PRIMARY ELECTION STRUCTURE AND POLITICAL BEHAVIOR THESIS FOR THE DEGREE OF Ph. D. MICHIGAN STATE UNIVERSITY TERRY BRUCE SMITH I 973 II I III III I I II IIIIII IIII IIII III IIII III III III III II II 3 12931 LIBRARY Michigan State University This is to certify that the thesis entitled Primary Election Structure and Political Behavior presented by Terry Bruce Smith has been accepted towards fulfillment of the requirements for Ph.D. degree mPolitical Science I/w/M [FWQR Major professor 0'2 7'33 ' I Date // I47 3’ o-ms ABSTRACT PRIMARY ELECTION STRUCTURE AND POLITICAL BEHAVIOR By Terry Bruce Smith The primary election, a purely American institution, has received much scholarly attention, most of which is descriptive, normative, or of the case-study genre. The important empirical studies, chiefly by V. 0. Key, deal with comparisons of states which nominate by primary with states which nominate in convention, with the influence of the primary on inter-party competition, or with the influence of incumbency or var- ious levels of inter-party competition on primary election outcomes. The present study investigates an additional important dimension of primary elections: structure. The basic research question is: is the behavior of voters and office seekers different in states which nom- inate in the open primary.than from states which nominate in the closed primary? ‘We find that behavior does vary, and in ways contrary to theo- retical expectations. Our data is all available gubernatorial primary and general elections held since 1900 (1533 of the 1600 total elections). Theoretically, we eXpect political behavior to be more volatile generally in states with cpen primaries, wherein the potential for voter "raiding" exists, and wherein candidates have a larger potential elector- ate to which to appeal. From.our theory, in cpen primary states we ex- pect to find more voters switching parties between consecutive primaries, and fewer voters switching parties between a general election and the following primary election, when compared to closed primary states. Terry Bruce Smith These expectations are substantiated. However, by assuming "raiding" to be the dominant motivation of the open primary voter who switches parties between the primary and the following general election, we ex- pected greater variation in cpen primary states and, finding the reverse to be true, conclude that when open primary voters switch, it is to en- dorse a preferred candidate in the opposition party and not to raid in an attempt to nominate a weak one. Also, from our theory, we expect more intra-party competitiveness in open primary states than in closed, assuming that the entry of non-party members into a partisan primary tends to intensify intradparty conflict. we construct various measures of intra-party competitiveness, derive hypotheses from our theory, and upon testing them find that intra-party conflict is less in open primary states than in closed, especially when one of the two parties is dominant. we find no difference between open and closed states in size of incumbent plurality and incumbent renomina- tion rate. One hypothesized relationship is supported -- fewer renomina- ted incumbents are reelected in open states than in closed. Difference of means and proportions tests are employed to test hypotheses. we also find regional and party variations in intra-party competi- tiveness. we find primary nominations to be carefully orchestrated by party organizations in the Northeast states but typically less tightly controlled elsewhere, least so in the Border and Southern states. Yet despite regional variations, during the period 1920-60 voters generally behave as though the primary were a meaningful event: we find them voting in the primary of the party whose nominee is most likely to capture the reins of government and therefore, as Governor, execute policy. we also find.Democratic candidates behaving in a theoretically predictable manner, typically competing for a nomination when it is worth having and leaving lerry Bruce Smith well enough alone when pursuit of the governorship is frivolous. 0n the other hand, for the period examined, we find Republican candidates behaving in theoretically unanticipated ways, vigorously contesting many primaries, the winner of which is likely to face defeat in the general election, and failing to contest many primaries, the winner of which is likely to win the general election. These findings evolve from correlat- ing measures of intra-party competition with a measure of inter-party competition. PRIMARY ELECTION STRUCTURE AND POLITICAL BEHAVIOR By Terry Bruce Smith A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Political Science 1973 ACKNOWLEDGEMENTS Because of the following people, this thesis is now part of the literature on political behavior: ‘Professors Joseph Schlesinger, Paul Conn, and David Meltz, who chaired and were members of my dissertation committee, respectively; Professor Spencer Wellhoffer, my fourth reader; Harold 01d, Paul Hain, and Marty Levine, who provided early criticism which helped sharpen up ideas; Dottie Hitchcock and Ted Tupper, who rendered keypunching and computer programming help; Pat Norris, an under- graduate student who independently hit upon my basic research question and in a paper gave preliminary confirmation to some of my hypothesis; my wife Janie, who is always nice to come home to; and finally, Robert Hebeler, a business associate of my father, who in 1968 inspired the whole idea when he advised my father, a Republican, to vote in the Demo- cratic Senatorial primary in Missouri by saying: "Since we aren't going to get the Republican elected, we might as well get the least obnoxious 3.0.8. in the Democratic Party nominated." I thank them all. ii Ibo. . k TABLE OF CONTENTS Chapter I. ImthCTIoN mm ‘1}i330hl' Introduction . . . . . . . . . . . . . . . . . Theory . . . . . . . . . . . . . . . . . . . Rationality . . . . . . . . . . . . . . . . Hypotheses . . . . . . . . . . . . . . . . . . Intervening Variables . . . . . . . . . . . Difficulty in Getting on the Ballot . . Preprimary Endorsements . . . . . . . . Inter-party Competition . . . . . . . . The Index of Inter-party Competition . Longitudinal Hypotheses . . . . . . . . . . Static Hypotheses . . . . . . . . . . . . Voter Behavior . . . . . . . . . . . Candidate Behavior .. . . . . . . . . Incumbency . . . . . . . . . . . . . Conclusion......oo........ 0 Summary . . . . . . . . . . . . . . . . . . . Backnotes, Chapter I . . . . . . . . . . . . Chapter II. THE IMPACT OF PRIMARY STRUCTURE: IntrOdUCtiOD o o o o o o o 0 Longitudinal Hypotheses . . . . . . Static Hypotheses . . . . . . . . . . . . Voter Behavior . . . Plurality . . . . Competitiveness and Fragmentation Dmmmmmy .. ... ... ... Candidate Behavior . . . . . . . . . Contesting, Unopposed Nominations, and Incumbent Challenges . . . . . . . Number of Candidates . . . . . . . . . Summagg.....ooo............ Backnotes, Chapter II . . . . . . . . . . . . . U) 000.000.000.000. 7-3 Chapter III. CUBERNATORIAL NOMINATIONS IN THE TWENTIETH CENTURY: REGIONAL AND LONGITUDINAL VARIATIONS IntrOdUCtion o o o o o o o o o o o Primagy'lypg and Political Bra . . Regional Behavioral Variations . . Contesting . . . . . . . . . . . Number of Serious Candidates . . Plurality............. Intra-party Competitiveness . . . . Regional and.Longitudinal Variations . Inter-party Competition 0 o o o o o Intra-party Competition: Democrats Intradparty Competition: Republicans SUIT“ ooooooooooooooooo BaCknOtes, Chapter III 0 o o o o o o o o iii TESTS Chapter IV. CCNCLUSION o o o o o o o o o o o o o o 0 Party Organization and Primary Elections . . . . Party Organization as an Independent Variable Party Organization as a Dependent Variable . conClUSionooooooooooooooooooo Suggestions for Further Inquiry . . . . . . . . . Backnotes, Chapter IV . . . . . . . . . . . . . . BIBLIOGPAPI‘IY......ooooooOOOOoooo APPENDICES . . . . . . . . . . . . . . . . . . . . . APPENDIX A: Type of Primary, for Selected Years APPENDIX B: Primary Type, by State and.Year . . APPENDIX C: Questionnaire Sent to Secretaries of for Information about Primary Structure . . . iv State LIST OF TABLES Table Page 1. The Index of Inter-party Competition . . . . . . . . . . lb 2. Average Inter-party Percent Variation in.Demo- cratic Percent of Total Two-party Vote, by PflmaryType............oo......... 33 3. Difference of Means Test of the Average between- Primary Percent Variation in Democratic Turnout, by Fri-mm Type O O O O O O O O O O O O O O O O O O 0 3h h. Average Percent Variation between Primary Election and Following General Election in Democratic Percent of Total Two-party Turnout, by Primary Type...oooooooooooooooooooooo 35 5. Difference of Means Test of the Average Variation between Primary and Following General Election in Democratic Percent of Total Two-party Turnout, byPrimaI'yTy‘pe................... 35 6. Average Variation between General Election and Following Primary Election in Democratic Percent of Total Two-party Vote, by Primary Type . . . . . . . 36 7. Difference of Means Test of the Average Variation between.General Election and Following Primary Election in Democratic Percent of Total Two-party 'I‘urnout,byPrinaIyType............... 36 8. Average Plurality of Leading Vote-getter in Primary, byPrimaryType...................110 9. Average Plurality of Leading Vote-getter in Primary, by Primary Type, Controlling for Inter-party Competition......o.............. ’40 10. Average Indices of Competitiveness and Fragmenta- tion, by Primary Type . . . . . . . . . . . . . . . . hl 11. Average Index of Competitiveness, by Primary Type, Controlling for Inter-party Competition . . . . . . . L1 12. Average Index of Fragmentation, by Primary Type, Controlling for Inter—party Competition . . . . . . . b2 Table Page 13. Percent of Nominations Won by Incumbent, and Average Plurality of Incumbent, by Primary fly-p6 O O O O O O O O I O O O O O O O O O O O O O O O O 143 lb. Percent of Nominations Won by Incumbent, by Primary Type, Controlling for Inter-party C 0mm titi on . . C . . C C . 0 . . . C O O O O O . . . b3 15. Average Vote Percent of Incumbents, by Primary Type, Controlling for Inter-party Competition . . . . Lb 16. Percent of General Elections won by Renominated Incumbent, by Primary Type . . . . . . . . . . . . . . Lb 17. Percent of General Elections won by Renominated Incumbent, by Primary Type, Controlling for Inter-party Competition . . . . . . . . . . . . . . . LS 18. Percent of Primaries Contested, by Primary Type . . . . L6 19. Percent of Primaries Contested, by Primary Type, Controlling for Inter-party Competition . . . . . . . L6 20. Percent of Nominations UnOpposed, by Primary Type O O O O I O O O O O O O O O O O O O O O O O O O C h? 21. Percent of Nominations Unopposed, by Primary Type, Controlling for Inter-party Competition . . . . h? 22. Percent of Incumbent Challenges, by Primary Type O O O O O O O O O O O O O O O O O O O O O O O O O ’48 23. Percent of Incumbent Challenges, by Primary Type, Controlling for Inter-party Competition . . . . h9 2h. Average Number of Candidates, by Primary Type . . . . . 50 25. Average Number of Candidates, by Primary Type, Controlling for Inter-party Competition . . . . . . . SO 26. Average Number of Serious Candidates, by Primary TM 0 O O O O O O O O O O O O O O O O O C O O O O O O 51 27. Average Number of Serious Candidates, by Primary Type, Controlling for Inter-party Competition . . . . 51 28. State Population Size and Average Number of Serious Candidates, Both Parties . . . . . . . . . . . . . . . Sh 29. Primary Type and Inter-party Competition Status . . . . SS Table 30. 31. 32. 33. 311 o 35 . 36c 37. 38. 39. L0. hl. Difference of Means Test of the Average Vari- ation between Change-of—party General Election and Following Primary Election in Democratic Percent of Total Two-party Turnout, Winner and Loser O O O O I O O O O O O O O O O O O O O O O 0 Difference of Means Test of the Average Vari- ation between Change-of-party General Election and Following Primary Election in Democratic Percent of Total Two-party Turnout, Winner and Loser, by Primary Type. . . . . . . . . . . . . . Democratic Percent of Two-party Turnout in Primary and General Elections for Governor in washington 3111031936ooooooooooooooooooo Type of Primary, by Region, for Selected Years . . Primary Contesting Rate, by Region, Both Parties . Average Number of Serious Candidates, by Region, BOth Parties 0 O O O O O O O O O O O O O O O O 0 Average Plurality by Winning Candidate, by Region, Both Parties 0 O O O O I O O O O O O O O O O O 0 Average Index of Competitiveness, by Region, BOthparI/iesoooooooooooooooooo Republican Election Turnout: A Comparison of Primary and General Election Percent of the Two- party Vote, by Region, 1920-1960 . . . . . . . . Democratic Intra-party Competitiveness: A Com- parison of Indices of Competitiveness and Fragmentation with Democratic General Election Turnout Percent, by Region, 1920—1960 . . . . . . Republican Intra-party Competitiveness: A Com- parison of Indices of Competitiveness and Fragmentation with Republican General Election Turnout Percent, by Region (Excluding the South), 1920-1960 0 I I O O O O O O O I O O O O O O O 0 0 Primary Type, by State and Year . . . . . . . . . . Page 56 57 S7 60 62 62 63 63 6h 67 69 (Appendix B) II LIST OF FIGURES Figure Page 1. Type of Primary, for Selected Years . . . . . . . (Appendix A) viii PRIMARI'ELECTION STRUCTURE AND POLITICAL BEHAVIOR By Terry Bruce Smith Chapter I Introduction and Theory Introduction Winston Churchill said it: "The first responsibility of a repre- sentative to his constituents is to get elected." In stating this, Churchill overlooks a less obvious but often more difficult responsi- bility of a representative: to get nominated. "Nomination" is the antecedent act; once nominated by his party, a politician often is easily elected, and reelected. On the other hand, because of the nature of the nomination struggle, politicians frequently lose general elections in which all other objective factors would indicate certain victory. This initial step on the general election ballot, des- pite its obvious importance, is often underemphasized by political re- searchers, many of whom see the decisive election as being the one in which voters choose among the parties (and their previously-designated candidates) and determine who will control the government. Schattschneider, however, was not beguiled by this excessively casual view of the electoral process: The nature of the nominating procedure determines the nature of the party; he who can make the nomination is the owner of the party. This is therefore one of the best points at which to observe the distribution of power within the party. (6h) The history of nominating procedure in the United States is a his- tory of expansive democracy. From.the legislative caucus to the nominat- ing convention and finally to the direct primary, we note a constantly increasing role played by the voter.1 No other democratic nation uses 1 2 the direct primary as a means of nomination.2 Given the peculiarity of the system, American political scientists noted a need for detailed study of this institution at the dawn of the behavioral era of political science: In the United States, more attention has been mid to nominating methods than in other countries. A nation- wide study of the relative merits of the delegate conven- tion system, the direct primary, the non-partisan primary, and other nominating devices needs to be made . . . . An analysis might be made of the operation of the primary where there is a short ballot and where there is a long ballot, where popular participation is high and where it is low, and where party traditions are strong and where they are weak. Standard specifications regarding petition requirements, tests of party allegiance and provisions against minority nominations should be worked out. (Merriam, 19308 33) Indeed, much literature on the topic of primary elections was avail- able even by 1930, and much was to follow. But until l9h9 this literature was either descriptive,3 and/or normative,h and that which dealt with the tapic empirically was of the case study genre, lacking theoretical im- portance and methodological rigor.S There were a few'useful contributions, in this preanalytic era in the study of primaries.6 Berdahl (191:2) takes the first systematic, com- parative look at the phenomenon of "raiding" -- the practice of sending or otherwise seducing voters normally affiliated with one party to vote in aunother's primary. He finds that, in Illinois and New Jersey during the 1930's, the degree to which party irregularity occurred seems to be a fm’lction of the ability of the urban party machines to entice "surplus" V°ters of the dominant party to vote in the minority party's primary. H13 Study has a "good-government" normative bent, and he ends on an anti- macmine prescriptive note, decrying the practice of raiding and calling f°r corrective legislation. Ewing (1919) takes note of the fact that primaries were the "real .... 5‘ 3 elections" in the South and follows his article with a book (1953) de- voted exclusively to the use of the primary election in the South as the alternative to two-party competition. V. 0. Key's Southern Politics (l9h9) is the first intensive analysis of the behavioral effects of primary elections on political systems. Key finds that in the South the primary did not serve as a satisfactory re- placement for two-party competition, in that it deprived the organizations of both the dominant and the minority parties of their basic function -- nominating suitable candidates -- thereby weakening parties as effective governing institutions (385-h62). Key later looks at the behavioral manifestations of primaries in non-South states as well in American State Politics (1956). Focussing on the problem of "party atrOphy," he finds a "tendency for popular interest to concentrate in the primary of the stronger party . . . ." (100), and " . . . . a marked decline . . . in the proportions of . . . contests settled by small margins Ett the general electioxa " (103-010. He also . notes that the primary is no guarantee of a genuine contest for the nom- ’ ination; fewer than two-thirds of the Republican primaries featured a candidate's winning with less than sixty percent of the vote, and only two Democratic primaries in five were similarly contested (116). Key shows that states which had retained the convention nominating system and states which had modified the primary with pro-primary endorsing procedures tended to maintain stronger party organizations and typically manifest a higher degree of inter-party competition than the states in which the primary by itself was the nominating vehicle (118-29). Other literature which follows Key's l9h9 study tends to substan- tiate his findings. Turner demonstrates that primaries do not provide a realistic alternative to inter-party competition in one-party Congressional h districts. Standing and Robinson find that, like voters, candidates also tend to enter the primary of the dominant party more often than that of the subordinate party. Outright and Rossi note that incumbents are inordinately favored in primaries due to superior personal organiza- tional resources and the name-rec0gnition factor, both factors weighing heavily in a political contest where the party identification cue is denied the voter. This analysis explains the earlier finding by Ewing (1953) that incumbents in the South win renomination more than ninety percent of the time (65), and partially explains why nominations for local offices in Iowa were frequently uncontested, as Porter finds. Some of this literature compares the behavioral influences among several states. Other portions of the literature compares states nomin- ating by primary and states nominating by convention. Nowhere does the literature deal directly with the problem of behavioral variance gmgpg states that have different primary structures. That is the problem analyzed in the present research. Primary structure is intuitively a key independent variable in the study of voting behavior and candidate behavior in primary elections. There are four distinct types of primaries: l. The blanket primary, wherein the voter receives a ballot for all parties and may vote in more than one party's primary for different offices. 2. The opgn primary, wherein the voter receives a bal- lot for all parties but may vote in only one party's primary. 3. The closed-challenge primary, wherein the voter publicly requests and is given a ballot for one party, and may be asked to demonstrate party loyalty by swearing to support the party's can- didates in the general election. h. The closed-enrollment primary, wherein the voter is given the ballot of the party with which he en- rolled when he registered to vote. 51 Intuitively, the differences among the above primary structures encourage greater or lesser party regularity among voters. Specifically, the more "closed" the primary, the more likely it is that voters will be constrained to vote in the primary of a party determined well before the election day, and the less likely it is that voters will be able to re- spond to campaign appeals and shifts in sentiment which take place dur- ing the period immediately preceding the election. It follows therefore that the more closed the primary, the more likely it is that primary election results for a given party will be determined by voters who regularly identify with that party, Conversely, The ease with which voters may move from party to party under the open primary doubtless creates uncer- tainties for the party leadership in its efforts to control nominations as well as in tests of strength between leadership factions. The primary of one party may be raided by the voters of another in order to assure the nomination of a weak candidate who can 7 be defeated in the general election. (Key, l96h: 391) In addition, the behavior of individuals seeking nomination should be affected by the structure of the primary in the state in which he seeks nomination. Specifically, the more open the primary, the more likely it is that a candidate's potential electorate will include more and more voters who do not regularly identify with his or any party. Therefore, a candidate's behavior -- whether or not to run, whether or not to challenge an incumbent -- will vary with the conditions imposed on his electorate by the primary structure. Bone neatly sums up the present line of inquiry when he asks: Does one method consistently produce demonstrably more capable nominees than another? Which system is the most "rational" from the standpoint of (l) the voter, (2) the candidate, (3) the party officialdom, and (h) party acti- vists and adherents? Can any type of primary impose a high degree of responsibility on a party organization for the purpose of selection of candidates yet keep cpen the chan- I ...}. .. a an. .yltrl~ 6 nels of recruitment to qualified persons? These ques- tions are far more easily raised than satisfactorily answered (268) In what follows we attempt to answer satisfactorily these, and other, questions concerning the relationship between primary structure and political behavior. Theogy The Use of Structural Variables in Election Research There is much precedent for examining structural influences on political behavior.8 Voter registration procedures and residency re- quirements have received attention.9 50 has the form of the ballot.10 Schlesinger (1966) has analyzed Opportunity structures for elective of- fices in the states. Various scholars suggest that primary election structure indeed may be an important independent variable. Bone indicates that "insur- gents, mavericks," and other outsiders might prefer the cpen primary to the closed, because the potential electorate is larger, more diffuse, and more heterogeneous: The real significance of the primary is that it helps to keep open and flexible the channels of recruitment to public office. It provides a method or alternative means of gaining power other than winning the support of leaders of the party. The open primary does this somewhat better than the closed. Note: Bone supplies no data to substan- tiate this claim; (281) Key (l96h) says much the same thing, although like Bone he provides no empirical evidence to support his statement: Who Open primary at times makes difficult main- tenance of orientations differentiating the two parties and probably handicaps the lesser party in those juris- dictions in which one party holds a substantial advan- tage. The voters of the lesser party may find it more 7 attractive to exercise a balance of power in the prim- ary of the major party than to engage in the trouble- some task of building up their own party. (392) In a case study of the washington blanket primary, Ogden argues that gross party irregularity -- massive crossover voting -- has not oc- curred. Of the thirty-four statewide contests he studies only four showed massive switching across party lines. It is clear, however, that the potential for switching exists, and where washington voters per- ceived a clear opportunity to make an impact on one party's primary outcome, they did so. Despite Ogden's contention that "cross voting has been the exception rather than the rule, and . . . has had the ef- fect of saving the regular parties from widespread raids on the part of non-members" (38), Bone (a professor at the University of washington for some years) asserts that "the fact that weak partisans and mavericks can often win in the primaries is a.major reason why the washington legisla- ture, despite pleading from party chairmen, will not adopt a closed primary" (272-73). Thus, we find argument and counterargument about the impact of primary structure, with little in the way of supporting evidence. Rationality One of the difficulties with the analyses cited above is their weak theoretical orientation, although admittedly the authors did not set out to construct a theory of political behavior in primary elections. By as- suming rational political actors--both voters and candidates -- we are able to overcome much of the theoretical deficiency which characterizes research in this area. Curiously, Turner hints at the notion of candi- date rationality in his early work, but no researcher has pursued his implicit line of inquiry, despite its promise: Cross-filing in California and the blanket primary 8 in washington, combined with rapid population in- creases in these states, may create a fluid condi- tion in which it is difficult for the potential candidate to calculate his chances. In the closed primaries of the Northeast, where the population remains much more stable, candidates seem better able to recognize impending defeat, and therefore withdraw. (207) "Calculating chances" is of course the prime motivational stimulus for the rational candidate for nomination, who asks: "Can I win?" or at the very least, "Can I make an electoral impact sufficiently large to extract policy concessions, patronage, etc. from the victor?"11 A can- didate's estimating of his chances in an election does not take place in vacuo, however. " CA]mbition for office, like most other ambitions, de- velops with a specific situation, that it is a response to the possibili- ties which lie before the politician." (Schlesinger, 1966: 8). The situation to which Schlesinger refers is the structure of opportunities in a given state for a given office, and although his context is different from the present one, his theoretical foundation is applicable: The central assumption of ambition theory is that a politician's behavior is a response to his office goals . . . . [tabs politician as office seeker engages in political acts and makes decisions appropriate to gaining office . . . . [0] ur ambitious politician must act . . . in terms of the electorate of the office which he hopes to win . . . . (6) we assume that the rational candidate seeks nomination in the prim- ary election because he wants to win, or minimally, because he expects to influence the eventual nominee's behavior. He therefore utilizes appro- priate information about his chances for nomination or influence of the nominee.12 An item of information which is of prime importance to his chances is the structure of the primary. Ceteris paribus, the more open the primary, the more incomplete the information, since a candidate's potential electorate tends to include a larger bloc of voters who do not 9 regularly identify with his or any party. The implications of this situational uncertainty are manifest in diverse ways, and are discussed in detail in the next section. V. 0. Key (1966) suggests that we may safely assume the existence of the rational voter: "EiIn the large the electorate behaves as ra- tionally and responsibly as we should expect, given the clarity of the alternatives presented to it and the character of the information avail- able to it." (7).13 He is of course discussing national presidential elections when he describes "an electorate moved by concern about central and relevant questions of public policy, of governmental performance, and of executive personality" (7-8), but it is probably more persuasive to assume Key's notion of rationality for the primary election voter than for the general election voter. In the primary there is no party cue to support the indolent voter, no "valence issues" with which a candidate may be associated, simply because he is of a certain party. Indeed, about the only meaningful information the primary voter does have concerns "personality and governmental performance." we therefore assume that the rational primary voter votes for his most preferred candidate, if he can. If that candidate is of the voter's own party, then he can vote for him in his party's primary; and if the primary is open, he can also indirectly vote for his preferred candidate by crossing party lines and voting for a "weak" candidate in the opposi- tion party ("raiding"). In the latter case the voter is usually reason- ably certain that his preferred candidate is snreto win the nomination of the voter's own party.1h If the preferred candidate is not of the voter's party, and the prhmary is open, the voter can cross party lines and vote for him ("switching" as Opposed to "raiding;" see note 23 for a detailed 10 discussion of these phenomena). If the preferred candidate is not of the voter's party and the primary is closed, the voter will either ab- stain or vote for a "weak" candidate in his own party. Specific mani- festations of the structural influence on primary voting behavior will be discussed in the following section.15 Hypotheses To investigate the relationships between primary structure and political behavior we examine all available primary and general elections for Governor16 and all states using the primary to nominate between 1900 and 1968. There are a total of 1533 primary elections in the data col- lection (23 blanket, 303 open, 360 closed-challenge, and 8h? closed-en- rollment ) .17 Intervening Variables Before developing the hypotheses, we must consider two related factors which might disturb otherwise strong relationships between primary structure and political behavior: a candidate's difficulty in getting on the ballot and pre-primary endorsement practices. Difficulty in Getting 2n the Ballot It might be argued that there would be fewer candidates running for Governor in states wherein many petitions signatures must be garnered than in states where all a candidate must do is simply submit his name to the Secretary of State. There is a slight but insignificant tendency for a greater number of candidates to be on the ballot in "easy" states than in "difficult" states.18 However, Penniman's comments are appropriate to this purported problem: 11 The excuse for requiring numerous signatures [pn nominating petitions] is that otherwise there would be a plethora of candidates. The argument is of doubtful validity. If a man is well-known and popular or if he is brought forward by the machine merely to draw votes from a dangerous reform candidate, he will not be de- terred by a somewhat oppressive requirement. On the other hand, if he is obscure and without backing and yet can offer himself because few signatures are required, his name will not add to the complexity of the ballot or injure the prospects of other candidates. The task of the voter is affected, not by the presence of other can- didates -- he will pick the man he knows and wants as easily from among a dozen candidates as from among three -- but by the multiplicity of elective offices. (367) Preprimary'Endorsements The pre-primary endorsing conventions are of two types: official (permitted by state statute) and unofficial (forbidden by state statute but held anyway by party "clubs" or "caucuses").19 The effect of both types on candidate behavior is about the same: far larger prOportions of nominations are unapposed, not seriously contested or contested by only two men in endorsement states than in states without this provision.20 Bone notes the broad impact of pre-primary endorsement systems: ETIhey have strengthened and vitalized the parties. Designations are denied to nonentities, and candidates expecting to make a serious bid for nomination must solicit and obtain support from party activists. Pre- primary endorsements help to balance popular control and party control; and, in Colorado at least, have in- creased party competition because nearly every position on the ticket is filled most of the time. The convention links local committees with the state organizations and thereby reduces organizational atrophy. (278) A number of the hypotheses developed below involve structural in- fluences on the number of candidates contesting a nomination and the fre- quency of serious intra-party contests. Since pre-primary conventions limit candidate entry and mitigate intra-party strife, and thereby in- hibit the behavior of rational candidates (many of whom.g§§i£g_a divisive primary election), we eliminate from our data those 136 primary elections m 12 preceded by pre-primary endorsing conventions. we are left with 1397 primaries (22 blanket, 276 cpen, 3R9 closed-challenge, and 7L9 closed- enrollment). Inter-party Competition There is one other factor which may intrude massively into relation- ships between primary structure and behavior: inter-party'competition (see notes 12 and 15). The effect of differing levels of two-party com- petition on voter turnout and candidate entry rate is well documented for every level of elective office, and is best stated by Key (1956): Such factors as urbanism and incumbency'may affect the frequency of primary competition through time but . . . in the long run the incidence of primary competition is a function chiefly of the prospects for victory in the general election. (179) Jewell (1960) notes that in Kentucky state legislative races, pri- mary contesting increases with party dominance (527) and finds the same evidence when looking at legislative contests throughout the South (1967: 22). Standing and Robinson note similar tendencies in Indiana state and local contests (1071-77). Key (1956) describes similar variations in state legislative primaries in northern and Border states (1969-93). Wolfe also notes a strong relationship between party dominance and high levels of primary contesting for Governor, Senator, and Congressman from 1952-61: (911-96). Clearly, levels of inter-party competition must be controlled for in analyzing the structure-behavior relationships. In the hypotheses which follow, two tests are made of each: (1) without, and (2) with controls for inter-party competition. Thg_lndex 9f Inter-party Competition A measure for inter-party competition which accounts for the rational behavior perspective described above. Rational voters and candidates look to at 1'1 13 to the competitive nature of a party system for cues leading to apprOpri- ate behavior. The voter asks, "In which party's primary is my vote most likely to 'count'?" and he registers and/or votes according to the response to his question. The candidate asks, "How well can I expect to do in the general election, if I win the primary?" or, "If my goals are limited to influencing the eventual nominee, how likely is it that the nominee will win the general election, and therefore be able to reSpond to my influence through policy concessions, patronage, etc.?" and he either enters the primary race or he does not, according to his response. Two objective indicators of the level of inter-party competition ex- ist: average percent of the vote for a given office over time, and the frequency with which a given party wins a given office over time. Neither of these measures, taken separately, are particularly useful (although both are used separately in otherwise capable research effort321); the former schemes fail to account for minority parties who win occasional elections, and the latter accentuates the occasional win while masking the long-term vote distribution. Schlesinger (1955) devises a method which integrates the two indi- cators of competition, and we use a variation of his two-dimensional index. Our inter-party competition variable designates the competitive situation of the two major parties in a state for the five general elections immed- iately preceding a given primary election. 1h TABLE 1 WW; I - _, ..TIFQLIINDEII..0F..:IUFERntPABIIrQQITEILTIQAIMM. ”......" ..- A “ “\ fl AWWKNMB 'Icl-p-i- Election Results for Five Preceding A Competitive Status Gubernatorial Elections minority 145% of 2-party vote; no wins semi-competitive h5% of 2-party vote: 1 or more wins, or h5% of 2-party vote; no wins competitive h5% but 55% of 2-party vote; 1 or more wins semi-dominant 55% of 2-party vote; win all, or 55% of 2-party vote, 1 or more losses dominant 55% of 2-party vote; win all .. .—- -—cm‘~~U‘—o=- . ‘.~‘w~n --—"w-c~-q—~<-~wg.- :w— u.--..—rvu---.‘----~'-A..... ‘rp—wzu-«u-v: ..-—r---—- "‘Ml‘~n This index has two advantages: (1) it permits looking at a given election as a product of electoral tendencies in the immediate past, not as a product of an average situation established by some arbitrary and unchanging base date, and (2) it follows from the reasoning that rational candidates and voters would apply to an electoral situation. Specifically, a recent win gives hope for future success to a minority party member (either voter or candidate), despite the other objective indication of minority status: vote percent. On the other hand, a recent loss by the majority party instills fear for the future (and perhaps dispels party lethargy), despite its otherwise majority status. The index as constructed captures the nuance of likely calculations which political actors employ when determining their action vis-a-vis the primary election. Longitudinal Hypotheses Wide changes in the ratios of primary participa- tion between the parties reflect fairly durable changes in the relative sizes of the groups of strong partisans, a matter of perhaps more significance than momentary shifts in voting strength in response to appeals of particular campaigns. (Key, 1956: 101-02) 15 Both candidates and voters are aware of the point of real deci- sion in a political system; where one party dominates, it is the primary of that party; where the parties are competitive, it is the general elec- tion. It is also true that shifts over time either away from or toward a competitive situation in the general elections are recognized by ra- tional political actors, who respond appropriately: " E38 the average Democratic proportion of the general election vote declines, the propor- tion of all primary voters who vote in Democratic primaries declines but at a more rapid rate." (Key, 1956: lOO).22 Key offers these findings on the empirical basis of looking at all primaries, without regard for the type of primary. Below we develop hy- potheses testing longitudinal variations in voter behavior as a response to the constraints placed on rational voting by primary structure. The size of that portion of the electorate which switches from party to party in primary elections as a response to shifts in the tun-party competitive situation is associated with the legal ease with which switch- ing is permitted. Open primaries place the fewest restrictions on the voter; closed, the most. we expect more switching of voters where switching involves the least cost. This switching is reflected in change in one party's percentage of the total two-party turnout from one primary election to the next. Thus, Hypothesis 1 For a given office, the more open the primary, the greater the variation from primary to the following primary in a party's percentage of the total two-party turnout. Voter switching also takes place between the primary and the gen- eral election which follows. we expect that in an Open primary more voters will participate in one party's primary and vote for the other winginuulflf .A. s.- ” .. .. V .E~ he l6 party's nominee in the general election than in the closed primary, given that they have the relative freedom to do 50. Thus, Hypothe sis 2 For a given office, the more open the primary, the greater the variation from a primary to the following general election in a party's percentage of the total two-party turnout. Also reflected by primary structure is variation toward or away from inter-party balance for an office over time. For example, we expect that in a former one-party state that is trending toward a two-party com- petitive status, party turnouts in the primary will tend to approximate each other sooner in open than in closed primary states, where the shift in party turnoutbalance in the primary will lag because of the registra- tion requirements which make party affiliation change more costly. Simi- larly, we expect that in a former two-parry state that is trending toward one -party dominant, party turnouts in the primary will tend to resemble turnout in the general election sooner in open than in closed situations; i.e., voters will turn to the primary of the dominant party in a shorter time span. Thus, Hypothesis 3 For a given office, the more cpen the primary, the less the variation from a general election to the following primary election in a party's percentage of the total two-party turnout. To test these hypotheses, we compare Democratic primary and general election turnout percentages for all states, controlling for primary type. we calculate the average deviation for all appropriate election pairs, eXpecting that the deviation will be highest in open states for Hypo- thesis 1, lowest for Hypothesis 2. 1? Static Hypotheses Voter Behavior Voters respond in various ways to the stimuli of a particular election, as well as over time. As noted, voters tend to participate in inordinate numbers in the primary of the dominant party. Also, given prOper conditions, voters periodically conduct forays into par- I ties with which they do not affiliate for the purpose of either nominat- ing preferred candidates (switching) or nominating weak candidates I (raiding).23 Clearly, a voter is more capable of switching/raiding in an cpen primary situation, whether this action is organized by party leaders or not. we expect, therefore, that behavioral manifestations of switdhing/ raiding will be more apparent in open primary election results than in closed (see Hypothesis 2 for an indirect indicator of this expectation).2h Specifically, these manifestations, as they should appear in cpen primary states when compared with closed states, are (1) smaller average plural- ities for the leading vote-getter, (2) smaller average percentage differ- ences between the leading vote-getter and the runner-up, (3) smaller average percent totals received by the two top vote-getters, (h) less incidence of incumbent renomination, (5) smaller average pluralities for an incumbent, and (6) less incidence of a renominated incumbent winning the general election. Thus, Hypothesis h For a given office, the more open the primary, the smaller the average plurality for the leading vote getter. For this and all subsequent hypotheses we test without and with controls for inter-party competition. For test purposes plurality is 18 operationalized as a simple percentage of the total vote of the leader's party. , Hypothesis 5 For a given office, the more open the primary, the larger the average index of competitiveness. The index of competitiveness measures the level of intensity of intra-party competition between the two leading candidates. It is oper- ationalized by subtracting from 100 percent the difference in the vote percent of the top two candidates. The larger the index number, the more vigorously contested the primary. Hypothesis 6 For a given office, the more open the primary, the larger the average index of fragmentation. The index of fragmentation is a measure developed by Key (19h9: £21) which he used to determine empirically the degree of fragmentation and factionalism in state Democratic parties in the South. It is calculated by subtracting from 100 percent the gpp_of the vote percent obtained by the two top vote-getters. The larger the index number, the more dis- persed or fragmented the party. Hypothesis 7 For a given office, the more open the primary, the smaller the percent of primaries won by incumbents seeking renomination. Hypothesis 8 For a given office, the more open the primary, the smaller the average plurality for an incumbent seek- ing renomination. Hypothesis 9 For a given office, the more open the primary, the 19 smaller the percent of general elections won by re- nominated incumbents. Candidate Behavior Ceteris pgribus, the rational candidate responds to the nature of his electorate as determined by primary structure. Information about the electorate in open primary states is more uncertain than about closed primary electorates. The more open a primary, the less predictable is a given party's turnout, both in terms of size and composition. A can- didate would thus be more uncertain of the effect of a campaign on his electorate because of its tendency toward increased heterogeneity. Hetero- geneity in itself is not a delimiting factor -- the victorious nominee would have to face a varied electorate in the general election -- but a more heterogeneous electorate in a primary election, §§p§_party cue, tends to make the candidate's perception of the primary outcome more random than would a campaign in a closed primary state. Incumbency In addition to the intervening influence of inter-party competition, we must control for an additional situational variable in the context of candidate behavior: incumbency; The literature is replete with discus- sions of the effect of incumbency on primary contesting, and most of it suggests that a given candidate's chances for nomination are greater in a primary in which the incumbent does not seek renomination. Turner notes that in safe congressional districts between l9hh and 1950, only 35 of the 908 incumbents seeking renomination lost (four per- cent) and R72, Sr 52 percent, were unopposed (208). Ewing (1953) finds that 90.6 percent of all incumbent executives in the South won renomina- 1 all.» 0.; . 20 tion (65), and that uncontested primaries with the incumbent running outnumbered those without the incumbent running by'a margin of seven to three (60). Jewell (1967) notes similar dampening influences of in- cumbency on contesting in state legislative races in Border and South states, and adds that competition for the nominations "varies inversely with the length of incumbency." (26) Key (1956) notes the inhibiting effect of incumbency on contesting of Missouri state representative pri- maries (175-76). 'wolfe finds that "for all senatorial and congressional primaries outside the South [I952-6N] the incumbency effect even results in less primary competition within the majority party than within the minority party," although he finds this not to be true for Governor (60), There are two dissentors to the above evidence about the suppressive effect of incumbency. Epstein finds that in Wisconsin incumbency does not dampen contesting for state legislative seats (131-32: 200),25 and Jewell (1960) notes that in Kentucky incumbency does not inhibit contesting when party competition is controlled for (528). However, no one offers find- ings supporting speculation that incumbency encourages primary challenges. he suggest that hidden within these practically unanimous findings are interesting variations associated with different primary structures. we expect to find more contesting of primaries generally and more chal- lenging of incumbents particularly in open primary states than in closed. Specifically, we expect that when comparing open primary states to closed, we will find (1) more "contested" primaries, (2) more candidates entering primaries, (3) more "serious" candidates entering primaries, (h) fewer unopposed nominees, and (5) a higher rate of challenging incumbents. In the hypotheses which follow, we test without and with controls for in- cumbency (and, as always, for inter-party competition). 21 Hypothesis 10 For a given office, the more Open the primary, the more likely it is to be contested. We operationalize "contesting" identically to Turner: Itfle will assume that any primary contest in which the winner garnered two-thirds of the vote for the top two candidates was sufficiently one-sided so that most voters and politicians alike realized before the primary that there was little chance of defeating the [Ieading candidate] . (203) we suggest that the nature of open primaries not only makes this pre-primary calculation less certain, it also provides the structural con- text for making post-primary evidence of the reasonableness of the pre- primary realization less persuasive; i.e., more Open primaries will indeed wind up "contested" than closed primaries, regardless of the initial sup- position about the vigorousness of the challenge. Hypothesis ll For a given office, the more Open the primary, the larger the average number of candidates likely to compete for the nomination. Given the relative incompleteness of information about Open primary situations, we expect that in those states more candidates will have rea- listic expectations of winning or influencing the nominee than in closed states. Hypothesis 12 I For a given office, the more Open the primary, the larger the number of serious candidates likely to compete for the nomination. A "serious" candidate is defined as one who receives at least twenty percent of his party's primary vote, unless the leading candidate has less than fifty percent Of the vote, in which case a serious candidate is de- 22 fined as one who receives at least ten percent (this qualification ac- counts for a number of bitterly contested primaries, especially in the South, where no candidate received as much as twenty percent, but where there were obviously several "serious" candidates). we expect more serious candidates in open primaries for the same reason we eXpect lower indices of dispersion: the relative situational ambiguity will encourage serious candidates for the governorship to join the race because of the existence of a large potential electorate to which appeals would be fruit- less in a closed primary, where much of this electorate is either com- mitted to vote in the other party's primary or kept from voting altogether by reason Of unwillingness to declare a partisan affiliation. Hypothesis l3 For a given office, the more open the primary, the fewer the percentage of elections with unOpposed nom- inees. This hypothesis is another indication of our expectation that poten- tial challengers of strong nominees calculate more Opportunity for a primary election upset in Open primaries, given the diffuse character Of the electorate, and are therefore more willing to contest an otherwise unopposed nomination. Hypothesis 11: For a given office, the more open the primary, the larger the percentage of elections in which the in- cumbent is challenged. The reasoning for this hypothesis is identical to that for Hypo- thesis 13, only "incumbent" is substituted for "strong nominee." Chapter II of the thesis reports the results of tests of the hypo- theses. 1551““?ng 3.“ 23 Conclusion Chapter III is fundamentally a descriptive chapter delineating (1) regional variations in primary structure and behavior, and (2) the post-1900 history of primary structural variation and the dynamics of inter-party competition, plus an historical develOpment, by party, of intpg-party competition for Governor in all states. Chapter IV is in two parts. First is a discussion of party organ- ization as both an independent and dependent variable in the general con- text Of primary elections. The literature on this topic is diverse and somewhat speculative about the impact of organizations on primaries; the evidence about the impact of primaries on party organization is less equiv- ocal -- more scholars believe party organizations are seriously disrupted by primaries than believe primaries have little or no effect -- but the evidence about ggg£§g_of impact, and indeed whether or not that impact is harmful in the long run, is the subject of much scholarly debate and will be considered in detail. The final part Of Chapter IV contains a concluding statement and suggestions for future research. Summapy The study of nominating procedure is an important area of the study of political parties, and the study of the primary election process in the United States has received a fair amount Of attention. Most of the analytical work either compares party systems using the primary with sys- tems using other methods of nomination, or treats primaries generally as an independent variable associated with varying degrees of inter- and intra-party competition. 2h The present analysis looks at the different structures of primary elections and hypothesizes that variance in voting and candidate behavior is associated with distinguishably different types of primaries. In general, we expect parties in Open primary states to be internally more competitive, and more volative electorally over time, than parties in closed primary states. we test for these expected variations, applying appropriate controls, and draw conclusions about the structural influence E; of primary elections on political behavior. 25 Backnotes,pChapter I 1For a brief history of American nominating practices, see V. 0. Key, American State Politics: An Introduction (New York: Alfred A. Knopf, 1956}, 85:97; and V. 0. Key, Politics,_Parties and Pressure Groups (New York: Thomas R. Crowell, 196D); 370-76. See Arthur C. welfe, The Direct Primary in American Politics (Unpublished'Ph. D. dis- sertation, University Of Michigan, 1966), lL-35, for a more extensive discussion of this topic, and for a history of early primary election legislation, see Charles Merriam.and Louise Overacker, Primary Elec- tions. (Chicago: University of Chicago Press, 1928), chap. V. F- 2welfe notes that two primary-type elections are recorded in Australia (1952) and Great Britain (1957). These were ad hoc events aimed at purg- ing dissident members of parliament and apparently have not been repeated. welfe: 55. 3See Annals of the American Academy of Political and Social Science, 106 (1923), no. 11; Leon E. Aylsworth, "Primary Election Legislation of 1909-1910," American Political Science Review, 6 (1912), 60-7h; Lamar T. Beman, The Direct Primary‘INew‘York: H. w; Wilson 00., 1926); Clarence A. Berdahl, "The Richards Primary," American Political Science Review, 111 (1920), 93-105; James H. Booser, "The Origin of the Direct Primary," National Municipal Review, 2h (1935), 222-23; Horace Flack, "Notes on Current Legislation," American Political Science Review, 6 (1912), 60-7h; Frederic H. Guild, "The Indiana Primary," National.Municipa1 Review, 11 (1922), 286-90; Charles Kettleborough, "Legislative News and Notes," Amer- ican Political Science Review, 13 (1919), 26h-66; Richard E. Manning,“ Parpy Affiliation (washington, D. C.: Library of Congress, Legislative Reference Service, December 13, 1932); Charles Merriam, Primary Elections: A Study of the History and Tendencies of Primary Election ngislation (Chicago: University of Chicago Press, 1908)} Charles Merriam and Louise Overacker,‘Primary Elections (Chicago: University of Chicago Press, 1928); Louise Overacker, "Primary Election Legislation Of 1926-27," American Poli- tical Science Review, 22 (1928), 353-61; Louise Overacker, "Primary Elec- tion Legislation of 1928-29," American Political Science Review, 2b (1930), 370-80; Louise Overacker, "Primary Election Legislation of 1930-31," Amer- ican Political Science Review, 26 (1932), 29h- 300; Louise Overacker, "Pri- mary Election Legislation of 1932-33," American Political Science Review, 28 (193R), 265-70; Louise Overacker, "Direct Primary Legislation in 1936- 39," American Political Science Review, 3h (19h0), h99-506; Ormund P. Ray, "Primary Legislation, 192E-25," American Political Science Review, 20 (1926), 3h9-52; and.0rmund P. Ray, "Recent Primary and Election Laws," American Political Science Review, 13 (1919), 26h-7h. hCongressional Digest (entire issue), "The Direct Primary System," 5 (1926):Ino. 10; Charles Evans Hughes, "The Fate of the Direct Primary," National Municipal Review, 10 (1921), 23-31; Oliver McKee, Jr. "Direct Primary: A Failure and a Threat," Atlantic Menthly, lb8 (1931), 185- 93; and T. Henry walnut, "Gifford Pinchot and the Direct Primary," National Municipal Review, 11 (1922). SSee Ralph S. Boots, "The Trend of the Direct Primary," American Eplitical Science Review, 16 (1922), h12-31; John E. Briggs, "The Iowa Primary Interpreted," National Municipgl Review, 11 (1922), 282-86; 26 Richard S. Childs, "Rhode Island Tries the Primary," National Municipal Review, 38 (19h8), 126-29; L. M. Holland, The Direct Primary_in Georgia (Urbana: University of Illinois Press, 19h9: Illinois Studies in the Social Sciences, 30, no. h); Boyd A. Martin, The Direct'Primgry in Idaho (Palo Alto: Stanford University Press, l9h7); Arthur C. Millspaugh, "The Operation of the Direct Primary in Michigan," American Political Science Review, 10 (1916), 710-26; Danie1.PL Ogden, "The Blanket Primary and Party'Regularity in Washington," Pacific Northwest Quarterly, 39 (19h8), 33-38; James K. Pollack, The Direct Primary in Michigan (Ann Arbor: University of Michigan Press, 1933: Michigan Government Studies, n. lb); and Kirk H. Porter, "The Deserted.Primary in Iowa," American Political Science Review, 39 (1915), 732.10. 6The reasons for the paucity of useful literature on primary 8‘ elections are effectively summarized by Outright and Rossi: "Primary elections are not the most dramatic of political phenomena, and much re- search has been directed toward that most glamorous of American elections, the general election of a President. But there are additional obstacles to the study of primary elections. For one thing, primary elections occur within each major party. Each party is a relatively homogeneous group, . as compared with the variation in a wide number of socio-economic char- L acteristics between parties. Hence, the factors which help to account for so much of the variation in the vote garnered by each candidate in a general election do not apply in primary elections." (262) 7Much more will be said about the true motivations of the primary voter later. Key's suggestion that voters "raid" another party's primary in order to nominate a weak candidate is only a partial eXplanation of voting behavior in "cross-over" situations. Indeed, Pollack suggests that most crossing is in support of a preferred candidate whom the voter in- tends to support in the general election. (60) 8For general discussion of the impact of structural variables, see Hugh Bone, American Politics and the Party System, hth ed., (New'York: MeGraw-Hill Book Co., 1971), h59-7l; Angus Campbell, The American Voter (New York: John Wiley and Sons, Inc., 1960), chap. 11; V. 0. Key, Jr., Politics, Parties and.Pressure Groupg, 5th ed., (New York: Thomas R. Crowell, l96h); chap. 23; Robert E. Lane, Political Life (New York: The Free Press, 1959), 307-17; and Lester Milbrath, Political Participation (Chicago: Rand, McNally 8. Co., 1965). 9See Ralph Goldman, "Move -- Lose Your Vote," National Municipal Review, h5 (1956), 6-10; Stanley Kelley, Jr., Richard E. Ayres, and William 0. Bowen, "Registration and Voting: Putting First Things First," American Political Science Review, 61 (1967), 359-79; Donald R. Matthews and James w. Prothro, "Political Factors and Negro Voter Registration in the South," American Political Science Review, 57 (1963), 355-67; Merrie Ogul, "Residency Requirements," Midwest Journal of Political Science, 3 (1959), 25h-62; and Report on Registration and Voting_Participation (washington, D. C.: United States Government Printing Office, 1963). 10See Robert C. Brooks, "Voter's Vagaries: The Value of Position on a Ballot," National MMnicipal Review, 10 (1921), 161-65; Henry M. Bain and Donald S. Hecock, Ballot Position and Voter's Choice: The Ar- rangement of Names on the Ballot and Its Effects on the Voter (Detroit: 27 Wayne State University Press, 1957); Angus Campbell and warren Miller, "The Motivational Basis of Straight and Split-Ticket Voting," American Political Science Review, 51 (1957), 293-312; and Jack Walker, "Ballot Form and Voter Fatigue," Midwest Journal of Political Science, 10 (1966), hh8-63. 11This latter calculation is most important in the primary election where such concessions are often granted by the winner to the losing can- didate(s) or faction(s) in the interest of party unity in the general election. 'Winners of general elections are typically not burdened by any such obligations, given the zero-sum nature of that election. 12The level of inter-party competition in a given jurisdiction is 5 a critical variable, because it is the most important aggregate predictor of general election outcomes. It has obvious impact on a candidate's assessment of his chances of winning the general election, and it there- fore intrudes upon his calculations about entering the primary contest to begin with. we will consider the effect of inter-party competition on behavior in detail below. However, this one is not present within the context of the primary election situation when viewed a§_an isolated event. And, of course, the L candidate must prevail in the intra-party contest before the forces of inter-party competition even begin to come into play, Since the inter- party competition factor is, in this sense, analytically removed from the candidate's sum of information about his chances, information is by definition more incomplete for any given candidate at the primary election stage than for any given candidate at the general election stage. 13Indeed the SRC group, the most noted eXpositors of the social de- terminism model of voting behavior, suggests that rational voting was clearly operative in the 1968 presidential election. Philip Converse, et a1, "Continuity and Change in.American Politics: Parties and Issues in the 1968 Election," American Political Science Review, 63 (1969), 1083-1105. 1hSee Berdahl, supra. 15Inter-party competition influences voters as well as candidates in their behavior with respect to primaries: "As the normal balance of electoral strength shifts to the advantage of one party, popular atten- tion tends to center in the direct primary of that party, the arena of governing decision in the politics of the state." (Key, 1956: 10h). Yet when one holds inter-party competition constant, we expect that variation in voter behavior will still be associated with structural variation. 16Gubernatorial nominations seem to feature the highest turnout of any statewide constituency. Even in the lhl primaries in which the nominee for Governor was uncontested and the nomination for the U.S. Senate seat was contested, the total number of votes cast for the lone gubernatorial nominee was higher than the total cast for all the senatorial candidates in.forty-seven percent of the elections, about the same in five percent, and lower in forty percent. 171 was unable to obtain results for sixty-seven of the total of 28 1600 primaries held for Governor between 1900 and 1968; this is slightly more than four percent of the total. Most of the gaps are pro-1920, when election data in many states, eSpecially in the South, are fragmentary at best. The only post-1920 data missing are in Georgia (l92h) and Nebraska (1936). There are 21b, occasions in which one party held a pri- mary for Governor and the other either nominated by convention or offered no candidate at all; most of these occurred in the South. My data are from a number of sources: Richard M. Scammon, ed., America Votes, 2-8 (washington, D. C.: Government Affairs Institute, 1956-68) is the best source for recent primaries. Alexander Heard and D. S. Strong, Southern Primaries and Elections (Tuscaloosa: University of Alabama Press, 1950) contains the county-by-county results of gubernatorial primaries in the old Confederacy from 1920-b8. Kentucky results were obtained from Malcolm E. Jewell, Kentucky Votes, 3 (Lexington: University of Kentucky Press, 1963). Montana results were obtained from Ellis Waldron, An Atlas of Mentana Politics Since 186h (Missoula: Montana State University Press, 1958). Nevada results are found in John Koontz, Political History of Nevada, 5th ed. (Carson City: Nevada State Printing Office, 1965). All other data are from state manuals and bluebooks, or were obtained through personal correspondence with secretaries of state, state elec- tion boards, state libraries, or departments of political science in.the state universities. The people in Nebraska were the most willing to help; the peOple in Utah did not even bother to answer my inquiries. Only Connecticut and Delaware have never held a direct primary election for Governor, and only Delaware has never had provision.for a gubernatorial primary. ‘ -T" ".1 8Obtaining accurate information about candidacy requirements is nearly an impossible task. Many states change their requirements with practically the sitting of each new legislature. There were but two periods for which I could find reasonably reliable descriptions of re- quirements in most states: 1926-28 (from Merriam and Overacker) and l95h-56 (from the National Municipal League, Compilation of the N8 Dir- ect Primary Systems). There are several gradations of difficulty in getting on the ballot: filing only; filing plus paying a fee or filing plus gathering a few petition signatures; and filing plus gathering a relatively large number of signatures (e.g., more than one percent of the party's vote total in the previous gubernatorial election). These requirements I designated as "easy." Filing, plus gathering a relatively large number of signatures with a geographical distribution (e.g., sig- natures from at least twenty counties), or any sort of pre-primary en- dorsement requirement, I designated as "difficult." For the years analyzed, there were hOO candidates and sixty-nine elections with "easy" requirements, for an average of 5.8 candidates per election, while in the "difficult" states there were forty-six candidates in eleven elections for an average of h.2 However, in seven of the pri- maries in "easy" states the largest number of candidates ever to run in the history of the gubernatorial primary appeared during the years ex- amined (e.g., fourteen in washington in l92h, sixteen in Ohio in 1926, and eighteen in Oklahoma in 195h). When these seven extreme cases are removed, the remaining sixty-two elections have 317 candidates running, for an average of 5.1, as compared to the h.2 in the "difficult" states. It also must be recalled that the difficult states include pre-primary endorsement states like Utah and Colorado, where by law'CUtah) or by 29 custom (Colorado) only two candidates may appear on a party's ballot. 190f the 1533 available primaries, 136 (about nine percent) were preceded by a pre-primary endorsing convention. Unofficial pro-primary endorsing conventions: California: California Republican Assembly since 19h2; California Democratic Council since 1953 Official pre-primary endorsing conventions (for both parties unless other- wise indicated: Colorado: since 19h2 Connecticut: since 1956, provides for a "challenge" primary for any losing candidate with more than twenty percent of the convention delegates. No one has requested a primary for Governor. Idaho: 1963-71 Iowa: Democrats only in 1966 Massachusetts: 1932-37 and since 1951 Minnesota: 1921-23 ‘ Nebraska:‘ l9hh-S6 New Mexico: 19h9-55 and 1963-67 Rhode Island: since 19h8 South Dakota: 1917-29. Since 1929, a post-primary nominating con- vention must be called if no nominee gets more than thirty-five percent of the primary vote. A convention has been called once, in 1930, and the leading vote-getter in the primary was not nominated in the convention. Utah: since 1937. Utah has a curious system; both parties tightly control the nominations by selecting in convention a maximum.of two candidates from each party, then they hold an open primary. 20The evidence on this point is persuasive. In California the Repub- lican Assembly endorsed thirty-five candidates for thirty-eight statewide contests between 19h2-58. Twenty-six were not seriously challenged (seventy- four percent). Of the twelve contests involving "serious" intra-party battles, seven endorsees won and two lost (six percent of the total). In that period endorsed candidates for Governor received an average of ninety— four percent of the primary vote. The two Democratic Council endorsees received eighty-eight and eighty-two percent of the vote (Rowe). In 1927 Colorado adopted its pre-primary system (wherein a potential candidate needs twenty percent of the delegate vote and three hundred pe- tition signatures to be placed on the ballot); ballot position is according to size of delegate vote. Eyre and Martin assert that the assemblies nom- inate most candidates, not the primaries. "Prospective candidates who do not have broad support within the party are either eliminated or discour- aged by the assembly system." (62) Bone notes that because of Colorado's system, three-fourths of the state house of representative nominations are uncontested (277). 21For the most noted example of the use of the first indicator, see Richard E. Dawson and James A. Robinson, "Interparty Competition, Economic Variables, and'welfare Policies in the American States," Journal of P011- 3122: 25 (1963), 265-89. Actually, Dawson and Robinson combine a number of executive and legislative offices into a composite index of competi- tion, concealing massive inter-office differences in many states. For an 30 example of the use of the second.indicator, see Standing and Robinson (1068-70). They call a constituency "safe" for one party if that party has won at least five consecutive elections, and "competitive" if one party wins one of three consecutive elections, but they do not account for vote percentage. For a systematic review of previous work on indices of inter-party competition, see David G. Pfeiffer, "the Measurement of Inter-Party Competition and Systemic Stability," American Political Science Review, (1967), h57-67. See also Paul T. David, "How Can an Index of Party Com- petition Best be Derived?“, Journal of Politics, 3h (1972), 632-38; and see David B. Meltz, "An Index for the Measurement of Inter-Party Compe- tition," (unpublished manuscript, Michigan State University, 1972), for a delicate measure of competition which seems to overcome many of the F“ theoretical failings of earlier indices. 22Turner notes a variation on Key‘s theme: ". . . the habit of primary competition develops only with long experience under one-party . rule, although former one-party states retain the competitive pattern after their change of status. Recently converted one-party states have not yet developed the competitive pattern in their primaries." (206) 23Sorauf's claim that there is "no evidence of raiding" (206) is overstatement. Berdahl documents considerable organized raiding in urban Illinois and New Jersey (38-50). Pollack notes that "Eilarge numbers of Democrats and independents have taken part in the well-attended Republi- can primaries" in Michigan (27). Millspaugh displays results from two Wayne County, Michigan wards in the 1912 gubernatorial contest (717-18): Pct. 1, ward 1 Pct. 1, ward 2 Primary General Primary General Republican 265 l 166 38 Democrat 12 259 2 1h? Penniman notes statewide results of three Minnesota gubernatorial con- tests (336-37): 1918 1920 1922 Primary' General Primary General Primary General Rep 125,115 155,789 360,263 366,2u7 500,620 367,929 Dem 28,3h0 112,576 22,h35 2h7,7h6 19,108 51,061 Martin cites the 1938 Idaho contest for Governor (70): Primagy General Republican 86,855 77,697 Democrat 30,398 106,208 To be sure, all of the above elections took place in open primary states. Aside from Berdahl, there is no evidence to either prove or refute Sorauf's initial point in the context of closed primary states. But Bone notes: "Party politicians and candidates complain about 'raiding' or a crossover of voters aimed to nominate a weak candidate. Some voters may admit to a crossover but declare that it was for the purpose of getting the ‘best man,‘ not the weakest, nominated. Many charges about raiding are hard to prove. It is contended by politicians and by many academicians that 31 the Open and blanket primaries permit raiding, and the latter to a greater degree than the former [see Ogden, supra]. Although this is true, the con- tention that no raiding is possible in a closed primary may be seriously questioned. Even in closed primaries, voters insist on switching their parties; and both candidates and parties will encourage raiding if they can find ways to facilitate it." (270-71) It is, of course, our contention that "facilitation" is easier in cpen primaries than in closed. 2hwelfe finds a larger percent of voters crossing parties in open primary states (eleven percent) than in closed-challenge (seven percent) or closed enrollment (six percent). He also notes forty-five percent of the cross-over voters reporting having stayed in the crossed-to party in the general election. ‘Wolfe's results are obtained from 1958 and 196k Survey Research Center data, and his N is very small (sixty-two of the combined sample of more than 3000). (111) 25Wisconsin is an open primary state. ., n. W“ "Iv-nu WW... 3 CHAPTER II THE IMPACT OF PRIMARY STRUCTURE: SOME TESTS Introduction In the previous chapter we developed hypotheses from notions of rational political actors responding to the structural constraints of primary election situations. we test these hypotheses in the sections which follow. As noted, there are four types of primary elections -- blanket, open, closed-challenge, and closed-enrollment. The closed-challenge type is elsewhere called "semi-open" (WOlfe: 111) and its neither-fish-nor- .fowl character presents us with an important analytical problem: do we consider it as open or as closed in testing our hypotheses? As it turns out, states which use the closed-challenge system manifest behavioral characteristics of closed states in some instances, open in others. There- fore, while we display voter and candidate behavior results of closed- challenge primaries in our tables, we test hypotheses excluding them. In our tests we combine the blanket and open results and compare them to closed-enrollment results. By thus dichotomizing we have two samples of election results to which we apply the Difference of Neans Test and the Difference of Proportions Test, as apprOpriate (Garrett: 127-38), to determine the significance of the variation in behavior be- tween the relatively "pure" open and closed primaries. Longitudinal Hypotheses Voters will vote in the primary of the party with the more meaning- ful and/or interesting contests, if they can. Ceteris paribus, rational 32 33 voters will participate in the primary of the party whose candidates are likely to win the general election or where an active contest is taking place for the nomination. However, all other things are not equal; different structural restrictions are imposed on the ability of the voter to vote in the most meaningful primary. 'we eXpect that more voters in open primaries will switch parties between elections than in closed pri- maries, since fewer impediments (the most important being the changing of party registration) confront them. Repeating Hypothesis 1, The more open the primary, the greater the variation from primary to the following primary in a party's percent of the total two-party turnout. The average between-primary percent variations, cited in Table 2, monoton- ically decrease with increasingly closed primaries.1 TABLE 2 AVERAGE INTER-PRIMARY PERCENT VARIATION IN DEMOCRATIC PERCENT OF TOTAL TWO -PARTY VOTE, BY PRIMARY TYPE »——-__-2.1__ . -. “— Primary Type 1 Average Variation Number of Election Pair blanket 3 17.76 . 10 open 3 9.67 i 126 closed-challenge Q 7.06 i 125 closed-enrollment ; 6.98 i 259 "857. 1 52—0 Testing the hypothesis, we find the difference between the average variation of open and closed primaries to be significant at the .01 level (Table 3). Thus we reject the null hypothesis that primary type has no influence on between-primary variation in voting turnout. 3h TABLE 3 DIFFERENCE 0F MEANS TEST OF'THE AVERAGE BETWEEN-PRIMARY PERCENT VARIATION IN DEIDCRATIC TURNOUT, BY PRIMARY TYPE -< -.—..x--;- '21_ ) Open Closed Mean 3 10.26 i 6.98 S.D. E 9.16 1 6.15 N g 136 g 259 t = 3.77 at 393 d.f. Further, we expect some voter switching to take place between the primary and the general election. A voter who switches has three possi- ble justifications for his behavior: (1) he wants to "raid" the other party in order to help nominate a weak candidate, (2) he wants to "switch" to the other party in order to help nominate a preferred candidate whom he intends to support in the general election, or (3) he simply wants to vote in the primary of the party whose candidates are most likely to win the general election. Which of these rationales we assume to be operative dictates the direction of our hypothetical relationship between primary and general election turnout variation. Assuming justification (1), we suggest Hypothesis 2: The more open the primary, the greater the variation from a primary to the following general election in a party's percent of the total two-party turnout. he find, in fact, the reverse to be true; i.e., Open primaries feature lggg variation (see Table h). 'We conclude that on an aggregate basis "raiding" occurs with less frequency than "switching" in open primaries (indeed, generally), and that in closed primary states a voter typically registers with the dominant party, votes in its primary, and votes his true party preference in the general election. The latter type of behavior is documented elsewhere for primaries generally but the 35 distinctions among primary types, especially the monotonic increment in variation that accompanies increasing "closedness," suggests that, while open primary states have the greater potential for gross voter volatility, they are actually more stable. On balance, in open states fewer voters abandon their primary election party in November than in closed states. TABLE h ‘AVERAGE PERCENT VARIATION BETWEEN PRIMARY ELECTION AND FOLLOWING GENERAL ELECTION IN DEIDCRATIC PERCENT OF TOTAL TWO-PARTY TURNOUT, BY PRIMARY TYPE Q Primary Type 1 Average Variation Number of Election Pairs .1- ‘ , - i . _ A _. . ...- blanket " 111,1 3 10 Open 1,4088 ‘ 133 closed-challenge 18.21 g 132 closed-enrollment 18.23 g 22§_ 1 16057 ‘ 571 Testing the hypothesis, we find the difference in average variations to be significant at the .01 level (Table 5). Thus we can simultaneously' reject the null hypothesis that primary type has no influence on primary- to-general election variation and our original alternative hypothesis that open primaries and relatively large variations are associated. TABLE 5 DIFFERENCE OF MEANS TEST OF THE AVERAGE VARIATION BETWEEN PRIMARY AND FOLLOWING GENERAL ELECTION IN DEIDCRATIC PERCENT OF TOTAL TWO-PARTY TURNOUT, BY PRIMARY TYPE ” '2 “ope; ‘ ‘ " Closed S.D. 10.57 i 10.62 N 1b3 g f 295 t - ”3.32 at A436 def. Finally, we expect many voters to respond to short-term change in the competitive status of the two parties as manifest by the variation 36 between the general election percent and the following primary election percent. Open primary states will tend to reflect a closer balance be- tween general and primary election results than closed states, and varia- tions should be smaller. Repeating Hypothesis 3, The more open the primary, the less the variation from a general election to the following primary elec- tion in a party's percent of the total two-party turnout. Table 6 indicates that the hypothesis is sound. TABLE 6 AVERAGE VARIATION BETWEEN GENERAL ELECTION AND FOLLOWING PRIMARY ELECTION IN DEMJCRATIC PERCENT OF TOTAL no- PARTY VOTE, BY PRIMARY TYPE Primary Type ) Average variation ' Number of Election Pairs blanket 10.87 E 10 open 1 1h.8h g 132 closed-challenge A 17.6h 3 13h closed-enrollment ; 17.81 g 221 16.96 ' S73 Testing the hypothesis, we find the difference in average variations significant at the .01 level (Table 7). we reject the null hypothesis that primary type has no influence on general-to-primary election varia- tion.2 TABLE 7 DIFFERENCE 0F MEANS TEST OF THE AVERAGE VARIATION BETWEEN GENERAL ELECTION AND FOLLOhHNG-PRIMARY ELECTION IN DEMDCRATIC PERCENT OF TOTAL THO-PARTY TURNOUT, BY PRIMARY TYPE ‘ Open A Closed I | r ‘ j ‘4 Mean ‘7 18.67 \ 17.6h S.D. ‘ 11007 i 9.92 N 7 111 291. t . 2.72 at (433 def. 37 In summary, we note that primary structure is an independent in- fluence upon short-term fluctuation of primary and general election voting. One aspect of the open-closed dichotomy which is intuitively of importance is the different physical nature of the voting process in states with the two types of primaries. In all states the general elec- tion ballot is secret -- barring fraud a voter's vote is known only to himself. In open primary states the primary ballot is also secret -- the voter votes in one party's primary and discards the unused paper in ballots or flips the voting machine lever which forever conceals his choice of party. But party affiliation is public information in closed primary states. A voter's candidate choices may be his information alone, but not his party. There are therefore probably a number of social and psychological pressures, some subtle and some not-so-subtle, brought to bear on the voter, especially at the time of party declaration during the voter registration state, in closed primary states. Many voters may resist these influences -- ethnic, religious, racial, marital, etc. Some will not. The important point is that the privacy of the ballot at pppp elections in open primary states penmits pyppy voter greater flexibility and independence in his behavior, and therefore encourages more rational action, than do closed primary states, wherein a crucial stage of the vot- ing process is quite public.3 Thus we should not be surprised that in open states more voters switch parties between primaries than in closed states, because it is easier to do so in the former than in the latter. Similarly, more voters switch parties between the primary and the general election in closed primary states than in open, because if switching is to take place, it can be privately consummated only in the general election polling booth.h In open primaries voters typically switch parties in order to vote for a 38 preferred candidate in the general election. It is possible, of course, that the larger variation noted in closed primary states can be attributed to "raiding." This is most unlikely, however, for one overriding reason: the organizational re- sources needed to entice a sufficiently large number of voters to change party registration to make a raid successful are huge at the 12231 level (Berdahl, 19h2: 12p. 913.). To mobilize such resources for a statewide incursion would probably not be possible, and there is no iron-clad guar- antee that the attempt would accomplish the desired goal: the "strong" candidate might win anyway, or the "weak" candidate might capture the imagination of the voters during the campaign and win the general elec- tion. An additional explanation for this finding is that since the closed primary of one party by its nature excludes a portion of the electorate from voting therein, the primary electorate is ceteris paribus smaller than the general election voter pool, and the larger difference found in closed states reflects an increase in the absolute number of voters between the two elections, relative to the increase found in open states, where the pool is theoretically equal for both elections. The switching which occurs between the general election and the fol- lowing primary is in part the reverse dynamic of the primary-general switching: voters return to their "home" party, and they do it in larger numbers in closed states than in Open. But the results of the general election also influence the switching. The voter in Open primary states seems to follow more closely the short-run tides of inter-party balance, using the previous general election outcome as one one to future primary behavior, whereas the voters in closed states, barring re-registration, finds his future primary behavior predetermined. 39 Static Hypothe sis In what follows we note interesting and frequently unexpected vari- ations in intra-election political behavior associated with different primary structures. We have already noted the controls we introduce: intra-party competition and incumbency. In the hypothesis tests we re- fine the Index of Competition slightly in order to control for competition in a relatively "pure" way; we eliminate the "semi-competitive" and "semi- dominant" categories, because the N's are rather small compared to the other classifications and because their exclusion clearly makes the re- maining three categories analytically discrete. Voter Behavior As described in Chapter I, when comparing open and closed primary states, we expect to find the following in open states: smaller plurali- ties for the leading vote-getter and incumbent, larger average indices of competitiveness and fragmentation, less incidence of incumbent renom- ination and less incidence of reelection of renominated incumbents. Anti- cipating the actual display Of the results, generally we find la_r_gp_r_ average pluralities for leaders in open primaries (but no difference for incumbents), 153s competitiveness and fragmentation, no difference in in- cumbent renomination rates, and less incidence Of incumbent reelection, the last finding being the only one consistent with a hypothesized rela- tionship. Plurality Expecting voter raiding/switching to accrue to the benefit of chal- lengers of leading candidates in Open primaries, while leading candidates in closed states will be faced with no such threat, we Offer Hypothesis )4: v ...' m Aura w—w-r I'll 80 The more open the primary, the smaller the average plurality for the leading vote-getter In Table 8 we note findings in clear contravention of the hypothesis. TABLE 8 AVERAGE PLURALITY OF LEADING VOTE-GETTER IN PRIMARY, BY PRIMARY TYPE v- . i . _ _. . . V— - a — . . ...1 .. 1 Primary Type i Average Plurality Number of Elections blanket i 66.1 22 Open i 68.1 . 276 closed-challenge I 65.2 389 closed-enrollment E 63.6 i 159 E 65.0 i 1397 In testing the hypothesis, we note in Table 9 the marked variation in average plurality where the parties are dominant which contributes to a significant overall difference.S While pluralities are actually slightly lower where the parties are not competitive in open states they are higher by more than ten percent in situations of dominance. Also note how plur- ality size decreases with increasing party fortunes. TABLE 9 AVERAGE PLURALITY OF LEADING VOTE-GETTER IN PRIMARY, BY PRIMmRY TYPE, CONTROLLING FOR.INTER-PARTY COMPETITION Primary Type 7 Minority Competitive Dominant**” M - AII;;_—fl— open 73.5 (SD) 68.8 (139) 61.h (56) 68.0 (298) closed 75.3 (87) 68.3 (297) 50.913051 63.6 (789) 1 78.6 (181) 68.3 (836) g 53.2 (261) g 68.9 (1087) Competitiveness and Fragmentation We anticipate more competitiveness and fragmentation in Open primary states for the same reason we expect lower pluralities: more raiding/ switching, and a more volatile electorate generally, in theory anxious to thrash the front-runner and disrupt, even embarrass, the other party. we 81 offer Hypotheses 5 and 6 and for both find closed primaries manifesting the characteristics eXpected of open: The more Open the primary, the larger the average index of competitiveness. The more-Open the primary, the larger the average index of fragmentation. In Table 10 we note the general tendency of intra-party competi- tiveness and fragmentation to increase, the more closed the primary. ' r33 TABLE 10 AVERAGE INDICES OF COMPETITIVENESS AND FRAGMENTATION, BY PRIMARY TYPE 1 I» :\ Primary Average Index of , Average Index of Number of Type Competitiveness ' Fragmentation Elections blanket 3 57.1 '2 10.7 22 Open g 53.6 § 10.1 276 closed-challenge ‘ 58.1 § 11.2 7 389 closed-enrollment ( 61.8 : 12:9 7 750 : 5 .0 7 11.8 ; 1397 The differences between high and low average are not stunning for either index but the general direction of increase is apparent, although not monotonic -- note the average indices for the blanket primary. The variations become more distinct as we test the hypotheses (see Tables 11 and 12). Again note the wide variance in index averages where the parties are dominant. Also note the direct relationship between increased intra- party competitiveness and fragmentation and party success level. TABLE 11 AVERAGE INDEX OF COMPETITIVENESS, BY PRIMARY TYPE, CONTROLLING FOR INTER-PARTY COMPETITION Primary Type V PTEnOrity Competitive Dominant** A1158“) _ -.- ll-l-. _ , -, . ‘ _ -1 .- _- , 32-3,“ -1 open a 85.0 (58) I 52.3 (139) 65.1 (56) 7 58.0 (298) closed 1 86.3 _(87) t 58.9 (297) ! 77.5.(205) 7 61.8 .(789) ' 85.8 (181) i 58.1 (836) E 78.8 (261) _j 59-2 (10871- I t L2 TABLE 12 AVERAGE INDEX CF FRAGMENTATION, BY PRINARY TYPE, CONTROLLING FOR INTER-PARTY COMPETITION Primary Type iinority 2 Competitive . Dominant** 3 A11 open 7.6 (5h) % 10.9 (139) E 12.0 (56) i 10.7 (298) closed 7.8 87 g 8.6 (297) g 20.7 (205) 7 12.0 (789) 7.5 (181) 9.3 (836) - -_..— 18.8 (261) 3 11.6 (1087) It is rather clear from all three measures of intra-party voter be- havior that strictly intra-party battles seem to be contested more bitterly when the nominating primary is limited only to voters who regularly iden- tify with a party, and especially when that party dominates a state's political system. Incumbency Elections are a form of "regularized recall." Periodically the re- cord of an incumbent administration is subjected to popular review. If the voters approve, the incumbent (or at least his party) retains power; if not, he is in a sense "recalled from office." Only in America does the electorate have a dual opportunity to scru- tinize an incumbent's conduct in office -- once at the primary, and (as- suming he prevails at this initial step) again at the general election. A number of states prohibit their Governors from seeking reelection. In those states where reelection is permitted, incumbents have been notably successful at getting renominated -- of the 835 incumbents seeking renom- ination, 373, or 85.7 percent, won their primary.6 we eXpect that the rate of incumbency success in primaries will vary with primary structure. Specifically, we expect that fewer incum- bents will win renomination in open primaries, because anti-incumbent voters will be more able, and therefore more likely, to register their 83 disapproval of the incumbent's administration at the primary rather than waiting until the general election. The more Open the primary, the smaller the percent of primaries won by incumbents seeking renomination. Repeating Hypothesis 7, Table 13 indicates that there is very little variation among primary types both in incumbency success percent and in incumbent's plurality. TABLE 13 PERCENT OF NOMINA‘I IONS WON BY INCUMBENT , AND AVERAGE PLURALITY OF INCUMBENT, BY PRIMARY TYPE 3 Primary Type A Percent an blanket : 90.0 open ‘ 83.8 closed-challenge ‘ 87.2 closed-enrollment ' 85.2 * 85.7 Average Plurality 68.8 69.2 70.0 69.2 69.6 Number of Elections 10 105 117 222 835 As a rule, incumbency renomination rate decreases as party dominance increases, but slightly fewer incumbents are renominated, and by slightly smaller pluralities, by dominant parties in closed states than in open states (see Tables lb and 15). Given the high rate of incumbency renom- ination and large incumbency pluralities, we thus conclude that even in open primaries much of the crossover voting is pro-incumbent. Whether this pattern continues through the general election is examined in Hypo- thesis 9. TABLE 18 PERCENT OF NOMINATIONS VDN BY INCUMBENT, BY PRIMARY TYPE, Primary Type Competitive— open : 92.6 (58) closed ; 91.1 (90) 91.7 (188) . ...—"......— CONTROLLING FOR.INTER-PARTY COMPETITION Dominant 77.8 (36) 78.0 (73) 75.2 (109) l 1 3 7 All 87.8 (111) 85.2 (203) 86.0 (318) 88 TABLE 15 AVERAGE VOTE PERCENT OF INCUMBENTS, BK PRIMARY TYPE, CONTROLLING FOR.INTER-PARTY COMPETITION Primary Type ! (Competitive I Dominant A All.“ 3 — IT ' open'u ‘ g '77.7 (58) i 61.8 (36) § 72.2 (111) closed 7 77.8 (90) ? 56.9 (73) i 69.8 (2032 , 77,5 (1gb) ; 58.8 (109) g 70.8 (318)’_, Earlier we noted that the incumbent renomination rate exceeds eighty- r?- five percent. The incumbent reelection rate is somewhat lower; of the 373 renominated incumbents, 270 were reelected (72.8 percent). A different political dynamic is at work here: we return to the inter-party struggle, still suspecting that primary structure will have an influence. ‘We anticipate that a disproportionate share of incumbents suffer general election defeat in open primary states, where anti-incum- bent forces (having failed to oust him in the primary) can covertly or overtly enlist the support of the opposition party in the general election. To be specific, intra-party feuds are more efficaciously carried over into the general election in open states than in closed. In Table 16 we find support for Hypothesis 9: The mOre open the primary, the smaller the percent of general elections won by renominated incumbents. TABLE 16 PERCENT OF GENERAL ELECTIONS WON BY RENOMINATED INCUMBENT, BY PRIMARY TYPE Primary'Type ‘ Percent won Number of Elections 2 l blanket 1 88.8 § 9 Open ’ 68.5 g 89 closed-challenge ; 72.3 E 102 closed-enrollment , 75.5 i 173 72.1A j _ 373 85 In the hypothesis test (see Table 17) we find small percent varia- tions between primary types, one of which is significant. Although our level of confidence is not breath-taking, we reject the null hypothesis that primary structure has no effect on incumbency reelection rate and accept our alternative hypothesis. TABLE 17 PERCENT OF GENERAL ELECTIONS WON BY RENOMINATED INCUMBENT, BY PRIMARY TYPE, CONTROLLING FOR INTER-PARTY COMPETITION l . I Primary Type { Competitive Dominant : All (*) _ open 7 56.9 (51) 85.7 (28) 66.3 (98) closed 63.8 (82) 92.6 (58) 75.7 $173) 60.9 (133) 90.2 (82) 72.3 (271) Candidate Behavior We predict the following in open states as compared to closed: more primaries will be "contested," fewer nominees will be unopposed, more in- cumbents will be challenged, and more candidates will enter primaries. In fact, with one exception, we find the reverse to be true for all can- didate behavior measures; glgsgd primary states feature the above char- acteristics. Contestin , Unopposed Nominations, and Incumbent Challenges We theorize that the compositional uncertainty of the open primary electorate encourages a relatively large number of citizens to contest nominations that would otherwise go unopposed and to challenge incumbents who would otherwise have a free ride, with the consequence of more genuine "contests" for the nomination (i.e., where the runner-up has at least half the front-runner's vote). Hypotheses 10, 13, and 18 follow therefrOm: The more open the primary, the more likely it is to be contested. 86 The more open the primary, the smaller the percent of elections with unopposed nominees. The more open the primary, the larger the percent of elections in which the incumbent is challenged. Table 18 delineates the actual frequency of contested gubernatorial nom- inations, and in Table 19 we test the hypothesis. Note that slightly more open primaries are contested by uncompetitive and competitive par- ties where no incumbent is seeking renomination, but that more closed primaries are contested in every other category. Also note the increased rate of contesting on two other dimensions: incumbent to non-incumbent and not competitive to dominant. TABLE 18 PEECEI-(T 0F PRDIAEIES CONTESTED, BY Phli-IARY TYPE . I ; ‘ ’ 1 Primary § Incumbent ; No Incumbent Type. ' Running E Running Total blanket ; 80.0 (10) N! 58.3 (12) E 50.0 (22) open ' 32.6 (105) 57.3 (171) 1 88.2 (276) closed-challenge 1 83.9 (117) ‘ 55.8 (233) i 51.6 (389) closed-enrollment . 38.9 (203) 61.0 (586) a 55.0 (789) 38.7 (835) 58.9 (963) _ 53.2 (1397) TABLE 19 PERCENT OF PRIMARIES CONTESTED, BY PRIMARY TYPE, CONTRCLLING FOR.INTER-PARTY COMPETITION A11 Primaries 8 Primary Type ( Minority Competitive Dominant (*) g All* i ' i : open E 80.0 (50) 85.7 (138) } 60.7 (56) 88.3 (298) closed 2 39.1 ((87) 88.8 (297) f 72.2 (205) 55.0 (789) g 39.8 (137) 87.8 (835) Q 69.7 (261) 53.1 (1087) All Primaries, IncumbentRunning - i in Primary Type % Minority Competitive ! Dominant All Open 5 -- 20.8 (53) g 55.6 (36) 33.9 (115) closed 2 -- 26.7 (90) E 57.5 (7}) :38-9 (203) 28.5 (183) 56.9 (109) 37.1 (318) 87 TABLE 19 (continued) All Primaries, Incumbent Not Running ; I I . I I Primary Type I Minority 3 Competitive I Dominant I All -, .-,.. 6 ,1 ,7 ,7 . , .__ I open I 80.0 (50) I 61.2 (85) I 70.0 (20) i 57.8 (183) closed I 39.1 (87) « 58.5 (207) i 80.3 (132) E 61.0 (586) I 39.8 (137) i 59.3 (292) i 78-9_(152) ,§_‘60-1 (729)11“- Table 20 diSplays the percent of unopposed nominations. Note that more unOpposed nominees are found in open primaries, but that the results found in Table 21 show several instances of lower percent unopposed for open primaries. This is because only one of the twenty-two blanket pri— maries (considered "open" for hypothesis-testing purposes) records a nominee without Opposition. Again, note fewer nominations going by de- fault when no incumbent runs and as party fortunes increase. TABLE 20 PERCENT CF NOMINATIONS UNOPPOSED, BY PRIMARY TYPE Primary I Incumbent No Incumbent I I . I Type 5 Running I Running g Total blanket ; 0.0 (10) f 8.3 (12) I 8.5 (22) Open a 26.7 (105) I 22.2 (171) i 28.2 (276) closed-challenge . 23.3 (117) é 19.3 (233) i 21.8 (389) closed-enrollment I 28.6 (203) g 15.9 (586) a 18.3 (789) 2 23.9 (835) i 17.7 (963) E 19.9 (1397) TABLE 21 PERCENT OF NOMINATIONS UNOPPCSED, BY PRIMARY TYPE, CONTROLLING FOR INTER-PARTY COMPETITICN 7 All Primaries .. .. ' I i 35.8 (137) f 22.8 (835) I 6-9 (261) I 19.5 (1087) Primary Type Minority Competitive g Dominant * I All Open 1 38.0 (50) 20.3 (138) f 18.3 (56) I 22.5 ‘(298) closed ; 38.5 (87) I 23.9 (297) ' 8.9 (205) I 18.3 (789) 88 TABLE 21 (continued) All Primaries, Incumbent Running Primary Type I Minority j Competitive I Dominant All Open -- 5 30.2 (53) I 16.7. (36) 328.3 (115) closed , -- . 37.8 (90) i 9.6 (73) I28.6 (203) ' _ 35.0 (183) _ I 11.9 (109) 328.5 (318) All Primaries, Incumbent Not Running 8 ‘8‘ Primary Type I Minority I Competitive I Dominant (*) I All (%)m 8. open I 38.0 (50) I 18.1 (85) I 10.0 (20) E 21.3 (183) closed ' 38.5 (87) I 17.9 (207) I 2.3 (132) I 15.9 (586) ..f g 35.8 (137) 3 16.8 (292) I 3.3 (152) f 17.3 (729) ' IFS-mi. -_'."«' . ' Finally, we note the frequency of incumbency challenges in Table 22 and the test of Hypothesis 18 in Table 23. Incumbency seems more perilous in closed primaries than in open; the variation is especially noteworthy when considering challenges by "serious" candidates. Note what inter- party competition level does to the challenge rate, and also note that no incumbent goes unOpposed in a blanket primary. TABLE 22 PERCEI-JT OF INCUMBENT CHALLENGES, BY PRIMARY TYPE . b _— More Than One More Than One Number of I Primary I I Type I Candidate I Serious Candidate I Elections _ _1.- A _ I _ . . ,1 -__- . I blanket I 100.0 I 80.0 I 10 Open I 70.3 I 52.5 I 105 closed-challenge I 78.8 I 59.5 I 117 closed-enrollment I 78.8 I 59.1 I 292 5 78.2 I 56.6 , 83S 89 TABLE 23 PERCENT OF INCUMBENT CHALLENGES, BY PRIMARY TYPE, CCNTROIIING'PCR.INTER-PARTY COMPETITION More Than One Candidate Dominant I All Primary Type I Competitive open I 57.8 (58) I 83.3 (36) 70.8 (115) closed 3 61.1 (901 I 89.0 (73) 5 78.8 (203) i 59.7 (188) I 87.1 (109) 73.0 (318) I” Mere Than One Serious Candidate Primary Type I Competitive I Dominant I All (*) 7 If open I 33.3 (58) I 75.0 (36) I 89.6 (115) 61636.1 85.6 (93) I 79.5 Q3) I 59.1 (203) ; I 81.0 (188) I 78.0 (109) i 55.7 (318) Thus for the three hypotheses tested above we note an increasingly familiar pattern: intraaparty disputes tend to elicit more pugnacious be- havior among both voters and candidates in closed primaries than in open. Number 2; Candidates we Speculate that the more uncertain electorate of the open primary tends to encourage larger numbers of citizens generally to enter the nom- inating fray, and brings out a larger number of office-seekers who take their politics seriously, as stated in Hypotheses 11 and 12: The more Open the primary, the larger the number of candidates likely to compete for the nomination. The more open the primary, the larger the number of serious candidates likely'to compete for the nomina- tion. The findings displayed in Table 28 have no clear pattern, but Table 25 shows curious differences among inter-party competition levels. In open primary states a relative abundance of candidates compete for seem- ingly quixotic nominations and relatively few compete for valuable ones. This especially occurs when no incumbent is running. 50 TABLE 211 AVERAGE NUMBER OF CANDIDATES, BY PRIMARY TYPE ‘ I 3 . _ -. .1 - . Primary I Incumbent I Incumbent Type Running I Not Running Total blanket 8.1 (10) I 8.0 (12) ' 8.0 (22) closed-challenge . 2.6 (117) 3-5 (233) . 3.2 (31-19) closed-enrollment g; I203) I 3.5 £5872 3.2 L789) 1 2.7 (835) 5 3.8 (963) 3.2 (1397) W TABLE 25 AVERAGE NUMBER OF CANDIDATES, BY PRIMARY TYPE, CONTROLLING FOR DEER-PARTY (EMPETTTION All Primaries Primary Type I Minority (3%) I Competitive Dominant-16* I All-8 open 2.67 (58) . 2.89 (139) 3. 81 (56) 2.92 (296) closed 2.07 {87) I 2.73 (297) 11.77 (205) 3. 25 (789) t 2.30 (181) 2 .78 (836) 8. 88 (261) I 3.16 (1087) All Primaries, Incumbent Running I 11.....— Primary Type I Minority I Competitive I Dominant I I All open I -- i 2.81 (58) 3. 39 (36) I2 2.77 (111) closed I —- I 2 .21 (9o) 3. 67 (73) I 2.72 (203) I ' 2 .29 (188) I 3.56 (109) I 2.78 (318) All Primaries, Incumbemt Not Running Primary Type I Minority (9+) Competitive I Dominant-ML I All-)6 ' i ' I ~ i open I 2.67 (58) E 3.26 (85) I 3.85 (20) I 3.08 (187) closed I 2.07 (87) I 2-95 (207) I 3.31 (132) I low—1 15116) : 2.30 (188) I 3.02 (292) Primary ‘ General r l Primary 1 General 1936 80.1 i 71.2 ‘; 1956 2 56.7 ' 58.8 1980 66.0 i 89.6 1960 i 82.2 I 50.7 T819 27.8% , 51.7 :i T938 38.7 ‘ 88.0 1988 51.0 i 88.3 N 1968 51.8 88.7 1952 1 62.2 g 87.8 7 A A I Years underlined - incumbent Democrat running in both primary and general election *unopposed Democratic nominee 3In the 1972 primaries in New Jersey and Illinois federal courts ruled that closed primaries were unconstitutional, thus permitting, as it turned out, Republican raids into the Democratic presidential primary in New Jersey and into the gubernatorial primary in Illinois, where the Daley machine candidate was upset by a reform candidate who Republicans cal- culated would not present as serious a challenge to their incumbent in November as would the Daley candidate. The Democrat won. 58 uUntil recently public Republicanism in the rural South was practi- cally felonious, as was public Democracy in much of the rural Midwest and upper New England. Since the closed-enrollment primary prevailed in these areas, the pressures to register with the dominant party were enormous, but there was little that could be done about the maverick voter who in- sisted on voting for a candidate of the outlaw party in the general elec- tion. There is a story about a town in New England which for years had never had a registered Democrat but always recorded one Democratic vote in the general election. One year there were two Democratic straight- ticket tallies, a phenomenon which moved an election judge to charge: "The son-of-a-bitch musta voted twice." 5For this and all subsequent tables the following key applies re- garding the statistical significance of the difference of the two means or preportions: (*) - significant at the .05 level, one-tailed test * a significant at the .05 level, two-tailed test ** - significant at the .01 level, two-tailed test In addition, the "all" category in the Index of Competition dimen- sion includes "semi-competitive" and "semi-dominant" states. 6Four incumbents were forced into a runoff primary, and all four lost. 7See Gordon S. Black, "A Theory of Political Ambition: Career Choices and the Role of Structural Incentives," American Political Science Review, 66 (1972), lthf but especially 188-50 for a theoretical discussion of "risk." 8See Black, lh6-h9, for a theoretical discussion of the impact of size of the electoral unit on behavior. CHAPTER III GUBERNATORIAL NORENATICNS IN THE TWENTIETH CENTURY: REGIONAL AND LONGITUDINAL VARIATIONS Introduction we have indicated that voting and candidate behavioral variation is associated with different primary types. Below we note that various as- pects of gubernatorial nominations differ over time and among regions of the United States.1 Primagy Type and Political Era During the present century the number of states conducting guberna- torial primaries has increased from zero in 1900 to a current forty-seven (the largest number ever). The number of states holding primaries, as well as the pr0portions of different types, has remained rather stable since 1920; during the decade prior to 1920 (the height of the Progressive Era), the number of states with primaries nearly doubled (see Appendix A, Type of Primary, for Selected Years; and Appendix B, Primary Type, by State and Year). Post-1920 structural changes are displayed in Table 33; note the re- gional variations. The Hississippi River appears to be a dividing line of sorts: to the east lays fairly strong party control of nomination, either through the convention or the closed primary, and to the west is both open country and relatively open nominating structure.2 Boots notes this ten- dency in 1922, contending that states are in two groups with respect to primary elections: (1) the populous urban eastern states "in which fairly stringent party tests have been maintained by law or tradition and in which the party organizations . . . still manage nominations to a consid- erable degree . . ." (826), and (2) the less populous rural west, 59 60 north-central, and mountain states, where there is possibly more individual thinking, where at any rate party ties sit more loosely, perhaps because of frequent insurgent movements, largely agrarian. The voter desires to vote in the primary of the party whose candidate will be elected. Primaries are by law, or become by practice, open. ‘Party labels are largely meaningless and the voter so regards them. (h28) TABLE 33 TYPE or PRIMARY, BY REGION, FOR SELECTED YEARS a- ___. .-.. . A .‘ 7V._..2A-—-»_-—-.h “m...“ 553%; ' * blanket Open challenge enrollment no primary Year 1920 NE - 1 2 6 3 MW - 3 3 3 - B - 2 h - S - - 2 8 - V _: .2. .3 .2 ..2 O 6 11 2S 6 19ho NE - - 2 S 5 MW - h 3 2 - B - - 2 h - S - - 2 8 - W' .1 .2 ll .9. {_: 1 7 10 25 5 1960 NE - - 2 7 3 MW - h 3 2 - B - - 2 u - S - - 2 8 - W .Jl .2. _l .12 _: 1 10 27 3 H Note: In many states the primary was optional and the minority party nom- inated by convention. In this table, only those states in which there is no provision for a gubernatorial primary for any party are included in the "no primary" category. In the half-century since Boots wrote, his observation (though per- haps not his rationale) seems to have been supported and extended. New York and Indiana dropped the primary and returned to the convention system 61 in the mid-twenties (each state having used it three times), while in the west the blanket primary was develOped (l93h in Washington and 1967 in Alaska). In 1968 all seven of the strictly open primaries were either in the Midwest or west regions. fiegional Behavioral Variations Numerous accounts of the politics of regions of the United States 5?! are available.3 The literature on regionalism generally notes three im- portant characteristics of American party politics: (1) relatively strong '5 party organizations in the Northeast, (2) relatively weak organizations 3 5;. west of the Mississippi, and (3) the dominance of the Democratic Party in the South and its semi-dominance in Border States. Following from the above, it is likely that nominations will be more closely controlled by party organizations in the Northeast (i.e., less contesting, fewer can- didates, larger winning pluralities, etc.), and that there will be more intra-party competition in South and Border Democratic primaries than elseWhere. Contesting An average of 52.6 percent of all 1397 primaries are contested (i.e., the runner-up had at least half the vote of the winner). There are clear regional variations in contesting (see Table 3b); more contesting occurs in South and Border states, due to the existence of the Democratic primary as the "real election" there. But in the regions with traditions of two- party competition, considerably fewer primaries are contested in the North- east than in the Midwest and west. Region MW W for each gubernatorial nomination, but there are major regional variations (see Table 3b). On the average there are slightly more than two "serious" candidates PRIMARY CCNlESTING P 13, BY REGION, BOTH PARTIES I v i Number of Elections 373 h23 270 152 179 1397 Number of Serious Candidates 62 lABLE 3h Number Contested 150 205 156 101 .122 735 Percent Contested Lo.2 h8.5 57.8 66.1-J» 68.7 52.6 Party organizations in the Northeast seem best able to re- strict serious challenge to organization candidates, and the Dixie Demo- crats, having no unified party organization, manifest considerably more party factionalism. AVERAGE NUMBER OF SERIOUS CANDIDATES, BY REGION, BOTH PARTIES Region be found in the size of the winning candidate's plurality. zations should be able to produce relatively large turnouts for the NE NW 'W B S U.S. i i 4 TABLE 35 373 h23 270 153 178 1397 Number of Elections Plurality I ‘ Average Number Serious Candidates 1.69 1.92 2.16 2.25 2.70 2.0h One indicator of party organization control of nominations should Strong organi- 63 favored candidate, when compared to weak organizations and especially intra-party factions. Again we find the Northeast manifesting this indicator most strongly in Table 36. TABLE 36 AVERAGE PLURALITY BY WINNING CANDIDATES, BY REGION, BOTH PARTIES . _ . | ‘ { , ‘ Number of g Average Region f Elections 5 Plurality f“‘ NE 373 E 73.h MW h23 f 68.h W 278 E 61.9 .. :1 B 153 56.5 ‘ s g 178 _ 50.9 f v.5. ‘ 1397 , 65.8 § 7 i A L IntraéParty Competitiveness A further indicator of organizational strength is not only how large a turnout the machine can obtain for the favored candidate, but also how much distance it can put between him and his closest rival. Our index of competitiveness measures this phenomenon, and our expecta- tion that primaries are less competitive in the Northeast than elsewhere is fulfilled (see Table 37). TABLE 37 AVERAGE INDEX OF COMPETITIVENESS, BY REGION, RDTH PARTIES A Number of I Average Index of Region : Elections ! Competitiveness 5 l NE y 373 f h6.6 W i 270 T 63.8 B 153 ‘ 71.2 S ‘ 178 a 77.3 U.S. ‘ 1397 ‘ 59.0 6h Regional and Longitudinal Variations Inter-party Competition Between 1920 and 1960 we note distinct trends over time and clear differences among regions in the primary and general election turnout (see Table 38).h TABLE 38 REPUBLICAN ELECTION TURNOUT: A COMPARISON OF PRIMARY AND GENERAL ELECTION PERCENT OF THE TWO-PARTY'VOTE, BY REGION, 1920-1960 ’ ) Region i 1920 { 1930 l9h0 E 1950 Northeast ; 3 A Q Pfimw sm5(nfl6us