© 1978 AN DREW DOUGLAS McN I TT ALL RI GHTS RESERVED AN EXAMINATION OF INTRA-PARTY COMPETITION: GUBERNATORIAL AND SENATORIAL NOMINATIONS IN THE UNITED STATES BY Andrew D. McNitt A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Political Science 1978 tj/ ABSTRACT AN EXAMINATION OF INTRAPPARTY COMPETITION: GUBERNATORIAL.AND SENATORIAL NOMINATIONS IN THE UNITED STATES By Andrew D. McNitt Multivariate techniques have not generally been used to analyze competition for nominations. As a consequence, much of the existing literature consists of a number of older studies each of which pays close attention to only one or two major variables. This study used multiple regression analysis to look at the relative influence of the major explanations of competition for nomi- nations on all gubernatorial and senatorial nominations made in the United States between 1954 and 1974. Incumbency, average party vote, office sought, social heterogeneity, opportunity rate, party name and degree of formal party organzation are related to both the number of candidates receiving 5% or more of the vote for nomination, and to the. percentage difference between the vote for the first and second place candidates for nomination. The results indicate that incumbency, average party vote and one of the measures of formal party organization, the degree of party control over statewide nominations, have an important influence on competition for nominations. Although no single variable is dominant, incumbency explains by far the largest amount of variance in competi- tion for nominations of any single variable. Andrew D. McNitt The effect of party name on competition for nominations changes over time. The results suggest that the decision to contest a nomina— tion may be influenced by short term expectations of success or failure in the general election because competition for nominations preceding the major post-war political landslides tends to be more intense for the nominations of the subsequently victorious party. Finally, competition for nominations appears to be a relatively unstructured activity. The entire set of indpendent variables explains a modest amount of variance in competition for nominations for the entire data set. Only when the equation is estimated for regional and yearly subsets of the entire data set is significantly more of the variance explained. ACKNOWLEDGMENT I am indebted to a number of people for the support and encour— iagement they gave me while writing this dissertation. My committee, Joseph Schlesinger, Charles Press and Ada Finifer, deserve special praise for their efforts on my behalf. Their perceptive insights into the mysteries of the nominating process, intellectual curiosity and emphasis on careful scholarship were of great help. I am also indebted to John Aldrich for methodological assistance; James Seroka for his encouragement; and Frank Codispoti and Bill Boyd for their comments. Finally, I would like to dedicate this dissertation to my wife, Paula. I am grateful for her patience and support. ii LIST OF LIST OF CHAPTER TABLE OF CONTENTS TABLES. FIGURES . I. SOME EXPLANATIONS OF COMPETITION FOR NOMINATIONS. Introduction. . . . . . . . . Why Study Competition for Nominations?. The Concept . . . . The Organizational Explanation. Nonorganizational Explanations. Summary . . . . II. THE DATA AND ITS MEASUREMENT. Introduction. . . Sources of Information. . Operationalization of Concepts. The Dependent Variable. Statistical Relationships Between the Independent Variables . III. AN ANALYSIS OF THE COMPLETE DATA SET. Introduction. . . . Presentation of the Model . Application of the Model. Average Party Vote. . Regional Analysis . Summary . . . . . IV. AN ANALYSIS OF COMPETITION FOR NOMINATIONS BETWEEN 1954 AND 1974 . Introduction. .. . . . . . . Changes in the Individual Variables Over Time . The General Equation Estimated Over Time. Summary . iii . 120 Page . vii ONL‘WNH (AN 32 32 35 35 42 50 50 SO 53 64 73 82 84 84 85 98 CHAPTER Page V. THE REDUCED MODEL . . . . . . . . . . . . . . . . . . . . . 122 Introduction. . . . . . . . . . . . . . . . . . . . . . . 122 Stepwise Regression . . . . . . . . . . . . . . 123 The Reduced Model Estimated Over Time . . . . . . . . . . 129 Summary . . . . . . . . . . . . . . . . . . . . . . . . . 140 VI. DISCUSSION. . . . . . . . . . . . . . . . . . . . . . . . . 142 Introduction. . . . . . . . . . . . . 142 A Final Summary of the Individual Hypotheses. . . . . . . 143 Implications for a Theory of Competition for Nominations . . . . . . . . . . . . . . . . . . 154 Implications For a Theory of Party Organization . . . . . 158 Conclusion. . . . . . . . . . . . . . . . . . . . . . . . 163 APPENDIX. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164 BIBLIOGRAPHY. . . . . . . . . . . . . . . . . . . . . . . . . . . 177 iv LIST OF TABLES TABLE 1. Correlation Matrix of Six Measures of Party Organization. . . . . . . . . . . . . 2. Regression of Measures of Party Organization on Themselves as a Test for Multicolinearity . 3. Correlation Matrix of Defined Variables . . . . 4. Multiple Regression of the Fully Operationalized Model on the Number of Candidates with 26% of the Vote (Pooled Data Set). . . . . . . . . . . . . 5. Multiple Regression of the Fully Operationalized Model on the Percentage Difference between the Vote for the First and Second Place Candidates (Pooled Data Set). . 6. Results of F Test to Determine if there is a Significant Difference between Linear and Curvalinear Estimators of Party Competition . . . . . . . . . . 7. Multiple Regression of the Fully Operationalized Model on the Percentage of Victory—Regional Split . . . . . . . . . . . . . 8. A Multiple Regression of the Fully Operationalized Model on the Number of Candidates Receiving 35% of the Vote for Nomination—Regional Split 9. Multiple Regressions of the General Equation on the Number of Candidates Who Received 36% of the Vote for Nominations. 10. Multiple Regression of the General Equation on the Percentage Margin of Victory. 11. Stepwise Regression Relating the General Model to the Number of Candidates Who Receivej:5% of the Vote for Nomination (listed in order of decreasing influence for variables significant at the p 3,05 level). . . . . . . . . . . . . . . . Page 43 44 47 55 61 71 75 76 . 100 . 111 124 TABLE 12. 13. 14. 15. 16. Stepwise Regression Relating the General Model to the Percentage Margin of Victory (listed in order of decreasing influence for Variables Significant at the p 3305 level) The Reduced Model Estimated Over Time Using the Number of Candidates Receiving 35% of the Vote for Nomination to Operationalize Competition for Nominations. . . . The Reduced Model Estimated Over Time Using the Percentage Difference between the Vote for the First and Second Place Candidates to Operationalize Competition for Nominations . Summary of Hypotheses Using the Number of Candidates Receiving >5Z of the Vote to Measure Competition The table indicates acceptance for Nomination. or rejection . Summary of Hypotheses Using the Percentage of Victory to Measure Competition for The table indicates whether or not the hypotheses are confirmed by Nominations. each of the equations. vi Page 125 . 130 . 132 . 144 . 145 LIST OF FIGURES FIGURE 1. 3A. 3B. 5A. 5B. 6A. 6B. 6C. 8A. 8B. 9A. A plot of the Estimated Relationship between Average Party Vote and the Number of Candidates Receiving 35% of the Vote for Nominations using both linear and nonlinear estimators ...................................... A Plot of the Estimated Relationships between average party Vote and the Percentage Margin of Victory using both linear and nonlinear estimators ..... ....... . ............. The Average Value of Several Measures of the Dependent Variable Competition for Nominations over Time ................ The Average Value of the Dependent Variable, Competition for Nominations, over Time ............................ ........ Average values for Measures of Party Organization for these States Making nominations by Year ....................... Avearge Opportunity Rate for Governors Over Time .............. Average Opportunity Rate for Senator Over Time ....... . ........ State Population by year for States Nominating Governors and Senators ........ . ........ ...... ...... . .......... Racial Diversity by Year for States Nominating Governors and Senators ....... . ........... . ........ ..... ....... Proportion of a State's Population Living in the Largest SMSA for States Nominating Governors and Senators . ............ Average Value for Office Variable by Year, Gov = l; Sen = 2 ......... ........................ .......... .. ..... ..... Average Party Vote by Year for States Making Gubernatorial and Senatorial Nominations ... ................... Standard deviation for Party Vote by Year for States Nominating Governors and Senators ...................... Incumbency by year; 0 = no incumbent running, 1 = incumbent running .................... . ..... ............... vii Page 66 67 86 86 88 9O 9O 92 92 92 93 95 95 96 FIGURE Page 9B. Incumbency by party by year; 0 = no incumbent running, 10. 11. l = incumbent running . . . . . . . . . . . . . . . . . . . . . 96 Selected Regression Coefficients for Variables From Table 9 Plotted Over Time for the # of Candidates and more than 5% of the vote for nomination . . . . . . . . . . . . 102 Regression Coefficients for Selected Variables From Table 10 Plotted Over time for the % Margin of Victory . . . . . . . . . 113 viii CHAPTER I SOME EXPLANATIONS OF COMPETITION FOR NOMINATIONS Introduction This is an empirical examination of competition for gubernator— ial and senatorial nominations. It is thus an exercise in the develop- ment of what Thomas Merton calls midrange theory (Merton, 1949:11). Rather than attempt to deduce a formal explanation of competition for nominations this study uses an inductive approach to weigh the relative importance of several potential explanatory variables. This process is one of empirical synthesis. A list of potential explanatory variables is developed from the literature on competition for nomination. Separate explanations are linked together in a single new model which is then evaluated with data from all gubernatorial and senatorial nominations made in the United States between 1954 and 1974. A comprehensive approach requires us to look at the reasons for studying competition for nominations, the ways in which the concept of competition for nominations is defined, and the existing literature on competition for nominations. Only then can we proceed to systemati- cally analyze the influence of the major explanatory variables on come petition for nominations. Why Study Competition for Nominations? There are a number of reasons for studying competition for nominations. ’Since one cannot be elected to office in the United States without first obtaining the nomination of one of the major par- ties, nomination is an important first step towards election. The nominating process is thus a great simplifying procedure where most of the candidates for office are eliminated from consideration. Those candidates who survive the struggle for nomination, directly control their party's policies (Schattschneider, 1942264, Sorauf, 1964:100). Party responsibility requires control over nomina— tions. The way in which candidates are selected influences the kinds of choices which are presented to the voters in the general election. Even relatively popular policies may not get a hearing if their advo- cates lose in either the primaries or conventions. In many places nomination is equivalent to election (Ewing, 1949:298; Turner, 1953:210). Although electoral competition has in- creased in the post-war period, a number of states are still dominated by a single party. In those states winning the nomination virtually guarantees winning the election. How a nomination is won affects a candidate's chances for suc- cess in the general election. The more divisive the primary, the harder it is for a candidate to win in the general election (Bernstein, 1977; 544). A divisive primary has a particularly strong impact on party activists. The losing candidate's campaign workers often either drop out or defect in the general election (Johnson and Gibson, 1974:77). The Concept Studies of competition for nomination have often been subsumed under more general analyses of party cohesion and factionalism. This approach however can create a great deal of confusion. The absence of competition for nomination is equivalent to neither party cohesion, nor factionalism. Competition for nomination is a simple concept. Party cohesion is a complex concept. Competition for nomination is the degree to which a single nomination is contested. Party cohesion involves much more than that; it also includes the extent to which a party's representa- tives vote as a bloc, the degree to which the party's cadre act as a single organization, and the extent to which a party's members remain loyal to its candidates. Competition for a nomination does not neces- sarily indicate a lack of cohesion. The presence of competition may be either an indication of deep-seated and highly emotional conflicts which threaten to tear a party apart or merely part of the normal poli— tical process. In the course of normal politics many conflicts are soon forgotten in orgies of good will and party unity. When we talk about competition for nominations, we look only at the quantifiable aspects of the situation. When we talk about party cohesion, we look at the qualitative aspects as well. Factionalism is another complex concept. Competition for nomin— ations has been confused with factionalism. However, the latter entails more than competition. Unlike factionalism, competition implies nothing about group structure. A faction is a loosely defined political group, and as such is usually unstable. Yet some factions, such as the Byrd machine of Virginia and the Long machine of Louisiana, have survived through several elections. To say that a party is factionalized is to claim that the party consists of a number of subgroups. The problem is how to identify these groups. Factions are hard to define. The supporters of a candidate may all belong to a single faction, or they may be a coalition of several separate factions. Competition for nomination is measured at a single paint in time. Factionalism is measured over time. Conceptual clarity requires that competition for nominations be studied as a separate phenomenon distinct from both factionalism and party cohesion. The Organizational Explanation Both the degree and type of party organization have been used to explain competition for nominations. Party organization has been viewed as both a recruiting and controlling agent. It is hypothesized that the more organized the party, the more the nominating process will resemble bureaucratic promotion. In bureaucratic systems all positions are filled with as little conflict as possible and competition for pro- motion is controlled in order to conserve scarce organizational re- sources. Early in this century scholars paid more attention to party organization than they do today. Highly descriptive and reformist in nature, these early studies related party organization to political behavior (Merrriam 1908; Merriam and Overacker, 1928). In the last twenty years, students of political parties shifted their attention from party organization to party identification and voting behavior (Henessey, 1968z6). Partially this is the result of the popularity of survey research as a method of analysis, and partially it is the result of a perceived decline in the influence of party organization in the United States. Survey research is a powerful tool for examining the attitudes and opinions of a large number of individuals. Initial fascination with the technique led to the intensive study of social backgrounds, attitudes and Opinions of the electorate (Lazersfeld, et al., 1964; and Campbell et al., 1960). Similar studies were also made of the social backgrounds, attitudes, and opinions of party activists (McClosky, et al., 1960; Conway and Feigert, 1968, 1974; and Wigins and Turk, 1970). Although fruitful in their own right, these studies ignored the struc- tural relationships that tie a party together. The apparent decline of party organization in the United States during the last half of the twentieth century also influenced scholar- ship. Party organization is ignored today because it is viewed as un- important. The list of criticisms is a long one. Political commenta- tors allege that the organization no longer controls nominations, that it has lost control of campaigns, and that it can no longer organize the government. Most of these allegations require a very narrow definition of party organization. Those who argue that party organization is dead, equate party organization with machine politics. While patronage based party organization is less common today than at the turn of the century, this does not prove that party organization has disappeared. Party organization has not died; rather it has changed. Party machines have been replaced by nonpatronage based party organizations. Broadly defined these nonpatronage organizations include candidate centered campaign staffs and associated volunteers, reform clubs, and the statutory or- ganization. Candidate centered organizations are the most common form of organization today. Even though these groups are not centrally con- trolled, their efforts on behalf of candidates bring money and volun- teers to the party's cause. These groups flesh out the party organiza- tion (Schlesinger, 1968; 430). "Amateur" or "Reform" clubs are part of the party apparatus. They are often formally affiliated with a party, control a number of party offices, and in many cases are as well organized as their more traditional competitors (Wilson, 1962). The formal legal structure is itself an important aspect of party organization. The law, by defining the formal structure, creates a need to fill the offices it authorizes. Many of the bitterest con- troversies in American politics arise over proposed changes in these party rules. Such bitter conflicts would not occur if these rules were unimportant. Party organization is a confusing concept because scholars have used two different definitions of party organization. The first defini- tion treated party organization as both structural arrangements and individual activists. The second definition treated party organization only as individual activists. The first definition implied that party organization was more than the sum of its individual members. While individuals were con- sidered important, party rules and structure were also part of the definition of a political party. The second definition de-emphasized all of these things. The second definition ignored the formal rela- tionship between activist members. It treated party organization as an undifferentiated collection of individuals. The party was viewed only as its individual activists. No attempt was made to look at the way in which these individuals were related. The failure to distinguish between these two different defini- tions of party organization created a great many problems. Those authors who concentrated on the structural aspects of party organiza- tion found that organization had a major impact on party cohesion (Schattschneider, 1942; Duverger, 1959; Key, 1964; Wolfe, 1966). Those authors who concentrated on party activists found that party or- ganization had little influence on competition for nominations in most, but not all, states. Studies done in Oregon, Iowa, North Carolina, Massachusetts, and Oklahoma found that party organization had little influence on competition for nominations (Seligman, 1961; Patterson, 1963; Eldersveld, 1964; Bowman and Boynton, 1966; and Paterson and Boynton, 1969), while a similar study done in Wisconsin, New Jersey, Pennsylvania and Illinois found that party organization had a strong influence on competition for nominations (Epstein, 1958; Ealau et a1, 1961; Sorauf, 1963; and Snowiss, 1966). The first definition of party organization is superior. It is better suited to an examination of intra-party behavior because it pro- vides a more complete picture of party activity, and because it focuses on the inter-relationship between individuals and structure. Any attempt to study the influence of political parties on competition for nomina- tions must not ignore the structural aspects of party organization. Party organization is a very difficult concept to specify be- cause it is so broad. Its measurement involves an examination of a number of very different types of phenomena. The problem is to identify which aspects of party organization affect intra-party competition. Organizational explanations of party cohesion have focused on structural arrangements and control over nominations as major explanatory variables. The basic hypothesis of the organizational studies is that the more bureaucratically organized a party, the greater its internal co- hesion. In its simplest form, this hypothesis can be traced to Michels, who argued that all forms of organization have a natural tendency towards oligarchy (Michels, 1914). Michels linked party organization to patterns of internal control, but his statement of the hypothesis was much too sweeping. Michels "iron law of oligarchy" is a tautology. Michels by defining democracy as direct democracy, made it impossible for any organization by its very nature to be anything, but an oligarchy. Michels theory does not distinguish between different degrees of organ- ization. It is difficult to apply the traditional bureaucratic model to party organization. A political party is a modern organization, but it is significantly different from other modern organizations. While a party is controlled by formal rules, and is hierarchically organized, the reward structure and sanctions available to political parties are very different from those available to business and governmental organ— izations. In the United States, political parties have very little control over their activist members because of an almost total lack of material incentives. Motivation and discipline are always a problem because volunteers, not paid workers, run the party organization. The bureaucratic model is not precise enough. Weber's theory of organization is the basis for much of our current analysis of bureaucracy. Weber defined bureaucratic organization as characterized by formal rules, technical evaluation of performance, a division of labor, hierarchical organization, impersonality, and efficiency (Blau and Meyer, 1971:29). Though Weber's typology distinguishes between radically different types of organizations, it does not differentiate between similar types of organizations. There are many different ways in which to organize a bureaucracy. There is in fact a great deal of variation within the bureaucratic type. Weber's ideal type does not address itself to an analysis of the effect of different bureaucratic traits on organizational behavior. Weber's theory can be reformulated by treating the bureaucratic model as a number of dimensions for analysis. This reformulation drops the assumption that the bureaucratic traits are necessarily corre- lated with each other. This change allows us to study the influence of organizational variation along a number of dimensions, but trans— forms the theory into little more than a technique for analysis. Instead of using the entire bureaucratic model as a guide for studying party organization, scholars have preferred to examine the influence of one or two key variables on political behavior. They argue that some organizational dimensions are much more important than others, and that most of the ways in which party organization differ from each other have no appreciable influence on competition for nomin- ation when compared to that of the key variable. Control over the distribution of resources is of critical importance to explanations of party cohesion and competition for nomin- ations. Of all the potential resources available, the ability of the formal party organization to legally control nominations has been considered of crucial importance for maintaining party cohesion. A number of authors have claimed that the direct primary fragmented party 10 organization in the United States by freeing candidates for office from the control of party regulars (Schattschneider, 1942; Key, 1964; Sorauf, 1964). Not only did they argue that primaries in general destroyed party cohesion, they argued that some types of primaries were more destructive than others. Run-off primaries were found to be associated with a higher level of internal competition than were single primaries (Key, 1949:419-471). Closed primaries were found to be less competi- tive than open and blanket primaries (Tobin and Keynes, 1975:678). Candidates nominated by convention and closed primary were found to be more closely associated with party leaders and more likely to have received preprimary endorsements than were candidates nominated by Open and blanket primaries (Tobin and Keynes, 1975:676—671). There was less competition in primaries in those states which made some nomina- tions by convention, than in those states where nominations were made exclusively by primary election (Wolfe, 1966:84). Registration requirements were also related to intra—party competition. Although American political parties are not mass member— ship organizations, membership is formally defined in a number of states. Party registration is a state regulated form of party membership. It requires a formal commitment to the party, and determines who can legitimately make a party's nominations. This stability by making it easier to predict the outcome of a primary should reduce competition for nomination.' Tobin and Keyens found less competition in Connecticut and Pennsylvania both of which use closed primaries than in Washington which uses a blanket primary (Tobin and Keynes, 1975; 678). 11 Paradoxically Terry Smith in his study of gubernatorial pri- maries produced entirely contradictory findings. Smith found that more candidates received at least 5% of the nominating vote in closed primaries than did in open primaries (Smith, 1973:83). Smith tried to explain this finding by suggesting that those voters who switch their votes from one party's primary to the other's tend to vote dispro- portionately for the most pOpular candidate in the contest. These floating voters then reduce primary competition because the registration system allows them to reinforce the winning margins of already advan- taged candidates. This model is consistent with the results that would be obtained if a large number of less politicized independents decided to vote in the Republican primary for a particularly popular incumbent governor. While Smith's explanation may be a correct interpretation of greater competition for nominations in closed as opposed to open pri- maries, it is also possible that the finding a statistical artifact. Smith was unable to fully control for the influence of a number of variables because of the limitations of contingency table analysis. This technique does not allow for an examination of the influence of more than a few variables on competition for nominations at any one time. The specialized studies of party activists suggest that members of the party organization have little influence on competition for nominations. Scholars studying four state legislatures found that party organization played a very limited role in sponsoring legislative can- didacies (Eaulau, Wahlke, Buchanan and Ferguson, 1961:226). Although more often mentioned by the minority party legislators, party was spon- taneously mentioned as a sponsor by less than half the members of the 12 legislatures in Ohio, California, and Tennessee. Only in New Jersey did a majority of the legislators state that their party played a major role in recruiting them (Eulau et al., 1961:227). Eulau, Wahlke, Buchanan and Ferguson had originally hypothesized that party would be more im— portant in the more competitive situations where the legislators would view it as a critical ingredient for electoral success. However, their findings of limited party influence in competitive situations in California and Ohio suggest that the party organization's influence recruitment is not solely dependent on competition. Patterson in a study of county chairmen and vice-chairmen in Oklahoma found that candidates for major office were not recruited from the party organization (Patterson, 1963:371). Most of the party leaders had not held any previous public office, most did not intend to run for office, and of those who had held public office, the office was almost always local. Patterson did find that the county leaders were interested in candidate selection--about half of them had encouraged someone to run, but less than 16% ever had tried to discourage anyone from running. Most of the leaders' efforts were directed towards recruiting candi- dates for races the party was likely to lose (Patterson, 1963:375). In Wisconsin, Epstein found that candidates were not recruited for the party apparatus. A hard core, about 40% of the party leaders, had no past, present, or future interest in running for public office. From this Epstein concluded that while the roles of party leader and candidate were not exclusive, a substantial number of activists were unlikely to ever become candidates. Epstein found that the party played a limited role in selecting candidates for local, but not state office (Epstein, 1958:930). At the 13 local level neither party endorsed county or state legislative candi- dates, although 83% of the party leaders, significantly more than in Patterson'sCklahomastudy, had encouraged candidates to seek nomination. At the state level, Epstein found that the Republican party's preprimary endorsements effectively controlled nominations. The Democratic party did not make preprimary endorsements at that time, however there was considerable interest in adopting a similar procedure because of its 'successful use by the Republicans. Eldersveld's study of Wayne County (Detroit) found that party leaders although dedicated to developing party organization paid little attention to candidate recruitment because of strong party opposition to preprimary endorsements. In an attempt to maintain party unity and avoid factional hostility political leaders adopted a "hands off policy" towards nominations (Eldersveld, 1964:341-343). Bowman and Boynton found that activists in North Carolina and Massachusetts did not consider participation in nominations a basic part of their role. "Less than 6 percent of the party officers were oriented toward nominating activities . . ." and though most had been consulted by candidates for office, there were almost always candidates- for local office (Bowman and Boynton, 1971:414). In another study of legislative recruitment, this time in Iowa, Patterson and Boynton found that although political parties played a more important role in recruiting candidates for public office, they did not play an exclusive role. Other interest groups and community leaders also played a significant role in recruiting legislative can- didates. Republican candidates were often recruited by local business interests (Patterson and Boynton, 1959:225-257). l4 Rossi and Cutright in a study of local primaries in a mid- western city found that while the party organization could mobilize enough voters to determine the outcome in close primary contests, the organization was often too divided to do so. They found that precinct committeemen split their support between the major contenders. They also found that party leaders refrained from endorsing candidates until just before the primary when the chances of backing a loser were rela- tively slim (Rossi and Cutright, 1959). Seligman found that the influence of party organization varied with the degree of inter-party competition. While party leaders seldom recruited or supported candidates in competitive or majority situations, the organization tended to recruit candidates when in the minority (Seligman, 1961:64). Patterson and Epstein also reported a similar re- lationship between recruitment and competitive status (Patterson, 1963; Epstein, 1958). Sorauf's study of Pennsylvania contradicts the argument that party organization has little influence on competition for nominations. Sorauf found that party leaders controlled the recruitment of legis- lative candidates, recruited candidates for minor offices, and arbitrated major intra-party disputes (Sorauf, 1963:54). Sorauf also found that both parties formally endorsed candidates for legislative nomination in a majority of districts, and that there was relatively little interference in the nominating process from nonparty groups. Snowiss nihis analysis of congressional recruitment in Chicago also found that the party organization influenced nominations. The Democratic party had aboslute control over congressional nominations within the city of Chicago, and was able to extend its influence into g ..p- .. Pd n3 15 suburban Cook county where it also controlled recruitment. However, the Republican organization was not quite as effective. Although it had some control over nominations within the city, it had a much harder time in the suburbs. In the suburbs, the Republicans were plagued by independent candidates, usually local businessmen, who were able to obtain the nomination on their ownnominate their candidates by state convention while their Democratic colleagues had to run in primary elections. State laws in the South allow a party to nominate by convention if its can- didates receive less than a specified percentage of the state wide vote. For years the Republican party fell well below this percentage. As Republican strength grew in the sixties Republicans began to nominate by primary election. In Virginia and South Carolina.however, the Republican party continues to nominate by convention because of more 18 liberal state laws which allow major parties to use either a nominating convention or a primary. Unfortunately, a number Of organizational practices occur in tOO few states to allow us to use them in this analysis. There is simply too little variance in the population to examine the influence Of such factors as formal membership, and centralized control over cam- paign financing. When we look at only those variables which can be measured, which are adequately represented in the sample, and which may be re- lated to competition for nominations, we are left with three major types Of variables. Those variables which measure a party's control over nominations, those variables which measure the extent to which the party organization is itself centralized and those variables that measure the extent to which the electorate controls a party's decision- .making process. The more resources that are controlled by a party's central decision~making body, the more leverage the party has tO control internal conflict. Internal conflict is eXpensive. It uses up the time and effort of volunteer workers, and wastes money which could be better used in the general election. Nominations are one of the few resources available to political parties which can be used to control internal conflict. The degree Of control overnominations varies from states where all nominations for statewide office are made by a convention to those where all nominations are made by an Open primary. A number Of states have mixed systems, major nominations are made by primary election and minor nominations by convention. Leege suggested that these minor nominations (i.e., Attorney General, Secretary of State, Treasurer, and 19 members Of administrative boards) are a major party building resource (Leege, 1970:519). Where a party controls nominations more is to be gained by accepting the party's choice than by contesting a primary. The formal hypothesis then is stated as follows. ,Hypothesis 1: The more control over nominations for state- wide Offices is restricted to the formal party activist organization, the less the competition for gubernatorial and senatorial nominations. Control over the selection Of presidential delegates is another major party resource. While the position of delegate to the national convention is not a full time position, and consequently not an adequate substitute for a gubernatorial Or senatorial nomination, the ability to controltflnaselection of presidential delegates is an indication of party strength. Until the recent proliferation Of presidential primary elections, the selection of presidential delegates was in a number Of states the only important nomination that could be made by the regular party organ- ization. Without the ability to nominate presidential delegates many state party organizations would lose their most important reason for existence. Only a weak or seriously threatened party would relinquish its control over these nominations. Although control over the selection Of presidential delegates may not be directly related tO the selection Of other kinds of nominees, it is a good indication of relative party strength. The stronger the party organization, the more likely it will be able to control compe- tition for nominations. 20 The hypothesis then is as follows. Hypothesis 2: When the formal party activist organization controls the selection Of presidential dele- gates, there will be less competition for gubernatorial and senatorial nominations. Yet the degree Of activist control over statewide nominations is insufficient to explain competition for nominations. The party apparatus must itself be unified if internal competition is to be controlled. TO the extent that legal regulations encourage the emergence of a separate centralized and unified activist organization, we would expect to find greater internal cohesion. The ease with which an individual can become part Of the party apparatus is one measure of the degree Of separateness of the party organization. The more per- meable a party organization, the easier it is for insurgencies to split the organization. Formally, the hypothesis is stated as follows: Hypothesis 3: The lower the barriers for admission to the formal party organization, the less likely there will be a unified activist organization and consequently the greater the competition for all of that party's nominations. The extent of indirect election is also a measure of the degree Of centralization. Duverger argued that the indirect election Of party leaders helped to maintain autocratic control over a political party (Duverger, 1954:138-142). The number of separate steps used to select members Of the state convention is one way to Operationalize the con- cept Of indirect selection. The more indirect the selection process, the more centralized the party and the less the competition for nomin— ations. 21 Formally, the hypothesis is stated as follows: Hypothesis 4: The more indirect the process for selecting members Of a state convention, the more hier- archically organized the party and conse- quently the less the competition for nomina- tions. Registration requirements also determine the permeability Of a party.‘ Closed registration makes for a more predictable election because it restricts control to a more stable electorate. Challenge, Open and blanket registration reduces that stability. Competition is more likely to occur in those situations where greater uncertainty allows more than one individual tO presume that they have a reasonable chance to be elected. Formally stated the hypothesis is as follows: ,Hypothesis 5: As registration requirements become more restrictive, competition for gubernatorial and senatorial nominations will decline. In addition to giving the rank and file member the authority to make nominations, some electoral systems allow him or her to make a greater number Of decisions. The run-Off primary differs from other types Of primaries because it allows the party voter to affect the selection process at more than one point in time. This changes the strategic environment of the primary election and tends to postpone final coalition formation. Formally, we will test the following hypothesis. Hypothesis 6: Those states which allow for a run-Off primary will be characterized by greater competition for gubernatorial and senatorial nominations. 22 Nonorganizational Explanations There are three major nonorganizational approaches that can be used to explain competitionibr nominations. Competition for nomina- tions can be explained as the result of sociological variables, indi- vidual variables, or collective political variables. There is a long tradition Of sociological explanations of political conflict. The sociological model explains political conflict as a direct reflection of group conflict. It is hypothesized that the more diverse a society, the more competition there will be for positions Of political leadership. Madison, in his discussion of an extensive versus a small republic, presented the basic pluralist argument. According to Madison a large republic is superior to a small republic because ". . the smaller the society, the fewer the distinct parties and interests, the more frequently will a majority be found in the same party; and the smaller the number of individuals composing the majority, and the smaller the compass in which they are placed, the more easily they will concert and execute their plans Of Oppression" (Madison 1789:22). This same argument has been used in the twentieth century to explain competition for power in both governmental and private organizations. Differences in the social substructure of these insti- tutions have beenlinked to varying degreesaoo oumum .msowumofiEOz Hmflucmmamoum HO>O Houuaoo ooo. mo. .mGOHumOHEOz Oumum uo>o Houucoo .mmmm huusm zumEHum «woncsm .o coaumuumfiwom mam mauumuoam sOHuam>soo Oumum macaumcfieoz Hmauconfimoum uo>o Houucoo .cowu coo. mo. tmuumflwom .mcoaumcweOz Oumum um>o Houucoo ommm muncm .m accumwOflm sowuso>sou Oumum mam humefium muonssm .mOOHumsfiEOz Hmfiusonflmmnm m:OHumcfleOz oo. «0. wo>o Houusoo .Ommm spasm .OOHumuumeom Oumum wm>o Houucoo .q mGOfiumsHEOz Hmfiusomwmmum um>o Houuooo mam mnoumumfim coauco>coo mumum .mcowumafiEOz Oumum ooo. «H. Ho>o Houucoo .Ommm zwucm .zumsaum uuoucom coaumuumfiwmm .m xcoumumfim :OHu:O>coo Oumum mom coaumuuwfiwom .mOOHumGHEOZ Oumum ooo. ca. um>o Houucoo .Ommm huusm .humefium wwOlcnm Hmfiuaomammum wm>o Houuooo .N :OHumuumeom can mcoaumcfiaoz Oumum uo>o Houusoo .OOHuOOHEOz Hmfiuaonfimoum 00. cm. uo>o Howucoo .Ommm xuusm .zquHum amoung Anoumwmfim coauco>oou Oumum .H mommOfimastm Nu moanmwum> usovconomcH oaamfium> uaochdoa mufiumocflaooauasz wow umma m an mo>aomEOLH so soauONficmwuo >uumm wo.mousmmo2 mo sOfimmowwom .N manna .'l 111 J‘- 45 control over nominations (Wolfe, 1966:39). It is also consistent with Ronald Weber's study of state party organization which found only low correlations among ten measures Of party organization. Weber further concluded on the basis Of a factor analysis of his variables that party organization isaimultidimensional, and not a unidimensional concept (Weber, 1969:17). Formal party organization in the United States is a hodge podge of inconsistent rules. A series of political reform movements have affected party organization in the United States. Each of the reform movements concentrated on one or two aspects of party organization and each managed to change party organizations in some Of the states. Unfor— tunately no distinction can be drawn between reformed and non-reformed states because of the inconsistent way in which these reforms were adopted. In the South, party organization has no control over nomina- tions for state Offices, but almost total control over the selection of presidential delegates. Party structure in that region is primitive, but registration requirements are among the most stringent in the country. In the North, four of the seven Open primary states allow their state conventions to play a significant role in nominating state officers. In Michigan the state conventions nominate candidates for the lesser state Offices. In Utah the state convention formally endorses candidates for nomination. If a candidate for statewide office receives 70% Of the vote in the endorsing convention no primary election is held. In Minnesota and Wisconsin, although state law does not provide for the formal endorsement of candidates, the state conventions make a practice of endorsing candidates for nomination. The diverse and perverse nature 46 of party organization in the United States suggests that it is probably wiser not to try to construct a single, all encompassing measure of the concept, but rather to treat each of the measures as separate variables. When the nonorganizational variables are correlated with each other (see Table 3), there is relatively little relationship. In no case do the simple correlation coefficients exceed .5. There is a weak association (r = -.35) between party name and average party vote. This can be attributed to the failure of the Republican Party to run candidates for major Offices in a number of southern states. There is a weak correlation (r = .31) between average party vote and incumbency. This is a reflection of the fact that a dominant majority party is more likely to have incumbent officeholders. Finally, there is a weak relationship (r = .31) between Opportunity rate for governors and racial fragmentation. This relationship can be explained by the stringent limitations placed on gubernatorial tenure in the South, which is also the most racially diverse region in the United States. There is also little relationship between the nonorganizational variables and the measures of party organization. The only exception is the relationship between racial fragmentation and the use of run-off primaries. The South is the most racially diverse region in the United States; unfortunately, it is also the home of the run-off primary. Oaklahoma is the only nonsouthern state to use run-off primaries. This correlation between racial fragmentation and run~off pri- maries is large enough to cause concern about the effects of multi- collinearity. Multicollinearity occurs when a close association between two or more variables makes it difficult to separate their independent effects. While the regression coefficients remain unbiased, at very high 47 ma «a ma NH HH oH o m m o m o m N H oo.H mo.l NH. oo. no. mo. mn.l oo. No.l ma. mo.| mH. oo. Hm.l NH.I sowumusoawmwm Hmfioom .mH oo.H mu.l mo. oo. ma. no. mm. Ha. no.1 HH.I mH.I 0H. oH. No.1 mouoowosmo oomam wsooom mam umwwm oomsu Ion mooowowmwo .qH oo.H oo.: oo.: m~.| «H.I mm.| oo.: oo. mH.I 0H. afi.l m~.l oo. mOumowoomo mo Homeoz .mH oo.H om. mo. mo.I oo. om.l o~.I No. no.1 oo. oH. No. znoumwoam coauoo>soo .N~ oo.H o~.| Ho.l mo.| oH.I mm.| oo. mo.l oo. om. no.3 Ommm human .HH oo.~ wo.l mo.t mm. 0H. oo.: o~.I oo. oo.: No. chfiuMGHEOz Oumum um>o Houusoo .oH oo.~ mo.l oo. oo.: oo.: NN.I Ho.l mo. «H. humsaum muoncsm .m oo.~ Ho. Ho.| HH.I Hm. oo. oo. No. mononesooH .m oo.~ Hm. do. oo.: Ho. NH.I oo.: m. zuwmm meuo>< .q oo.H oo. oo. ooawwo .m oo.H mo.l muoouo>oo wow mumm muacauuommo .N oo.H muouocom wow Oumm zufiasuuoooo .H moanmwwm> moowmoo mo xauumz OOHumaouwoo .m manna 48 levels of multicollinearity the coefficients become unstable. At these levels small changes in the equation or the data set can cause variation in the estimated coefficients. At more modest levels, multicollinearity increases the standard error of the estimates (Johnston, 1963:160). A larger standard error means that the significance levels of the coeffi- cients associated with racial fragmentation and run-Off primaries will be depressed. This increases the risk of underestimating the importance of these variables. A more stringent test for multicollinearity is provided by systematically regressing the entire set of independent variables on each other. This tests for the possibility that some of the variables are linear combinations of each other. When all of the independent variables are regressed on each other, the findings suggest that there is some multicollinearity. In three cases, average party vote (R2 = .25), state population (R2 = .27), and degree of convention hierarchy (R2 = .26), the squared multiple cor- relation coefficients exceeds .25. This moderate level of multicollin- earity complicates, but does not destroy the analysis. Caution is required in interpreting the results because of the inflated standard errors. Although the coefficients themselves are still the best unbiased estimators of the relationship, some of the coefficients which appear to be insignificant may not be so. To summarize, an analysis of the relationship between the inde- pendent variables suggests that the measures Of party organization should be treated as separate independent variables. Party organization appears to be too complex to be measured by a single variable. The analysis also suggests that the effects of multicollinearity will be of moderate 49 significance. While there is enough multicollinearity between the independent variables to require caution in interpreting the results, the degree of relationship is not high enough to cause undue concern. CHAPTER III AN ANALYSIS OF THE COMPLETE DATA SET Introduction An analysis of the entire data set provides a general understand- ing of those factors which influence competition for nominations. This style of analysis summarizes the relationship between the variables by averaging the effects of temporal trends and consequently cannot reveal how competition for nominations changes during the postwar period. How- ever, the larger number of cases in the combined data set makes it much easier to detect weak associations between variables and thus allows for a more sensitive analysis of specific relationships. Presentation of the Model To ascertain the relative influence of several variables on competition for nominations I use a multiple regression model. With multiple regression it is possible to test several individual hypotheses while at the same time studying the combined influence of the independent variables on the dependent variable. The regression coefficients in a multiple regression describe the relationship between a given independent variable and the dependent variable while controlling for the influence of all other independent variables in the equation. The standardized regression coefficients are equation specific. They measure the relative contribution of each 50 51 independent variable in the equation to the amount of variance in the dependent variable explained by the entire regression model. The standardized coefficients can only be used to describe the relationship between the independent variables within a specific model applied to a specific set Of data. It is invalid to compare standard— ized coefficients in one model with similar coefficients in another model. The unstandardized coefficients, however, are not equation speci- fic. They can be compared across both different equations and different data sets (Blalock 1967:136). The unstandardized coefficients indicate how much the dependent variable will change for each unit change in the independent variable. The study is designed to test the following theoretical model: y=a+Bxl+By2+Bx3+Bx4+ByS+€ Where: y competition for nomination a = the intercept X1 = Party organization X2 = Social fragmentation X3 = Incumbency X4 = Average Party Vote X5 = Opportunity rate E Error term The process of operationalizing the independent variables changes the model in several ways. Party organization, social fragmentation, and Opportunity rate require several separate measures to Operationalize each of the concepts. Also, in order to Obtain a large number of cases, data were collected over a 20-year period from 1954 through 1974. The rotating 52 nature of American elections complicates this analysis, and raises the possibility that the rate of change in competition for nominations may be caused by nothing more than the rotating election schedule. Three additional variables are required to control for the effects of these electroal cycles. The presidential cycle is one of the most prominant features of American politics. The process of nominating and electing a President increases both interest and participation in politics. A number of gubernatorial terms have been adjusted to avoid presidential election years. A dummy variable is included in the equation to determine if the processes of nominating a president has any influence on competition for gubernatorial and senatorial nominations. Senatorial elections are conducted on a rotating basis. Every two years one-third of the United States Senate is elected from 33 of the 50 states. The rotating nature of senatorial nominations may also account for some of the variance. Two dummy variables are added to the equation to control for the effects of senatorial class. The operationalized model, then, is specified as follows: y = a + blxl + bzx2 + b3x3 + b4X4 + bsx5 + b6x6 + b7x7 bsxa + b9x9 + blOXlO + bllxll + bllez + b13"13 b14x14 + blSXlS + b16x16 Where: y = Competition for gubernatorial and senatorial nominations a = The intercept x1 = Party control over state nominations x2 = Registration x = Entry base 53 x4 = Degree of convention hierarchy x5 = Use of a Run-off primary x6 = Party control selection of presidential delegates x7 = Proportion of state's pOpulation living in the largest SMSA in a state x8 = State population x9 = Metropolitan population xlo = Presence of an incumbent xll = Opportunity rate for governor x12 = Opportunity rate for senator xl3 = Average party vote x14 = Presidential year x15 = First senatorial class xl6 = SecOnd senatorial class e = Error term Application of the Model The first step in the analysis is the application Of the fully operationalized model to all gubernatorial and senatorial nominations made in the United States between 1954 and 1974. This procedure is used because it vastly increases the number of cases allowing for a more sensitive analysis of the relationship between the independent variables and competition for nominations. However, caution is required when interpreting the results because estimating the equation for the entire data set can Obscure temporal trends. This procedure requires all relationships be measured in a consistent direction over time by averaging all changes which occur in a relationship over time. This procedure also increases the possibility that autoregressive disturbances 54 might affect the significance levels of the individual coefficients. While autocorrelation does not bias the individual regression coeffi- cients, it can alter their significance in either direction. The greater sensitivity which is gained by pooling the entire data set then comes at the expense of disguising temporal trends and increasing the probability of autoregressive disturbances. Table 4 relates the fully Operationalized model to the number of candidates who receive EEK or more of the vote for nomination when the equation is estimated for the entire data set. By operationalizing com- petition for nominations as the number of candidates receiving 5% or more of the vote, the model focuses on the extent to which serious candidates are discouraged from seeking the nomination. The table indicates that party organization has an influence on the number of candidates for nomination. Two of the five measures of formal party organization, control over state nominations and registra- tion requirements, have a significant influence on competition for nomina— tions. As the regular party's control over statewide nominations increases the number of candidates for nomination decreases. This finding is con- sistent with Wolfe's Observation that primary competition is less in those states which made some of their nominations by convention than in those states which made all nominations by primary election (Wolfe, 1966: 39). The finding is also consistent with Leege's argument that lesser nominations are an important organizational resource for political parties (Leege, 1970:219). As registration requirements become more restrictive, the number of candidates for nomination increases. This finding is contradictory to Ranney's argument that registration requirements have no affect on 55 Table 4. A Multiple Regression of the Fully Operationalized Model on the Number of Candidates with 15% of the Vote (Pooled Data Set) Variable B b Sig Racial Fragmentation .05 .004 .18 State Population .08 .00002 .01 % in Largest SMSA -.06 .003 .04 Opportunity Rate Gov. -.10 .018 .00 Opportunity Rate Sen. .00 .001 .82 Average Party Vote .16 .012 .00 Incumbency -.28 .67 .00 Office -.11 .25 .00 Party Name -.09 .20 .00 Control over State Nominations -.16 .119 .00 Selection of Pres. Del. .02 .002 .41 Registration .06 .068, .05 Entry Base -.07 .051 .18 Convention -.02 .015 .44 Run-off .03 .07 .45 Presidential Year .02 .038 .53 Senatorial Class #1 .04 .113 .25 Senatorial Class #2 -.01 .004 .75 Constant 2.734 R2 18.5% F 14.467 Sig. = .001 56 on party cohesion (Ranney, 1974:76). The finding confirms Terry Smith's Observation that competition for nomination decreases as registration requirements become more Open, a hypothesis which violates the logical expectation of decreased competition (Smith, 1973:83). Smith's study is based on an examination of all gubernatorial nominations made after 1900. This analysis is based on an examination of all gubernatorial and senatorial nominations made between 1954 and 1974. While there is some overlap between the two data sets, the findings in Table 4 provide addi- tional evidence supporting Smith's contention that competition is greater in closed as Opposed to open primaries. While none of the four remaining measures of party organization is significantly related to competition for nomination, three of the four insignificant relationships are in the hypothesized direction. Greater party control over entry into the activist organization is associated with fewer serious candidates for nomination. Greater hierarchy in the state convention is associated with fewer serious candidates for nomina- tion. Run—Off primaries are associated with a greater number of serious candidates for nomination. Centralized control over the selection of presidential delegates is insignificantly associated with an increasing number of serious candi— dates for nomination. This finding contradicts hypothesis 2, which con- tends that there is a relationship between control over the selection of presidential delegates and control over the selection of gubernatorial and senatorial candidates. The finding suggests that there is a great deal of separation between national and state level nominations, and that the ability to control one group of nominations does not necessarily aid in controlling the other. 57 Several nonorganizational variables have a significant influence on the number of candidates contesting a nomination. In this case incumb- ency and average party vote are the most important nonorganizational variables. The presence of an incumbent is associated with a large decline in the number of candidates for nomination who receive 5% or more of the vote. Increasing average party vote is associated with a signifi- cant increase in the number of candidates who receive 5% or more of the vote. Incumbency has the greatest influence of any of the variables in the equation on competition for nominations. In particular incumbency has a greater influence on the number Of serious candidates than does average party vote. The linear relationship between average party vote and competition for nominations is strong, but not overwhelming. This finding contradicts Patterson's contention that average party vote plays a dominant role in explaining competition for nominations (Patterson, 1976). Two of the three measures of social fragmentation, state popula- tion and metrOpOlitan domination, have some significant influence on the number of candidates. The larger a state's population, the greater the number of candidates for nomination who receive 5% or more of the vote. This observation supports Madison's contention that political institutions in a large republic will be less likely to be dominated by a single fac- tion than in a small republic (Madison, 1789), and contradicts Terry Smith's contention that there will be less competition for nomination in a large state because of the greater cost of campaigning (Smith, 1973). As the proportion of a state's population living in its largest metropolitan area increases, the number of candidates for nomination 58 decreases. This relationship is consistent with the social diversity hypothesis. It suggests that competition is associated with a multi- plicity of population and media centers in a state. Increasing racial diversity is also associated with a greater number of candidates for nomination who receive 5% or more of the vote, but not significantly so. However, caution is required in interpreting this result because of the correlation between the variables measuring racial fragmentation and the use Of run-Off primaries. The most racially diverse states are also the states with run-off primaries. Party name is significantly related to the number of candidates who receive 5% or more of the vote. There are more serious candidates for Democratic nominations than for Republican nominations. This finding is consistent with, but does not provide a critical test of the social diversity hypothesis. The results indicating that party name has a significant influence when party competition is controlled for contra— dicts Morehouse's observations that the different competitive positions of the parties explain greater competition for the Democratic nominations (Morehouse, 1976). Office sought is also significantly related to the number of can— didates who receive 5% or more of the vote. There is more competition for gubernatorial than for senatorial nomination. Evidently more is involved than just the attractiveness of the office. The larger number of potential gubernatorial candidates and the greater vulnerability of the governor seem to result in greater competition for that Office. The relationship between Opportunity rate and the number of can- didates for nomination is puzzling. The opportunity rate for governor is significantly related to competition for nominations while the 59 Opportunity rate for senator is not. For both governor and senator, increasing Opportunities are associated with a declining number of can— didates for nomination. Although the direction of the relationship between opportunity rate and competition for nominations is as hypothesized, the signifi- cance of gubernatorial opportunity rate, but not of senatorial opportunity rate is disturbing. The finding does not appear to be the result of multicollinearity because of the low correlation between the two variables (r = .06). Nor does the finding appear to be the result of a lack of variation in senatorial opportunity rate. The standard deviations for senatorial Opportunity rate is much too close to that for gubernatorial Opportunity rate to sustain this interpretation (3 = 5 vs, 3 = 6). Although the difference between the mean opportunity rates for governor, 12, and the mean rate for senator of, 6, indicates that there are con— siderably more opportunities to become governor. The relative insignificance of senatorial vs. gubernatorial Oppor- tunity rate can be interpreted in two ways. One, it may be the result of differences in length of gubernatorial term and tenure limitations rather than turnover rate that causes the difference in significance levels. Two, the finding may also indicate that the governorship is the major outlet for pent up political ambitions at the state level. The governor- ship's reputed vulnerability may attract politically ambitious individuals who choose that office as the easiest of all the major offices to obtain. None of the variables which control for the electoral cycles is significant. Presidential year is insignificantly related to an increas» ing number of candidates for nomination. The variables for the first senatorial class is insignificantly related to an increasing number of 60 candidates, while the variable for the second senatorial class is insignificantly related to a decreasing number of serious candidates. The low levels of significance suggest that these electoral cycles do not have a great deal of affect on competition for nominations. To summarize, the generally modest multiple correlation coeffi- cient for the equation suggests that competition for nominations is relatively unstructured. Of all of the variables incumbency, average party vote, and control over state nominations are the most strongly related to the number of candidates contesting the nomination who receive 5% or more of the vote. Racial fragmentation, method of selecting presi— dential delegates, opportunity rate for senators, entry base, convention hierarchy, run-Off primaries and all of the cyclic variables do not appear to be significantly related to competition for nominations when measured by the number of candidates receivingj:5% of the vote. Table 5 relates the model to the percentage difference between the vote for the first and second place candidates for nomination. This measures the extent to which a single candidate is capable of dominating the nominating process. A positive coefficient indicates an increasing percentage margin of victory and a negative coefficient indicates a decreasing percentage margin. Only one of the measures of formal party organization, control over state nominations is significantly related to this measure of com- petition for nominations. As a party's control over state nominations increases, the margin of victory of the leading candidate for nomination also increases. Registration is not significantly related to competition for nominations when competition is measured by the percentage margin of victory. However, the direction of the relationship is still consistent 61 Table 5. Multiple Regression of the Fully Operationalized Model on the Percentage Difference between the Vote for the First and Second Place Candidates (Pooled Data Set) Variable B b Sig Racial Fragmentation -.04 -.094 .30 State Population -.05 -.0005 .ll % in Largest SMSA .11 .195 .00 Opportunity Rate Gov. ..09 .052 .00 Opportunity Rate Sen. -.01 -.096 .61 Party Competition -.17 -.446 .00 Incumbency .32 25.33 .00 Office .15 11.7 .00 Party Name .08 6.016 .00 Control Over State Nominations .12 2.94 .00 Selection of Pres. Del. -.60 -.017 .87 Registration -.04 -1.80 .13 Entry Base .01 .332 .80 Convention HIR .04 .614 .35 Run-Off .03 2.654 .43 Presidential Year -.01 -.643 .76 Senatorial Class #1 -.02 -2.228 .50 Senatorial Class #2 -.01 -.769 .83 Constant 32.645 R2 18.5% F 14.457 Sig. = .00 62 with Smith's Observation of increasing competition as registration requirements become more Open. As in Table 4, control over the selection for presidential delegates is insignificantly associated with greater con- trol over state nominations. Entry base and degree of convention hier- archy are related as hypothesized to competition for nominations, but are insignificant when competition for nominations is measured by the percent— age margin Of victory as well as when measured by the number of candidates who receive ;:5% of the vote for nomination. Finally, not only is the relationship between run-off primaries and the margin of electoral vic- tory insignificant, it is also exactly the reverse of the results which are Obtained when competition was measured by the number of candidates receiving 35% of the vote for nomination. Once again, incumbency and party competitiveness are the most important nonorganizational variables in the equation. The presence of an incumbent is significantly associated with an increasing percentage margin of victory. Greater party competition is also associated with a declining percentage margin of victory. The relationship for the measures of social fragmentation, that is, racial diversity, state pOpulation, and proportion of a state's population in its largest metropolitan area, and competition for nomina- tions, is essentially the same when competition is measured by the percentage margin of victory as when measured by the number of serious candidates. The only substantial difference is that state population is barely insignificant in the first case (p = .11) and statistically significant in the second case (p = .01). The relationship between party name and competition for nomina- tions indicates that there is more competition for Democratic nominations 63 than there is for Republican nominations. This is the same finding measuring competition as the percentage margin of victory as was obtained by measuring competition by the number of serious candidates. This further contradicts Morehouse's arguments about the lack of a relation- ship between party name and competition for nominations (Morehouse, 1974). Once again there is more competition for gubernatorial than for senatorial nominations. There is a strong and highly significant rela- tionship between Office sought and difference between the vote for the first and second place candidates for nomination. The relationship between Opportunity rate and competition for nominations is substantially the same for Table 5 as for Table 4. A restricted opportunity rate for governor is significantly associated with increasing competitiveness. The relationship between competition for nominations and senatorial opportunity rate is also insignificant for both tables. The only minor difference is the reversal of the insignificant relationship between competition for nominations and senatorial OppOrtunity rate when the percentage margin of victory is used instead of the number of candidates receivingj:5% of the vote to Operationalize competition for nominations. 'None of the dummy variables used to control for the effects of electoral cycles is statistically significant. Again, presidential year is insignificantly associated with closer competition for nominations. A finding which further reinforces the notion of a separation between state level and presidential level nominations. Finally, both of the variables for senatorial class are associated with decreasing margins of electoral victory, but neither is statistically significant. 64 To summarize, the use of percentage margin of victory to measure competition for nominations produces results which are generally similar to those obtained when competition is measured by the number of candi- dates receiving 5% or more of the nominating vote. Incumbency, average party vote, and office sought were the variables which were most strongly related to the percentage margin of electoral victory. Party organiza- tion continues to have an influence on competition for nominations. Control over state nominations remains the most important organizational factor controlling competition although registration requirements no longer have a significant influence. Finally, the modest multiple correlation coefficient for Table 5 continues to suggest that the nom- inating process is relatively unstructured. Average Party Vote Average party vote has been used as a major explanation of competition for nominations(Patterson,1976). It is hypothesized that as the chances of winning an election improve, competition for nominations increase (Key, 1956). While the analysis of the entire data set indi- cates that average party vote is an important variable, average party vote does not play a dominant role. This finding, however, may be caused by incorrectly specifying the nature of the relationship between competition for nominations and average party vote because average vote may not be related to competition for nominations in a linear manner. If average party vote is not linearly related to competition for nominations, then substituting nonlinear estimators of average party vote into the general equation should significantly improve the explana- tory power of the equation. Alternate hypothesis 11 suggests that the 65 relationship between average party vote and competition for nominations although positive is not linear because of a decline in competition for nominations when the two parties are closely competitive. This hypothesis is tested by including two additional variables in the general equation which estimate the relationship between party competition and average party vote as a third degree polynomial. The general equation is modi- fied by substituting bx + bx2 + bx3 for the simple linear estimator of average party vote, bx. When the new equation is estimated and the relationship between average party vote and competition for nominations plotted for a third degree polynomial (see Figures 1 and 2), the results are not consistent with alternate hypothesis 11. There is no indication that competition for nominations declines when inter—party competition is relatively even, i.e. when average party vote is approximately 50% of the two party total. Consequently alternate hypothesis 11 is rejected. The results however do suggest a curvilinear relationship. When the relationship between average party vote and competition for nomina- tions is plotted as a third degree polynomial the amount of competition declines when average party vote exceeds 60% of the two party vote. A curvilinear relationship between average party vote and compe- tition for nominations is plausible if there is a threshold effect. If the important factor in explaining competition for nominations is the ability to assume that once the nomination is obtained victory in the general election is assured, then average party vote would have less of an influence on competition for nominations once a party's average vote rises above the minimum amount required to assure electoral success. In 66 Ib(1og x) bx + bx2 + 6x3 1.00 bx 0 1o 20 30 4o 50 60 7o 80 90 Average Party Vote as % of Total Vote Average Party Vote R2 for General Equation Linear estimate 18.5% Second degree polynomial 19.0% Third degree polynomial 19.6% Log linear estimate 18.2% Figure l. A plot of the Estimated Relationship between Average Party Vote and the Number of Candidates Receiving:;5% of the Vote for Nominations using both linear and nonlinear estimators 67 00 10 20 30 40 50 6O 7O 80 90 100 0 10 20 30 4O 50 60 70 80 90 100 Party Vote as % of Total Vote Average Party Vote R2 for the General Equation Linear estimate 18.4% Second degree Polynomial 19.5% Third degree Polynomial 19.7% Log linear estimate 18.7% Figure 2. A Plot of the Estimated Relationships between average party Vote and the Percentage Margin of Victory using both linear and nonlinear estimators 68 practice this would mean that the number of active candidates would not continue to increase at the same rate once average party vote passes the 50% point. In addition to the third degree polynomial, a curvilinear and a log linear model were also estimated for average party vote to take account of different degrees of drop-off in the rate of competition for nominations after the minimum winning point had been reached in the general election. The curvilinear model is estimated by treating average party vote as a second degree polynomial. The general equation is modi- fied by substituting bx + bx2 for the simple linear estimator, bx, of average party vote. The log linear model is estimated by substituting the loged estimator, b (log x), into the general equation for the linear estimator, bx, of average party vote. The question is do any of the curvilinear models provide a better estimate of the relationship between average party vote and comp— etition for nominations than does the linear model, and if so is any one of the curvilinear models clearly superior to the others. This is not an easy question to answer because it mixes both theoretical and statis- tical concerns in its interpretation. Individual significance levels are not helpful for evaluating polynomial equations because of the high degree of multicollinearity between the estimators. The variables, in the polynomial expression are mathmatical transformations of each other. This often results in a close correlation between coefficients which inflates their standard errors resulting in statistical insignificance (Kementa, 1972:452). However, it is possible to use the multiple correlation coeffi- cients for the entire equation to determine if the use of nonlinear 69 estimators produces a significant increase in the general equations explanatory power. When the linear estimator of average party vote is used to estimate the general equation for the number of candidates receiving2_5% of the vote for nomination the r2 for the entire equation is 18.4%. All but one of the nonlinear models explain slightly more of the variance in this measure of competition for nominations when included in the general equation. However, the amount of improvement is very small. The r2 for the general equation in 10.6% when the third degree polynomial is used, 19.5% when the second degree polynomial is used and 18.2% when the log linear model is used. When the linear estimator of average party vote is used in the general equation to estimate the percentage margin of victory the r2 for the entire equation is 18.4%. Once again all of the nonlinear models explain slightly more of the variance when they are included in the general equation. However, the amount of improvement is very small in this instance as well. The r2 for the general equation is 19.7% when the third degree polynomial is used, 19.5% when the second degree poly- nomial is used and 18.7% when the log linear model is used. Are these slightly improved estimates for the nonlinear models statistically significant? To answer this question an F statistic is calculated for the difference between the estimates of the general equation using each of the nonlinear measures of average party vote. The log linear model produces the smallest amount of change in the r2 for the general equation. The log linear model uses only a single variable to estimate the relationship between competition for nomination and average party vote. The log linear model is evaluated by using a difference of variance test to calculate the F statistic which 70 tests the null hypothesis that the coefficients for the linear and log linear models are of equal value (Winkler and Hays, 1975:386). The general formula used to calculate the F statistic is as follows: '11 II W U) I—‘NNN Where S Variance of the general equation using a log linear estimator of party competition. 03 MN ll Variance of the general euqation using the linear estimator of the competition for nominations. The results of the F tests (see Table 6) allow us to immediately reject the log linear model. The log linear model is not a significantly better estimator of the influence of average party vote on either measure of competition for nomination than is the linear model. The two polynomial models explain a slightly greater amount of variance than do the linear and log linear models. The F statistic for the difference between the linear and polynomial models is calculated in a slightly different manner to adjust for the different number of coefficients required to estimate the general equation when second and third degree polynomials are included in the equation. The formula for the F statistic in this instance determines if the difference between 2 . . . . the r for the general equation u31ng a linear estimate of average party vote and the general equation using the polynomial estimates of average party vote is statistically significant. The general formula used to calculate the F statistic is as follows: 2 2 (Rem-(R1,) F: 2 l-R ‘17 q xi 71 Table 6. Result of F Tests to Determine if there is a Significant Difference between the Linear and Curvalinear Estimators of Party Competition # 15% of Vote % Margin of Victory Actual Needed* Actual Needed* bxl + bxi 17.26 1.75 7.59 1.75 Bxl + bxi + bx? 7.046 1.75 9.765 1.75 b (log K) .0008 1.00 .00075 1.00 *F value required for significance at the p .05 level 72 Number of cases Number of coefficients in new model Number of coefficients in the old model - Squared multiple correlation coefficient for the new model Squared multiple correlation coefficient for the original model Where: .OWN FPNW'D D l The results of the F test for the two polynomial equations (see Table 6) indicate that the two polynomial models do a significantly better job of estimating the relationship between average party vote and competition for nominations for both measures (If competition for nominations. However, this does not necessarily mean that the curvi- linear model of the relationship between average party vote and competi- tion for nominations is definitely superior to the linear model. The polynomial models do only a slightly better job of estimating the relationship between average party vote and competition for nominations. Further, this difference could be little more than a statistical arti- fact. Both of the measures of average party vote have an upper and lower limit, the number of candidates who receivej:5% of the vote for nomination ranges from one to 20, and the percentage margin of victory ranges between zero and 100. When a variable is limited in this manner, it tends to produce slightly bowed estimates similar to those obtained for both of the polynomial models. While the results of this portion of the analysis are clearly ambiguous a few tentative conclusions are possible. Alternate hypothesis 11, which predicted a decline in interparty competition when a party's average vote was around the 50% level can be rejected. Evidently any tendency toward greater internal solidarity in closely competitive situations is not reflected in decreased competition for party nomina- tions. 73 The second, and perhaps most obvious conclusion is that although a curvilinear model could be used to estimate the relationship between average party vote and competition for nominations, it would not make much of a difference in the explanatory power of the general equation. The relatively modest r2 for the general equation then is not the result of incorrectly specifying a linear instead of a curvilinear relationship between average party vote and competition for nominations. Finally, and most tentatively of all, the amount of improvement in the predictive power of the general equation which results from using curvilinear estimates of the relationship between average party vote and competition for nominations while significant is not great enough to rule out the possibility that these findings are not due to statisti- cal problem associated with the use of bounded measures. Consequently for reasons of simplicity and methodological conservatism the linear model should continue to be used until a clearly superior fit can be obtained forzinonlinear alternative. This examination of competition for nominations has not developed enough evidence to justify such a change. Regional Analysis The Mason-Dixon line is more than a line on the map. The South is still the most "regional” of all of the regions in the United States. Political, social, and institutional differences separate the South from the rest of the country. Although the South's 100 years of political isolation may be slowly ending, there are still a number of good reasons to estimate the general equation separately for the South. The South is the home of 74 more than one kind of peculiar institution. The run-Off primary, limita- tions on gubernatorial succession, exceptionally severe registration requirements and one-party rule are disproportionately Southern. While Southern society has changed many of the institutional differences con- tinue to influence political behavior. From a methodological point Of view, the close association between institutional and social variables in the South makes it diffi- cult to determine which is more important. Specifically, the correlation between run-off primaries and racial diversity makes it difficult to analyze their influence on competition for nominations in a national sample. Most of this correlation is the result of including the South in the data set. Only by splitting the data set into a Southern and nonsouthern sample can a clear idea be Obtained of the relationship between racial diversity and run-off primaries to competition for nomina- tions. Several interesting things happen when the data is split into two regional groups, and the general equation is re—estimated for each data set (see Tables 7 and 8). The general equation does a much better job of explaining competition for nominations for the South than it does for the North. The squared multiple correlation coefficient for the entire equation estimated for the South when competition for nominations is measured by the number of candidates receiving 5% or more of the vote is 32.6% compared with 17.4% for the nonsouth, and the squared multiple correlation coefficient for the South when competition is measured by the difference between the percentage vote for the first and second place candidates is 34% compared to 17.5% for the nonsouth. 75 Table 7. Multiple Regression of the Fully Operationalized Model on the Percentage of Victory - Regional Split South* Nonsouth Variable b Significance b Significance Racial Fragmentation .122 .68 -.217 .07 State POpulation -.0017 .20 -.0004 .28 % in Largest SMSA -.223 .64 .216 .00 Opportunity Rate Gov. -.371 .62 .541 .00 Opportunity Rate Sen. .188 .63 -.130 .56 Party Competition .0529 .78 -.565 .00 Incumbency 31.974 ,00 25.030 .00 Office 8.375 .15 12.513 .00 Party Name 49.48 .00 4.699 .04 Control Over State Nominations .034 .98 3.272 .00 Selection of Pres. Del. .285 .35 -.042 .70 Registration -7.167 .04 —1.567 .23 Entry Base 1.212 .73 .606 .68 Convention Hierarchy 1.444 .33 .599 .47 Run-Off 13.014 .07 -5.726 .30 Presidential Year 1.198 .78 -1.528 .52 Senatorial Class #1 1.49 .84 -3.367 .38 Senatorial Class #2 2.54 .76 -l.037 .80 Constant 41.508 28.927 R2 36.6% 17.6% F 5.783 Sig. = .00 10.843 Sig. = .00 *There is a great deal of multicollinearity in the Southern data set, especially between incumbency, party name, and party competition. 76 Table 8. A Multiple Regression of the Fully Operationalized Model on the Number of Candidates Receiving3;5% of the Vote for Nomination - Regional Split South* Nonsouth Variable b Significance b Significance Racial Fragmentation .001 .93 .0088 .01 State Population .00006 .22 .00002 .09 % in Largest SMSA .005 .78 -.0044 .01 Opportunity Rate Gov. -.0251 .34 .016 .00 Opportunity Rate Sen. -.0055 .69 .002 .71 Party Competition -.002 .78 .016 .00 Incumbency -.770 .00 -.644 .00 Office -.711 .00 -.130 .15 Party Name -1.38 .00 -.168 .01 Control Over State Nominations -.007 .90 —.131 .00 Selection of Pres. Del. .004 .73 .0017 .56 Registration .170 .16 .075 .04 Entry Base .112 .37 -.064 .12 Convention Hierarchy -.079 .13 .0044 .85 Run-Off -.028 .26 .415 .01 Presidential Year .218 .18 .175 .79 Senatorial Class #1 -.003 .99 .100 .35 Senatorial Class #2 .109 .70 -.076 .50 Constant 5.793 2.275 R2 32.6% 17.6% F 5.428 Sig. .00 10.811 Sig = 00 77 The intriguing question is why does the general equation do a better job of explaining competition for nominations for the South that for the rest of the country. The general equation is an attempt to measure the structured elements of competition for nominations. The greater its explanatory power, the more "structured" the situation. The results suggest that competition for nominations is more controlled by constant political, social, and organizational factors in the South than in the rest of the country. A finding which at first seems to contradict all conventional interpretations of Southern politics. The lack of party organization, multiple candidacies, and close competition for nominations in the South has been interpreted as evidence of a wide Open political system (Key, 1949). Structured behavior is not expected from the "wild men" of American politics. Structure, however, does not necessarily require organization. Political circumstances such as incumbency, party history, and Office sought can also explain competition for nominations. The lack of conflict over political issues among white politi- cians in the South may have produced a greater amount of stability. Southern politics has been shaped by the race issue. An issue so salient that almost all politicians were forced to take an identical position on the question while submerging their other policy differences. Racial issues were particularly salient during most of the 19503 and 603. It should be no surprise to find a more structured component to competi- tion for gubernatorial and senatorial nominations in the South during this period because with only one position that could be taken on the most important issue of the day, less advantaged candidates for nomination 78 were deprived of a major resource for wresting control away from their opponents. The higher squared multiple correlation coefficient for the general equation in the South can also be explained as the result of a much simpler system. The lack of genuine two-party competition elimi- nates the necessity of the candidate trying to calculate who his Opponent will be in the general election. Hence, a one-party system reduces the amount of uncertainty and restricts a candidate's strategic calculations to his own party. The next question is what happens to the relationship between the independent variables and competition for nominations when the general equation is estimated separately for the South and for the rest of the country. This is a much harder question to answer than is at first indicated by an examination of Tables 7 and 8. Splitting the data makes it easier to look at some of the relationships outside of the South, but increases the problem of multicollinearity for the Southern sample. Incumbency, party name, and average party vote are all multi— collinear for the Southern sample. This makes it very difficult to rely on any of the significance levels that are reported for these coefficients. Although the estimators are still unbiased, it is likely that some of them are insignificant because of the multicollinearity. The results for the Southern sample when competition for nomina— tions is operationalized by the number of significant candidates suggests that incumbency, party name, and office explain most of the variance. Incumbents have a tremendous advantage in obtaining renomination. There is more competition for Democratic nominationsthanfor Republicannomina— tions, and there is more competition for gubernatorial as opposed to 79 senatorial nominations. However, the multicollinearity between incumb- ency, party name, and average party vote affects the results. The fact that average party vote is not significantly related to competition for nominations for the Southern sample is probably the result of this multicollinearity. When competition for nominations is operationalized by the per- centage margin of victory, the results for the South suggest that incumb- ency, party name, and registration procedures explain most of the variance. Once again, incumbents have a greater advantage of obtaining renomination and there is closer competition for Democratic as opposed to Republican nominations. However, when party registration is operationalized by the per- centage margin of victory, party registration becomes significant. In the South the principal difference in registration requirements is between those states which use challenge registration and those states which use more formal systems of closed registration. Consistent with Smith's argument there is more competition for those states with the more restrictive registration requirements than there is in the less restrictive states (Smith, 1973:83). The close association between incumbency, party name, and average party vote suggests that party Competition should be closely related to the margin of electoral victory in spite of the lack of statistical significance except for the multicollinear nature of this relationship. The problems of multicollinearity are particularly severe for the average party vote, incumbency, and party name in the South. A problem which makes it extremely difficult to say exactly which of these variables is more important in any given instance. 80 It is much easier to interpret the results for the rest of the country than it is for the South. The results for the nonsouth are very similar to those for the original equation when competition for nominations is Operationalized by the number of candidates receivingELSZ of the vote. However, outside of the South, there is less difference in the rate of competition for governor and senator. Also when the South is excluded, racial fragmentation becomes significantly related to greater competition for nominations, and run-off primaries become significantly related to more competition for nominations. The finding relating run-off primaries to greater competition for nominations is, however, not as interesting as it at first might seem because the find- ing is based only on cases drawn from a single state, Oklahoma. Con- sequently, the relationship should not be regarded as an overly impor- tant deviation. The results for the national sample and the nonsouthern sample are also similar when competition for nominations is measured by the difference between the vote for the first and second place candidates. Most of the relationships remain the same, however, there are a few interesting changes when the South is excluded from the sample. Racial fragmentation becomes a significant factor. Increasing racial diversity is associated with closer competition for nominations outside of the South, but evidently not for the Southern states themselves. In the South racial diversity drew the politically active elements together during the 19503 and 19603. Matthews and Prothro found a close rela- tionship between the number of Blacks in an area and the extent of racial disenfranchisement in the South during that period (Matthews and Prothro, 19662115). This suggests that the difference in the 81 influence of racial diversity in the North and the South may be the result of White politicians coalescing to continue the formal exclusion of Blacks from the electorate during the late 19503 and early 19603. Also, while many of the Southern states were racially diverse during this era, their electorates tended to be homogeneous because of the systematic exclusion of the Black voters. The removal of the South from the sample changes some of the relationships between party organization and percentage margin of victory. The margin of victory becomes significant in this instance and increases as entry into the party organization becomes more restricted. This may be the result of removing some of the least develOped party organizations from the analysis when the South is dropped. The lack of a correspondence between formal rules and actual practices may have been great enough in that region to confuse the analysis of the entire data set. The run-off primary does not appear to be as influential as Key hypothesized (Key, 1949:417). Run—off primaries are either signi- ficantly related to competition for nominations or are related in a contradictory manner to competition for nomination when the sample is split, i.e., characterized by less competition in one sample and by more competition in the other sample. This suggests that whatever the actual relationship, it is much weaker than originally suspected. To review, the separate analyses of the two regions indicates that competition for nominations is more predicatable in the South than the rest of the country, suggests that the relationship between the independent variables and competition for nominations is not signi- ficantly altered when the South is excluded from the analysis; suggests 82 that the relationship between competition and racial diversity is dif- ferent in the South from the rest of the country, and suggests that run— off primaries do not increase competition for nominations as much as was previously suspected. Finally, the high degree of multicollinearity for the Southern sample makes it difficult to adequately assess the impact of the independent variable for that subsample. Summary This chapter analyzed the data on competition for nominations by pooling all information about nominations made between 1954 and 1974 and treating it as a single data set. This procedure allowed for a more sensitive preliminary analysis of the individual relationships, but at the cost of increasing the possibility of autoregressive dis- turbances affecting the estimates and making it impossible to deal with the effects of temporal trends. The analysis demonstrates a relationship between competition for nominations and formal control over state nominations. A finding which suggests that at least some aspects of party organization affect competition for nominations. The findings also suggest that incumbency, office sought, and average party vote are of major importance in explain- ing competition for nominations, although no one variable plays a domi- nant role. Particular attention is paid to the relationship between average party vote and competition for nominations. Average party vote when estimated as a linear relationship does not play a dominant role in determining the amount of competition for nominations. However, the relationship between competition for nominations and average party vote may not be entirely linear. While alternate hypothesis 11 is rejected, 83 the relationship between competition for nominations and average party vote can be estimated by some form of a curvilinear model. The three curvilinear models all produce slightly better results than the linear model, but the differences are very small and can be explained as statistical artifacts. Consequently, because of the lack of either a clearly superior fit or a strong theoretical justification for the non- linear model, the linear model is cautiously accepted. Finally, there is a significant regional difference in the explanatory power of the general model. The model does a better job of explaining competition for nominations in the South than in the rest of the country which suggests that intra—party competition in the South, although greater than in other regions, is also much more standardized. CHAPTER IV AN ANALYSIS OF COMPETITION FOR NOMINATIONS BETWEEN 1954 AND 1974 Introduction The probability of autoregressive disturbances is greatly increased and many interesting trends are obscured by treating this data set as though all of the nominations occurred at a single point in time. In order to obtain greater sensitivity and summarize the relationship between the independent variables and competition for nominations, the fully operationalized equation was estimated for the entire data set in Chapter III. However, in this chapter the general equation is estimated separately for each of the eleven nominating periods that occur between 1954 and 1974. By estimating the equation separately for each time period the possibility of autoregressive disturbances is greatly diminished, and it becomes easier to discover any trends which may exist in the data. However, there is also a tradeoff involved in opting for a year by year analysis because the sensitivity of the analysis is reduced relative to that of the pooled data set. 84 85 Changes in the Individual Variables Over Time Before estimating the general equation for each of the nomina- ting periods, it is useful to look at how each of the individual variables change over time. This procedures can provide valuable infor- mation about changes in the party system which have occurred since the Second World War. This in turn aids in interpreting any changes which may occur in the significance level of these variables when the general equation is itself estimated over time. Competition for nominations increased slightly between 1954 and 1974. Figure 3 indicates that the number of candidates who receive at least 5% of the vote for nomination has been slowly increasing. Figure 3 also indicates that the percentage margin of victory, the percentage vote for the leading candidates, and the combined percentage vote for the first and second place candidates have all been slowly declining. While these changes suggest that there is slightly more internal competition for nominations, there has not been a radical change in the amount of competition for nominations during the post- war period. This suggests that if political parties have been losing their control over gubernatorial and senatorial nominations during the twentieth century, this loss has not been associated with any major change in the rate of competition for those nominations. With a few exceptions the independent variables also tend to be relatively stable during the post-war time period. All of the measures of party organization except for the method of selecting presidential delegates and the number of nominations made by run-off primary remain relatively stable when their average value is graphed over time (see Figure 3). 86 100% 90% W A 807. N7 707. combined Z vote for the first two candidates Z vote for first place candidate Z margin of victory 54 Figure 3A. {‘0 56 58 60 62 64 66 68 7O 72 74 The Average Value of Several Measures of the Dependent Variable Competition for Nominations over Time # of candidates receiving : 5% of the vote for nomination 54 Figure 3B. 56 58 60 62 64 66 68 7O 72 74 The Average Value of the Dependent Variable, Competition for Nominations, over Time 87 The average value of the variable measuring the degree of party control over the selection of presidential delegates declines sharply in 1972. This decline reflects the McGovern Commission reforms in presidential selection procedures. While the McGovern Commission dealt with the internal practices of the Democratic party, many of the changes that it inspired required the passage of new state laws which also changed the way in which the Republican party had to make its nominations. The McGovern Commission reforms constituted a major change in the system of selecting presidential delegates, but not in other facets of party organization. The results for the use of run-off primaries fluctuates over time. The pattern of fluctuation suggests that the rotating nature of the senatorial election calendar is influencing the results. An unusu- ally large number of run-off primaries were held in 1972. This difference could inflate the significance level of subsequently estimated coef4 ficients for run-off primaries. As a result care is required in inter- preting the coefficients when this variable is included in an equation. Although there is a very slight decline in the average oppor- tunity rate for governor and senator these variables remain relatively constant throughout the entire period (see Figure 5). The opportunity rate for governor is higher than that for senator and remains consis— tently higher until 1974 when there is a sharp decline in gubernatorial opportunity rate. The decline is consistent with changes lengthening gubernatorial terms and removing nonsuccession clauses which were made during the late sixties and early seventies. Opportunity rate for senators has also been slowly declining between 1954 and 1974, a find- ing which indicates a declining turnover rate in the senate and is 88 Registration 5 __ A A ,__;A 2.5 54 56 58 60 62 64 66 68 70 72 74 Control over State Nominations W 54 56 58 60 62 64 66 68 70 72 74 Figure 4. Average values for Measures of Party Organization for these States Making nominations by Year 89 Control over the Selection of Presidential Delegates 25 20 10 54 56 58 6O 62 64 66 68 70 72 74 The Number of Nominations Allowing for a Run-Off Primary made by Year 30 20 10 S4 56' 58 60 62 64 66 68 7O 72 74 Figure 4. Continued 90 10 54 56 58 6O 62 64 66 68 70 72 74 Figure 5A. Average Opportunity Rate for Governors Over Time 10 54 56 58 6O 62 64 66 68 70 72 74 Figure SB. Average Opportunity Rate for Senator Over Time 91 consistent with other observations made about the increasing dominance of incumbents in Congress. The measures of social diversity are not nearly as stable as those of the other independent variables (see Figure 6). State popu- lation, metropolitan concentration and racial diversity all increase during the post-war period. However, the gubernatorial electoral schedule may have changed sufficiently by 1972 to introduce some bias into the sample. The mean values for state population, racial frag- mentation and metrOpOlitan concentration fluctuate much more widely during the 19703 than they did during any of the previous time periods. The relative number of governors and senators running for nomination at any one time might also affect the results. In this data set, candidates for gubernatorial nomination were identified by a numerical code of one and senatorial candidates were identified by a numerical code of two for the variable called office. The year by year average of numerical value of the office variable is one way to measure the relative number of governors and senators running for nomi- nation. When the average value of the office variable is plotted over time the value is remarkably stable until 1977 when it begins to fluc- tuate very slightly in a two-year cycle (see Figure 7). This is the result of changes in gubernatorial term which begin to make a difference in the relative number of governors and senators seeking nomination in any given year by the late sixties. A fact which may complicate the analysis by causing some of the regression coefficients for social diversity to fluctuate for the 1970, 1972 and 1974 data sets. Although presidential year proved to be insignificant when estimated for the 92 40 35 30 25 20 50 54 56 58 60 62 64 66 68 70 72 74 Figure 6A. State Population by year for States Nominating GoVernors and Senators 25 20 15V— 10 54 56 58 60 62 64 66 68 70 72 74 Figure 6B. Racial Diversity by Year for States Nominating Governors and Senators 35 30 25 54 56 58 60 62 64 66 68 70 72 74 Figure 6C. Proportion of a State's Popualtion Living in the Largest SMSA for States Nominating Governors and Senators 93 2 1.5 — \/\/\ 1.0 54 56 58 6O 62 64 66 68 70 72 74 Figure 7. Average Value for Office Variable by Year, Gov = l; Sen = 2 94 entire period 1954 through 1974, the electoral schedule may only be beginning to influence results. Unfortunately the two nominating periods, 1968 and 1972, which are the most atypical cannot be inter- preted only in terms of a changing gubernatorial election schedule. There is no way to separate the effects of historical circumstances for these two periods from the effect of a changing electoral schedule. Several additional nomination periods will be required before the degree of bias caused by a changing electoral schedule can be appro- priately estimated. The average party vote does not fluctuate very much between 1954 and 1974 (see Figure 8). The floating average for Democratic candidates for governor and senator is slightly over 50% for the entire period. The floating average vote for the Republican candidates is slightly under 50% for the same period. However, the standard devia- tion for both Democratic and Republican candidates declines from 1954 to 1974 indicating narrowing of party competition for governor and senator. There is a much less variation in party competition on the state level in 1974 than there was in 1954, reflecting a decline in the number of one—party states. The number of incumbents running for renomination fluctuates only very slightly between 1954 and 1974 (see Figure 9). However, when the number of incumbents running for renomination is averaged by party several interesting differences appear. The rate at which incumbents seek renomination anticipates national political trends. Incumbents are more likely to seek renomination in those years when their party appears to have an exceptionally good chance of winning in the general election, and much less likely to seek renomination when it appears as 95 54 56 58 6O 62 64 66 68 70 72 74 Figure 8A. Average Party Vote by Year for States Making Gubernatorial and Senatorial Nominations 54 56 58 60 62 64 66 68 70 72 74 Figure 8B. Standard deviation for Party Vote by Year for States Nominating Governors and Senators 96 54 56 58 60 62 64 66 68 70 72 74 Figure 9A. Incumbency by year; 0 = no incumbent running, 1 = incumbent running i—‘NUJJ-‘UI 54 56 58 60 62 64 66 68 7O 72 74 Figure 9B. Incumbency by party by year; 0 = no incumbent running, 1 = incumbent running 97 though their party might suffer a serious defeat. The average number of incumbents seeking renomination peaks for the Democratic party in 1964 and 1974. Both the 1964 and 1974 elections appeared likely to be Democratic landslides several months prior to the actual election day. The Republican average peaks in 1956, 1962, 1966, and 1972. Although only 1972 appeared to be a landslide for the Republicans before the election, 1956, 1962 and 1966 also seemed to offer several advantages to potential Republican candidates. An extremely popular Republican president, Dwight Eisenhower, was running for reelection in 1956. The 1962 election followed the very narrow defeat of the Republican presi- dential candidate in 1960, and offered the opportunity for Republican candidates to run in an election that would not be affected by the presence of President Kennedy at the tOp of the Democratic ticket. The 1966 election offered the Republican candidates chance to run against a number of weakly entrenched Democratic opponents who had managed to win office during the Johnson landslide. Exceptionally bad political prospects also seem to lead to a reduction in the number of incumbents seeking renomination. The average number of Republican incumbents who sought renomination plummeted in 1964 before the Goldwater debacle, and the average number of Democratic incumbents seeking renomination fell in 1972 prior to the McGovern disaster. While candidates for nomination can never be certain that their party will be defeated in the general election, it is often pos- sible to estimate one's chances in light of current political circum- stances. Premonitions of coming electoral defeat probably reduce an individual's interest in seeking a nomination. There were very strong prior indications that the Republican party would lose badly in 1964 98 and that the Democratic party would suffer serious loses in 1972. Evidently a number of incumbent governors and senators anticipating a hard general election chose not to run for renomination. To summarize, both the dependent variable and most of the inde- pendent variables are relatively stable over time. However, decreasing party competition at the state level may be reducing the role of that variable as an explanation of competition for nominations during the late sixties and early seventies. When the number of incumbents seek- ing renomination was graphed over time by party, the number was found to anticipate national political trends. Fewer incumbents sought renomination prior to their party's major electoral defeats than did before more nomal election years. All of which suggests that politi- cians look at their short term chances of electoral success as well as at more long term trends. The General Equation Estimated Over Time The next step is to estimate the general equation separately for each of the eleven election periods. The unstandardized regression coefficients can then be traced over time to indicate any trends that might exist in the data. If an independent variable is actually re- lated to competition for nominations, the unstandardized coefficient should reveal a consistent and significant relationship. If the un— standardized coefficients are inconsistent and appear to randomly fluctuate, then there probably is not a significant relationship between the independent variable and competition for nominations. If the coefficients are consistent, but insignificant, then this suggests that the relationship is probably not very strong or much too indistinct to be picked up by these equations. 99 The equation is modified for this analysis by dropping those variables which were originally included to control for the affects of electoral cycles. The year by year analysis makes it impossible to include these variables in the equation because of the lack of variance. When the new equation is estimated for each year operationaliz- ing competition for nominations by number of candidates who received 5% or more of the nominating vote, the results differ in several ways from those obtained by pooling the entire data set (see Table 9). Control over state nominations has by far the greatest influence of any of the measures of party organization on competition for nominations. Centralized control over nominations is associated with fewer signifi- cant candidates for nomination. All but one of the regression coef- ficients are significant at least at the p 31.10 level. The finding is consistent with Wolfe's observation that there is less competition for nominations in those states where at least some of the major candidates are nominated by convention (Wolfe, 1966:84). It is also consistent with Leege's argument that the less important nominations serve as an important party-building resource (Leege, 1970:519). When the unstandardized regression coefficients for control over state nominations are plotted over time, they also reveal an increasingly influential relationship. While this relationship could be interpreted as the result of a greater willingness of state party organizations to use their resources to control major nominations, a more likely explanation focuses on the changing role of party competi- tion. The difference between the average Democratic vote for governor and senator and the average Republican vote for governor and senator declines during this period. As there is less and less variation in 100 OO. NO. OO. OO. NO. ON. .ONO OON.N OOO.N OON.N NOO.N OOO.N OOO.N O NON. NON. NON. NON. NNN. NON. NO Os. ONN.- Om. ON.- OO. OOO. OO. NOO. ON. OOO. NN. OOO. NuouOOO NN. mo.» OO. OOO. NO. NO. OO. NOO.- OO. NOO.- OO. NOO.- NOONONON: OONOOO>OOO NO. OOOO. ON. OOO.- OO. OON.: ON. NOO.I OO. OON.- Om. OOO.- OOOO ONOOO ON. ONN. Os. NON. OO. NNO.n Om. ON. NO. OON.. NN. OOO.| OONOONOONOOO ON. ONO. ON. NOO.- ON. OOO. NN. ONOO.a OO. ONO. OO. OONO.| .NOO .OONN No OONOOONOO NO. ONN.I OO. OON.- OO. NON.I OO. OOO.- OO. ONO.: OO. NNN.- OOONOOONaoz mumum um>o Houucoo Om. NN.- OO. ONO.I OO. NON.- NO. NOO.I OO. ONO. NO. NON. msmz OONOO NN. .NNO.: NN. ON.: Oq. OON.I OO. NN.I NO. NNN.- Om. NNN. M...ONOOO NO. ONO.- OO. OOO.- OO. ONO.- OO. OON.- OO. OOO.- OO. OOO.- OOOOOEOOON OO. ONO. NO. OONO. OO. OONO. NO. NONO. OO. OOOO. OO. OOOO. OOOO OONOO OOONO>< ON. NOOO.| ,OO. OOOO. NO. NO. Om. ONO. OO. OOOO.- OO. ONOO. .OOO OOOO OONOOONOOOO ON. NNO.I NO. _ OOO.- OO. OOOO. ON. ON.w NO. OO.| OO. ON.: .>OO OOOO OONOOONOOOO ON. OOOO.- ON. OOOO.- OO. OOOOO.- NO. ONOO. NO. OONO.u NO. OOO.- OOOO OOOONOO ON N ON. OOOOOO. NO. OOOOOO. ON. NOOOO. OO. OOOOOO. ON. NOOOO. OO. OOOOO. OONOONOOOO OOOOO OO. NNOO.s ON. OOOO. NO. No.1 NO. NNOO.- OO. ONO.| NN. OOO.: OONONO>NO NONOOO ONO O ONO O ONO O ONO O ONO O ONO O ONOONNOO OOON NOON OOON OOON OOON OOON In macaumcwaoz How muo> onu wo Nm Aco>Hmomm 0:3 mmumpfiwcmu mo NOLEDZ mnu co cowumsvm Hmuocoo ocu mo mdoammmuwmm oHaHuHDZ .m oHan 101 OO. OO. NO. OO. OO. .ONO ONO.N OO0.0 ONN.N OOO.N OOO.N O ON. OO. OO. ON. OO. NO ON.- ON. OO.- ON. NN.- OO. NNN. NO. OOO. OO. OOO. NNO-:sO OOO. OO. OOO. NO. ON.- OO. ONN.: OO. OON. NO. ON.: .NNO OONOOO>OOO OO.: OO. OOOO. NO. NNN.- NO. OON.: OO. OOO. ON. OON. OOOO NNOOO NO. ON. ON. ON. ONN. NO. NO0.0 NO. OON. OO. NON. OONOONOONOOO OO.: OO. OOO.- OO. ONO.. NO. OOO.: OO. NOO.- OO. ONO. .NOO .OONO Oo OOONOOONOO OO.: OO. OON.- NO. NON.- NN. ONN.- OO. OON.- NO. NON.- OOONOchNaoz wumum Hm>0 HOHUCOU OO.- OO. OOO.- NO. NOO. OO. ONO.- OO. OON.| ON. ON.- mamz NONOO OO.- ON. NON.- ON. OO.- NN. OON.- OO. ONN.- OO. ONN.- OONNNO ON.- OO. NO.- OO. OO.- OO. OOO.- OO. OO.: NO. ONO.- NucmOEOOON NN. ON. NNO. OO. OOO. OO. NNOO. OO. ONOO.- ON. NNI. muo> NONOO OOONO>< OO. OO. NNO. OO. ONOO.- ON. OOO.- OO. NNO. ON. NOOO.- .OmO OOOO NONOOONOOOO OO. OO. NNNO. OO. OOOOO.- OO. ONO.- OO. NOOO. OO. ONOO.- .>OO OOOO NONOOONOOOO NN. OO. NN.- OO. ONOOO. N. NOOO. ON. ONOO.- ON. OOOO.- OOzO OOOONON ON N NO. OOOOO. OO. ONOO.- OO. NOOOO.- NN. OOOOO. ON. NOOOO. .OOO ONOOO OO. ON. NN. OO. NOO.- ON. OOOO. OO. OOOO. NO.- OOO. NONON0>NO NONOOO Hmmw ONOONNO> ONO O ONO O ONO ONO O ONO O mNOONNOO NOO O ONON NNON ONON OOON OOON QQDCHHGOU . m 0H nah. 102 Incumbency 0 -.25 -.50 -.75 54 56 58 60 62 64 66 68 70 72 74 Average Party Vote .05 .05 .10 54 56 58 60 62 64 66 68 70 72 74 Office -54_”-_56‘-V-58_-r 60“. _62t 7‘64 1” 66 ‘68_ . 7O . 72MM 74 Figure 10. Selected Regression Coefficients for Variables From Table 9 Plotted Over Time for the # of Candidates and more than 5% of the vote for nomination. Control Over State 103 Nominations 0 -.l -.2 -.3 54 56 58 60 62 64 66 68 70 72 74 Registration .3 .2 .l 0 —.1 -.2 -.3 54 56 58 60 62 64 66 68 7O 72 74 Figure 10. Continued 104 Party Name .4 More competition for Republican .2 Nominations 0 -.2 More competition for Democratic Nominations -.4 -06 Racial Diversity .10 .05 -.05 -.10 Figure 10. Continued 105 average party vote competition its importance as an explanation of competition for nominations will decline causing other variables to increase in importance. The other measures of party organization; convention hierarchy, run-off primaries, registration requirements, control over the selec- tion of presidential delegates, and entry base do not appear to be related to competition for nominations. The coefficients for the degree of convention hierarchy tend to be statistically insignifi- cant and change direction several times between 1954 and 1974. Although the use of run-off primaries is associated in seven of eleven instances with an increase in the number of significant candidates running for nomination, none of those coefficients is statistically significant. Increasingly restrictive registration requirements are asso- ciated in eight of eleven cases, with an increase in the number of candidates for nomination who receive 5% or more of the vote, but the results are statistically significant in only one case. The direction of the relationship provides some additional support for Smith's earlier observation that competition for gubernatorial nominations increases when registration requirements become more severe (Smith, 1973:83). However, the lack of statistical significance when the equation is estimated over time, but not when it is estimated for the pooled data set, support Ranney's contention that registration requirements have relatively little influence on competition for nominations (Ranney, 1974:76). If the difference in the influence of registration proce- dures is relatively small, then the switch from a pooled analysis with over 1100 cases to a year by year analysis with about 100 cases each, 106 might very well result in the significance levels changing from signi- ficant to insignificant. Taken together the evidence suggests that if there is any relationship between competition for nominations and registration requirements that it is at best not very strong. There is little relationship between the method of selecting presidential delegates and competition for gubernatorial and senatorial nominations. Although seven of the eleven coefficients indicate that the number of significant candidates for nomination increases as party control over the selection of presidential delegates becomes more cen- tralized, only two of those coefficients are even weakly significant (p:.10). A The relationship between the number of candidates receiving 5% or more of the vote and the requirements for joining the professional party organization is also relatively weak. In seven of the eleven years the coefficients suggest that the more restrictive the require- ments for joining the statutory party organization, the fewer the number of candidates for gubernatorial and senatorial nominations, but only three of those coefficients are significant. Although there is less fluctuation in the value of these coefficients, the six rela- tionships are all relatively small. The measures of social diversity, state population, racial di- versity and the proportion of a state's population living in the larg- est metropolitan area are not strongly related to competition for nominations. The coefficients for state population in over half the cases are associated with greater competition, but in only one instance, 1968, are any of the coefficients significant. The size of a state seems to have little influence on the number of candidates 107 running for nomination. Smith's argument that greater campaign costs will result in less competition for nominations in the more populous states is not confirmed by these findings which, in fact, suggest the Opposite relationship (Smith, 1973:54). Madison's contention that the larger states will be characterized by greater competition for power is supported by the direction of the coefficients, but not by their significance levels (Madison, 1789:22). Racial diversity is associated with more competition for nom- inations after, but not before 1964. Although the finding suggests that non-whites are playing an increasingly important role in internal party politics in the United States, the coefficients are generally insignificant. The degree to which a state's population is concentrated in a single metropolitan area is not strongly related to the number of can- didates. Although eight of the eleven coefficients suggest that increasing metropolitan concentration is associated with a reduction in the number of candidates, the relationship is significant at the g) 1.10 level in only two of the cases. Consequently, the relationship, if it exists at all, is probably not a very strong one. Party name is generally not related to the number of candidates running for nomination who receive at least 5% of the vote. However, there are several interesting deviations from this general pattern which correspond with major political crises in post-war American poli- tics. This finding differs from the relationship that was discovered between party name and competition for nominations when the equation was estimated for the entire data set. It also contradicts Morehouse's argument that there will be little association between party name and 108 the rate of competition for nominations when the average party vote is controlled for (Morehouse, 1976:214). Candidates for nomination seem to be anticipating major elec- toral shifts. The direction of the unstandardized regression coeffici- ents for party name measures the relative amount of competition for Republican and Democratic nominations. Although the results indicate slightly more competition for Democratic nominations, the coefficients are significant only in 1958, 1970 and 1974. The 1958 election was a major defeat for the Republican party in Congress, while the Watergate crisis combined with a generally bad economy hurt Republican candidates in the 1974 election. In only one instance, prior to the Republican landslide year of 1972, was there more competition for Republican nomi- nations than for Democratic nominations. The 1964 election is the one case where the amount Of competition for Republican nominations does not seem to be changing in response to an anticipated electoral defeat. There should have been even fewer candidates seeking Republican nomina- tion in 1964 considering how poor the prospects were for success in the general election when gubernatorial and senatorial candidates were being nominated. Either the Republican candidates did not correctly estimate the potential for electoral disaster or they chose to ignore any suggestions that 1964 would be a Democratic year. Incumbency is the most consistently significant variable in the analysis. The presence of an incumbent is association with a reduction in the number of candidates running for nomination who receive 5% or more Of the vote. In all eleven years the coefficients for incumbency are significant at least at the p_:.lO level. The data also indicates that the influence of incumbency on competition for nominations increased 109 between 1954 and 1974. A finding which is consistent with studies of Congress which indicate a general increase in the ability of incumbents to Obtain reelection (Ballock, 1972 and Fiorina et al., 1975). This increased ability of incumbents to discourage competitors does not seem to be associated with a decline in the influence of party organization. On the contrary, the most significant measure of party organization con— trol over state nominations also increases in influence during the same period. The increasing influence of incumbency may also be related to decreasing party competition at the gubernatorial and senatorial levels. This supposition is partially supported by the relationship between average party vote and competition for nominations. Increasing average party vote is associated with an increase in the number of signi- ficant candiates for nomination. Four of the six coefficients measuring the relationship of average party vote to competition for nominations between 1954 and 1964 are significant, while only one of the coefficients measuring the relationship after 1964 is significant. An interesting feature Of the relationship between average party vote and competition for nominations is the greater values for the coefficients in 1964 and 1972. These deviations correspond with the two post-war presidential landslides. This suggests that candidates are much less willing to take unusual risks during times of political turmoil. The possibility of a major presidential landslide evidently vastly increases the salience Of relative party position for potential candidates. The relationship between opportunity rate and competition for nominations is only slightly clarified by the year by year analysis. Senatorial opportunity rate is not related to competition for nomina- tions. The coefficients fluctuate over time and none of them are llO significant. Gubernatorial opportunity rate does not appear to be a major factor influencing competition for nominations. Although increas- ing Opportunity rate for governor is associated with a decline in compe— tition for nominations, only two of the coefficients are statistically significant. Office, the final variable, is also only weakly related to com- petition for nominations. More candidates running for gubernatorial nomination receive at least 5% Of the vote than do those running for senatorial nomination, but none of the coefficients are significant. The results, at best, indicate a weak relationship between Office sought and competition for nominations. The next major step in the analysis is to estimate the general equation over time using the percentage margin of victory instead of the number of candidates who receive over 5% of the nominating vote to measure competition for nominations (see Table 10). When competition for nominations is measured by the percentage margin of victory control over state nominations remains the most important measure of party organiza- tion. Increasing party control over state nominations is associated with a larger percentage margin of victory for the winning candidate in a nomination contest. However, the number of significant coefficients is much smaller when the competition is measured by the percentage mar- gin of victory instead of by the number of serious candidates. Again this suggests that party organizations are probably more effective at disuad- ing minor candidates from contesting a nomination, than they are at delivering the nomination. The other measures of party organization; convention hierarchy, run-off primaries, registration, method of selecting presidential delegates lll OO. OO. OO. OO. NO. NN. .ONO NMo.N NNM.N OON.N NMm.H mMo.N OmM.H m NM. NN. MN. ON. NN. om. mm MN. NNo.O MN. Oom.ON oo. MMM.OI MM. NNM.N mm. MMo.I MM. oNo.MI wwOIasm oN. NNN. MM. MMM.NI MM. MNM.NI MM. oNN.N MN. wo.m NN. MN.O NOONONONm cowuco>cou Nm. NNN.I OM. MNM.M NM. MNN.~ Mm. Nmo.l ON. O~.N MM. NN.NI Ommm Nuusm No. MON.~I OM. Nmo.OI MM. ONO.OI om. moN.MI Mo. m.MI NM. Noo.NI OONumNumNmom MM. oNM.I MN. OOH. we. NON.I mo. NON. NN. NMM.I MN. omM.I .NOQ .mmum «0 :ONOOONOM MN. NNM.M NM. MNO. mo. Ncm.M MN. MMN. ON. NN.I ON. M0.0 OOONNOOOEOZ mumum NO>O Houucoo NN. OOO.oN NM. MMN.OI mm. owe. oo. MNO.NN MN. NMM.N ON. NN.N OEmz Nuuwm NN. ONo.NN Oo. OMM.MN MN. mNm.oN mo. ONo.M om. NN.oN NN. OM.oN OONNNO No. MOo.MM oo. ch.mN oo. NNN.NM oo. NNo.NN oo. OM.MN oo. mm.MN OocmnesocH No. MNN.I oo. NON.I Mo. mNM.I No. ON©.I NN. MN~.I MM. MMo.I ONO> Ouumm ommuo>< NM. NCO. No. Omo.l mo. Mom. MN. QMN.I MO. MOM.N mm. mMN.I .cmm mama NuNcsuuoaao ON. NMm. NN. New. ON. MoN.N OO. OOM. No. MON.N mm. mwo.| .>oo oumm NuNcsuuoamo oo. MMN. NN. mNo. oo. Mmo. MO. NNO.! Oo. OOO. NO. NNN. ¢mzm ummwumq ON N mo. NNOO. MM. Noo. OO. mooo.l MN. MNoo. om. Oooo.n OM. Noo.| :ONumasaom Oumum MO. COM. oo. MON.I MN. coo. NO. com. mo. OHM. NO. NNM.| NuNmNO>NQ Hmwomm ONO. O ONO O ON O ONO O ONO O ONO O mNOONNO> OomN NOON OOMN meN MMmN OMON NNONOO> mo :qumz meucmoumm Ocu co :ONumsdm Nmumcoo msu mo :OHmmONmmm ONQNNHOE .OH OHan 112 OO. OO. OO. ON. NO. .ONO NO0.0 NON.O NOO.N ONO.N NNO.N O NO. OO. NO. NN. OO. NO OO. OO. OOO.O ON. ON.ON OO. ONN.O: NO. OOO.ON OO. NOO.O NOO-OOO NO.- NN. ONN.N- OO. NO.O NO. ONO.N NN. ONN.O- NN. OOO.N NOUNONONO OONOOO>OOO NO. OO. NNN.N ON. OO.O NO. OOO. OO. NO.O- NN. OO.O- OOOO NOOOO NN.- ON. NOO.OI ON. O0.0- ON. OO0.0 ON. ON.N- NO. OON.- OONOONOONOOO OO. NO. OOO. ON. OOO.- NO. NNO. NN. ONN. NO. ONO.- .NOO .OOOO No OONOUONOO NO. OO. NOO.N NO. NN.N OO. OOO.- NN. NO.O NO. OON.O OOONOOONOOZ Oumum NO>O Nouucoo NO. OO. OOO.NN NO. NON.O ON. ONO.NN NO. NOO.O OO. O0.0 OOOz NOOOO NN. NO. NOO.ON OO. OON.O NN. NOO.O ON. OO.NN OO. OO.ON OONOOO ON. OO. ONO.NO OO. O.NO OO. OOO.ON OO. OON.ON OO. NOO.ON NOOOOOOOON OO.- OO. ONO.- NO. NO.N- NO. NOO.- NO. OO.- OO. NOO.- Ono> NOOOO OOONO>O ON.- ON. NNO.N: OO. OO. NO. OOO. ON. OOO.- OO. NON.- .OOO OOOO NONOOONOOOO ON.- OO. ONN. NO. ONNO. OO. OOO.- ON. OOO.N OO. OOO.- .>oO OOOO NONOOONOOOO ON. OO. OOO. NO. ONN. NN. ONO. NO. OOO. OO. ONO. OOOO OOOONON ON N OO. OO. OOOOO. OO. NOOO. OO. NNOOO. NN. ONOO.- ON. ONOO.- OONOONOOOO OOOOO OO.- OO. NON.- NO. OON.N- OO. NOO.- OO. NOO.- OO. NNN.: NONONO>NO NONOOO HMO? mNOONOOO ONO O ONO O ONO O ON O ONO O ONOONOO> NON O ONON NNON ONON OOON OOON OOOONOOOO .ON ONOON 113 Incumbency 30 20 10 54 56 58 60 62 64 66 68 70 72 74 Average Party Vote 54 56 58 6O 62 64 66 68 70 72 74 Office 15 10 5 0 54 56 - 58 6O 62 64 66 68 70 72 74 Figure 11. Regression Coefficients for Selected Variables From Table 10 Plotted Over time for the % Margin of Victory. 114 Party Name 10 More Democratic Competition 5 0 _5 More Republican Competition -0 54 56 58 6O 62 64 62 68 7O 72 74 'Racial Diversity 1.0 54 56 58 60 62 64 66 68 70 72 74 Registration 5 0 -5 10 54 56 58 60 62 64 66 68 70 72 74 Figure 11. Continued Control over State Nominations 10 -10 115 54 Figure 11. 56 58 Continued 60 62 64 66 68 70 72 74 116 and entry base are not closely associated with competition for nomina— tions when measured by the percentage margin Of victory. The degree of convention hierarchy is not consistently related to the margin of victory, although the only two significant coefficients, those for 1970 and 1972, suggest a weak relationship between greater hierarchical organization and an increasing margin of victory. The effect of run-Off primaries on competition for nominations is much harder to estimate. All of the regression coefficients are insigni- ficant and there is a great deal of fluctuation, but the majority of the coefficients indicate a relationship between an increasing margin of victOry and the use Of run—off primaries. This clearly contradicts Key's competition for nominations (Key, 1949:19). However, the lack of signi- ficance is so total that it would be foolish to go to the opposite extreme and argue that the positive correlation is therefore a theoreti- cally significant relationship. The variable measuring registration requirements is associated with a declining margin of victory in ten of the eleven years, but is significant in only one of them. Smith's contention that competition for nominations increases as registration requirements become more restrictive is supported by these results as well as by those in Table 9, but once again the relationship is largely insignificant (Smith, 1973). The lack of significance suggests that whatever the relationship it is not a very strong one if it exists at all. The degree of control a party has over the selection of presiden- tial delegates does not seem to be related to the margin of victory. The coefficients are inconsistent and only two Of them, 1966 and 1974, are significant. These results further reinforce the notion that there is as great deal of separation between national and state level politics. 117 Entry base does not appear to be related in any way to competi- tion for nominations. The coefficients are both inconsistent and entirely insignificant. Making the requirements for joining the statutory party organization more stringent probably will not affect competition for nomi— nations. The measures Of social diversity; state population, racial diver- sity and proportion of a state's population living in its largest metro- politan area are somewhat more closely related to the percentage margin of victory than to the number of candidates receiving 5% or more Of the vote. State population is not related to the percentage margin of vic- tory. The coefficients fluctuate throughout the period and in only one year, 1964, is one Of the coefficients even weakly significant. Racial diversity is an interesting case. Racial diversity appears to be related to a decreasing margin of victory after 1974, but not before. This relationship may actually reflect increasing political participation of racial minorities in American politics during the late twentieth century. However, since only three of the coefficients are significant and there is a tendency for the coefficients to cancel each other out, this relationship would not appear as significant when the entire time period is analyzed as a single data set. Increasing metropolitan domination is associated with a greater percentage margin of victory for the winning candidate. Ten of the eleven coefficients fall in the appropriate direction and four of them are significant at at least the p:_.10 level. While metrOpOlitan con- centration does not seem to prevent the emergence of Opposition candi- dates (see Table 10), the existence of a single dominant urban center does seem to reinforce the position of the winning candidate. A 118 dominant city provides a target for those running for nomination with the Democrats concentrating their attention on the city prOper, while the Republicans fight it out in the suburbs. The candidate who has the most appeal to the electorate in the dominant city in a state has a tremendous advantage in winning nominations. In some areas such as Cook County, the remnants of an urban machine are a centralizing force in party affairs. In other places the existence of a single dominant urban center leads to the centralization of the communications network within the state. When a state's news media is concentrated, it is harder for less favored candidates to appeal to the electorate because information is channeled through only one or two major sources. The relationship between party name and the percentage margin of victory is affected by political trends. Although there is generally slightly less competition for Republican nominations than there is for Democratic nominations, the difference is exaggerated prior to major political landslides. There is much less competition for Republican nominations in 1958, 1964, 1970 and 1974. The 1958, 1964 and 1974 elections were particularly difficult elections for Republican candidates. While there is less competition for Democratic nominations as compared to Republican nominations in 1962, the coefficients also drop in 1954, 1956, 1968 and 1972, all years in which the Democrats either did not control the White House, appeared likely to lose a presidential election or seemed to be about to suffer a major defeat in the Congressional elections. This suggests that while most of the time party name does not ‘have much influence on competition for nominations, the possibility 119 of a major political landslide can vastly increase the influence of this variable. Incumbency is once again the most consistently significant vari- able. All of the coefficients indicate that the incumbent has a major advantage in seeking renomination. This advantage on the average amounts to about 25 percentage points over the nearest rival. The renomination advantage of incumbents does not appear to change significantly over time when we look at the margin of victory, although it does when the general equation is estimated for the number of candidates who receive 5% or more of the nominating vote. Evidently the increasing advantage of incumbents seems to consist mostly of the ability to discourage relatively light weight opponents. Party competition is also related to the percentage margin of victory. All of the coefficients indicate that the margin of victory decreases as a party's average vote increases. All of the coefficients fall in the hypothesized direction, and seven of the eleven coefficients are significant at the p3_.10 level. This is a much stronger relation- ship than exists when party competition is correlated with the number of candidates with 15% of the nominating vote. Office continues to be related to competition for nominations. The percentage margin of victory is much higher for senators than it is for governors. Senators have about a 10 percentage point advantage in seeking renomination. The stability of the coefficients during this period suggests that the relative noncompetitiveness of senatorial as opposed to gubernatorial nominations has not changed during the latter half of the twentieth century. This is consistent with Schlesinger's observation that senatorial Opportunity structure is much more stable 120 than gubernatorial opportunity structure (Schlesinger, 1966:196). The findings also suggest that structural and not policy difference accounts for the greater gubernatorial vulnerability. The final set of independent variables, those measuring oppor- tunity rate, continue to present contradictory and confusing results. The percentage margin of victory decreases slightly as the opportunity rate for governor increases. However, the coefficients for gubernatorial opportunity rate are not entirely consistent and only one is significant. Senatorial Opportunity rate is unrelated to the percentage margin of victory. The coefficients for senatorial opportunity rate are highly consistent and almost entirely insignificant. The results for opportunity rate are among the most difficult to interpret because of this split. Although caution is required, the evidence suggests that, at best, oppor- tunity rate as measured here is only weakly related to competition for nominations. Summary This chapter analyzed competition for nominations between 1954 and 1974. When the general equation is estimated over time, incumbency, average party voteanuicontrol over state nominations stand out as major explanations of competition for nominations. Party name also becomes important during landslide years although it is not an important variable during more normal times. The explanatory power of the equation varies considerably over time (r2 from .20 to .55). The occasional significance of party name explains some, but not all of the variation. The relatively modest squared multiple correlation coefficients for both the measures of 121 competition for nominations suggest that the behavior is relatively unstructured for both measures. However the slightly better R2 for the number of candidates receiving over 5% of the vote for nominations sug- gests that the major influence of the independent variables lies in reducing the number of candidates for nominations rather than in insur- ing victory in that struggle. CHAPTER V THE REDUCED MODEL Introduction The last stage of this analysis involves the simplification of the general model. It is necessary to simplify the general equation because of a large number of insignificant variables. Their inclusion makes it difficult to comprehend the results, reduces the significance of marginally associated variables and may even change the direction of some of the relationships if the independent variables are closely inter-related. Consequently, it is necessary to drop some of the insignificant variables from the analysis in order to increase our understanding of competition for nominations. How should the variables be selected for exclusion? While it is possible to exclude all of the insignificant variables as a group from the equation this could create problems in those cases where a variable is only marginally insignificant. Such marginal variables are less likely to be insignificant in a reduced equation than they are in an over-specified equation. This means that if all of the insignificant variables are dropped at once, some of the excluded variables may actually be significant in more reduced versions of the equation. What is required is some way to insure that all of the potentially signifi— cant variables will be included in the equation. 122 123 Stepwise Regression Step-wise regression is one way to proceed. Step-wise proce- dures provide a method for ordering variables in an equation according to their relative influence on the dependent variable. The independent variable which explains the greatest amount of variance is entered first into the equation, then the independent variable which explains the greatest amount of the remaining variance is entered next and so on until all of the independent variables have been included in the equation (Nie et. al., 1975:345). This produces a rank ordering of all of the independent variables which can be used to provide a more sys- tematic statistical justification for excluding insignificant variables from the equation while allowing us to observe the consequences of that exclusion on the equation at each step of the process. Step-wise regression has recently been criticised because of its tendency to discriminate against those regressors entered last into the equation (Wonnacott and Wonnacott, 1970:312). In practice, this means that in those cases where two independent varaibles are related to each other, the procedure will inflate the value of the first variable entered into the equation at the expense of the second. Consequently, caution must be used in interpreting the results of a step-wise regression. Step— wise regression should only be used in conjunction with other techniques to provide additional information about the stability Of a relationship, but never as the sole method for estimating a relationship. When step-wise regression is used to estimate the general equation the results are slightly different for the two measures of competition for nominations (see Tables 11 and 12). When competition for nominations is opeartionalized by the number of candidatesvfluareceive:>5% of the 124 Table 11. Stepwise Regression Relating the General Model to the Number of Candidates Who Receive3;5% of the Vote for Nomination (listed in order of decreasing influence for variables significant at the p:.05 level) Unstandardized Standardized Regression Regression Variables Coefficient Coefficient Sig. Incumbency -.670 -.28 .00 Party Competition .013 .16 .00 Control over State Nominations -.118 -.16 .00 Opprotunity Rate for Gov. -.228 —.13 .00 Office Sought -.220 —.10 .00 Party Name -.l96 -.09 .00 Run-Off Primaries .206 .07 .01 Registration .064 .05 .05 Constant = 2.643 F = 30.44 R2 = 17% Sig = .00 125 Table 12. Stepwise Regression Relating the General Model to the Per- centage Margin of Victory (listed in order of decreasing influence for Variables Significant at the p 3305 level) Unstandardized Standardized Variables Coefficient Coefficient Sig. Incumbency 25.811 .32 .00 Party Competition -.452 -.17 .00 Office Sought 10.84 .14 .00 Control over State Nominations 3.04 .12 .00 Opportunity Rate for Gov. .55 .09 .00 Proportion living in the Largest Metropolitan Area of State .172 .10 .00 Party Name 6.158 .08 .00 State Population -.0007 -.07 .02 Constant = 29.295 F R = 18% Sig. = 32. 175 .00 Re th 31': 126 nominating vote the results indicate that only eight of the fifteen variables eXplain a significant amount of the variance (p 3;.05). Incumbency, party competition, control over statewide nominations, gubernatorial opportunity rate, office sought, party name, run-off primaries and registration requiremetns are all significantly related to the number of candidates running for nomination. Entry base, popula- tion living in the largest SMSA, state population, racial diversity, method of selecting presidential delegates, convention hierarchy and opportunity rate for senators are not related to the number of candi- dates running for nomination. Incumbency is the most important single factor in the anslysis. The presence of an incumbent is associated with a decline in the number of candidates receiving at least 5% of the vote for nomination. Party competition, control over state nominations and guber— natorial opportunity rate also play important explanatory roles. Increasing party competition is associated with an increase in the number of serious candidates. Increasing party control over state nominations and increasing gubernatorial Opportunity rate are associated with a decline in the number of serious candidates for nomination. Office sought, party name, run-off primaries and registration requirements have a relatively modest influence on the number of candi- dates when the relationship is estimated using step-wise regression. There are fewer candidates for senatorial nomination than there are for gubernatorial nominations and there are fewer candidates for Republican nominations than there are for Democratic nominations. Also, the use of run-off primaries and more controlled registration procedures are associated with an increasing number of candidates for nominations. 127 In short, all of the eight significant coefficients in the stepwise equation continue to be related to competition for nomination in the same way as in the original model when the equations are esti— mated for the entire data set using the number of candidates who re- ceive Z_5% of the vote as the dependent variable. This finding sug- gests that the relationships between the significant independent varia- bles and the dependent variable are relatively stable even when the composition of the variables included in the equation is changed by eliminating the insignificant variables. When step-wise regression is used to estimate the general equation using the percentage margin of victory to Operationalize com— petition for nominations the results are slightly different than when the number of candidates receiving.:5% of the vote was used. Although there is some overlap, only political factors have a significant influence on the number of candidates running for nomination, while some sociological factors have an influence on the percentage margin of victory. These findings suggest that party organization and political circumstances are more important for explaining who runs than for de- termining how close the election will be. On the other hand, sociolog- ical factors seem to be more important for explaining the closeness of an election. Many of the studies which focus on inter-party competition have looked at social factors as an explanation of inter-party compe- tition, intra-party competition as well may be related to the same social diversity. If social groups are the building blocks of elec- toral coalitions, then, greater social diversity may well lead to greater internal as well as external competition. Once a candidate decides that conditions are right to contest a primary, the kind of 128 coalition he can build seems to be related more to the existing social conditions than to political and organizational factors. In a socially diverse society there may be a natural tendency towards political di- vision which is relatively easy to awaken. Ineaheterogeneous society conflict over leadership positions easily becomes group conflict, while in a homogeneous society coalition formation must proceed at a slower pace because there are no antagonisms to exploit. In a heterogeneous society coalitions can be built by adding groups of individuals to- gether. In a homogeneous society coalitions must be built by adding single individuals together. The step-wise estimates of the relationship between the per- centageumrgin of victory and the general equation (see Table 12) are also similar to the original estimates. Incumbency, party competition, office sought, control over state nominations, opportunity rate fOr governor, metropolitan domination, party name and state population are significantly related to competition for nominations. Registration, degree of convention hierarchy, use Of run—off primaries, entry base, degree of control over presidential delegates, the opportunity rate for senators and all of the variables designed to measure electoral cycles are insignificant. Again, the presence of an incumbent is the most important single variable. The margin of victory is also greater for senators, for majority parties, for Republican nominations, when there are more Opportunities to run for governor, and when a single city domin— ates a state. Finally, the greater a state's population the smaller the margin of electoral victory. State population has a significant influence when step-wise regression is used to reduce the total number 129 of variables in the equation, but is not significant in the original equation. Evidently reducing the number of variables increases the degrees of freedom enough to produce a significant relationship between state population and competition for nominations. When step-wise regression is used to estimate the general equation for the entire data set all of the significant independent variables continue to be related to competition for nomination in the same way they were before this procedure was used. The relationship between state population and competition for nomination is the only relationship which changes as a result of using step—wise regression, but this change involves only a small increase in significance level. The direction of the relationships remains stable. The relationship between competition for nominations and the independent variables was not greatly affected by changing the number of variables in the equa— tion. These results replicate the original findings. The Reduced Model Estimated Over Time Next, the reduced equation is estimated over time to see if the relationships are both consistent and significant in a number of different settings. While step—wise regression was used to reduce the equation, the procedure is not used to estimate the new reduced equation over time because of the potential for biasing some of the results. The eight variables which were significant at at least the p 3_.05 level were used to create new simplified eight variable models. These reduced equations were then estimated using normal least squares procedures for each election year. When the reduced equations are estimated for each nominating period (see Tables 13 and 14) four variables, incumbency, average 130 00. No. No. 00. MN. Oo. .me NN0.0 O0.0 NO.N ON0.0 NOO.N ONO.N O NON NON NON NON NNN NON NO ON. ON. NO. NO. OO. OO.- ON. ON. OO. ON.. NN. OO.- OONOONOONOOO NO. ON.- OO. ONO.: OO. ON.- NO. OO.- OO. NO.- OO. ON. OOOz NOOOO OO. NN.- NN. OOO.- ON. OO. ON. NN. OO. NN. ON. NO. NNOONNO ONouOOO ON. NO.- OO. OO.: NO. NO.. ON. NO.- NO. OO.- OO. OO.- NOONO>OO Now mumm Nuwcsuuoaao OO. NN.- OO. OO.- ON. NN.- NO. OO.- OO. OO.- OO. ON.- OOONNOONOOO mumum Hw>0 HOHuGOU ON. NO.: ON. ON.- OO. NN.- OO. ON.- NN. OO.- ON. OO.: 8NNNO OO. ONO. NO. ONO. OO. ONO. NO. ONO. ON. NOO. NO. NOO. mso> NONOO OOONO>< OO. NN.- NN. NO.- OO. OO.- OO. ON.- ON. OO.- OO. OO.- NOOOOOOOON ONO O ONO O ONO O ONO O ONO O ONO O . ONOONNO> OOON NOON OOON OOON OOON OOON I. OOONNOOOEOZ Now OOOOONOOBOU ONNNOOONNONOQO ou OONuchaoz NOw Ouo> mnu mo NMAchN>NOOOO mmumwwwcmo mo NOQEOZ ecu mafia: OaNH um>o wmumeNumm Homo: OOOOOOO OSH .MN ONOOH 131 ONO.O NOO.O ONO.N NNO.N OON.O O OO. OO. NO. OO. OO. .ONO NNN NNO NNN .NON NON NO OO. NO. OO. OO. OO. ON. NN. OO. OO. OO. OONOONNONOOO OO. OO.- NO. OOO. NN. OO.- OO. ON.- ON. ON.- msmz OONOO OO. NO. OO. OO. OO. OO. NO. ON. OO. ON.- NNOONNO NOO-OOO OO. OO.- OO. NON.s NO. NO.- NO. OOO.- NO. NO.- NOONO>OO NON mumm huwcsuuoaao OO. NN.- NO. ON.- ON. OO.- NO. NN.- OO. NN.- OOONNOONOOO mumUm HON/O HOHuGOU ON. ON.. ON. OO.- NO. ON.- ON. ON.- NO. NO.- OONNNO ON. NNO. OO. OO. NO. NOO. NO. OO0.0 OO. OOO. ONOO NOOOO OOONO>O OO. NOO.- OO. OO.- OO. OO.- NO. OO.- NO. OO.- OOOOOOOOON ONO O ONO O ONO O ONO O ONO O ONOONNO> ONON NNON ONON OOON OOON OOOONOOOO .ON ONOON 132 NmN.M Mo.M mMM.N Mwm.~ NMN.N NNN.N m OO. OO. NO. OO. OO. ON. .ONO NNO NON NON NNN N0.0N NON NO NN. NOO.- NO. NOO. OO. NOO.- NO. NOO. ON. NOO. OO. OOO.- OONOONOOOO OOOOO NN. OO.NN NO. ON.O: OO. NO. ON. NN.NN NO. OO.N OO. ON.O OOOz NONOO OO. ON. NO. OO. OO. NN. OO. ON.- ON. NN. ON. NN. OOOO OOOONON _, ON OON>NN N ON. ON. NO. OO. OO. OO. NO. NOO. OO. OO.N OO. NO. NOOOO>OO NON Oumm NuNcsuNomao ON. ON.O OO. OO. ON. O0.0 NO. ONO.N OO. ON.N NN. ON.O OOONOOONOoz mumum uo>o Houucoo NN. O0.0N NO. N0.0N NN. OO.NN OO. ON.O ON. ON.O ON. ON.ON 8NNOO OO. ON.- NO. OO.- OO. OO.: NO. OO.: OO. NN.- OO. NN.- muo> NONOO OOmuo>O OO. 0.00 OO. O0.0N OO. OO.NO OO. O0.00 NO. ON.ON OO. ON.NN NOOOOOOOON ONO O ONO O ONO O ONO O ONO O ONO O . m.NOONNOO OOON NOON OOON OOON OOON OOON OOONNOGNEOZ Now OONNNOOOEOU ONNHOOONNONOOO ou moumancmu oomNm wsooom mam umNNm onu pom Ouo> onu amoauon mocmuowwfin meucooumm mnu MONO: OeNH um>o woumaNumm Homo: wooovom 058 .ON OHOOH 133 NN0.0 NO0.0 ONO.N OOO.N NNO.O O OO. OO. NO. OO.. OO. .ONO NON NNO NON NNN NNN NO ON. NOO.- ON. NOO.- ON. NOO.- ON. NOO.- OO. NOO.- OONOONOOOO OOOOO OO. OO.NN OO. NN.O ON. O0.0 NO. O0.0 OO. ON.O OOOz ONNOO OO. OO. ON. OO. ON. ON. OO. OO. OO. ON. OOOO OOOONON ON OON>NN N OO. OO. NN. OO. NO. OO.: OO. OO. OO. OO.- NOONO>OO How Oumm Nansuuomao OO. O0.0 NO. N0.0 OO. NN.NI ON. 0.0 NO. O0.0 OOONNOONOOO Oumum uo>o Houucoo OO. N.NN NO. ON.O ON. O0.0 ON. O0.0N NO. O0.0N 8NNNO OO. OO.- NO. OO.: ON. OO.: NN. ON. OO. OO.- OOOO NNNOO OOONO>O OO. O0.00 OO. O0.00 OO. NO.ON NO. N0.0N OO. O0.0N. OuOOOOOoON ONO O ON O ONO O ON O ONO O ONOONNOO ONON NNON ONON OOON OOON OOOONOOOO .ON ONOON 134 party vote, gubernatorial opportunity rate, and party control over state level nominations, have a consistent influence on competition for nominations. While all of the variables in the reduced equation have a significant influence on the combined data set, the year by year analysis produces different results because of the great reduction in sample size (i.e., from 1100 to approximately 100 each). When the num- ber of observations is radically reduced only the most important rela- tionships continue to be significant. While this procedure reduces the sensitivity of the anslysis it also greatly reduces the possi— bility of autoregressive disturbances and allows us to look more closely at political trends. The results when competition for nominations is measured by the number of candidates receiving j:5% of the vote for nomination for the reduced equation are similar to the results for the full equation. The presence of an incumbent is still the single most important variable. All of the regression coefficients for this variable are negative and the vast majority of them are also significant (eight of the eleven at p _>_.05 and ten of eleven at p __>_.10) indicating a tendency for in— cumbency to reduce competition. Although this is a slightly weaker relationship than when the full equation was estimated over time the substantive interpretation remains unchanged. Average party vote continues with only a few minor differ— ences to play an important role in the reduced equation. In ten of the eleven years the regression coefficients for average party vote indicate a positive relationship with the number of candidates receiv- ing :_5% of the vote although the coefficients are significant in only five of the eleven years at the p :_.05 level. With the exception of 135 1972, average party vote tends to be significantly related to competi- tion for nominations more frequently prior to 1964 than afterwards. This relationship is consistent with the assumption that because of greater competition at the state level average party vote has had less of an influence on competition for nominations in recent years. How~ ever, the coefficients have their greatest values in 1964 and 1972, the two post-WWII presidential landslide years. The average vote for a party's previous condidates for governor and senator may have a greater influence on a potential candidates' estimate of the value of a nomination when he believes that there will be a landslide in the coming general election. While it is never possible to be certain that a landslide is going to occur, Johnson's victory in 1964 and Nixon's victory in 1972, were easily predictable when gubernatorial and sena- torial nominations were being made. When a presidential landslide appears imminent, it is much harder for candidates for statewide office to disassociate themselves from their party. Consequently, the influ- ence of the average party vote on competition for nomination is greater prior to a landslide election. The degree of control over state nominations is also, with a few minor variations, negatively related to the number of candidates running for nominations. As with the full equation, the relationship is more likely to be significant after 1962 than before although for some of the years (1970 and 1972) the coefficients are less signifi- cant than with the full model while in others (1964) the coefficients are more significant when the reduced equation is estimated over time. The relationship between gubernatorial opportunity rate and tflne number of candidates running for nomination becomes much clearer 136 when the reduced model is used. The coefficients are more consistent. An increasing gubernatorial Opportunity rate is associated with a decline in the number of candidates for gubernatorial and senatorial nominations in ten of the eleven years with the reduced equation compared with seven of eleven years with the full equation. The coefficients for gubernatorial opportunity rate in the reduced equation are significant at the p-: .10 level in only four of those years. This is still twice the number of times as for the full equation. Although the reduced model has improved the statistical measurement of the re- lationship between gubernatorial opportunity rate and competition for nominations it does not help the theoretical interpretation. The cur- ious split between significant relationship for gubernatorial opportun- ity rate and an insignificant relationship for senatorial Opportunity rate raises too many questions about the role of different Opportunity rates to allow for any easy interpretation of the relationship. Office, party name, run-Off primaries and registration pro- cedures are generally insignificant when the reduced model is esti- mated for each year. Office is negatively related to competition for nominations in a consistent, but insignificant manner. These results suggest that while governors have slightly more competition for nominations than do senators, the difference is proabably not a very important one. The relationship between party name and the number of candidates continues to fluctuate with party fortunes. Although there are slight differences, the results fit the pattern found when the full equation was estimated over time. The relationship between providing for a run-off primary and the number of candidates running for nomina- tion also becomes more consistent when we look at the reduced equation. 137 The presence of a run-off primary is positively related to competition for nominations in nine of eleven years although that relationship is significnat at the p :_.10 level once once. It seems that the reduced equation rescues the hypothesis that run-off primaries are related to increasing competition for nominations, but at the cost of suggest- ing that the relationship is not a very important one after all. Registration also is slightly more consistent when the reduced equation is used. Increasingly Open registration procedures are in- significantly associated with slightly more competition for nominations in nine of eleven years with the reduced model as compared with eight of eleven years with the full model. In general the reduced model produces estimates that are very similar to those produced by the full model when competition for nomin- ations is operationalized by the number of candidates receiving 3_5% of the vote. Although the differences are minor there is slightly less consistency for incumbency, party competition and control over state nominations fortfluareduced model than there was in the full model. This, however, is compensated for by slightly more consistent estimates for some of the less significant relationships, i.e., office, use of run-off primaries and registration requirements. Finally, the reduced equation also helps to clarify the influence of gubernatorial oppor— tunity rate on competition for nominations. The regression coefficients for gubernatorial opportunity rate are much more consistent in the reduced model than for the full. When the reduced model is estimated for the percentage margin of victory the results are also similar to those Obtained using the full model. Incumbency, party competition and to a lesser extent the 138 degree of centralized control over state nominations and gubernatorial opportunity rate continue to be significant in the time series analysis. The presence of an incumbent is strongly associated with less competition when the reduced equation is estimated over time. Average party vote, with a number of slight variations, is positively related to the percentage margin of victory. Yet for both the full and reduced equations, average party vote is not nearly as significant as often as was incumbency, suggesting that average party vote is a less important factor in determining the amount of competition for nominations than is incumbency. Control over state level nominations is more consistently rela- ted to an increasing percentage margin of victory for the reduced equa- tion than it was for the full equation. However, for both the reduced and full equations, centralized control over state level nominations is not as closely related to the percentage margin of victory as it is to the number of candidates for nomination. An increasingly gubernatorial opportunity rate is more consis- tently related to an increasing percentage margin of victory when the reduced equation is used than with the full equation. Evidently the exclusion of several of the insignificant variables results in greater consistency for some of the marginally significant variables. The relationship between Office sought and margin of victory is significant for more years when the reduced model is estimated over time instead of the full model. Senators tend to win by a greater margin than do governors. Party name, metropolitan domination and state population are largely insignificant when the reduced model is estimated over time. 139 Party name continues to fluctuate in value with general electoral trends as the potential electoral advantage shifts back and forth be- tween the parties in approximately the same way it did for the full model. The greater proportion of a state's population concentrated in a single urban area the greater the percentage margin of victory. Although the coefficients for the degree of metropolitan domination are consistent, only the 1974 coefficient is significant at the p_: .05 level. The relationship between the percentage margin of victory and state population is significant more often when the reduced model is estimated for each year than when the full model was estimated for each year. The greater a state's pOpulation the smaller the percentage margin of victory for gubernatorial and senatorial nominations, how- ever, none Of the coefficients is significant at even the p_: .10 level. When the reduced equation is estimated for the percentage margin of victory the results are also essentially the same as when the full equation was used. The major differences between the two models is that the reduced model produces slightly more consistent estimates with somewhat greater significance levels. In several cases this amounts to moving marginally insignificant coefficients over the magic line of statistical acceptability. In one case, 1954, the im- provement is enough to increase the overall F so that the entire equation becomes significant. 140 My In this chapter step-wise regression was used to rank order the independent variables in terms of their relative influence on competition for nominations. The entire data set was analyzed using step-wise regression and those variables which were significant at the p Z_.05 level were used to create two new reduced eight variable equations. The step-wise procedure split the original equation into two slightly different reduced equations for each of the measures of the dependent variable. The reduced equation for the number of candidates receiving :_5% of the vote consisted only of political variables while the equation for the percentage margin of victory included several measures of social diversity. The reduced equations produced essential- ly the same estimates as did the full equation for their common vari- ables. The two reduced equations were estimated for each nominating period. The results generally replicated earlier findings, although several of the variables were more consistently estimated in the reduced equations than for the full model. The only really new findings pro- duced by the time series analysis of the reduced equations were the clarification of the roles of gubernatorial opportunity rate and run— off primaries. Gubernatorial opportunity rate is more consistently related to both the number of candidates and the margin of victory in the reduced models. The relationship between run-off primaries and the number of candidates for nomination but not the percentage margin of victory is less erratic when the reduced equation is used, but also less singificant than in the full equation. 141 The major contribution of the reduced equation to an under- standing of competition for nominations is as a check on the stability of the earlier findings. The relationships first demonstrated by the full equation were not greatly changed when the equation was al— tered by discarding a number of the insignificant variables. More confidence can be placed in the theoretical implications Of this work because of the stability of the relationships over time and in the differently specified models. CHAPTER VI DISCUSSION Introduction This dissertation started with a number of hypotheses which were used to develop a multiple regression model of competition for nomina- tions. The model was then tested with data for all gubernatorial and senatorial nominations made between 1954 and 1974. The equation was estimated in a number of different ways. It was estimated for the entire data set, for each of the eleven separate nominating periods, for the North and the South, and with linear and non—linear estimators of average party vote. The entire equation was also estimated using step—wise regression to develop new reduced models, which were then reestimated for each of the individual years. Repeated estimation of similar models under a variety of conditions provides several different views of the relationship between the independent variables and competition for nominations. This procedure allows for a more reliable estimation of the actual relationships. The results test the original individual hypotheses, contribute to general understanding of competition for nominations and serve as a point of departure for further speculation about the nature of party organization. 142 143 A Final Summary of the Individual Hypotheses The first question to be answered is: Are the major hypotheses confirmed by this study? When all of the evidence is considered can any of these hypotheses be rejected or accepted, or must some more qualified statement be used to describe the actual results. Tables 15 and 16 sum- marize the results of each of the separate analyses. The first six hypotheses relate the measures of party organization to competition for nominations. All of these hypotheses are based either on prior empirical studies or on theoretical speculation about which aspects of party organization influence competition for nominations. The major question which these hypotheses are designed to answer is: Do any aspects of party organization in the United States still influence compe— tition for nominations. Hypothesis one states that the greater the control the formal party organization has over the state-wide nominations.the less competi- tion there will be for gubernatorial and senatorial nominations. The evidence suggests that this hypothesis should be accepted for both measures ,of competition for nominations. Increased party control over nominations is associated with decreased competition for nominations for all of the pooled data sets, and for the post 1964 nominating periods when the rela- tionship is measured over time. There is only one equation where control over state-wide nomina- tions does not have an influence. The variable is insignificant when the general equation is estimated separately for the South. The high degree of multicollinearity between a number of the major variables in the Southern sample and the general lack of variation in the degree of party 144 OOONODOO :N omvoaoofl uoc ONLONNO> n O O oz oz oz oz oz oz NONOOOM mmw mmz mm» ammozzoo mmw oz mm» NOcNo>oo mumz NNHOONNOQQO O O O oz Noooz NOOONNcoz CONuHumoEoo Noumm oumONOuN< mm» mm» may mm» mm» oz mmz Noooz NOOONNIIOONuNuooEOU Noumm oz mmz mm» oz oz may may OONwwo mm% mm% mm» mm» mm» mmw mm» NoconsoocH oz OOO OOO oz OOO OOO OOO NOOOz NNNOOV oOOO HmNoom m.>uumm mo NuNmNO>No O oz oz oz OOO oz OOO OONoONOOOO ONOOO O oz oz oz OOO oz OOO OONOONoOooOoo OOONNOOONOOO O oz oz oz OOO oz oz OONmno>NO NONOOO NuOmNm>No NmNoom oz mmw mm» oz mmz oz oz muoncsm oz mHHmommo mHHmommo oz meHmommo oz mHHmomzo OucoaONNovoz OONumNumemm O oz oz oz oz oz oz OONuomNom NOONNOOH O oz oz oz oz oz oz OONumuNcmwuo Noumm ocu ON :ONOONEo< No>o Nouusoo O oz oz oz oz oz oz OoOOOONmO NONOOOO INmmNm mo OOONOONOM uo>o Nouucoo OO. Hmom mm% mm» mm» mm» mm» oz mmw OOONumaNaoz Oumum uo>o Houusoo OENuuo>o ooaoom Omflsooum oENH :uuoz nuoom mumm mumo mmmonuomzz OOOOOOO OOOOOOO No>o muom mumo ooNoom Hmconom .OOONOOmON No Oocmuaooom moumoaosN ONnmu may .OONumsNaoz pom OONNNNOOEOQ ousmmoz ON ouo> mo NMumch>Nmoom moummNoomu mo Nonaoz ocu chm: momozuoozm mo zumaasm .MN ONOOH 145 cowumovo ozu :N ooozaocN No: OHQONNO> n O oz oz oz oz oz oz Noumowm oz mm» mmw mm» mmw mm» NOONo>oo oumz NONOONNOOQO O O O O O O oz Hoooz NOOONNcOz OOOuNuOOsOU NuNmm mumououa< mmw mm» mmw may mmw oz mmw Hoooz NOOOONIIOONNNNOOBOU Noumm mm» mm% mm» mmz N oz mmw OONMNO mmz mmw mmw mmz mmw mm% mm» NoaonaoooH oz OOO ON; oz OOO OOO OOO. $st 5.3.: OON.O HmOoom m.zuumm mo zuHmNO>NQ O mm» mm» oz oz oz mmw coaumaooom mumum O Maw may mm» mmz oz mmz OOHumNucoocoo cmuNaooouuoz oz oz oz oz oz oz zuNmNo>Oo HmNomz zuNmNm>HQ HmNoom O oz oz oz O oz oz mmOIcoz « oz oz oz oz oz oz mucoaouwsooz CONumNumOwoz O oz oz oz oz oz oz coauooaom uoouNosH O oz oz oz oz oz oz OONNONNOOONO ONO.N mzu ou OONmmNEo< NO>o Houucoo O oz oz oz oz oz oz mOumwOHoo NmOucop INOONm mo OONOOONOM um>o Houucoo oo. Nouw< zaxmoz mmw mm% mm% mmw oz may OOOONOOOEOZ Oumum No>o Houusoo oENH ooaoom omNoom oENH zusomcoz zusom muom mumo momocuoazz uo>o . woosooz mmNBQONm NO>O muom mono ooaoom omooooz HOCONmoz .mOONumooo OSN «0 50mm Np ooENchoo mum ammonuoazz mnu No: no nosuona OONOOHOOO magma one .OOONOOOOBOZ you GONuNuooaoo muammoz ou NNOOON> mo owmuooouom onu chmD mononuomzz mo NNOEEOM .ON ONAOH 146 control over statewide nominations in the South, however, easily explain this exception. The confirmation of the first hypothesis lends support to Schattschnieder and Key's contention that control over nominations is of critical importance for maintaining party discipline (Schattschnieder, 1942, Key, 1956). The results are also consistent with Wolfe's observa- tion that those states where the formal party organization still makes some nominations by convention have less primary competition than those states which nominate exclusively by primary election (Wolfe, 1966). Finally, these findings support Leege's contention that convention nomi- nations for second-level offices are a major party building resource (Leege, 1970). The second hypothesis states that when the activist organization controls the selection of presidential delegates there will be less compe— tition for gubernatorial and senatorial nominations. The results for both measures of gubernatorial and senatorial competition for nominations clearly suggest that this hypothesis should be rejected. The variable is insignificant in all of the equations which included it. Although control over the selection of presidential delegates is one of the few tasks left to the state party organizations, it evidently is not a very impor- tant organizational resource. The ability to select non-professional, unpaid, and temporary officers as presidential delegates gives a party little leverage over its office-oriented members. The position of pre- sidential delegate is probably more of an incentive for the amateur activists and party workers who are more interested in prestige than in seeking higher office. 147 The lack of relationship between control over the selection of presidential delegates and competition for state-level nominations fur— ther underlines the extent of separation between state and national level politics. The ease with which a number of state parties gave up their control over the selection of presidential delegates by instituting state— wide preference primaries rather than by reforming their own internal organization adds additional support to this interpretation (Ranney, 1974). It is possible that the presidential party is so separate from the rest of the state party as to exert little influence on it. The third hypothesis states that the easier it is to join the formal party organization, the greater the competition for nomination. Again, the evidence suggests that this aspect of party organization is not significantly related to competition for nominations. The variable used to test this hypothesis ranks the requirements for joining the official party organization in terms of difficulty. The lack of a signi- ficant relationship suggests that factional conflict cannot be controlled by making it more difficult to become part of the formal party organi- zation. Consequently the third hypothesis is rejected. The fourth hypothesis states that the more indirect the selec- tion procedures for a state convention, the less will be the competition for gubernatorial and senatorial nominations. The results suggest that the degree of organizational hierarchy, as measured by the number of separate steps required to select a state convention, has no significant influence on either measure of competition for nominations. Hypothesis four is rejected. The rejection of both the third and fourth hypotheses implies that a political party cannot prevent internal competition by attempting 148 to insulate the formal party organization from external pressures. Isolat- ing the decision makers by increasing the admissions requirements or by developing a more hierarchical structure is ineffective because conflicts over nominations seem to arise more from disagreements among established members of the organization than from infiltration by less professional outsiders. The Goldwater and McGovern movements, which generated a number of nomination contests which pitted professional politicians against their supposedly amateur rivals, are exceptions. Yet even these party insurgencies would have been far less effective had they not won the sup- port of many party regulars. While outsiders play a role, the truth is that whatever the amount of infiltration, success seems to require a great deal of internal cooperation. The nonconfirmation of hypotheses three and four also suggests that the formal requirement for a separate hierarchical party organiza— tion is not sufficient to create a stable organization. A party organi- zation does not function in the same way that a bureaucracy does. The structure of a party is less important than that of a bureaucracy because of the periodic nature of its organizational life. One meeting every two years, or even one meeting every month is inadequate to produce a sense of "organizational consciousness." The fifth hypothesis states that the more open the registration requirements are, the greater will be the competition for gubernatorial and senatorial nominations. The results make it easy to reject the hypothesis for both measures of competition for nominations. Registra- tion requirements are related to competition for nominations in a manner opposite to the original hypothesis and that relationship is at best, only marginally significant. The results provide limited support for 149 Smith's contention that the amount of competition for nominations actually increases in more structured settings (Smith, 1973). However, the weak and often insignificant nature of the coefficients suggests that the rela- tionship is not an extremely strong one and in this respect confirms Ranney's SuSpicion that registration requirements have relatively little influence on competition for nominations. The sixth hypothesis states that the possibility of a run-off primary will result in greater competition for nominations. The results, when competition for nominations is measured by the number of candidates who receive at least 5% of the vote for nominations, are mixed. The dummy variable for run—off primaries is statistically significant only for the stepwise and reduced models when applied to the entire data set. In all of the other instances the variable is either statistically insignificant or substantitally uninterpretable. These findings suggest that the relationship between competition for nominations and run-off primaries is probably not very important. The results are much clearer when the percentage margin of victory is used to operationalize competition for nominations. Hypothesis six does not hold for this measure of competition for nominations. None of the equations produce any significant supporting evidence. It is easy to overestimate the influence of run-off primaries on competition for nominations after reading Key and Duverger (Key, 1949, Duverger, 1954). Whatever influence the institution may possess lies in slightly increasing the number of minor competitors for nomination, but not in reducing the margin of victory of the leading candidates. In fact, Key's own results are better when he relates the number of 150 candidates for nomination with the adeption of the run—off primary than when he looks at the percentage margin of victory (Key, 1949:421). The seventh hypothesis states that the greater the degree of social diversity, the greater will be the competition for nominations. When aggregate measures of racial diversity, metrOpOlitan concentration, and state population are used to operationalize the degree of social diversity, the results are mixed. Racial diversity was a significant factor only when the fully operationalized equation was estimated for the nonsouthern data set, using the number of candidates receiving 5% of the vote to operational- ize competition for nominations. However, this variable should not be dismissed from consideration because, when estimated over time, the influence of racial diversity on competition for nominations appears to reverse itself. The coefficients for racial diversity changed from a negative relationship before 1964 to a positive relationship after 1964. The 1964 and 1965 civil rights acts and increasing rates of black parti— cipation are probably responsible for this change. A number of primary contests in both the North and the South reflect this new black assertive— ness. Richard Austin's campaign for the Senate nomination in Michigan, the Wallace-Flowers contest for the gubernatorial nomination in Alabama, and the Carter-Sanders contest for the Georgia gubernatorial nomination all involved intra-party competition between candidates who were closely allied to the competing racial and ethnic groups in their states. Of all of the measures of social diversity, the degree of metro- politan concentration is the most consistently related to competition for nominations. In general, the more concentrated is the state's population, the greater the margin of victory for the winning candidate. 151 However, the extent to which a state's population is concentrated in a single city has a greater influence on the percentage margin of victory than on the number of candidates running for nomination who receive at least 5% of the vote. State population is not closely related to either the number of candidates for nomination or the margin of victory. Only when the full equation is estimated for the entire data set is there a significant relationship between state population and competition for nominations. While slightly more candidates run for nominations in the more populous states and the percentage margin of victory is smaller for the more populous states the coefficients are generally insignificant. If there is any relationship between state population and competition for nomina- tions the evidence suggests that it is not a very strong one. The results do not allow for any conclusive statement about the seventh hypothesis. Aggregate measures of social diversity (racial fragmentation, metropolitan concentration and state population) have some influence on competition for nominations especially when competition for nominations is operationalized by the percentage margin of victory, but the relationship is often statistically insignificant. When party name is related to competition for nominations there is more competition for Democratic nominations in every instance and for both measures of competition for nominations. However, when the party name is related to competition for nominations over time there is a great deal of fluctuation in the value of the coefficients. This fluctuation seems to be associated with major presidential landslide elections. While there is almost always slightly more competition for Democratic candi- dates, it is difficult to attribute these results only to differences in 152 the-social composition of the two parties because of the high degree of fluctuation. Social diversity may have some effect, but it appears to be overshadowed by changing party fortunes (see Figure 10). Curiously while the lack of a relationship between presidential year and competition for nominations, and the method of selecting presi- dential delegates and competition for nominations suggests a strong separation between state and presidential politics, the fluctuating nature of the relationship with party name suggests the opposite. While national politics may not have much of an influence on competition for state nominations under normal circumstances unusually disruptive presi- dential campaigns may be an exception. Bickering and infighting at the national level, if they threaten to devalue a party's nomination, can cause a number of otherwise interested politicians to drop out of the race. While hypothesis eight can be neither accepted nor rejected as a consequence of these results, the findings suggest that more attention should be paid to candidate estimates of their short term chances of success. As party voting declines as a consistent habit of the American electorate single election swings from one party to the other become more important. The greater the likelihood of the electorate to deviate from past habits the more sensitive the potential candidatesvfiJJ.be to situa- tional variables. The ninth hypothesis states that the presence of an incumbent reduces competition for nominations. This hypothesis is strongly con- firmed for both measures of competition for nominations and for each of the equations. Interestingly enough former incumbents try to recapture a gubernatorial or senatorial nomination, and in those few cases where 153 a former incumbent did try the attempt was almost always unsuccessful. Once momentum has been lost in American politics it is almost impossible to regain it. Former incumbency does not seem to provide any of the advantages of incumbency suggesting that more than name recognition is involved in giving an incumbent the advantage. This also suggests that a constitutional limitation on concurrent terms iseulentirely effective way to break the hold of incumbents on their offices. The tenth hypothesis formally states that there will be more competition for gubernatorial nominations than for senatorial nominations. The results confirm this hypothesis for both measures of competition for nominations, but most especially for the percentage margin of victory. The results using the number of candidates receiving 5% of the vote, although the coefficients fall in the appropriate direction, are not significant for the year by year analyses for the nonsouthern data set. The eleventh hypothesis states that the greater the average party vote, the greater the competition for nominations. The hypothesis postu— lates a linear relationship between competition for nominations and average party vote. The results confirm the hypothesis for both of the measures of competition for nominations and for all of the equations, except for the analysis of the Southern data set. Average party vote definitely does influence the amount of competition for nominations. While the influence of average party vote on competition for nominations was treated as a linear relationship, it could also have been estimated as a curvalinear relationship. While a curvilinear model might have produced a slightly stronger estimate, it was not used because lFormer incumbents attempted to obtain their party's nomination again in only 16 of 1162 cases. Their success rate was very low. In only five of the sixteen cases did a former incumbent actually manage to recapture the nomination. 154 there was no strong theoretical justification for its use and because the actual increase in explanatory power is much too small to be relied on by itself. Either way competition for nominations plays an important, but not a controlling role in influencing competition for nominations. The twelfth hypothesis formally states that the less the oppor- tunities to run for higher office, the greater the competition for nomi- nations. The results are contradictory. Senatorial opportunity rate F} has no influence on either measure of competition for nominations while wo1 gubernatorial Opportunity rate does. The office of governor may be the key safety valve in a state's political system. A high gubernatorial '~« turnover rate may well protect a state's senators from primary competi— tion. The split results break down the logic of the theory. Perhaps some other variable which is associated with gubernatorial, but not senatorial opportunity rate is responsible for this relationship. Per- haps constitutional limitations on gubernatorial tenure make it easier for potential candidates to estimate the actual opportunities for seek- ing higher office. From a theoretical perspective the results are highly unsatisfactory. The only conclusion that can be reached about the influence of opportunity rate on competition for nominations is that it is more complicated than first hypothesized and consequently more study is required. Implications for a Theory of Competition for Nominations Taken as‘a whole, these results tell us a great deal about compe- tition for nominations. While strictly speaking, any generalizations can only be applied to the set of gubernatorial and senatorial nomina- tions actually included in the study, the findings have a number of implications for a general understanding of competition for nominations. 155 Although theoretical issues cannot be resolved solely by looking at empirical findings, empirical findings do have a number of theoretical consequences. I The first and most obvious finding is that the variables which have traditionally been considered as the explanation of competition for nominations, do not explain that much of the variance. Competition for nominations is not a highly structured kind of activity. Although a number of variables influence competitionikn:nominations the amount of variance explained by the standard set of variables fluctuates widely over time, the squared multiple correlation coefficient for the entire equation never rises above .50. This finding suggests that more idosyncratic, strategic, policy related and personality centered types of variables are also playing a major role in explaining competition for nominations and that only a small portion of the variance can be explained by the less time bound variables found in the general equation. The second major finding is that it is incumbency and not average party vote that has the greatest influence on competition for nominations. Key and Patterson overestimated the importance of average party vote on competition for nominations (Key, 1956, Patterson, 1976). Average party vote is an important variable, but it is far from a dominant variable. Whether estimated as a linear or as a curvalinear relationship average party vote simply does not explain as much of the variance as does incumbency. The intriguing question is has incumbency always been the most important variable influencing competition for nominations or has it only recently begun to play such an important role. A number of other studies have pointed to the increasing tendency for incumbents to be returned to 156 office, might not this tendency also be accompanied by a similar increase in the ability of incumbents to obtain nomination? While it is impossible to be certain because these data did not include any nominations made before 1954, the post-1954 results show only a slight increase in the relative influence of incumbency on competition for nominations over time. This would suggest, but alas cannot prove, that incumbency has probably always played a major role. In fact, it seems logical to assume that under the more controlled conditions that existed earlier in the century political parties might have been even more loath to give up electoral advantage and consequently less likely than today to allow a challenge to an incumbent. The third important finding is that at least one aspect of party organization, control over state level nominations, does have an influence on competition for nominations. Even in the United States, the home of some of the most decadent party organizations in the world, one of the formal powers of the party organization still plays a significant role in controlling competition for nominations. It is the ability of the party organization to make nominations rather than the structural configuration of the formal organization that is related to competition for nominations. Social fragmentation does not seem to have very much influence on competition for nominations. Group structure is relatively less important in explaining competition for nominations than the more directly political variables such as incumbency, average party vote, and the degree of party control over the nominating process. The pluralist hypothesis is insufficient in and of itself to adequately explain competition for nominations. Political conflict is not a direct expression of group conflict. Political circumstances and 157 electoral rules play a vital intermediary role. Although social groups may serve as a basis for political coalitions, intervening decision rules and situational factors seem to control the expression of that conflict by prescribing the size and number of potentially viable elec- toral coalitions each of which may be composed of more than one social group. Future students of competition for nominations would do well to focus on the role of incumbency, average party vote, and the degree of control a state party organization has over nominations. These variables are clearly the most consistently important ones in this analysis. Also, the perplexing nature of the results for Opportunity rate suggest that a more intensive study of this concept might prove useful. Opportunity rate was_measured as a combination of potential and actual turnover rates for both the offices of governor and senator. Other measures of the potential opportunities for statewide elective office might also be tested. Opportunity structure is a difficult concept to measure at the state level because frustrated candidates can always go to other arenas if their ambitions are blocked within their home state. Turnover at the federal and local level can also relieve some of the political pressure within a state. More research needs to be done on the strategic expectations of candidates running for nomination. The shifting nature of the relation- ship between party name and competition for nominations suggests situa— tional rather than structural factors are having the greatest influence. At the present time theories about competition for nominations focus almost exclusively on the static or at best very slowly changing kinds of variables which affect candidates' perceptions of their chances of 158 success. More work should be done with the effects of short term factors such as economic recession, political scandal and war on the stability of party leadership in the United States. Finally, it might be interesting to look at qualitative measures of competition for nominations as well as quantitative measures. There are probably great differences between the behavior of the professional politicians and their nonprofessional colleagues. Whether or not major officeholders choose to contest a nomination or leave that conflict to those with limited political experience might offer fruitful ground for future research. Implications For a Theory of Party Organization While this is a study of competition for nominations, it also focuses on party organization. I spent a great deal of time trying to use several measures of party organization to explain competition for nominations. The most important finding is that party organization can be better understood as a combination of several conceptually related variables rather than as a single variable. There is very little cor- relation between the different measures of formal party organization. The measures of organizational separateness and hierarchy are not closely associated with each other, and the measure of control over statewide nominations and control over the selection of presidential delegates are not closely associated with each other. Only when the organizational measures were regressed on each other is even a moderate relationship discovered between one of the variables and all of the rest in additive combination. 159 In the United States, state level political parties are charac- terized by a great deal of organizational diversity and inconsistency. The results of this study and Ronald Weber's analysis of state party organization (Weber, 1969) suggests that party organization should be treated as a number of separate and not necessarily related dimensions. Organization, even formal organization, is too complex Of a phenomenon to be treated as a single variable. When the measures Of party organization were related to competi- tion for nominations only one Oftfiuan,the degree of party control over state level nominations, proved to be both significantly and consistently related to competition for nominations. All Of the other measures, degree of control over the selection of presidential delegates, the degree Of convention hierarchy, use of run-off primaries, Openness of registration procedures and entry base had little effect. A party is tOO free form of a political organization to be under- stood by a simple bureaucratic model. Structural variables tell us rela- tively little about how the organization behaves. The intermittent nature of party activity and the lack Of correspondence between formal organiza- tion and informal behavior make it very difficult to use structural measures to explain competition for nominations. The one measure that was consistently related to competition for nominations measured the degree of control the formal party organization had over statewide nominations. The ability to make nominations is the ability to distribute an important benefit or resource. Perhaps the kind of resources controlled by a party may be a better way to understand and evaluate party organization than in party structure. 160 While the structural functional analyses of party organization are no longer in vogue, structural descriptive and structural behavioral approaches are still used to analyze political parties in the United States. The structural descriptive and structural behavioral analyses make the assumption that organizational form is linked to organizational behavior. Key's concentric party model and Schlesinger's candidate- centered model of party organization explain party organization in terms of its structural characteristics (Key, 1958; 179—183, Schlesinger, 1968; E13 430). Although descriptively rich, neither model can be easily used to evaluate party strength. r. Duverger developed a more behaviorally oriented theory Of party id] structure (Duverger, l954). Durverger analyzed party organization in terms Of its structural units and their degree Of coherence. His theory provides a number of dimensions which can be used to evaluate party organ- izations in terms of their relative complexity. Although this model allows for the evaluation of different types of party organization along structural lines, it has not proved to be highly successful in relating political behavior to party structure. Perhaps another approach should be tried. If we start from the assumption that a political party is an Office-seeking organization, then we can evaluate party organization in terms of the resources necessary to obtain those Offices. By evaluating party strength in terms of the resources controlled by a party organiza- tion, it is possible to break away from the assumption that the party structure is related to specific kinds of behavior. This new method Of analysis shifts the understanding of party organization to an analysis of 161 the resources which can be used to influence political, electoral and governmental Operations. The resources available to a party could be measured by five variables, control over nominations, financial support, party workers, loyalty of adherents and control over governmental Offices. While some of these are harder to measure than Others because of the lack Of data all are at least potentially measurable. Control over nominations can be either centralized, decentralized or externalized. Schlesinger's scale offers a reasonably good measure Of this variable with respect to state Offices (Schlesinger, 1974). Similar measures could also be developed to extend this type Of scale to other classes of office. Financial resources can be measured by the proportion of total party expenditures that are controlled by the formal party organization. Changes in the reporting requirements now make it much easier to Obtain access to more accurate estimates Of a party's financial resources. A state by state comparison might be made of the two major parties to determine if there is any significant difference between these parties in the degree to which the formal organization controls the expenditure of party funds and if this has any effect on their internal cohesion. The number of party workers available can be measured either by the number of paid workers per state party or by the number Of volunteer workers associated with a state party organization. While this is a hard figure to estimate the number of individuals attending the state conventions Of each of the major parties could also be used to estimate this number. 162 Party loyalty is important because it provides us with some information about the degree to which a party's leadership has to be concerned with the potential desertion of their supporters. The degree of split-ticket voting in a state can be used to measure the degree of party loyalty. Finally, control over government is also an important party resource. The degree of control over a state government can be esti- mated as the proportion of partisan offices held by a party at any one time. While this variable is probably highly related to average party vote in the United States and consequently hard to measure here, it should be more useful in other countries. In any multiparty system where the relationship between control over offices and average vote is much lower, the prOportion of government offices held is a useful measure of actual party strength. Once party resources have been measured only two additional assumptions are required to related party organization to political be- havior. The first assumption is that the more resources that are con- trolled by a political party, the greater its strength. The second is that the more centralized the control over these resources within a party, the greater the ability Of that party to maintain internal discipline. While this study was written from a combination of structurally oriented and nomination centered theories of party organization, the results suggest that it might be better to take a more resource oriented approach to evaluating party organization. After all, the one measure of party organization which had any relationship to competition for nominations fits nicely into a resource-based model of party organization. Future research should look more closely at other kinds of resources as well. 163 Conclusion While this project raises almost as many questions as it resolves some progress has been made in shifting the grounds of inquiry. Compe— tition for nominations offers a fertile field for future research. Compe- tition for nominations cannot be completely understood by continuing to rely on the traditional explanations. While a respectable amount of the variance is explained by incumbency, avearge party vote and the degree of party control over statewide nominations a great deal of room is left for other kinds of explanations. Party organization continues to be a difficult and elusive con- cept to deal with because party organization cannot be treated as a single variable. Even the formal characteristics of a party organization are not closely associated with each other. Structural measures do not seem to add to our understanding of competition for nominations. Party structure is so free form, segmental and periodic that it is extremely difficult to measure. Instead of grasping at a phantom which does not seem to be highly related to political behavior, it appears that more can be gained by trying to evaluate parties in terms Of the resources they actually control. Resources may be more directly related to political behavior than is party structure. APPENDIX Control Over State Nominations Table A.l. Year 1956 1958 1960 1962 1964 1966 1968 1970 1972 1974 1976 1954 State Alabama — D Alaska Arizona Arkansas 165 California Colorado Connecticut Delaware Florida QC! Georgia - Hawaii Idaho Ill - D 1nois Indiana Iowa Kansas l66 NHNNNNNLHN NNNNQQMN NHNNNNNWN NNNNQ’KDMN Nr—INNNNNLON NNNNx‘T’KDMN NHNNNNNLHN NNNNQOMN NHNNNNNWN NNNNQ’OMN NI—INNNNUWNN NNNNQ’KOMN NHNNNNWNN NNNNQ’OMN NHNNNNNNN NNNNQOMN NHNNNNNNN NNNNQ’OC’WN NHNNNNNNN NNNNQ’KOMN 60 N H m N N N N l\ N NNNNQOMN (\o NNNNQOMNONHMNNNWNN muoxmo zuuoz mafiaoumo :Ouoz xuow 3oz OOHsz 3oz Ammumh 3oz Ouwzmoamm 3oz mom>mz mxmmubmz mcmucoz Husommflz Hoofimmflmmfiz muomoccwz amwzzoaz muummocommmmz ocmazumz moan: mommfisoq zxuoucmz ommz quad NmmH ommz wood oomfi qomz Noafi coma mmmz omma qwma pom» Oumum omsoaucoo .H.< «Home Continued Table A.l. Year 1956 1958 1960 1962 1964 1966 1968 1079 1972 1974 1976 1954 State Ohio Oklahoma Oregon Pennsylvania - D 167 Rhode Island South Carolina - D South Dakota Tennessee D Texas Utah Vermont D irginia — V Washington West Virginia Wisconsin Wyoming .qmma umnouoo .ummcwmmazom nomm0h zn OOOOHO>OQ mamom vmzmfiHnOOO: .OwOOHHmso ou uummbom uOO .OOHOOO>OOO ma mOowumOHEOO HHO .m .OOfiuOO>Ooo >2 mumsuo .zumafluo zn mOowOmOHEOO uOmOfiwHOwfim meow "Emumzm omxwz .o .OOHOOO>OOO um om>mflnom ma N OOHMflOOom «a zumefiua m Ofi OwOOHHmzo ou OomnnOm .OOHumOflEOO OOHOOO>Ooo .m 168 .uoaamn Oo Hmouo OH mOEmO wOHOmHO mm zoom .OOHOOO>OOO Ho OOOOHEEOO Oumum now OOHOOOOM zumEHuOIOum .q .zpmsauo OH N m O>onm OO>HOOOH OOO OO uOO>O Ofi OOOOOHOOOO HmOfiw Ono wOHmoozo ..m.O mumH>Huom zuumo MOM OHOO zumafiHOIOOOO uOmOHwfiOwmea no Acummu Ozonua3v uOOEOOOOOOO zumafiualmuo How OOfimH>OHO man .zumEHuO zn ommono mOumOHOOmO HHO .m .zumafluo mmmoao mo Show meow an Omozo mOuOOHoOmO HH< .N .zumawuo Ommo OH Ommono mOumOHwOmo HH< .H "mOofiumOwOOz Oumum HO>O Hoquoo mo mamom omsafiuooo .H.< magma 169 Table A.2. Method of Selecting Presidential Delegates Year State 1948 1952 1956 1960 1964 1968 1972 1976 Alabama-D 2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 —R 3.5 3.5 3.5 3.5 3.5 3.5 2.2 2.2 Alaska 3.5 3.5 3.5 3.5 3.5 Arkansas 4.6 4.6 4.6 4.6 4.6 4.6 2.6 2.6 3.5 35. 35. 35. 3.5 3.5 California 1.0 1.0 1.0 1.6 1.6 1.6 1.6 1.6 Colorado 3.5 .5 3.5 3.5 3.5 3.5 3.5 3.5 Connecticut 3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0 Delaware:: convention 3:; 3:; Florida . 1. 1. 1.6 1.6 1.6 1.6 1.6 1.6 Georgia—D 5.0 5.0 5.0 5.0. 5.0 5.0 3.5 1.5 -R 3.5 3.5 3.5 3.5 3.5 3.5 3.5 Hawaii 3.5 3.5 3.5 3.5 3.5 Idaho 4.9 4.9 4.9 4.9 4.9 4.9 4.9 4.9 Illinois 2.1* 2.1* 2.1* 2.1* 2.1* 2.1* 2.1* 2.1* Indiana 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 Iowa 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 Kansas 4.9 4.9 4.9 4.9 4.9 4.9 4.9 4.9 Kentucky 3.0 3.0 3.0 3.0 3.0 3.0 3.0 1.7 Louisiana-D 5.0 5.0 5.0 5.0 5.0 5.0 3.5 3.5 -R 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 Maine 3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0 Maryland 1.7 1.7 1.7 1.7 1.7 1.7 1.5 1.5 Michigan 4.4 4.4 4.4 4.4 4.4 4.4 1.7 1.7 Minnesota '3.5 1.0 3.5 3.5 3.5 3.5 3.5 3.5 Mississippi 3.0 3.0 3.0 3.0 3.0 3.0 3.0 3.0 Missouri 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 Montona 2.4 2.4 1.0* 2.4 2.4 2.4 2.4 2.4 Massachu- 1.6 1.6 1.6 1.6 1.6 1.6 1.6 1.6 SettS 170 Table A.2. Continued Year State 1948 1952 1956 1960 1964 1968 1972 1976 Nebraska 1.6 1.6 1.6 1.6 1.6 1.6 1.6 1.6 Arizona 4.7 4.7 4.7 4.7 4.7 4.8 3.0 3.0 3.5 4.7 4.7 4.7 4.7 4.7 3.5 3.5 New Hampshire 1.6 1.6 1.6 1.6 1.6 1.6 1.6 1.6 New Jersey 1.6 1.6 1.6 1.6 1.6 1.6 1.6 1.6 New Mexico 3.5 3.5 3.5 3.5 3.5 3.5 1.7 3.5 New York 2.2 2.2 2.2 2.2 2.2 2.2 1.6 1.6 N. Carolina 3.5 3.5 3.5 3.5 3.5 3.5 3.5 1.7 N. Dakota 4.4 4.4 4.4 4.4 4.4 4.4 3.5 3.5 Ohio 1.5 1.5 1.5 1.5 1.5 1.5 1.5 1.5 Oklahoma 4.9 4.9 4.9 4.9 4.9 4.9 3.0 3.0 Oregon 1.6 1.6 1.6 1-6 1.6 1.6 1.6 1.6 Pennsylvania 2.3* 2.3* 2.3* 2.3* 2.3* 2.3* 2.3* 2.3* Rhode Island 4.6 4.6 4.6 4.6 4.6 4.6 1.6 1.6 S. Carolina 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 S. Dakota 1.6 1.6 1.6 1.6 1.6 1.6 1.6 1.6 Tennessee 3.5 3.5 3.5 3.5 3.5 3.5 1.7 Texas-D 3.5 3.5 3.5 3.5 3.5 3.5 3.5 1.6 -R 3.0* Utah 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 Vermont 3.0 3.0 3.0 3.0 3.0 3.0 3.0 2.5* Virginia 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 Washington 3.5 3.5 3.5 3.5 3.5 3.5 3.5 3.5 W. Virginia 1.7* 1.7* 1.7* 1.7* 1.7* 1.7* 1.7* 1.7* Wisconsin v1.0 1.0 1.0 1.0 1.0 1.0 1.0 1.0 Wyoming-D 4. 4. 4. 4. 4. 4. 3.0 3.0 -R 3.5 4.5 Nevada 3.5 3.5 3.5 3.5 3.5 3.5 3.5 1.7 *Advisory primary, not binding. .5 subtracted from original value. 171 Table A.2. Continued Presidential Nominations Scale Locus of Power to Name Presidential Delegates 1. Lack 1.0 1.5 1.6 1.7 of Hierarchy Open primary chooses all delegates who are committed Closed, challenge primary chooses committed delegates Closed, enrollment primary chooses committed delegates Binding primary delegates selected in another manner 2. Casual participant chooses in timely setting, but low levels of hierarchy create need for continuous activists 2.0 2.2 2.3 2.4 2.6 2.7 Open primary chooses uncommitted delegates Closed, challenge primary chooses uncommitted delegates Closed, enrollment primary chooses uncommitted delegates Open primary chooses delegates to convention or committee which in turn selects delegates Closed challenge primary chooses delegates to convention or committee which in turn selects delegates Closed enrollment primary selects delegates to convention or committee which in turn selects delegates 3. Entry confined to activists by means of caucus, convention, or mass meeting in a timely setting, with assorted types of hierarchy. 3.0 3.5 Caucus chooses delegates to a meeting which chooses national delegates Caucus chooses delegates to meeting which in turn chooses delegates to one higher meeting 4. Entry is at untimely point, usually two or more years before the actual selection of convention delegates, with various types of entry and hierarchy 4. 4. 8 9 172 Open primary entry--non-hierarchic selection Closed challenge primary--non-hierarchic selection Closed enrollment primary, non-hierarchic selection Open primary, hierarchic selection Closed challenge primary, hierarchic selection. Closed enrollment primary-—hierarchic selection Caucus entry, non-hierarchic selection Caucus entry, hierarchic selection 5. Obligarchic selection--choice is made by party or elected officers or by a committee appointed by them. Note: Source: 0.5 is subtracted from the scale score if there is a non-binding primary. The above is a modification of Joseph Schlesinger's "Presidential Nomination Scale," unpublished 1974. 173 Table A.3 through A.20. Variations in State Party Organization for 1974. These tables indicate the number of state parties in the United States which are characterized by the following organizational features. Table A.3. Uses Pre-Primary Table A.4. Has a Separate Endorsement. Amature Organization Yes 17 Yes 6 No 83 NO 94 Total 100 Total 100 Table A.5. Method Used to Table A.6. Type of Primary Nominate Gover- nors and Senators Run Off 20 Convention 7 Single Primary 74 Challenge Primary 6 Challenge Primary 6 Primary with Privisions for a Convention in Case Total 100 of Deadlock 4 Primary 83 Total 100 174 Table A.7. Officeholders Auto— matically Included on State Committee Table A.8. Officeholders Auto- matically Included in State Convention Yes 14 Yes 26 NO 84 No 66 . Not Applicable 3 DK '-;; DK 5 Total 100 Total 100 Table A.9. Method of Voter Table A.10. Formal Membership Registration Closed 58 Yes 6 Open 16 Challenge 21 No 94 Blanket 2 DK 3 Total 100 Total 100 Table A.11. Selects State ' Table A.12. Major Nominations Committee Made by State Con- vention Primary 26 Governor 6 State Convention 16 Senator 6 County or District Caucus 24 Major Executive Officers 8 Congressional District Trustees and Minor Convention or Committee 10 Officers 6 Composed Automatically Congressmen 2 of Officeholders 2 No Public Officers Governor 2 other than Electors Unclear 20 and delegates 82 Total 100 No Convention 7 175 Table A.13. Number of Separate Table A.14. Degree of Hierarchy Hierarchies in the Required to Reach Party Organization State Committee 1 41 l 25 2 53 2 25 3+ 2 4 32 don't know or 4+ 10 unclear 4 DK, and Total 100 not clear 8 Total 100 Table A.15. Degree of Hierarchy Table A.16. Method of Entry for Required to Reach the Party Organi- State Convention zation (Schlesinger, 1974) 1+ 15 2+ 39 1. Open primary 6 3+ 28 2. Closed challenge primary 14 Not Applicable 9 3. Closed enrollment primary 34 DK ——2' 4. Caucus, convention, Total 100 or "mass meeting" 42 DK 4 Total 100 176 Table A.17. Control over State Table A.18. Number in State Nominations Convention (Schlesinger, 1974) None 8 1 4 1 - 500 31 2 46 501 - 1000 ll 3 26 1001 - 2000 19 4 12 2001 + 11 5 Not known or 6 unclear 20 7 __4_ -- Total 100 T°tal 100 Table A.19. Number on State Table A.20. Method Used to Committee Select Delegates to National Con- vention (Schlesinger, 1974) O - 50 28 51 - 100 30 101 - 500 38 1 3 501+ __13 2 35 Total 100 3 26 4 30 5 6 Total 100 BIBLIOGRAPHY Axelrod, Robert. "Where the Votes Come From: An Analysis of Electoral Coalitions, 1952-1968" in American Political Science Review LXVI (March 1972) pp. 11-20. Bernstein, Robert A. "Divisive Primaries Do Hurt: U. 3. Senate Races, 1956-1972" in American Political Science Review LXXI (June 1977) pp. 440-446. Blalock, Hubert M. "Causal Inference and Measures of Association" in American Political Science Review LXI (March 1967) pp. 130-136. Blau, Peter and Meyer, Marshall. Bureaucracy in Modern Society (N.Y.: Random House, 1971). Bowman, Lewis and Boynton, G. R. "Activities and Role Definitions of Grass Roots Party Leaders" in William E. Wright A Comparative Study of Party Organization (Columbus: Charles E. Merrill Publishing Company, 1971). Bullock, Charles. ”House Careerists: Changing Patterns of Longevity and Attrition" in American Political Science Review LXVI (Dec. 1972) pp. 1295-1300. Campbell, Angus; Converse, Philip; Miller, Warren; and Stokes, Donald. The American Voter (N.Y.: John Wiley & Sons Inc., 1960). Conway, M. Margaret and Feigert, Frank B. "Motivation, Incentive Systems, and Political Party Organization" in American Political Science Review LXII (Dec. 1968) pp. 1169-1183. Conway, M. Marget and Feigert, Frank "Motivations and Task Performance Among Party Precinct Workers" in Western Political Quarterly 27 (Dec. 1974) pp. 693-709. Cowart, Andrew. "Electoral Choice in American States" in American Political Science Review LXVII (Sept. 1973) pp. 835-853. Cutright, Phillip and Rossi, Peter H. "Party Organization in Primary Elections" American Journal of Sociology 44 (Nov. 1958) pp. 262- 269. Curry, R. L. and Wade, L. L. A Theory of Political Exchange (Engle- wood Cliffs: Prentice Hall, 1968). l77 178 Dahl, Robert. Who Governs (New Haven: Yale University Press, 1961). Duverger, Maurice. Political Parties. (N.Y.: John Wiley and Sons, Inc. 1954). Eldersveld, Samuel J. Political Parties: A Behavioral Analysis. (Chicago: Rand McNally & Company, 1964). Epstein, Leon. Politics in Wisconsin. (Madison: University of Wisconsin Press, 1958). Eulau, Heinz; Wahlke, John C.; Buchanan, William and Ferguson, LeRoy. "Career Perspectives of American State Legislators" in Dwaine Marvick (ed) Political Decision-Makers (Glencoe, Ill: The Free Press, 1961). Ewing, Cortez. "Primaries as Real Elections" Southwestern Social Science Quarterly. 29 (March 1949): pp. 293-298. Fiorina, Morris; Rohde, David and Wissel, Peter. "Historical Change in House Turnover" in Norman J. Ornstein (ed) Congress in Change Evolution and Reform (N.Y.: Prager, 1975) pp. 72-88. Francis, Wayne. Legislative Issues in the Fifty States. (Chicago: Rand McNally Co., 1967). Hennesy, Bernard. "On the Study of Party Organization" in William J. Crotty (ed) Approaches to the Study Of Party Organization (Boston: Allyn and Bacon, 1968). Hunter, Floyd. Community Power Structure (Chapel Hill: University of North Carolina Press, 1953). Johnson, Donald Bruce and Gibson, James. "The Divisive Primary Revisited: Party Activists in Iowa" in American Political Science Review LXVIII (March, 1974) pp. 67-77. Johnston, J. Econometric Methods. (2nd ed) (N.Y.:McGraw—Hill Book Company, 1963). Jonas, Frank. Western Politics. (Salt Lake City: University of Utah Press, 1961). Kementa, Jan. Elements of Econometrics (N.Y.: The Macmillan Company, 1971). Key, V. 0. Southern Politics (N.Y.: Vintage Books, 1949). Key, V. 0. American State Politics: An Introduction (N.Y.: Alfred A. Knopf, Inc., 1956). Key, V. 0. Politics, Parties, and Pressure Groups (4th ed). (N.Y.: Thomas Y. Crowell Company 1958). 179 Kostrowky, Warren Lee. "Party and Incumbency in Post War Senate Elections: Patterns and Trends" American Political Science Review LXVII (Dec. 1973) pp. 1213-1234. Lazarsfeld, Paul; Berelson, Bernard; Gaudet, Hazel. The People's Choice N.Y.: Duell, Sloan & Pearce, 1944). Leege, David Calhoun. "Control in the Party Convention Nominating System: The Case of Indiana", pp. 200-228 in James B. Kessler Empirical Studies of Indiana Politics (Bloomington, U. of Indiana Press, 1970). Lipset, Seymour Martin; Trow, Martin; Coleman, James. Union Democracy. (Garden City: Anchor Books, 1962). Lipset, Seymour Martin. Political Man: The Social Base of Politics. (New York: Doubleday, 1963). Madison, James. "The Federalist Papers, NO. 10" in The Federalist Papers Alexander Hamilton, James Madison and John Jay, Roy D. Fairfield (ed). (Garden City: Anchor Books, 1966 (1789)). Matthews, Donald R. and Prothro, James W. Negroes and the New Southern Politics. (N.Y.: Harcourt, Brace and World Inc., 1966. McClosky, Herbert; Hoffman, Paul J. and O'Hara, Rosemary. ”Issue Con- flict and Consensus Among Party Leaders and Followers" in American Political Science Review LIV (June, 1960), pp. 406-427. - Merriam, Edward C. Primary Elections. (Chicago: The University of Chicago Press, 1908). Merriam, Edward C. and Overacker, Louis. Primary Elections. (Chicago: U. of Chicago Press, 1928). Merton, Robert K. Social Theory and Social Structure. (Toronto: The Free Press, 1949). Michels, Robert. Political Parties. (N.Y.: Crowell-Collier Publish- ing Co., 1962 (1914)). Mileauer, Jerome and Suzner, George. Campaigns for the Massachusetts Senate. (Amherst: University of Massachusetts Press, 1974). Morehouse, Sarah McCally. "The Governor as Political Leader" in Herbert Jacob and Kenneht N. Vines Politics in the American States (Boston: Little Brown and Company, 1976). Nie, Norman H., Hull C. H., Jenkins, J. G., Steinbrenner, W., and Bent, D. Statistical Package for the Social Sciences (2nd ed). (N.Y.: McGraw—Hill Book Company, 1975). 180 Olson, David M. "Attributes of State Political Parties: An Explora- tion of Theory and Data" pp. 123-158 in James A. Riedel (ed) New Perspectives in State and Local Politics (Waltham, Mass., Xerox College Publishing, 1971). Patterson, Samuel. "Characteristics of Party Leaders" in Western Political Quarterly XVI (June, 1963) 332-52. Patterson, Samuel and Boynton, G. R. "Legislative Recruitment in a Civic Culture" Social Science Quarterly (Sept. 1969) pp. 243— 263. Patterson, Samuel "American State Legislatures and Public Policy" in Herbert Jacob and Kenneth N. Vines (3rd ed.) in Politics in the American States (Boston: Little, Brown and Company, 1976). Rae, Douglas W. and Taylor, Michael. The Analysis of Political Clevages. (New Haven: Yale University Press, 1970). Ranney, Austin. "Changing the Rules of the Game" in James David Barber Choosing the President (Englewood Cliffs: Prentice Hall, 1974). - Ranney, Austin. "Parties in State Politics" in Herbert Jacob and Kenneth N. Vines. Politics in the American States. (Boston: Little Brown and Company, 1976). Rohde, David. "Risk-Bearing and Progressive Ambition: The Case of Representatives in The U. S. House of Representatives". A paper given Aug. 1974 at the Conference on Uncertainty, Political Processes and Public Policy in San Diego. Schlesinger, Joseph. Ambition and Politics: Political Careers in the United States (Chicago: Rand McNally & Company, 1966). Schlesinger, Joseph. "Party Units" in David L. Sills (ed.) International Encyclopedia of the Social Sciences Vol. 11 (N.Y.: The Macmillan Company and the Free Press, 1968) pp. 428-452. Schlesinger, Joseph. "Scale for defining formal locus of power to name and control from states to national party nominating convention" Unpublished, Michigan State University, October. 1974. Schlesinger, Joseph. "Control Over State Nominations." Unpublished Scale, Michigan State University Oct. 1974. Schlesinger, Joseph. "Entry Base Scale" Unpublished, Michigan State University, Oct. 1974. Schattschneider, E. E. Party Government. (New York: Rinehart and Co., 1942). 181 Seligman, Lester. "Political Recruitment and Party Structure a Case Study" in American Political Science Review LV (March 1961) pp. 77-86. ' Snowiss, Leo M. "Congressional Recruitment and Representation" American Political Science Review, LX (Sept. 1966); pp. 627-639. Sorauf, Frank. Party and Representation, Legislative Politics in Pennsylvania. (N.Y.: Atherton Press, 1963). Sorauf, Frank J. Political Parties in the American System. (Boston: Little Brown & Company Inc., 1964). Smith, Terry. Primary Election Structure and Political Behavior. Unpublished dissertation, Michigan State University, 1973. Standing, W. H. and Robinson, J. A. "Inter-Party Competition and Primary Contesting" American Political Science Review LII (December 1958) 1066—77. Stokes, Donald and Miller, Waren. "Party Government and the Saliency of Congress" in Public Opinion Quarterly. 26 (Winter 1966) pp. 531-546. Tobin, Richard J. and Keynes, Edward. "Institutional Differences in the Recruitment Process: A Four-State Study" American Journal of Political Science. XIX (Nov., 1975) 667-682. Turner, Julius. "Primary Elections as the Alternate to Party Competi- tion in'Safe'iiistricts" Journal of Politics 15 (May 1953) 197-210. Watts, Thomas M. "Party Leadership and County Candidate Selection" pp. 174-199 in James B. Kessler Empirical Studies of Indiana Politics (Bloomington, U. of Indiana Press, 1970). Weber, Ronald E. "Competitive and Organizational Dimensions of American State Party Systems" unpublished paper delivered at the Northeast Political Science Association Nov. 1969. Wigins, Charles and Turk, 'William "State Party Chairmen: A Profile" in Western Political Quarterly 23 (June 1970) 321-32. Wilson, James Q. The Amateur Democrat (Chicago: University of Chicago Press, 1962). Winkler, Robert and Hays, William. Statistics: Probability, Infer- ence and Decision. Second edition (N.Y.: Holt Rinehart and Winston, 1970). Wolfe, Arthur Coffman. The Direct Primary in American Politics. An unpublished dissertation, University of Michigan, 1966. 182 Wonnacott, Ronald J. and Wonnacott, Thomas H. Econometrics. (N.Y.: John Wiley and Sons, Inc., 1970). "I11111111111111111'“