic items on the "welfare" scale (as Lenski had relied on economic items), significant differences do not appear. The authors further state that liberalism is not a unidi- mensional concept, and that liberalism on one dimension, while related to liberalism on other dimensions, cannot be accurately predicted from it. The results also indicate that perceived inconsistency is not a better indicator of attitudes than actual inconsistency. Kelly 8 Chambliss conclude that social class membership and ethnicity are more important determinants of political attitudes than the degree of status inconsistency. This study, in addition to the Brandmeyer and Kenkel reports, indicates that a general definition of status inconsistency does not suffice. The theory and model must be more completely specified in order to explain the phenomenon of political liberalism. Further Research by Lenski Although other researchers found no relationship between liberalism and any specific patterns of inconsistency, the basic assumption that various types of inconsistency are differentially related to liberalism still appealed to Lenski. Using this assumption, he conducted further research which sought to clarify the theory of status inconsistency 20 (Lenski, 1967). Lenski performed a re—analySis of the results of twenty-five separate surveys from Australia, Great Britain, Canada, and the United States discussed by Alford (1963). Lenski investi- gates the effect upon voting behavior of inconsistencies between occupational class and socio-religious group. His test of the hypothesis that inconsistencies between these two statuses lead to increased liberalism was supported in 21 of 25 cases in the four countries. Reiterating his belief that all types of inconsistencies do not have equally potent effects, Lenski notes that the analysis in this study includes one achieved status variable (occupational class), and one ascribed status variable (socio-religious group). He claims that this type of inconsistency, between achieved and ascribed . variables, is the type most likely to produce effects on political behavior. His conclusion is based on these findings as well as some earlier work done by Jackson (1962). This emphasis on inconsistencies between achieved and ascribed variables is a natural development of Lenski's and Brandmeyer's contention that not all types of inconsistencies produce the same effects. Whether or not this is actually how status inconsistency operates, however, can only be determined by further research. Segal and Knoke The relative importance of inconsistencies between achieved and ascribed status variables is partly the focus of a study by Segal 8 Knoke (1968). Although their major concern is how people vote as a consequence of their mobility from the working class to the middle class (or vice versa), Segal 8 Knoke also test the hypothesis that 21 inconsistencies between achieved and ascribed statuses have stronger effects than inconsistencies between only achieved or only ascribed variables. Segal 8 Knoke's results, however, fail to confirm this hypothesis. They find that the effects of achieved—ascribed inconsistencies are greater than those resulting from inconsistencies between achieved statuses, but the highest mean inconsistency effects are found between ascribed statuses. Segal 8 Knoke conclude that persons who are inconsistent as a result of social mobility will tend to support the Democratic party. They believe that in this case the important variable for status inconsistency is not the achieved-ascribed dimension as - Lenski suggested, but rather, the effects of social mobility. Segal 8 Knoke's study, while it seems to refute the claim that achieved-ascribed inconsistencies are most important, sustains the belief that status inconsistency is related to political liberalism. This study provides a new direction for consistency theory and research: determining the effects of inconsistencies which result from social mobility. Broom and Jones The achieved-ascribed emphasis in inconsistency theory did not completely die out with Segal 8 Knoke's study, however. Three achieved statuses (occupation, education, and income) and one ascribed status (religion) are the variables focused on in a study by Broom 8 Jones (1970). As in the previous studies, the dependent variable is voting behavior. Broom 8 Jones test several different hypotheses in their analysis. The first hypothesis states that political liberalism is 22 inversely related to achieved socioecomomic status. This means that those persons with the lowest scores on the occupation, education, and income hierarchies are the most politically liberal while those individuals with high rankings on these variables are the least politically liberal. The second hypothesis states that after controlling for the effects of the status variables, those persons with inconsistent statuses are more liberal that consistent individuals. The second hypothesis tests the general definition of status inconsistency. This represents a valuable replication by Broom 8 Jones which lends further evidence to the question of the adequacy of a general definition of status inconsistency. The results confirm the first hypothesis: the achieved status variables are inversely related to liberal voting behavior. The second hypothesis, however, is not confirmed. Adding the status inconsistency score as another variable does not contribute to the understanding of voting behavior. Thus further evidence that an overall status inconsistency score does not suffice as a predictor of political liberalism is provided. Broom 8 Jones conducted further analysis to determine if some specific patterns of inconsistency produce effects which are masked by less important types of discrepancies. They hypothesize that after controlling for the effects of the achieved statuses, individuals who rank high in educational investment but low on rewards are politically liberal. Likewise, they also suggest that after controlling for status, persons with high rewards but low educational investment are politically conservative. Neither of these hypotheses is supported by the data. These hypotheses are derived, theoretically, from Homans' theory 23 of distributive justice (Homans, 1961). The theory of status inconsistency can be elaborated through the use of the concepts of Homans' theory. Various status ranks, such as education, are investments for the individual. Other ranks, income and occupation for instance, are rewards. According to the principle of distributive justice, a person's rewards should be proportional to his investments. When rewards are less than investments, the response is likely to be hostility or anger. When rewards are greater than investments, the individual will probably feel guilty. In either case, the situation is equivalent to a state of status inconsistency. Broom 8 Jones are hypothesizing then, that when investments (education) and rewards (occupation or income) are out of line and the principle of distributive justice is violated, individuals respond to this inconsistency by voting for a liberal political party. However, Broom 8 Jones' results do not substantiate this proposed relationship. Three more hypotheses are tested in this research by Broom 8 Jones. The fifth hypothesis states that after controlling for the achieved status variables, Catholics are more liberal than non—Catholics. This proposition is confirmed by the data and the inclusion of this variable improves the prediction of voting behavior. This represents an additive effect, however, and not an inconsistency effect. The sixth hypothesis suggests that after controlling for the effects of the achieved statuses and religion, Catholics (low ascribed status) of high achieved status are more liberal than other Catholics. This is a test of the inconsistency arising from the differences between ascribed and achieved status variables. This hypothesis is confirmed, 24 as an interaction term to represent Catholics with high achieved status increases the explanatory power beyond that of the purely additive model. The final hypothesis suggests that after taking into account the additive effects of the achieved statuses and of religion, Protestants (high ascribed status) with low achieved status are more liberal than other Protestants. This proposition is also a test of the achieved-ascribed type of inconsistency. The data do not support this hypothesis; there appears to be no increase in liberalism among inconsistent Protestants. Thus Broom 8 Jones find that the results of achieved—ascribed inconsistencies are important in one instance, but not in the other. They cannot demonstrate that whenever these two types of statuses are out of line, political liberalism is increased. However, they do show that a general definition of status inconsistency does not suffice and that further study should be carried out on the effects of various patterns of inconsistencies. The discussion of this study by Broom 8 Jones concludes our description of the theoretical developments relating status inconsistency to political liberalism. The theory has developed from a most general form to increasingly more specific conceptualizations. However, it is readily apparent that theorists have not yet hit upon a theory of status inconsistency which can adequately explain the political behavior of people in the United States. Outlining the past developments in status inconsistency theory, as we have done, points out the need fer replication in this area of study. If the theory is to be further 25 enhanced, new ideas must be tested and retested through replications. Section II: The Measurement and Analysis of Status Inconsistency This chapter concludes with a discussion of the various changes and improvements in methodology which have occurred along with the theoretical developments of the concept of status inconsistency. We begin by examining the procedures used by Lenski in his first report on status inconsistency. Lenski Lenski's initial study (195k) measured each respondent on four vertical hierarchies: occupation, education, income, and ethnicity. The status crystallization score is created by first constructing a standard score on each of the status variables, and then taking the square root of the sum of the squared deviations from the mean of the four hierarchy scores for each person and subrtacting the result from one hundred. In order to control for the effects of the status variables themselves, Lenski dropped a number of subjects from the analysis so that the mean scores of the inconsistent individuals are higher on each of the four variables than those of the highly crystallized subjects. Disregarding inconsistency effects, it seems likely that the group with the lower mean scores on the status variables (the consistents) will be more liberal in political party preference than the group with higher average scores (the inconsistents). Although this control reduces the margin of difference, Lenski still finds inconsistents (higher mean scores) to be significantly more liberal than consistents in two of the three elections analyzed. Lenski's results are substantially the same when he tests the relationship on a 50% subsample, operationalizing 26 liberalism this time in terms of the respondents' views on government- sponsored health insurance, price controls, and a general extension of governmental powers. He finds that inconsistents are more likely to be politically liberal than consistents on all three of these issues. Kenkel As noted above, Kenkel's attempt to substantiate these findings has the same theoretical base as did the Lenski article. However, the methods employed by Kenkel are substantially different than those utilized by Lenski. Kenkel measures his subjects on the following four hierarchies: occupation, education, rental value of dwelling, and dwelling area prestige. He constructs the inconsistency score in the same manner as Lenski and divides his sample of 300 respondents into two groups. The 150 with the highest crystallization scores are designated the most inconsistent while the other 150 are the consistent group. This method of dividing the sample into equal halves is a deviation from Lenski's original method and may account for the different results. Lenski divided his sample at a "natural breaking point" which contrasts the most crystallized three-quarters of his subjects with the least consistent one-quarter. There is no such natural break in Kenkel's sample. Political liberalism is operationalized in Kenkel's study on the basis of responses to questions concerning the Taft-Hartley law, foreign trade, government care for the needy, strikes during wartime, price controls, government ownership of aircraft factories, and strictness of labor laws. Using these procedures and variables, 27 Kenkel finds no relationship between status inconsistency and political- liberalism. Kenkel concludes that his results are in basic disagree- ment with Lenski's. However, we cannot actually say that Kenkel disproved any of Lenski's results. The two studies are not similar enough to warrant such a judgement. We can say that inconsistency as operationalized by Kenkel does not seem to be related to political liberalism, but this does not reflect upon Lenski's formulation. Kenkel has shown us that at least one definition of inconsistency does not work, but he has not refuted Lenski's claim.. There are any number of explanations which may account for the contradictory results, e.g. the different cutting points, the failure to include ethnicity, the different status variables, the different questions on liberalism, the failure to control for the effects of the status variables (Lenski, 1956a). Thus, although these two studies have the same underlying assumptions, their results are not necessarily contradictory. To the extent that they both claim to test the effects of status inconsistency, the general applicability of the theory may be questioned. But we cannot say that Kenkel's study refutes Lenski's findings. We have found an instance in which the theory does not hold, but we have garnered no further evidence attesting to the appropriateness of Lenski's formulation. Evidence confirming his findings can only be collected through replication. Brandmeyer Brandmeyer's study (1965), however, is a replication of the Lenski article. Brandmeyer measured his subjects on the occupation, education, and ethnicity hierarchies, disregarding the income 28 hierarchy on Lenski's advice. Brandmeyer's methods of ranking the respondents and constructing inconsistency scores are the same as Lenski's. The sample is divided so that the lower one-fourth (inconsistents) is contrasted with the upper three—fourths of the sample (consistents). Political liberalism is defined in terms of party preference and responses to questions on jobs for the unemployed, old age insurance, medical care, guaranteed college education, minumum annual income, and government housing. Brandmeyer found no relationship between inconsistency and liberalism. Even when he compared the most consistent one-fourth of his respondents with the least consistent one-fourth, disregarding the middle half of his sample, he failed to get significant results. Brandmeyer also attempted to reconcile his results with Lenski's by investigating the effects of certain patterns or types of inconsistencies. None of the specific patterns are associated with political liberalism. Because this article by Brandmeyer is a replication of Lenski's earlier work, the differing results raise crucial doubts about the value of the theory of status inconsistency. Since the methodology is substantially the~same in both studies, the results seem to indicate that status inconsistency does not always operate toward political liberalism as Lenski suggests. Kelly and Chambliss The Kelly 8 Chambliss study (1966) is in some respects a replication, while in other areas they formulate new methods of their own. Their subjects are measured on the income, occupation, and education hierarchies. Standard scores for each individual are 29 constructed on the status variables following Lenski, but the measure of status inconsistency is computed in a slightly different fashion. Political liberalism is determined on four separate scales instead of one and they divide the sample into the most consistent one-half, the middle consistent one-fourth, and the least consistent one—fourth. This type of break enables them to compare the most consistent half of their sample with either the least consistent one-half or one—fourth of the respondents. The results of this study do not support the hypothesis that persons who are status inconsistent are more liberal than individuals with consistent status rankings. The results are the same whether the least consistent one-half or one-fourth of the sample is used. Kelly 8 Chambliss do not use a measure of ethnicity, but claim that if that is the reason for their failure to support the hypothesis, then ethnic status alone would be the determinant of attitude formation and not status inconsistency. Further Research by Lenski Earlier in this chapter we saw that the theory of status inconsistency shifted its emphasis from a general definition to a more specific concern with the consequences of particular patterns of inconsistency. Lenski is one of the main advocates of this shift in emphasis and has conducted research in this area. In his study of inconsistencies between ascribed and achieved variables, Lenski uses a type of analysis which is totally different than that used in the other studies of inconsistency and liberalism (Lenski, 1967) He tests for inconsistency by using Alford's 2 x 2 30 tables in which subjects are divided into middle and working class, Protestant and Catholic. The Protestant-middle class and Catholic— working class cells are defined as consistent status configurations, while the Protestant-working class and Catholic-middle class cells contain inconsistent individuals. A comparison of the sum of the percentage of subjects voting for a liberal party in the two consistent cells is made with the sum of therresponding percentages in the two inconsistent cells. If the sums of the two pairs of cells are the same, then no inconsistency effect is present. But if the sum of the percentages of liberal voters in the inconsistent cells is greater than the corresponding percentages in the consistent cells, the hypothesis that inconsistency increases liberalism is supported. This was the result in 21 of the 25 tests of the hypothesis in Lenski's study. This study represents a new emphasis in both theory and methodology. Theoretically, it is a move in the right direction. Methodologically, it is difficult to evaluate. The analysis does not seem to be very rigorous, but Lenski claims that the test of each hypothesis is an independent event and the probability of the hypothesis being confirmed in 21 out of 25 tests is .0005. The adequacy of these new developments in theory and methodology will have to be determined by further replications. Sggal and Knoke Segal 8 Knoke (1968) use this type of analysis to test Lenski's claim that inconsistencies between ascribed and achieved statuses result in political liberalism. As reported previously, they failed to 31 substantiate this hypothesis. They found that people who are inconsistent vote Democratic as a result of social mobility and not because of achieved-ascribed inconsistencies. Thus, employing the same method of analyzing 2 x 2 tables as Lenski had used, Segal 8 Knoke obtain results which contradict Lenski's reformulation of the theory. It is obvious that further studies are needed to assess the importance of discrepancies between achieved and ascribed statuses in inconsistency research. Broom and Jones The final important development in the methodology employed in status inconsistency research is the use of regression analysis with dummy variables. The previous studies we have reviewed all have one common fault. They fail to adequately differentiate between the main effects of the status variables and the effects of the inconsistency. As a consequence, it is not possible to determine if the results are due to status inconsistency or if they are a result of the status variables. Using regression analysis with dummy variables is not new to inconsistency research. However, Broom 8 Jones (1970) are the first to use it in analyzing how inconsistency and liberalism are related. This method solves the problem outlined above by letting the status variables operate first, before the effect of the inconsistency is determined. The inconsistency score is subsequently introduced into the regression equation and hypotheses are tested by looking at the increase in the amount of variance explained, over and above that attributable to the status variables. 32 Broom 8 Jones test the hypothesis that persons with inconsistent statuses, in general, are more liberal than people with consistent statuses. This hypothesis is not confirmed. The equation including an inconsistency term, broadly defined, does not explain a significantly greater amount of the variance than the equation containing terms only for the status variables. They do, however, find support for the presence of inconsistency effects among Catholics (low ascribed status) with high achieved statuses. The results of the other hypotheses tested in this study are reported earlier in this chapter. What is important to note is that this article by Broom 8 Jones shows us that this type of statistical analysis is probably the most useful method available for determining the effects of status inconsistency. What is needed now are more studies using this type of analysis so that their results may be compared with those obtained by Broom 8 Jones. To summarize, the theory which links status inconsistency with political liberalism has gradually evolved from a concern with general status discrepancies to an emphasis on particular types of imbalances. ResearCh has been conducted on the effects of achieved—ascribed inconsistencies and investment-reward discrepancies, but the results are inconclusive. Methodologically, the statuses which have been employed and the method of creating the inconsistency scores has been substantially the same in the studies we have reviewed. However, the procedures used to analyzetjugdata have varied among the studies. This lack of replication makes comparison of the results problematical. In addition, a majority of the studies are plagued by the inability to distinguish between the 33 main effects of the status variables and the inconsistency effect. This dilemma can be solved by using multiple regression analysis. This procedure allows us to determine, specifically, the effect of the A status inconsistency term. Taking into account these previous developments, the present research effort tests the effects of a general inconsistency score, as well as the effects of achieved-ascribed and investment-reward inconsistencies on Democratic party preference. Regression equations, with dummy variables, are employed to test our hypotheses. CHAPTER III HYPOTHESES, DATA, AND METHODS In this chapter we formulate the hypotheses to be tested, describe the data upon which the study is based, and explain the procedures we use throughout the research effort. The chapter concludes with a dis- cussion of a preliminary analysis which investigates the relationship between party preference and a number of background variables. Hypotheses As noted above, this research endeavor is a virtual replication of the Broom 8 Jones study (1970). We perfOrm a replication because we are interested in determining the reproducibility of prior results in incon- sistency studies. According to Finifter, the goal of a virtual repli— cation is "...to repeat an original study not identically, but rather 'for all practical purposes' to see if its results 'hold up' against chance and artifact (Finifter, 1972, p. 121)." Therefore, the hypotheses listed below are taken directly from the Broom 8 Jones article (1970). As outlined in the preceeding chapter, Broom 8 Jones' results indicate that a general definition of status inconsistency does not suffice and that it is necessary to investigate the effects of specific patterns of inconsistency. The hypotheses they employ, and that we also test, are described below. Hypothesis #1: Political liberalism is inversely ‘ related to achieved socioeconomic status. As the sub'ects' rankin s on the occu ation, education, and income 3 8 P hierarchies rise, the liberalism scores are expected to decline. The 35 regression equation used to test this hypothesis is the simple additive model which predicts the value of the dependent vari- able from only the scores on the status variables. Hypothesis #2: After controlling for the additive effects of the achieved statuses on political liber- alism, status inconsistent individuals are more liberal than status consistent persons. Broom 8 Jones' data fail to confirm this hypothesis. The amount of the residual variance which is explained by the incon- sistency term is not significant. This hypothesis represents a test of the original Lenski formulation (195”) which states that any inconsistency is related to political liberalism. As noted in the review of the literature, other researchers have also failed to find empirical support for this type of hypothesis. The regression equation used to test the hypothesis includes all the terms present in the first equation, plus an in— teraction term to represent the respondent's degree of status inconsistency. Hypothesis #3: After controlling for the additive effects of the achieved statuses, respondents with a high education investment but low rewards will tend to be politically liberal. Hypothesis #H: After controlling for the additive effects of the achieved statuses, people with high rewards but low educational investment will tend to to be politically conservative. These two hypotheses are an attempt to discover if these specific types of inconsistency are related to political party preference. Previous research has indicated that a general 36 definition of inconsistency may not suffice and that the effects of specific types of inconsistency should be investigated. Both hypotheses are tested through the same regression equation. It con- tains all the terms of the additive model, plus one term represent— ing high investment-low reward inconsistency, and a different term for low investment-high reward individuals. Since in this case we are adding two terms to the additive model, we will look not only at the total increase in the amount of variance explained, but also determine the significance of each of the two coefficients individ- ually. Hypothesis #5: After controlling for the additive effects of the achieved socioeconomic statuses, Catholics are more liberal than non-Catholics. This hypothesis is confirmed by Broom 8 Jones. However, this finding does not say anything about status inconsistency. It shows that an equation containing terms for occupation, education, income, and religion explains a significantly greater amount of the variance in voting behavior than an equation including terms for only the first three variables. The equation used to test this hypothesis is, of course, an additive model and differs from equation number one only by including the independent effects of religion upon political liberalism. Hypothesis #6: After controlling for the additive effects of the achieved statuses and of religion, Catholics with high achieved status are more liberal than other Catholics. Hypothesis #7: After controlling for the additive effects of the achieved statuses and of religion, Protestants with low achieved status are more liberal than other Protestants. 37 Like the second, third, and fourth hypotheses, these propositions represent an inconsistency effect. Since Catholicism is a low ascribed status, a Catholic with a high achieved status will presumably experi- ence status inconsistency. Historically, Protestants have been consider— ed high on the ranking of religious preference. Therefore, when a Protestant holds a low socioeconomic status an inconsistency between his achieved and ascribed status levels is present. According to the theory outlined previously, the result of this inconsistency between the achieved and ascribed statuses should be an increase in political libera- lism. Hypothesis #6 is supported by Broom 8 Jones' data. Catholics with high achieved statuses are more liberal than consistent Catholics. The seventh hypothesis, however, is not confirmed by Broom 8 Jones. Protes- tants with low achieved statuses are not more likely to vote liberally than consistent Protestants. The equation used to test these hypotheses contains terms for occupation, education, income, religion, inconsistent Catholics, and inconsistent Protestants. To determine the combined effects of these two types of inconsistencies we compare the R2 from this equation to the R2 from an equation including only the four status variables. The significance of the coefficient associated with each type of inconsistency is also computed so that we may determine their effects individually as well as collectively. The results of the tests of these hypotheses are reported in the following chapter. We now proceed to a description of the data set employed in this research effort. Data In this research, a secondary analysis of a body of date originally 38 collected by the Gallup organization is performed. The survey was conducted in the summer of 1970 for the Department of Sociology at Michigan State University. The instrument used to gather the data was constructed by J. Allan Beegle and others in the Sociology Department and administered by the Gallup organization. The subjects represent a random sample of the rural population of Michigan. Three-hundred and forty-three respondents were chosen from thiry—four randomly selected sampling points in the state. Each sampling point represents a rural township. Census Bureau criteria are used to determine which districts qualify as rural townships. Approximately ten sixty-minute interviews were conducted in each of the thiry-four districts sampled. Because this is a sample of the rural population of Michigan, it contains a larger proportion of older persons than one would find in a sample of the entire population of the state (See Table 1). There is also a greater number of females in the sample than males. Table 2 and Table 3 give the distribution of the sample by sex and social class. There is no reason to suspect that the degree of inconsistency should be any different for persons in rural areas than it is for urban residents. In fact, Goertzel (1970) presents evidence which supports the contention that the degree of inconsistency in rural areas is generally equal to that found in urban areas. This is not surprising when one considers the rapidity with which differences between urban life and rural life are disappearing. Koebernick 8 Beegle (unpublished manuscript) also note that while this is a sample 39 TABLE 1 DISTRIBUTION OF SAMPLE BY AGE AGE N % Under 25 35 ' 10.2 25-29 27 7.9 30-34 30 8.7 35-39' 33 9.6 uo-uu #9 12.8 “5-99 ' 23 6.7 50-54 33 9.6 55-59 ' 23 6.7 60+ 9” 27.“ No Answer 1 0.3 Total 393 99.9 TABLE 2 DISTRIBUTION OF SAMPLE BY SEX SEX N % Female 181 52.8 Male 162 “7.2 90 TABLE 3 DISTRIBUTION OF SAMPLE BY SOCIAL CLASS SOCIAL CLASS N % Upper l 0.3 Upper-Middle ’ 19 5.5 Middle 176 51.3 Working 90 26.2 Lower—Middle an 12.8 Lower 10 2.9 No Answer 3 , 0.9 141 of rural residents, many of the subjects have no direct connection with agriculture, but only live in rural areas. This characteristic of the respondents serves to make their attitudes even more similar to those of urban residents. The variables focused upon in this study are occupation, education, income, religion, and political party preference (See Tables u through 8 for the distribution of these variables in the sample). The first four variables are the status indicators upon which each respondent is scored. The scores are used not only to indicate one's position on a specific hierarchy, but also to construct the various types of status inconsistency scores. For instance, education, occupation, and income are used in the determination of the general measure of inconsistency. Religion is employed in conjunction with the other three statuses to measure achieved-ascribed discrepancies. The inconsistency scores, as well as the status indicators themselves, are independent variables in our analysis. The latter variable, political party preference, is the dependent variable in this study. As noted above, a preference for the Democratic party is interpreted as a liberal political response, or equivalently, as support for programs of social change. We hypothesize that status incon- sistency increases one's political liberalism. Therefore, our prediction of Democratic party preference should be significantly better when we use the measures of inconsistency in addition to the status variables, as compared to when we use only the status indicators. These variables are included in this study because they are the ones most often employed in status inconsistency research. Broom 8 Jones also used these variables, and since our work is a replication of theirs, 142 TABLE 4 DISTRIBUTION OF SAMPLE BY RELIGION RELIGION N % Protestant 240 70.0 Catholic 76 22.1 Agnostic, Athiest 3 0.9 Sectarian groups 1 0.3 No religious preference 1? 5.0 No Answer 6 1.7 ---———‘———-_—-——_——_—e—--——_—__——_-‘-_-—-—_——_—_——_—--——-—-—-——---—~~_~_fl_—_. 43 TABLE 5 DISTRIBUTION OF SAMPLE BY OCCUPATION OCCUPATION N % High status occupations Professional, technical, and kindred 22 6.4 Managers, office workers, and proprietors: Self-employed 12 3.5 Managers, office workers, and proprietors: Salaried 8 2.3 Managers, office workers, and proprietors: General 1 0.3 Medium status occupation Craftsmen and foremen 52 15.2 Clerical and kindred 30 8.7 Farmers and farm managers 18 5.2 Low status occupations Operatives and kindred 57 16.6 Service workers 19 5.5 Sales Workers 7 2.0 Laborers and mineworkers 5 1.4 Farm laborers and farm foremen l 0.3 Private household workers 1 0.3 Never worked 83 24.2 No answer 27 7.9 an TABLE 6 DISTRIBUTION OF SAMPLE BY EDUCATION EDUCATION N % Primary education Grades 1-4 ' 7 2.0 Grades 5,6,7 17 5.0 Grade 8 n? 13.7 Secondary‘education High school incomplete 66 19.2 High school graduate 117 34.1 Post-secondarygeducation Technical, trade, or business school 22 . 6.4 College (incomplete) 46 13.4 College graduate 21 6.1 1+5 TABLE 7 DISTRIBUTION OF SAMPLE BY PARTY PREFERENCE PARTY N % Independent 7111 32.4 Democrat 109 31.7 Republican 106 30.9 Other 2 0.6 No Answer 15 4.4 ----——-—-———_———.-c------—-—————---———--—————-—---—----————————-—-———---_--—— 46 TABLE 8 DISTRIBUTION OF SAMPLE BY INCOME * o\° INCOME PER YEAR N Low Income Under $1000 4 1.2 $1000—1499 2 0.5 1500-1999 1 0.3 2000-2999 ' o 0.0 3000-3999 0 0.0 uooo-u999 1 0.3 Medium Income $5000-5999 15 ' 4.4 7000-9999 38 11.1 High Income_ $10,000-1u,999 33 9.5 15,000+ 10 2.9 No Answer or Woman** 239 69.6 Total 343 100.0 *This table contains income data for male respondents only **The income data for females who hold jobs outside the home are not available L17 it is appropriate for us to include occupation, education, income, religion, and party preference in the present analysis. The items used to collect data on these variables are described below. The occupational data were obtained by asking the subjects a two—part, open—ended question: "What is your occupation?”. The reSponses are classified according to the fourteen job types normally used by the Bureau of the Census and are then collapsed into three categories: low, medium, and high prestige occupations. The low status occupations are sales workers, operatives and kindred workers, service workers, farm laborers and farm foremen, laborers and mineworkers, and private household workers. The medium status occupations are farmers and farm managers, clerical and kindred workers, and craftsmen and foremen. The high status jobs are professional, technical, and kindred workers; and managers, office workers, and proprietors who are salaried, self-employed, or in general categories of these occupations. Educational background is determined by the subjects' responses to the question "What was the last grade or class you completed in school?". The answers to this question are classified in eight different categories representing various amounts of educational attainment and are then collapsed into three categories. Low educa~ tional attainment includes grade eight and below. Medium educational attainment is defined to include from some high school through high school graduates. High educational attainment includes the categories technical, trade, or business school; college incomplete; and college graduate. as To obtain a measure of income, the interviewers handed each respondent a card with ten income categories on it. The incomes were listed as annual totals and as the corresponding weekly equivalent of each total. Every income category had a different letter assigned to it and the subjects were instructed as follows: "I would like you to look at this card. What would you say your yearly income is from your present job? Please tell me your answer by letter." Incomes were originally coded into ten categories and then collapsed as follows: low income-below $5000 per year, medium income—$5000 to $9999 per year, high income-$10,000 per year or more. The respondents' religious affiliation was determined by the question "What is your religious preference?". Protestants are considered to be high on religious prestige, Catholics low on this status indicator. Catholics are assigned the lower status position for several reasons. Historically, Catholics have predominated among ethnic immigrant groups. These groups usually have a lower educational level, occupy the lower prestige occupations, and are located in lower income brackets than white, Anglo-Saxon Protestants. Their immigrant status has tended to further depress the overall status level of Catholics. These factors have combined to create an association of Catholicism with the lower social classes which has survived to the present time. Catholics have been unable to erase this stigma and as a result, we assign them positions on the religion hierarchy which are lower in status than those occupied by Protestants. Political party preference is measured by the question "Regardless 49 of how you voted last time, what do you consider yourself politically.". The response categories are Democrat, Republican, Independent, and Other. If the response was Independent, the subjects were asked "Toward which party do you lean?". We will see later how the response to this latter question is included in the analysis. The general status inconsistency score is developed in the following manner. First each respondent's score on the education, occupation, and income hierarchies is converted to a standard score to make them compara- ble with one another. Then the standard deviation among these three scores is computed for each individual. This figure is multiplied by ten to yield the status inconsistency score. A high score indicates a wide variation in scores on the three status variables and thus a high degree of status inconsistency. A low score indicates comparable values on the status indicators and thus a consistent individual. The range of status inconsistency scores is from one to twenty-four. Broom 8 Jones report a range of from zero to twenty—fbur in their inconsistency score. The high investment-low reward variable includes people who have at least a high school diploma (high investment) and either earn less than $5000 per year or are in the low occupation status category (low reward). The low investment-high reward variable consiSts of persons with an eighth grade education or less who are also in the highest in- come or occupation category. We have discussed previously the theoreti- cal importance of this type of inconsistency. The high ascribed-low achieved variable includes persons who are Protestants (high ascribed)and are in either the lowest education, occupation, or income category (low achieved). 50 The low ascribed-high achieved variable consists of Catholics who are also either high school graduates or are in the highest categories of income or occupation. This concludes our discussion of the variables employed in this research effort. We now proceed to the description of the analysis procedures utilized in our study. Method The method used to analyze the data and test the hypotheses is taken from three previous studies of the effects of status incon- sistency (Broom 8 Jones, 1970; Jackson 8 Burke, 1965; Treiman, 1966). In all three of these reports regression analysis with dummy variables is used to discern the effects of status inconsistency. Only the Broom 8 Jones report deals with political liberalism and how it is affected by inconsistency, but the method of analysis in all three studies is salient to the present research. We are using dummy variable regression analysis because it solves one of the main problems found in the majority of research in this area. It adequately differentiates between the main effects of the status variables and the inconsistency effect. This procedure enables us to account for some of the variance in the dependent variable which is not explained by the status variables. A simple additive regression model containing terms only for the status dimensions does not produce a complete explanation of political behavior. By modifying the additive model with the addition of terms for status inconsistency, a signifi- cantly greater amount of the variance in the dependent variable will ‘hopefully be explained. However, it is necessary to remove the 51 independent effects of the statuses before measuring the effects of the inconsistency. The critical question is not how much of the total varia- tion in political party preference can be explained, but whether the equa- tion including the inconsistency terms explains a significantly greater amount of the variance than the additive model. To develop the simple least-squares equation each category of the status variables is first converted into a separate dummy variable. For example, the three categories of occupation; low, medium, and high; gene-' rate three dummy variables. If a respondent has a low status occupation, the dummy variable for that category takes on the value of one; while the other two dummy variables for occupation are set equal to zero. Each dummy variable is weighted by a separate regression coefficient. If the dummy variable equals zero, the result of multiplying it by its coeffi- cient is zero and that term drops out of the prediction equation. If the dummy variable equals one, the result of multiplying it by its weight gives us the value of the coefficient. This term is added to the constant term, thus predicting the score on the dependent variable. The simple additive model, then, predicts the dependent variable on the basis of the dummy variables for education, occupation, and income. However, if we include all nine of the dummy variables in the equation, the least-squares procedure breaks down (Suits, 1957). Therefore, it is necessary to "constrain" one dummy variable for each of the status variables used in the equation. In this research the low category for each variable is constrained. This means that the equation includes six terms plus the constant; two dummy variables for each status. The effects of these constrained categories are collected in the constant term. The regression coefficients in the prediction equations for the various 52 dummy variables can be interpreted as the influence of those categories over and above that of the constrained category. The numerical value of the coefficient for the medium class of occupa- tion, for instance, represents the difference between being in this category as opposed to being in a low status job. For each hypothesis tested, a separate regression equation is used. The value of R2 for the equation indicates the amount of the variance in the dependent variable accounted for by the independent variables. The values of R2 for two different equations can be compared by an F-test to determine if one of the equations explains a significantly greater amount of the variance than the other. To test for inconsistency the R2 from the 2 from an equation containing simple additive equation is compared to the R those terms, plus a term representing status inconsistency. If the latter equation explains a significantly greater amount of the variance we con- clude that status inconsistency does have an influence on the dependent variable, over and above that effect generated by the status variables. This is the strategy of analysis used in this study. At this point a brief comment on the problems associated with this method of analysis is in order. First there is the difficulty of multi- collinearity. Regression analysis encounters problems of interpretation when the intercorrelations among the independent variables are high. How- ever, as can be seen from Table 9, this is not a problem because the inter— correlations among the independent variables are low. The high inter— correlations among the categories of a single variable are expected since they are mutually exclusive categories. A second problem arises because many of the variables used in this 53 0.H 0:.- 0m.- mm. Hm.- m0.- mm. 00.- 0H.- .asooo swam 0.H 00.- m0.- 00. 00.- 0H.- as. 00. .dsooo .00: 0.H 0H.- NH. s0. m0.- as.- as. .asooo 30; 0.H ms.- Hm.- 0m. 00.- ma.- msooeH swam 0.H 00.- 0H.- 0H. :0.- decodH .00: 0.H 00.- ma.- em. maoocH :04 0.H 0s.- 0m.- .osem swam 0.a mm.- .050m .00: 0.e .osem ace .asooo .msooo .mdooo oeoocH oEoocH osoonH .osom .ofiom .osom amen .00: son swam .00: see swam .00: 30a mmqm BzmmzmmezH mmk ozoz< mhzmHlohmOU ZOHB