AN ANALYSIS OF NET DEVIATEON VOTING WOFMW What-W 1W1 This is to certify that the thesis entitled AN ANALYSIS OF NET DEVIATION VOTING presented by ORAL EUGENE PARKS has been accepted towards fulfillment of the requirements for PH.D. degreein POLITICAL SCIENCE Major professor MW 7/. 0-169 LIBRARY Michigan State University ABSTRACT AN ANALYSIS OF NET DEVIATION VOTING BY Oral Eugene Parks Net deviation from straight party ticket voting is the difference between the percentages of the votes cast for two candidates of the same party for two different offices, and is the result of the differential attractive- ness of the two candidates, relative to their respective opponents. It is a measure of the way in which the charac- teristics of an election interact to produce deviations from outcomes expected if straight party ticket voting were the sole determinant of the vote. It is the net directional effect of split-ticket voting and selective non-voting (incomplete ballots). Two different analytical approaches are used in this study in an attempt to explain the causes of net deviation voting in Montana. The first approach is to test a model consisting of a multiple regression equation designed to determine the effects on net deviation voting of five political variables, two socio-economic variables, and nine Oral Eugene Parks demographic variables, all eXpressed as aggregate statis- tics. The presidential election years from 1944 to 1964 are analyzed, testing the elections for the office pairs of Governor/President, Governor/U. S. Senator, and Governor/Congressman. No significant relationships are found between any of the variables and net deviation voting, and it is concluded that net deviation voting is not shown to be the result of social and demographic factors. The second approach is a historical case study. The political history of the state is examined to see what forces have influenced voting behavior in a period of seventy-five years. The historical patterns of changing corporate partisan political activity are shown to correspond to the patterns of changing partisan control of state and national offices. It is concluded that in a highly competitive state such as Montana, where the partisan division of the vote has for some 75 years been almost evenly split, political activity by corporations to protect their interests by supporting Republican candi- dates for state offices and Democratic candidates for national offices results in the state having a high level of net deviation voting. AN ANALYSIS OF NET DEVIATION VOTING BY Oral Eugene Parks A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Political Science 1971 ACKNOWLEDGMENTS I am deeply indebted to Dr. Rufus P. Browning, whose contributions of time, energy, and ideas went far beyond those usually made by a dissertation Chairman, and to Dr. LeRoy C. Ferguson, who gave me guidance at a critical period in the development of this study. TABLE OF CONTENTS LIST OF TABLES O O O O O O O O O O O O O 0 LIST OF FIGURES O O O O O I O O O O O O 0 Chapter 1. INTRODUCTION 0 O O O O O O O O O Approaches Used in the Study Purpose of the Study The Concept of Net Deviation From Straight Party Ticket Voting Summary SPLIT-RESULTS ELECTIONS AND NET DEVIATION VOTING . . . . . . . . Introduction Split-Results Elections in the United States and Montana Summary A SURVEY OF RELATED LITERATURE . . Introduction Literature in the Field Summary THE REGRESSION MODEL . . . . . . . Introduction The Regression Model The Dependent Variable in the Regression Model The Independent Variables in the Regression Model The Political Variables The Demographic Variables The Method of Multiple Regression Analysis iii vii 12 26 43 Individual Behavior and Net Deviation Voting Application of the Regression Model Summary 5. TESTING THE HYPOTHESES . . . . . . . . . . 77 Introduction Testing the Regression Model Testing the Hypotheses in the Model The Regression Coefficients The Union Influence Hypothesis A New Approach Summary 6. CORPORATE POLITICS IN MONTANA . . . . . . 101 Introduction A New Approach Incidence of Split Control Over State and National Offices Corporate Influence in Montana Politics The Corporate Influence on Split Control Five Periods of Corporate Partisan Activity Progressisivm in Montana The Relationship Between Split Control and Corporate Strategy Summary 7. SUGGESTIONS FOR FURTHER RESEARCH . . . . . 154 BIBLIOGRAPHY. O O O O O O O O O O O O O O O O O 159 iv Table 1. LIST OF TABLES Per Cent Actual of All Possible Splits in Election Results for Pairs of Offices, United States, 1920—1968, Presidential Election Years . . . . Per Cent Actual of All Possible Splits in Election Results for Pairs of Offices for Montana, Grouped by Time Periods, 1920-1968, Presidential Election Years . . . . . . . . . . . Congressional Districts with Split Election Results: Districts Carried by a Presidential Nominee of One Party and by a House Nominee of Another Party, United States and Montana, 1920-1964 . . . . . . . . . Frequency Distribution of Net Deviation Voting Between the Offices of President and Governor . . . . . . . Frequency Distribution of Net Deviation Voting Between the Offices of U. 8. Senator and Governor . . . . . Frequency Distribution of Net Deviation Voting Between the Offices of Congressman and Governor . . . . . . Net Deviation in Montana, for Designated Pairs of Offices, 1944-1964 c o o o o o o o o o o o o 0 Per Cent of Voters Who Split Their Tickets, by Occupation, in 1952 and 1956 . . . . . . . . . . . . . . Origins of the Foreign Born in Montana 0 O O O O O O C O O Q 0 O C O V Page 15 15 19 23 23 24 24 56 61 Table 10. ll. 12. l3. 14. 15. 16. R2 Values . . . . . . . . . . . . . Number of Regression Equations in Which Variables Were Retained, by Year . . . . . . . . . . . . . Range of Coefficients of Correlation Between Net Deviation Voting and the Independent Variables . . . . Regression Coefficients (byx) . . . Counties in Which Dominant Occupations are Highly Unionized, Ranked by Net Deviation Voting . . Partisan Results of Montana Elections, 1890-1968 . . . . . . . Incidence of Split Control in Montana Elections, 1890-1968 . . . vi Page 79 81 84 86 95 105 150 Figure 1. LIST OF FIGURES Split-Results Elections Between the Offices of President and Governor, United States and Montana, 1920-1968 . . . . . . . . Split—Results Elections Between the Offices of U. 8. Senator and Governor, United States and Montana, l920-l968 . . . . . . . . Split—Results Elections Between the Offices of Congressman and Governor, United States and Montana, 1920-1968 . . . . . . . . Per Cent of Congressional Districts Carried by a Presidential Candidate of One Party and a House Candidate of Another Party, United States and Montana, 1920-1968 . . . . . . . . Incidence of Split Control in Montana, 1890-1968 . . . . . . . . vii Page l6 l7 18 20 150 Chapter 1 INTRODUCTION This study is an attempt to determine the vari- ables in elections which influence voters to deviate from straight party ticket voting. It is a study of elections in one state-—Montana--in which such deviations have been especially common. If all people voting for candidates for each of a pair of offices vote a straight party ticket, the per cent Democratic of the total votes for each of the two offices will be equal. If there are some deviations from straight party ticket voting, either in the form of ticket-splitting or in the form of selective non-voting, but the Democratic candidates for the two offices are equally affected by these deviations, the Democratic vote percentages for the two offices will still be equal. However, if there is some directional effect to these deviations from straight party ticket voting—-if both parties are not equally affected by the total ticket- splitting and selective non-voting which occurs-~there will be a net deviation from straight party ticket voting. If the Democratic candidate for one office is more attractive to the electorate than the Democratic candidate for another office, relative to their respective Republican Opponents, and if voters either split their ballots or do not vote for the less attractive candidate, this differ- ential attractiveness between the two Democratic candidates will result in a difference between their respective per- centages of the votes cast for the two offices. The difference between the votes received by the Democratic candidates for two offices is here termed net deviation from straight party ticket voting, or simply net deviation. It is a measure of the way in which variables in an election--e. g., the candidates, the issues, the national- level and lower-level structures of political parties, and the attitudes of the voters toward each level of party organ- ization--interact to produce deviations from the outcomes that analysts would have expected if overall party identi- fication were the sole determinant of a party's share of the vote for different offices, or if deviations from straight party ticket voting affected all of a party's candidates for different offices equally. It is, precisely, the percentage of the two-party vote received by the Democratic candidate for the national office of a pair, minus the percentage received by the Democratic candidate for the state office. Net deviations for three pairs of offices are analyzed in this study: President and Governor; U. S. Senator and Governor; and U. S. Congressman and Governor. There are three important reasons for studying net deviation in voting behavior: 1. Net deviation voting is substantial and has consistently increased over the last fifty years, as will be shown below in Chapter 2. In 1968, voters in 38 per cent of the states voted for a President of one party and a Governor of the other party; 16 per cent divided their votes in this way in the elections for U. S. Senator and Governor; 44 per cent divided their votes thus in the elections for U. S. Congressman and Governor; and voters in 48 per cent of the nation's congressional districts divided their votes this way in the elections for President and U. S. Congress- man. The causes of this amount of deviation from straight party ticket voting seem to deserve careful investigation. 2. This increase in net deviation constitutes a significant trend in the relationship of voters to the party structure which raises important questions about the relation— ship of the party structure to issues salient to voters across offices at different levels of government. An increase (appearance) of issues on which it seems rational to voters to vote for different parties at different levels of government is a source of strain on party identification and could cause a decline in the effect of party identification on the vote, or at least on the consistency of the influence of party identification at all levels of government. This would result in a weakening of the existing relationship between the voter and the party structure which could be potentially danger— ous to the party system as the stable attachment the voters have had with the parties weakens. 3. The phenomenon of net deviation from straight party ticket voting is virtually unexplained in the research literature. Studying the correlates of net deviation from straight party ticket voting can give us insights into the voters' effective party loyalties, their perceptions of the different levels of party organization and the ability of the different levels to handle problems unique to their level of government, and changes in voters' attitudes toward the parties over time. Approaches Used in the Study Three different approaches are used in the study. The first attempt is to develOp a model of net deviation from straight party ticket voting by regressing net deviation statistics for Montana counties on five political variables, two socio-economic variables, and eight demographic variables. Since net deviation voting is the net result of the voting behavior of an electoral unit, the independent variables must be eXpressed as ’3me — aggregate statistics for the electoral unit. In this study, the county is the electoral unit and the unit of aggregation for the independent variables. Ten hypotheses linking the independent variables to net deviation voting are tested with data from the Presidential election years of 1944, 1948, 1952, 1956, 1960, and 1964. It is concluded that no model could be develOped using aggregate data which would consistently eXplain a significant amount of net deviation from straight party ticket voting. None of the ten hypotheses was supported by the analysis. The second approach is an attempt to determine whether the degree of unionization of the dominant occupa— tion in a county was related to the amount of net devia- tion from straight party ticket voting. The hypothesis tested was: Counties in which the dominant occupations are highly unionized have the least amount of net devia- tion voting, while counties which have few or no unionized occupations have the most net deviation voting. It is concluded that there is no apparent relation— ship between highly unionized occupations in a county and the amount of net deviation from straight party ticket voting which occurs. The hypothesis is not confirmed. The third approach is a historical case study. The political history of the state is examined to see what forces have influenced voting behavior in the state over a long period of time. The activities of the two dominant interest groups in the state, the Montana Power Company and the Anaconda COpper Company, are examined to determine if such activities could be shown to have contributed to net deviation from straight party ticket voting over a period of seventy-five years. It is concluded that in a highly competitive state such as Montana, where the partisan division of the vote is almost evenly split, political activity to protect corporate interests by their supporting conservative Republican candidates for state offices and liberal Democrats for national offices results in the state having a high level of net deviation voting. This pattern is caused by the flexibility of corporate partisan support and the success of the corporations in convincing the voters of Montana that state interests require a split of party support between state and national levels. The more successful the corporations are in convincing the voters that such a configuration of interests exists, the higher the incidence of net deviation from straight party ticket voting. To illustrate this conclusion, the historical patterns of changing corporate partisan political activity are shown to correspond to the patterns of changing partisan control of state and national political offices. Purpose of the Study The general purpose of this study is to investi- gate the phenomenon of net deviation from straight party ticket voting, a phenomenon which is becoming more preva- lent in modern elections and thus more important to the student of politics who seeks to understand every facet of the behavior of the electorate. The specific purpose of this study is to determine if a predictive model of net deviation voting can be developed from the relationship between such voting and a number of theoretically relevant political, socio-economic, and demographic variables. Net deviation from straight party ticket voting has received no attention from students of voting behavior. In spite of the magnitude of net deviation voting which occurs, as shown below, there have been no studies reported in the literature devoted to this phenomenon. There have been a few studies of ticket—splitting by individual voters, but these are two different phenomena. This study is an attempt to fill part of the gap that exists in our knowledge about the causes of net deviation voting and the impact of such voting on the political system. The Concept of Net Deviation from Straight Party Ticket Voting Net deviation voting is different from "split- ticket voting," a concept commonly misused in referring to the behavior of the electorate in electing a candidate from one party to one office and the candidate of a different party to another office in the same electoral unit. Split—ticket voting is the act of voting for one or more candidates of each of two or more political parties on the same ballot. The term should be used only when such actions by the individual voters are known to have occurred, either by an examination of the ballot or from a statement by the voter that he did split his ballot. The term should not be used when analyzing aggregate voting statistics, because the amount of Split-ticket voting is obscured in the aggregate results by both the canceling effect of ticket—splitting in both directions and the extent to which incomplete ballots are included in the results. If one person votes a straight Democratic ticket except for Governor and does not vote for a candi- date for Governor, and another person votes for the Republican candidate for Governor and does not vote for any other office, the aggregate result would appear to be one split—ticket, where in fact there were two incomplete ballots. Another variation would be for the second voter to vote only for the Democratic candidate for Governor, making the aggregate result appear to be one straight ticket. In dealing with aggregate statistics, we have no way of knowing if either of these acts do occur, or how often they occur. There is no way to accurately measure "split—ticket voting" from aggregate statistics. Net deviation voting, on the other hand, is the net result of all such deviations from straight party ticket voting. It shows the net effect of ticket- splitting and incomplete ballots in the aggregate unit, making possible a comparison of aggregate units and an analysis of the variation of this net effect between units. The sign of the net deviation statistic indicates the direction of the deviation. The offices are paired in this study so the first office is always the national office and the second office is always the state office. A positive score indicates that the Democratic candidate for the national office did better than the Democratic candidate for the state office, and a negative score indicates that the Democratic candidate for the state office did better than the Democratic candidate for the national office. The net deviation score thus becomes a measure of the differential attractiveness of the Demo- cratic candidates at the two levels of governmental struc- ture and party organization. To attempt an analysis of every type of net deviation voting is beyond the sc0pe of a single study. The offices chosen for analysis here were selected to illustrate splitting between national and state offices, the most prevalent and consistent type of splitting that 10 occurs. The frequency of net deviation voting varies considerably from region to region and within regions. The area chosen for analysis was a single state--Montana-- in which net deviation is especially high. To reduce the units of analysis to the smallest unit and encompass as much variation as possible, yet keep the units politically meaningful, counties were chosen as the basic units of analysis. Counties had the added advantage of having statistical data readily available for them. A single state was chosen for the purpose of hold- ing the factor of ballot form constant. Several studies have concluded that, eSpecially for weakly motivated voters, a ballot which encourages them to vote a straight ticket, which can be done in many states by a single mark or pull of a lever on a voting machine, will usually result in a high rate of straight party ticket voting. Montana uses a system which rotates the names of the candidates alphabetically and which requires every name to be marked separately. Whatever the effect of ballot form on net deviation voting, it would only be on the amount of net deviation which occurs in the state, not upon the variation between counties within the state. This study does not assume that a statistical eXplanation of the variation in net deviation voting by the independent variables means that reliable predictions can be made of net deviations in the future solely on the 11 basis of this explanation. The stability or predictability of a set of variables cannot be determined in a single study. It is hOped that the application of the predictive model to six election years will, through a small amount of repetition, increase the reliability of the model. The primary aim of the study, however, is to formulate the problem and select variables which are theoretically relevant so that later repetitions of the study will produce a reliable predictive model. Victory or defeat of a candidate is not considered in this score. A large amount of net deviation voting may occur even though a party wins both offices, loses both, or wins one and loses the other. The eventual outcome of the election does not actually reflect the amount of deviation from straight party ticket voting which occurs. Summary This chapter has stated the purpose of this study, defined the concept of net deviation from straight party ticket voting, and discussed the method of measuring the concept and the contribution to the understanding of Voting behavior that can be made by studying the concept. Chapter 2 SPLIT-RESULTS ELECTIONS AND NET DEVIATION VOTING Introduction In this chapter, the national and Montana patterns of split—results elections are discussed to show the amount and increasing incidence of this phenomenon. This chapter also describes the pattern of split—results elec— tions and net deviation voting in Montana counties. Split-Results Elections in the United States and Montana This section shows the patterns of Split—results elections, one of the results of net deviation voting, in the United States and in Montana for the three pairs of offices analyzed in this study. A split—results election occurs when one party's candidate wins one office and another party's candidate wins another office in the same constituency. It occurs only as a result of net deviation from straight party ticket voting. When the frequency of occurrence of split—results elections is considered, there can be no doubt that a 12 l3 serious gap exists in our knowledge of voting behavior. Table 1 shows the amount, consistency, and increase in split-results elections in the United States from 1920 to 1968 between the offices of President and Governor, U. S. Senator and Governor, and U. S. Congress- man and Governor. Table 8 in Chapter 6 shows the partisan results of Montana elections for these four offices between 1899-1968. Table 2 shows the same results for the state of Montana but, because of the small number of such elections each year--one for each pair of offices—-the time periods have been grouped into three time periods, 1920 to 1936, 1940 to 1952, and 1956 to 1968. Figure 1 illustrates the pattern of split-results elections between the President and the Governor, Figure 2 illustrates the pattern of Split-results elections between the U. S. Senator and Governor, and Figure 3 illustrates the pattern of split—results elections between the U. S. Congressman and Governor. In order to classify those elections in which the Governor was not elected at the same time as the President, U. 8. Senator, or member of the U. S. House of Representa- tives were elected, a split—results election was considered to have occurred when the federal office holder was elected from a party different from the incumbent Governor. Table 3 shows the number and percentage of Con- gressional districts carried by the presidential nominee 14 of one party and the Congressional nominee of a different party in the U. S. and in Montana, 1920-1964. This example of Presidential and Congressional Split-results elections is used only to illustrate the increase in split ticket voting and will not be analyzed in this study. Figure 4 illustrates these patterns. These data make it clear that the incidence of split-results voting is rising both in the United States as a whole and in Montana for all three pairs of offices. The mean incidence of Split-results elections in Montana is considerably higher than the means over all states for these office pairs. When 42 per cent of the states have Split election results between the offices of President and Governor (1964), 16 per cent between the offices of U. S. Senator and Governor (1968), 44.1 per cent between the offices of U. S. House of Representatives and Governor (1968), and 48 per cent of the nation's congressional districts have Split-results between the offices of U. S. House of Representatives and President (1968), there is sufficient Split—results voting to justify research to examine this phenomenon, although it is not suggested that such research should reach a conclusion similar to that of the Ohio politician who reported as early as 1943 that his party was planning an entire campaign to suit the "rising Table 1. 15 Per Cent Actual of A11 Possible Splits in Election Results for Pairs of Offices, United States, 1920-1968, Presidential Election Years Year President & Senator & Congressman & Governor Governor Governor 1920 8.3 3.1 23.3 1924 20.8 7.3 25.3 1928 33.3 9.3 18.6 1932 16.7 4.1 34.2 1936 10.4 6.2 30.1 1940 35.4 8.0 30.8 1944 33.3 10.4 30.1 1948 33.3 6.3 27.6 1952 18.8 7.3 32.1 1956 39.8 8.3 38.7 1960 40.0 9.0 40.0 1964 42.0 15.0 39.1 1968 38.0 16.0 44.1 Source: Congressional Quarterly, Congress and the Nation, Vol. I, 1945-1964, Vol. II, 1965—1968 (Washington, D. C.: Congressional Quarterly Service, 1965 and 1969), Vol. I, pp. 67, 71-87; Vol. II, pp. 27, 33, 36-44. Table 2. Per Cent Actual of A11 Possible Splits in Elec— tion Results for Pairs of Offices for Montana, Grouped by Time Periods, 1920-1968, Presidential Election Years Time Number of President & Senator & Congressman & Period Elections Governor Governor Governor 1920—1936 5 40 0 20 1940-1952 4 50 66.7 62.5 1956—1968 4 50 100 62.5 Source: Waldron, Ellis, An Atlas of Montana Politics Since 1864_(Missou1a, Montana: Montana State University Press, 1958), pp. 176—8, 196—200, 220—2, 240-2, 246, 264—6, 270, 286-8, 306-8, 326—8, 356—8, 384—6 and Richard Scammon, America Votes, Vols. 4, 6 and 8 (Washington, D. C.: Governmental Affairs Insti— tute, 1966, 1968, and 1970), Vol. 4, pp. 235-7; Vol. 6, pp. 240-2; Vol. 8, pp. 223-7. 16 100 -——-— All states H Montana 90 80 7O 60 50 +- no 30 20 10 1920 1924 1928 1932 1936 1940 191M} 19148 1952 1956 1960 196‘} 1968 Figure l: Split-Results Elections Between the Offices of President and Governor, United States and Montana, 1920-1968. Source: see Table 1. 17 All states H Montana 90 100 80 7O 6O 50 1&0 30 20 10 1920 19215 1928 1932 1936 1940 19414 1948 1952 1956 1960 1964 1968 Figure 2: Split-Results Elections Between the Offices of U. S. Senator and Governor, United States and Montana, 1920-1968. Source: see Table 1. 18 All states H Montana 100 90 80 70 6O 50 1+0 30 20 10 1920 1924 1928 1932 1936 1940 19% 1948 1952 1956 1960 196“ 1968 Figure 3: Split-Results Elections Between the Offices Congressman and Governor, United States and Montana, 1920-1968. Source: see Table l. 19 Table 3. Congressional Districts with Split Election Results: Districts Carried by a Presidential Nominee of One Party and by a House Nominee of Another Party, United States and Montana, 1920—1964. Year and Party Number of of the Winning Districts Presidential Number of With Split Candidates Districts Results Per Cent Montana 1920 R 364 11 3.2 0 1924 R 356 42 11.8 50 1928 R 359 68 18.9 50 1932 D 355 50 14.1 0 1936 D 361 51 14.1 0 1940 D 362 53 14.6 50 1944 D 367 41 11.2 0 1948 D 422 90 21.3 50 1952 R 435 84 19.3 50 1956 R 435 130 29.9 100 1960 D 437 114 26.1 100 1964 D 435 145 33.3 50 1968 R 435. 209 48.0 100 Total 4903 1088 20.4 46.2 Source: Cummings, Milton, Congressmen and the Electorate, Elections for the U. S. House and the President, 1920—1964. New York: The Free Press, 1966, p. 32. Statistics for the election year 1968 and Montana have been added to the table and the totals brought up to date. Waldron, pp. 178, 200, 222, 242, 266, 288, 308, 328, 358, 386. 100 20 All states H Mont ana 90 80 7O 60 50 40 30 20 10 1920 192“ 1928 1932 1936 1900 191M 1948 1952 1956 1960 196“ 1968 Figure 4: Per Cent of Congressional Districts Carried by a Presidential Candidate of One Party and a House Candidate of Another Party, United States and Montana, 1920-1968. Source: see Table 20 21 tide of Split-ticket voting."1 The net deviation is important not only as a kind of voting behavior to be studied, but for the insights it can give into the attitudes the voters have toward the political parties and, through them, the nature of the federal system of government. An extensive discussion of these attitudes is beyond the scope of the present study, but the relationship between the net deviation effect and an attitude about the political system relevant to the analysis of Montana voting will illustrate this kind of attitude. Net deviation voting may be related to a rational policy-oriented attitude toward the political system. Members of economic interest groups, for example, may feel that one party's policy orientation is acceptable to them and corresponds to their economic interest orientation at all levels of government, resulting in consistent interest group support for the party's candidates. Organized labor groups are typical examples of this kind of consistent straight party support. Other economic interest groups may feel that the policy orientation of one party corre- sponds to their economic interest orientation at one level of government but not at another. Members of these lEllis Waldron, An Atlas of Montana Politics Since 1864 (Missoula: Montana State University Press,—l958), p. 4. 22 groups would feel that their interests would best be protected by supporting that party's candidates when they appear likely to support the groups' economic interests, and withholding support, or voting for the other party when such support does not appear to be forthcoming, resulting in a high level of net deviation voting. Farmers groups are often typical examples of this incon- sistent party support. In both cases, the voting behavior of the groups is the result of a rational policy—oriented attitude toward the political parties. Tables 4, 5, and 6 illustrate the amount of vari- ation which occurs among the counties in Montana by show- ing the frequency distribution of Net Deviations between the three pairs of offices, President/Governor, U. S. Senator/Governor, and U. S. Congressman/Governor, for each of the elections analyzed in this study. Table 7 shows the mean net deviation for the three pairs of offices between 1944—1964. As the tables show, there was a substantial amount of variation in net deviation among the counties in the state. This makes Montana a suitable state for testing the regression model. Its counties show high variation in net deviation voting over all counties between a single pair of offices in a single election, between different pairs of offices, and from one election year to another. Table 4. between the Offices of President and Governor 23 Frequency Distribution of Net Deviation Voting Per Cent of Counties with a Net Deviation of 1944 1948 1952 1956 1960 1964 —30 or less 14.3 5.4 —29.9 to —20 3.6 26.9 5.4 -19.9 to ~10 (8.9 44.6 19.6 — 9.9 to 0 34.1 14.3 41.0 17.9 0 to 9.9 3.6 32.2 23.2 53.5 3.6 10 to 19.9 12.5 12.5 5.4 25.0 33.9 20 to 29.9 50.0 8.9 3.6 44.6 30 and over 33.9 17.9 Table 5. between the Offices of U. S. Frequency Distribution of Net Deviation Voting Senator and Governor Per Cent of Counties with a Net Deviation of 1944 1948 1952 1956 1960 1964 -30 or less 1.8 —29.9 to -20 1.8 1.8 -19.9 to -10 8.9 10.6 - 9.9 to 0 25.0 32.2 7.1 0 to 9.9 41.1 35.8 28.6 1.8 10 to 19.9 16.0 16.0 51.8 1.8 20 to 29.9 5.4 1.8 12.5 23.1 30 and over 1.8 73.3 TflfleG. between the Offices of Congressman and Governor 24 Frequency Distribution of Net Deviation Voting Per Cent of Counties with a Net Deviation of 1944 1948 1952 1956 1960 1964 —30 or less 5.4 14.2 -29.9 to -20 8.9 26.8 ~19.9 to —10 19.2 26.8 1.8 1.8 — 9.9 to 0 21.2 19.7 5.2 1.8 21.2 0 to 9.9 3.6 10.7 8.9 37.5 26.8 68.1 10 to 19.9 10.7 14.3 3.6 26.9 55.4 8.9 20 to 29.9 21.3 14.3 19.7 16.0 30 or over 64.4 5.4 8.9 Table 7. Net Deviationl 1944-19 64. in Montana, for Designated Pairs of Offices, Year President & U. S. Senator & Congressman & Governor Governor Governor2 1944 26.96 36.22 1948 2.15 3.26 - 5.17 1952 —17.67 7.50 -13.65 1956 — 4.85 3.95 1960 6.72 11.50 13.24 1964 23.61 33.22 3.87 1Difference between percentage of votes received by the Democratic candidates for two offices. 2Average of the percentages in two Congressional districts. 25 It may be characterized as a highly competitive two—party state in which the voters choose freely between the two parties. Summary In this chapter, the patterns of split-results Voting and net deviation from straight party ticket voting in the United States and Montana have been presented. It has been established that the state of Montana has a relatively high rate of split-results elections and net deviation voting, and is a suitable state for testing the regression model. Chapter 3 A SURVEY OF RELATED LITERATURE Introduction A survey of the related literature in the field is reported in this chapter. Since nothing has been written about either net deviation from straight party ticket Voting or differential attractiveness, the survey is restricted to literature on the topics of split-ticket voting, party defection, and research using methods of analysis similar to the present study. Literature in the Field Three articles in the professional journals have been devoted specifically to Split-ticket voting. Two utilize the survey data gathered by the Survey Research Center at the University of Michigan. The more significant of the three is the article by Campbell and Miller.1 After identifying four voting types, the Indifferent Straight—Ticket Voter, the Indif— ferent Split—Ticket Voter, the Motivated Split-Ticket lAngus Campbell and Warren E. Miller, "The Motiva- tional Basis of Straight and Split Ticket Voting," American Political Science Review, 59 (1957), 293—319. 26 27 Voter, and the Motivated Straight-Ticket Voter, Campbell and Miller conclude that ticket-splitting depends on the strength of motivation of the voter, and that the most important factor motivating the voter to vote a straight ticket is the strength of his party identification. They also argue that the ease with which a straight ticket can be voted because of the form of the ballot facilitates straight—ticket voting. In considering differences in straight and Split—ticket voting of socio-economic groups, which range from 42 per cent to 69 per cent for straight— ticket voting and from 11 per cent to 42 per cent for Split-ticket voting, they conclude, Although there are differences of some size between certain of the socio—economic groups in the North in the prOportion who vote a straight ticket, in general the data give the impression of greater similarity than difference. One does not get the impression a . . . satisfactory understanding of why some pe0p1e vote straight and some Split can be arrived at from a knowledge of socio-economic position.2 Three weaknesses in the Campbell and Miller argu- ments raise doubts about the validity of their conclu- sions: 1. It is implied that strong motivation is synonymous with straight party ticket voting. For example, the Motivated Straight-Ticket Voter ". . . uses his vote to implement his political orientation . . . the more 2 Ibid., p. 296. 28 strongly motivated he is, the less he is willing to weaken his vote by splitting between two parties."3 The split— ticket voter could be as strongly motivated toward a political orientation which may be implemented only by splitting his ticket, such as the desire to avoid being a "party hack," or the conviction that the political system should be "balanced" between Democrats and Republicans. Yet the Motivated Split-Ticket Voter, according to Camp- bell and Miller, has "political motives which are in conflict and splits his ticket to compromise."4 His actions may appear to be in conflict--at least with the idea that straight—ticket voting is the normal form of behavior of strongly motivated voters--but his motives may be highly consistent with what he believes to be the pr0per way to cast his ballot. A farmer, for example, may take a rational policy—oriented attitude toward poli- tics which leads him to vote for a conservative Republican candidate for Governor because he feels such a man would protect his interests when policy is made at the state level on such actions as prOperty tax assessments, but which also leads him to vote for a liberal Democrat for President who advocates a high level of price supports for the crOps he produces. His motives in each case would 31bid., p. 303. 41bid., p. 311. 29 be consistent-—to protect his economic interests--and his ticket-splitting would not be a compromise but a rational action to achieve his preferences. 2. A more serious weakness in the article is the intermixing of levels of data in explaining Split—ticket voting. The questions on preference and motivation were asked in the context of national elections, yet much of the ticket—Splitting that is eXplained by these prefer- ences and motivations occur between the state and national levels of elections. For example, national candidates are referred to by name, but state candidates are not; party preferences are asked about the national party, but not the state party; and issue opinions are asked on national issues, not state issues. These preferences should not be cited in explaining Splitting between the two levels of government without citing comparable informa- tion on state party preferences, issues, or candidate orientation for state candidates. It is possible that much Splitting may be explained by voters having Split orientations to different levels of government in terms of party loyalty, candidate loyalty, and local issue orienta- tion. 3. In dismissing the eXplanation that straight- ticket voting is the product of indifference and lack of concern, Campbell and Miller state "We do not find . . . any significant tendency for the higher educational and 30 occupational groups to Split their tickets as this theory would lead us to expect."5 Straight party ticket voting may be the product of both indifference and lack of con- cern at one level and at another level be the product of intense concern. When the ballot form makes it easier to Vote a straight ticket than a split ticket, voters who are indifferent and are not concerned will vote an incomplete ballot. (Campbell and Miller do not specify whether incomplete ballots are considered something less than straight ticket voting.) When the ballot form makes it difficult to vote a straight party ticket, only those who are concerned enough to go to the trouble to mark a difficult ballot will vote a straight party ticket. The fact that highly interested voters vote a straight party ticket for one reason does not exclude the possibility that indifferent voters vote a straight party ticket for another reason. Ogden's article,6 based on the same survey data, also contains two similar criticisms of the Campbell and Miller approach: The fact that each American party contains within its ranks prominent leaders of almost every persuasion sIbid., p. 301. 6Daniel M. Ogden, Jr., "A Voting Behavior Approach to Split—Ticket Voting in 1952," Western Political Quarterly, XI (September, 1958), 492. 31 on virtually every substantial public question of recent times severely limits the use of . . . "issues partisanship" as a measure of party partisanship.7 This is especially true when the national party has, for example, a dominant liberal orientation and the local party has a dominant conservative orientation. . . . because we have no data on candidate partisan— ship below the presidential level, and because issue partisanship data are of limited validity, all analysis of either straight or split-ticket voting in terms of party direction is actually dependent upon either the party identification which the respondent assigns himself or upon the principal party vote which he says he cast. To Speak of conflicting motives in casting a long ballot, from such limited data, . . . is to jump to a conclusion which the available data will not support.8 Ogden's approach differs from Campbell and Miller's in that he assigns party labels to split-ticket voters on the basis of their vote for four principal offices (President, U. S. Senator, Congressman, and Governor) compared with their vote for state and local offices, classifying voters by party and direction of Split, rather than by level of government as do Campbell and Miller. A split-ticket voter is labeled a Democrat if he voted Democratic for two or more of the four principal offices and all of the state and local offices; if state and local offices were split, he would be labeled a Democrat only if he voted Democratic for three or more of the principal 71bid., p. 492. 81bid., p. 492. 32 offices; if he voted Republican at the state and local level, he would have to vote Democratic for all four principal offices. The reverse would be true for assign- ing a Republican label. All split-ticket voters who could not be labeled by this system would be Independents. The result of this system was to label 50.9 per cent of the Split-ticket voters Republican, 46.2 per cent Demo- cratic, and only 2.9 per cent Independents. Ogden then uses the chi square method to test for similarity between split—ticket voters and straight—ticket voters on the basis of eight groups of data generated by the questions in the Survey Research Center survey on party loyalty, attitude toward ticket-Splitting, etc. The eighth group was formed by demographic data, of which only two variables were found to be Significant in dis- tinguishing between the two types of voters: race and occupation. Ogden concludes (a) that "the myth that Republi— cans are more likely to split their tickets than are Democrats is now eXploded";9 (b) a great majority of ticket-Splitters are not significantly different from straight-ticket voters.lo These conclusions are not surprising, Since by Ogden's system of classification of 91bid., p. 489. lOIbid., p. 489. 33 voters a person could vote Republican for every office but Congressman and not be labeled a Republican, while another person could vote Democratic for all state and local offices but one, vote Democratic for President, and still be labeled a Republican. With this kind of inconsistency in the categories, practically no chi square would be significant. Jerrold G. Ruskll attempts to analyze the effects of an institutional prOperty of the electoral system--the Australian Ballot reform--on Splitfiticket voting patterns in America between 1876 and 1908, and advances a theory to predict the ways in which ballot might be eXpected to affect split-ticket voting. The value of Rusk's SOphisticated methodological analysis is destroyed by two mistakes: 1. Split-ticket voting is measured by subtracting the lowest percentage from the highest percentage of the two party vote cast for the Democratic party among the array of statewide races in any given state and election year—~a measure authored by Walter Dean Burnham. This is not a measure of Split-ticket voting, but rather a measure of the net effect of split tickets and incomplete ballots-- 11Jerrold G. Rusk, "The Effect of the Australian Ballot on Split Ticket Voting, 1876-1908," American Political Science Review, 64 (December, 1970), 1220—1238. 34 or, as I have defined it, net deviation from straight party ticket voting. This is not just a semantic problem, because in the article Rusk discusses——as a different finding--the contribution of the Australian ballot vari- able in eXplaining "roll—off," or the tendency of the electorate to vote in greater numbers for high than for low prestige offices on the ballot. "Roll-off," of course, results from incomplete ballots, and is a part of the variable, Split-ticket voting, as defined by Rusk. This confusion about the dependent variable is the result of Rusk's failure to recognize that the incidence of individ- ual ticket-splitting cannot be measured from aggregate statistics. 2. The use of an independent variable which occurs in only one election to develop a predictive model for explaining a dependent variable which occurs in all elections seems to be wasted effort. If the Australian Ballot reform explains the increase in ticket—splitting in 1876—1908, as Rusk maintains--which is Open to serious doubt-~what then explains the ticket-splitting which occurred--and occurs at an increasing rate each election—- between 1908 and 1970, when there were no ballot reforms? Split-ticket voting-~or net deviation voting--is caused by several factors, of which one may by an institu- tional property such as the Australian Ballot reform, but Rusk has not proved it. If, as Rusk argues, each election 35 has a different combination of variables which explain split-ticket voting, then the effort eXpended to show the relationship between split-ticket voting and a variable which occurs only once does nothing to test a predictive model, unless the prediction to be made is that if the Australian Ballot reform is ever attempted again, it will be related to an increase in ticket—splitting. Rusk's effort is of doubtful value. A theoretical model must be applicable to more than one election if it is to contribute to our understanding of the electoral process. The only other major attempt to analyze split- ticket voting to be found in the literature is a chapter in a book by Milton Cummings.12 His data, however, are from aggregate statistics, so he is actually analyzing net deviation from Split—ticket voting. The form of splitting is that between President and Congressman. Cummings is primarily interested in the incidence of ticket—splitting and the effect it has on the political system. He does attempt to explain why voters split their tickets, but relies almost completely on Campbell and Miller's article cited above. The criticisms of that 12Milton C. Cummings, Jr., The Congressman and the Electorate, Elections for the U. S. House and the Presi- dent, 1920-1964 (New York: The Free Press, 1966), pp. 28-55. 36 article are equally appropriate for Cummings' analysis of motivation. In the process of analyzing his charts on the incidence of ticket—splitting, Cummings makes some puzzling statements. Discussing a table listing the percentage of Congressional districts with Split election results from 1920 to 1964 which shows an increase for each presidential election year over the preceding presidential election year for nine of the twelve elections, and an overall increase from 3.2 per cent to 33.3 per cent, Cummings notes ". . . there has been no regular, steady increase in 13 Also, "The three most ticket—Splitting since 1920." recent presidential years of 1956, 1960 and 1964 were years when ticket-splitting was unusually wideSpread. But so also were 1924 and 1928."14 1956, 1960, and 1964 were high splitting years, but only one of the twelve years (1944) had less splitting than 1924, and every year since 1944 has had more than 1928.15 In an interesting article on the causes of party defection, Boyd found that cross pressures relate mono- tonically to voting defection—-the greater the cross 13Ibia., p. 46. 14Ibid., p. 46. lSIbid., p. 32. 37 pressures, the greater the likelihood of party defection.l6 This conclusion was not surprising, considering that cross pressures are defined as conflicts between attitudes toward political parties, the nominees of those parties, and the policies associated with the political parties. Boyd does establish that the three sets of political attitudes combine interactively, rather than additively, in their impact upon the voting act. Conflicts between any of these attitudes reinforce one another in their impact on the voting act, resulting in a higher rate of party defection than one would expect from adding together the separate effects of each of the attitudes.17 Boyd failed to establish the causes of cross—pressures. Status inconsistency, inter—generational status mobility, and geographical mobility were unsuccessfully tested in trying to explain cross—pressures. No causes of cross—pressures were found in Boyd's study.18 An article by Press19 analyzes the question of variation in motivation underlying voting behavior. He 16Richard W. Boyd, "Presidential Elections: An Explanation of Voting Defection," American Political Science Review, 63 (June, 1969), 498-514. 17 Ibid., p. 509. 181616., p. 509. 19Charles Press, "Presidential Coattails and Party Competition," American Political Science Review, 52 (June, 1958), 1041-1051. 38 concludes that the "coattails effect" is most crucial in competitive districts; that when the ballot makes it easy to vote a straight ticket, the indifferent voter will do so; that the importance of party, issue, and candidate factors change from election to election; and, most importantly, the competitive pattern of the district has a strong effect on voter motivation. In a brief research note, Schoenberger, in assessing the effects of party loyalty to an unpopular nominee for President on the outcome of Congressional elections, found that the displacement of popular hostility to Goldwater onto the Congressmen who supported him most strongly in 1964 cost the party a good many seats in Congress.20 He concludes that these findings "underscore the limits of empirical usefulness of models that assume that party leaders always act in common."21 The remaining references to net deviation from straight-ticket voting in the literature are found as incidental remarks in studies of voting in general,22 20Robert A. Schoenberger, "Campaign Strategy and Party Loyalty: The Electoral Relevance of Candidate Decision-Making in the 1964 Congressional Elections," American Political Science Review, 63 (June, 1969), 519. 21 Ibid., p. 519. 22Angus Campbell, Philip E. Converse, Warren E. Miller and Donald E. Stokes, The American Voter (New York: John Wiley and Sons, Inc., 1964). 39 references to ballot form,23 and inter—party competition.24 Several studies have been made using ecological variables to eXplain political behavior. Most relevant to this study are the ones which relate ecological variables to voting behavior, especially those which use multiple regression analysis. Simmons25 developed a scale of three dimensions of urban socio—economic variation (social class, urbanization, and segregation) and related them to party support in an urban election. He found that 31 per cent of the total variation in support of the three parties could be accounted for by these three dimensions. Cutright and Rossi26 did a combination ecological and survey study, using a multiple regression equation to link the voting records and a precinct's social charac- teristics which provided an estimate of how each precinct 23Leon D. Epstein, "Size of Place and the Division of the Two-Party Vote in Wisconsin," Western Political Quarterl , 9 (March, 1956), 138—151; Jack Walker, "Ballot Forms and Voter Fatigue: An Analysis of the Office Block and Party Column Ballots," Midwest Journal of Politics, 4 (1966), 448-459. 24J. A. Robinson and R. E. Standing, "Inter—Party Competition and Primary Contests," American Political Science Review, 52 (1958), 1066—1079. 25James W. Simmons, "Voting Behavior and Socio- Economic Characteristics: The Middlesex East Federal Election, 1965," Canadian Journal of Economics and Politi— cal Science, 33 (1967), 389—402. 26Phillips Cutright and Peter H. Rossi, "Grass Roots Politicians and the Vote," American Sociological Review, 23 (1958), 171—179. 40 would vote if its social factors were the only ones which influenced the vote. Kenney and Butts27 used multiple regression analysis in a slightly different way to reveal that fifty- one variables of social, political, or economic nature accounted for 91 per cent of the variation in pOpulation change in Mississippi from 1950 to 1960, and for 81 per cent of the variation in Negro voter registration in 1966. Hofferbert28 established strong relationships between ecological variables and public policy variables. Bogue and Harris29 used multiple regression analysis to relate metropolitan growth in Illinois with certain socio- economic and demographic variables. A study which did not use multiple regression but analyzed state party strength in terms of ecological variables was done by Golembiewski30 in which fifteen 27David Kenney and William Butts, "Variables Associated with POpulation Change and Negro Voter Regis- tration in Mississippi," Public Affairs Bulletin, Vol. 1, NO. 1 (1968)! PP. 1‘5. 28R. I. Hofferbert, "Ecological Development and Policy Change in the American States," Midwest Journal of Politics, 4 (1966), 464-472. 29Donald J. Bogue and Dorothy L. Harris, Compara- tive POpulation and Urban Research via Multiple Regression and Covariance Analysis (Chicago: Scripps Foundation for Research in Population Studies), #8, 1954. 30R. T. Golembiewski, "A Taxonomic Approach to State Political Party Strength," Western Political Quarterly, 11 (1958), 494—505. 41 ecological variables, predominantly pOpulation and growth variables, were analyzed using charts and t—tests for Significance. Ten of the fifteen variables were found to be significantly related to party strength. Robinson and Dawson31 utilized economic variables in a study relating inter—party competition, economic variables, and welfare policies. The relationships lost some of their validity because the index of inter—party competition did not take the cyclical nature of party competition into considera— tion. These studies indicate the interest in ecological factors as useful variables in explaining political behavior, and Show the different ways in which they have been used to analyze elements of the political process. They also illustrate the ingenuity with which statistical tools have been utilized in these analyses. They have been valuable to the present study in defining some of the problems encountered in the use of aggregate ecological statistics. The problem of the ecological fallacy, for example, was faced in one form or another by most of the analysts. The solution presented below in Chapter 4 was the result of an evaluation of their successes and failures. 31R. E. Dawson and J. A. Robinson, "Interparty Competition, Economic Variables and Welfare Policies in the American States," Journal of Politics, 25 (1963), 42 Summary No literature exists on the tOpic of net deviation from straight party ticket voting, so relevant literature on the tOpics of Split—ticket voting, inter-party competi- tion, and party defection were reviewed in this chapter. Research using methods of analysis similar to those used in this study were also reported. Chapter 4 THE REGRESSION MODEL Introduction This chapter is devoted to the development of the regression model and a discussion of the statistical methods used in the design and testing of the model. The independent variables are defined, the methods of obtain- ing their values are eXplained, the hypotheses relating them to the model are stated, and the theoretical linkages between net deviation voting and each of the independent variables are established. The method of multiple regres— sion analysis, its application to this study, and the statistics for interpreting and testing the multiple regression equation are discussed. The problem of making inferences about individual behavior from aggregate statistics is discussed, and the kind of inferences made in this study are justified. The Regression Model The theoretical model tested in this study con~ sisted of a multiple regression equation showing the association between net deviation from straight party 43 44 ticket voting and two political variables, two socio— economic variables, and nine demographic variables. Ten hypotheses link these variables to the dependent variable, net deviation voting. The usefulness of the regression model derives from the precision with which it shows the total variance in net deviation voting that is associated with the vari— ance in all the independent variables acting together, and the amount of variance in net deviation voting that is explainable with the variance of a given change in any one of the independent variables with the remaining independent variables partialed out. If a substantial amount of the variance in net deviation voting can be shown to be associated with the independent variables, then one might be able to predict the amount of net deviation voting in future elections from the values of the independent vari— ables at that time. Since the most competitive counties have the lowest IPC scores, a high correlation between IPC and net deviation voting will be shown by a negative correlation coefficient. The Dependent Variable in the Regression Model The dependent variable, net deviation from straight party ticket voting, is the measure of the differential attractiveness between the two Democratic candidates for 45 the pair of offices being analyzed. Three different net deviation variables are used in this study: NDl, between the offices of President and Governor; ND2, between the offices of U. S. Senator and Governor; and ND3, between the offices of Congressman and Governor. The Independent Variables in the Rggression Model The political variables are Inter—Party Competi— tion and Voting Turnout. The socio-economic variables are Age and Education. The demographic variables are POpula— tion, POpulation Density, Mobility, Per Cent of the POpulation Living in Urban Areas, Per Cent of the Popula- tion Not Employed on Farms, Country of Origin (Scandi— navia), Country of Origin (British Islands), Total Foreign Born, and Non—white POpulation of the County. The values of the independent variables were obtained from the decennial reports of the U. S. Bureau of the Census.1 The six presidential years in this study took place during three different census reporting periods, so each of the independent variables had three different values. Their values for the 1944 and 1948 elections came from the 1940 Census; for the 1952 and 1956 elections, the values came from the 1950 Census reports; and for the 1U. S. Government Printing Office, Bureau of the Census (Washington, D. C., 1960). 46 1960 and 1964 elections, the values came from the 1960 Census reports. The values of the Country of Origin variables were combinations of the individual country statistics reported by the Census Bureau and represent the per cent of the counties' pOpulation born in the Country of Origin. The values of the remaining variables were taken from the corresponding categories listed in the Census report. The dependent variable, net deviation voting, varies over the counties in the state according to the Variation in the differential attractiveness of the two candidates from the same party running for two different offices relative to their respective Opponents. This differential attractiveness between the two candidates results in a difference between their percentages of the Votes cast for the two offices. The amount of net devia— tion voting is the measure of this differential attractive- ness, and varies over the counties as the differential attractiveness of the two candidates varies from county to county. A positive net deviation score indicates that the candidate for the national office had more differential attractiveness than the candidate for the state Office. A negative net deviation score indicates that the candidate for the state office had the most differential attractive— ness. 47 All the independent variables used in these hypotheses are aggregate measures of characteristics of individuals hypothesized or shown in earlier research to 'be associated with individual ticket—splitting. Differ— ences in these independent variables over the counties are to be used in the regression analysis to explain differ- ences in net deviation voting. As explained earlier, net deviation voting is not a direct measure of individual ticket-splitting, but rather a measure of differential attractiveness of a party's candidates for two different offices. The hope is that in the presence of some differential attractiveness overall, it will be possible to associate the independent Variables with differences between the counties in their response to this differential attractiveness——a reSponse that depends on the rate of individual ticket—splitting and incomplete balloting, in the direction of the differ- ential attractiveness of the candidates. The Political Variables Hypothesis 1: The amount of net deviation voting in a county increases as Inter—Party Competition in the county increases. The Inter-Party Competition scale (IPC) is one 48 developed by Hofferbert2 to improve on the index of party competitiveness develOped by Schlesinger3 to measure both overall and cyclical competitiveness in a state. Hoffer- bert's classification system is especially apprOpriate for this study because it is a single ordinal rank index and is more easily adapted for use in a multiple regression equation than a multi—rank index would be. The IPC scale allows all counties to be ranked from most competitive to least competitive, with no ties. The most competitive county receives a score (rank) of one, and the least competitive county receives a score of fifty-six. It also allows for the measurement of competi— tiveness even when one party wins all Offices in all elections, which makes it especially appropriate for the purposes of this study. Competitiveness in a county should increase the amount of net deviation voting in the county. The more competition there is between the two parties, the more utility there will be to the voters in ticket—splitting and selective non-voting. If there is little or no compe- tition between the parties, the utility to the voters of 2R. I. Hofferbert, "Classification of American State Party Systems," Journal of Politics, 26 (1964), 3Joseph A. Schlesinger, "The Structure of Competi— tion for Office in the American States," Behavioral Science, 5 (1960), 197—210. 49 voting for a candidate of a party which has little chance of winning is reduced. Given that the voters are attracted in different degrees to a party's candidates for two offices, they have much greater incentive to express this difference in preferences when competition is closer, hence the probability that each voter's vote will deter- mine the outcome is higher. Inter—party competition, then, increases the utility to the voter of ticket—Splitting and selective non-voting which, in the presence of differential attractiveness, increases net deviation from straight party ticket voting. Counties which experience higher turnout are likely also to be counties with a higher proportion of Weak or non—party identifiers among those actually voting. Because individual ticketvsplitting is known to be associ- ated with weakness of party identification, an increase in the number of weak identifiers will result in an increase in ticket—Splitting and, consequently, in net deviation voting. Hypothesis 2: Net deviation voting increases as the voting turnout for the offices in the pair increases. The turnout variable is measured as the per cent of persons Over twenty—one years of age in the country who Voted for the office of President. 50 Increases in voting turnout are associated with increases in the prOportion of voters who are either marginal party members or non—party members. These mar— ginal voters are usually attracted by either the appeal of a particular candidate or by a particularly salient political issue. They are not likely to be strongly attached to either party, so their presence in the electorate will increase the amount of split-ticket voting and selective non-voting which occurs in the election. Campbell and Miller found that straight party ticket voting is higher among strong party identifiers.4 "A straight ticket is after all a straight party ticket, not a candidate or an issue ticket."5 Hypothesis 3: Net deviation voting decreases as the median age of the county increases. Older voters, having observed the political parties over a longer period of time, have develOped a richer and more personally meaningful association with their party than the younger person who sees the parties in terms of what they have been in his eXperience. To the older voter, 4Campbell and Miller, "The Motivational Basis for Straight and Split-Ticket Voting," p. 306. 5Ibid., p. 305. 51 for example, the Democratic party is not only the party of Humphrey, Johnson, Kennedy and Stevenson, but also the party of Truman and Roosevelt. The young voter sees the party in terms of what it has done recently, in the Cold War period, while the older voter sees it in terms of what it has done over a longer period of time embracing the Depression of the 1930's, the Second World War, and the Cold War. Strength of party identification increases as the length of the voters' psychological commitment to the party increases, not just in terms of number of years but in terms of the prOportion of one's life that has been committed to it. Since most voters stay with the same party most of their lives, strength of party identifica— tion increases with age. A political party signifies little more to the young voters than its current leaders, so they are psychologically free to shift their vote from party to party.6 Older voters, then, tend to be party oriented, while younger voters tend to be candidate oriented. Party orientation reduces Split—ticket voting and selective non—voting and, consequently, net deviation from straight party ticket voting. Campbell, 2E.21' reported a decrease in ticket—Splitting correlated with increase in age in the 6Campbell, EE gr., The American Voter, p. 264. 52 presidential election years of 1952 and 1956.7 The more older voters a county has, and the higher its median age, the less net deviation voting it will have. Hypothesis 4: Net Deviation voting increases as the median years of school completed in the county increases. Education is positively correlated with net deviation voting because educated voters have more interest in, and are more involved, in politics.8 As the level of education in the county increases, the level of involvement in politics increases. The more involved the Voter is in a political campaign, the more information he has about the parties and the candidates, which in turn increases his ability to discriminate between those candidates who are more attractive to him and his political ideas and those who are less attractive to him. The educated voters' increased interest and information about the campaign leads to a higher rate of ticket—splitting and selective non- voting which, given the condition of differential attrac- tiveness, will result in more net deviation from straight party ticket voting. 7Campbell and Miller, pp. 297—8. 8Campbell, gr 31" The American Voter, p. 253. 53 Campbell, §r_§l., found that "There is no Signifi- cant relationship between strength of party identification and formal education, but the educated stand out with great clarity on the motivational dimensions we call 'sense of political efficacy' and 'sense of citizen duty.'"9 Campbell and Miller found that, in the presiden- tial election years of 1952 and 1956, the per cent of voters who voted incomplete ballots declined as amount of . . 10 education increased. The Demographic Variables Rural-Urban Characteristics: Hypothesis 5: Net deviation voting is more frequent in rural counties, less frequent in more urban counties. Four variables are used as alternative measures of the Rural-Urban Characteristics: A. The population of the county. Counties with large populations are more urban than counties with small pOpulations, so the scale is from least rural for large pOpulation to most rural for small population. 91bid., p. 253. 10Campbell and Miller, "The Motivational Basis for Straight and Split-Ticket Voting," p. 297. 54 B. Population density, measured as people per square mile. Counties with the highest pOpula- tion density are least rural, and counties with the lowest population density are most rural. C. Per cent of the pOpulation that lives in urban areas. Urban areas in Montana are not com— parable to urban areas in most other states because there are only two urban areas with more than 50,000 pOpulation. Since the rural— urban contrast is the important factor, in this study an urban area is one which exceeds 5,000 peOple. Counties with the highest per cent of its pOpulation living in urban areas will, of course, be the least rural counties, and counties with the lowest per cent of its pOpulation in urban areas will be the most rural counties. D. Per cent of the population employed on farms, including owners of the farms who actually work on them. The counties with the highest per cent of the farm employees are the most rural, and the counties with the lowest per cent of farm employees are the least rural. These measures are defined so that high values of each measure, with the exception of (D), indicate the less 55 rural, more urban counties. Hence, the general hypothesis would be confirmed by negative correlations in between (A), (B), and (C), and a positive correlation with (D). This hypothesis is explained by the nature of the relationship between the rural voter and the political campaign. In a rural county, with fewer peOple, sparsely settled, who have lived in the county for many years, and with fewer organized grOUps to intervene between the voter and the candidate, there is more personal contact between the voter and the candidates for state offices.11 The rural voter is more oriented to the candidate, because of his personal contact with him, than he is to the party. Candidate orientation is associated with ticket-Splitting and selective non—voting. Farmers split their tickets more than any other occupational group. Table 8 shows the per cent of voters who Split their tickets, by occupation, in the presidential election years of 1952 and 1956. The Variability of party choice of the farmer has been established by Campbell, gr al.12 Counties with stronger candidate orientation will, given the condition of differential attractiveness of the candidates of the same party, have a higher rate of net deviation from straight party ticket voting. llIbidc, pp. 297—80 12Campbell, gr gr.,'The American Voter, p. 211. 56 Table 8. Per Cent of Voters Who Split Their Tickets, by Occupation, in 1952 and 1956 Occupation Eisenhower voters Stevenson voters Professional 45 29 Businessmen and Managers 39 40 White Collar Workers 37 19 Skilled and Semi—skilled 37 27 Unskilled 46 24 Farm Operators 51 47 Extracted from Tables II and III, pp. 297—8, Campbell and Miller, "The Motivational Basis of Straight and Split— Ticket Voting," APSR, LI (June, 1957), 293-312. 57 Hypothesis 6: Net deviation voting is higher in counties with high mobility, lower in counties with low mobility. Mobility is measured by the per cent of the people in the county who have lived in the county for five years or more. People who have recently moved into a county have not become as strongly identified with the local political parties as people who have lived in the county for longer periods of time. Until they establish a stronger identi- fication with the party, mobile voters are more candidate oriented than party oriented. Candidate oriented voters have higher rates of ticket-splitting and selective non- Voting, so counties with higher rates of mobility will, in the presence of differential attractiveness of two candidates of one party, have more net deviation from straight party ticket voting. Hypothesis 7: Counties with high concentrations of Scandinavian immigrants have more net deviation voting than counties with few or no Scandinavian immigrants.* *The Scandinavian countries are Norway, Sweden, and Denmark. 58 Immigrants from Scandinavian countries come from a culture which is very different from the culture in the U. S. and Montana. The ease with which they are assimi- lated into the political culture of the U. S. is hindered by the barriers of language, living style, and political history. They are rarely familiar with elections as con— ducted in Montana, especially the widespread participation of organized interest groups in political party affairs, and are not converted immediately into strong partisan voters, or, in fact, voters who turn out with high fre— quency. Voting turnout increases as their eXperience in the political system increases and, of course, as they acquire the rights of citizenship when their status as immigrants is changed. They have settled into the pre— dominantly rural areas of Montana. The interaction of the factors of low voting turnout, rural residence, and imperfect assimilation into a political party culture results in Voters who are candidate oriented, rather than party oriented. Candidate orientation, as mentioned above, is associated with higher ticket—Splitting and more selective non-voting. Counties with higher concentrations of Scandinavian immigrants will, because of their stronger candidate orientations, in the presence of differential attractiveness of the candidates of the same party, have a higher rate of net deviation from straight party ticket voting. 59 Hypothesis 8: Counties with high concentrations of immigrants from the British Islands have less net deviation voting than counties with few or no immigrants from the British Islands and Canada.* Immigrants from the British Islands are more easily assimilated into the American political culture than are immigrants from most other countries. Not only do they speak almost the same language, but their politi- cal history Shares common elements with U. 8. political history. The participation of organized interest groups, especially labor unions, in political party affairs is not new to them. They have settled predominantly in the mining regions of Montana, where the miners' unions are very active in partisan politics and strongly Democratic. These areas are also urban in nature. The interaction of the factors of high voter turnout, urban residence, ease of assimilation into a political party culture, and member- ship in labor unions which encourage party loyalty results in voters who are party oriented, rather than candidate oriented. Party orientation is associated with low ticket— Splitting and low selective non—voting, so the counties *Immigrants from the British Islands are from England, Ireland, Scotland, and Wales. 60 with high concentrations of immigrants from the British Islands and Canada will have lower rates of net deviation from straight party ticket voting. Hypothesis 9: Net deviation voting increases as the per cent of foreign-born immigrants in the counties increases. Although voters born in the British Islands are exceptions to this hypothesis, for reasons discussed above, most foreign born voters are similar to the Scandinavian immigrants. Table 9 shows the per cent of the total foreign born population in the state, broken down by the major Census Bureau ethnic groupings, for the three decades covered in this study, and the per cent of the state's pOpulation born in all foreign countries. The British Islands ranks third in 1940 with 15.8 per cent of the state's foreign born; ranks fourth in 1950 with 11.9 per cent; and ranks fifth in 1960 with 12.6 per cent. Since the immigrants from the British Islands constitute a small portion of the total immigrants in Mon— tana, on the whole the immigrant population will vote more like the Scandinavians for the same general reasons. 61 Table 9. Origins of the Foreign Born in Montana 1940 1950 1960 Per Cent of State Population Foreign Born 9.9 7.3 4.5 Country of Birth of Foreign Born: Eastern EurOpe 19.0 19.0 20.0 Scandinavia 24.7 23.6 19.4 Russia & Finland 10.3 11.0 10.9 British Islands 15.8 11.9 12.6 Southern EurOpe 8.1 6.8 5.6 Canada 15.4 17.3 18.3 Other , 6.7 10.4 13.2 Total 100.0 100.0 100.0 United States Bureau of Census Reports, Characteristics of the Population, Part 28, Montana (Washington: Government Printing Office, 1960), Table 40, p. 89. 62 Hypothesis 10: Net deviation voting will be high— est in counties with the highest concentrations of non-whites, and lowest in the counties with the lowest concentrations of non- whites. This is a special category in Montana and refers more to American Indians than any other race because of the large number of Indians who live on reservations in the state. The fact that reservation Indians are wards of the federal government makes this a potentially signi- ficant variable. Voting turnout is low in the Indian pOpulation, especially for state offices. Interest in state elections is low because the state has little to do with the well—being of the reservation Indians. The fed— eral government, on the other hand, is an important source of the necessities of life to most of the reservation Indians. The tribes who have income of their own from non- governmental sources are exceptions to this dependence upon the federal government, and their voting patterns are different from the dependent tribes, both in terms of turnout and support for incumbent national administrations. However, they constitute a small portion of the Indian pOpulation. Most of the reservation Indians support incum— bent administrations at the national level, but vote 63 infrequently for state Offices. This selective non-voting, in the presence of differential attractiveness of candi— dates at different levels of government, results in a high rate of net deviation voting. The Method of Multiple Regression Analysis Multiple regression analysis is a powerful and efficient statistical method of estimating values of a dependent variable from values of multiple independent variables.13 The multiple regression equation is not only a more exact method of determining these estimates than linear equations obtained by any other method, but it allows the researcher to state the association between differences in the dependent variable and the differences in each of the independent variables in a condensed form for a large number of Simultaneous observations. The multiple regression equation takes the form Y = a+blxl+b2x2+...bixi+e where Y_is the dependent variable; 3 is a constant which estimates the value of Y which may be expected when each 13Multiple Regression analysis is discussed extensively in Mordecai Ezekiel and Karl A. Fox, Methods of Correlation and Regression Analysis (3rd Edition (New York: John Wiley and Sons, Inc., 1959); Donald J. Bogue and Dorothy L. Harris, gp_grr., pp. 1-18; and Statistical Series Description No. 7, Michigan State University EXperiment Station, April, 1968. 64 of the other independent variables has a value of zero; 2 is the regression coefficient which shows the units increment in Y associated with one unit increment in the independent variable when allowance has been made for the other independent variables; and the r's denote independent variables. The multiple regression equation is a test for the hypothesis that the behavior Observed for a dependent variable may be accounted for in terms of the independent variables in the equation. NO inference of causation is made in the statistical relationship; such inferences must be made from theoretical interpretations of the statistical results. The most useful characteristic of multiple regres- sion analysis, particularly in the development of a predicative model, is the precision with which the results Show how much of the variance of the dependent variable is associated with the independent variables in the equa- tion. Five statistics calculated about the multiple regression equation are particularly important in inter- preting the equation: 1. The coefficient of determination (R2), which Shows what prOportion of the variance in the values of the dependent variable can be eXplained by, or estimated from, the 65 concomitant variation in the values of the independent variables. This is usually expressed as the percentage of the variation eXplained. The standard error of estimate is a measure of the accuracy of the estimated values, or how closely they agree with the actual Values observed for the variable being estimated. F statistics for testing: a. that coefficients correSponding to the subsets Of independent variables are Simultaneously zero (i.e., do not add to the explanation of the dependent vari- able); and b. hypotheses regarding the individual least squares coefficients. The highest order partial correlation coeffi— cients between the dependent variable and each independent variable, a measure of the extent to which that part of the variation in the dependent variable which was not eXplained by the other independent variables can be explained by a single independent variable. The residuals, the difference between the actual value and the estimated value of the 66 dependent variable for each observation. These are important in testing the accuracy of the regression model for predictive pur— poses. A stepwise regression method was used in this study. A stepwise deletion computer program was used which deletes variables from the regression equation if they do not meet Specific significance criteria.* It first cal- culates an initial least squares (or multiple regression) equation using all of the independent variables. The statistically least significant variable is then deleted from the equation and a new least squares equation is estimated. A second statistically least significant Variable is deleted and the least squares equation is recalculated. The procedure continues until a variable selected as a candidate for deletion meets the stopping criterion. The stOpping criterion used in this study was the significance level of the F test, and all the variables remaining in the least squares equation were significant at the .05 level. Variables may be retained by this procedure which would have been discarded if only one equation had been calculated, Since the elimination of one *The program used was the Stepwise Deletion of Variables from a Least Squares Equation (LSDEL), prepared by the Michigan State University Agricultural Experiment Station. ‘ 67 variable from the equation often changes the regression coefficients of the other independent variables in the equation. Individual Behavior and Net DeviatiOn Voting Robinson raised the question of the validity of making inferences about correlations between variables, using individuals as units, on the basis of correlations based on groups as units when he argued that ecological correlations, in which the statistical Object is a grggp of persons, cannot be validly used as substitutes for individual correlations, in which the statistical object is indivisable, because there need be no correspondence between the individual correlation and the ecological correlation.l4 Inferences about individual correlations are indirectly based on aggregate correlations in the regres— Sion model used in this study, so the problem of the ecological fallacy must be dealt with here. Net deviation voting, which is a measure of the concept of differential attractiveness between two candi— dates of the same party, is a collective statistic. It is the net result of individual actions. Individual 14W. S. Robinson, "Ecological Correlation and Behavior of Individuals," American Sociological Review, 15 (1950), 351. 68 voters cannot "net deviate vote," so it cannot be an individual action, but is the result of a group of indi— vidual actions. Using the regression model to predict net deviation voting does not involve making inferences about individual actions; it is predicting a group (col— 1ective) action from_group (aggregate) statistics. However, the variables which are used to construct the regression model are individual statistics, so the hypotheses linking these individual statistics to the regression model are making inferences about individual behavior on the basis of aggregate statistics. The structure of the regression model depends upon these inferences, so the model is, in predicting an aggregate statistic, indirectly relating it to individual behavior. The relationship between net deviation voting and individual voting behavior is complex. A high net devia- tion score means that one candidate is more attractive, relative to the Opposing party's candidates, to the voters than another candidate from his party. A low net devia— tion score, however, indicates a more complex relationship. This relationship is shown by the following scale of the contribution Of the individual to the vote totals used to calculate the net deviation statistic: +2 +1 0 —l -2 T-S S N-V S-T S N-V T-S for l for 1 D or R for 2 for 2 or No vote 69 T-S for 1 means the voter split his ticket, voting for the Democratic candidate for office one and the Republican candidate for office two; S N—V for 1, selective non— voting in favor of candidate one, means the voter voted for the Democratic candidate for Office one but did not vote for office two; S—T, D or R, or no vote, means the voter either voted a straight Democratic party ticket or a straight Republican party ticket, or did not vote for either office one or Office two; S N—V for 2, selective non-voting in favor of candidate two, means the voter Voted for the Democratic candidate for office two, but did not vote for office one; and T—S for 2 means the voter split his ticket by voting for the Democratic candidate for office two and the Republican candidate for office one. The scale values are based on the contribution to the net deviation score made by each type of voter response. Since the net deviation score is a measure of the differential attractiveness of the candidates, there is no net deviation score if both candidates get the same percentage of the two-party vote cast, i.e., there is no differential attractiveness. If the voter votes a straight Democratic ticket or a straight Republican ticket, both candidates get the same percentage of the vote. If the voter does not vote for either office, the difference between the percentage each candidate gets of the total vote is not affected. In all three cases, the 7O contribution to the net deviation score is zero, so the scale score for these responses is 0. If the voter votes Democratic for office one and does not vote for office two, he has contributed one unit in favor of candidate one to the difference between the percentage of the vote the two candidates get, SO the scale score for this response is +1. If the voter Splits his ticket by voting for the Democratic candidate for office one and for the Republican candidate for office two, he adds two units to the difference between the percentage of the vote the two Democratic candidatesget: one unit for voting for candi— date one, and one unit for voting against candidate two. A split ticket thus contributes twice as much to the net deviation score as selective non-voting so the scale score for this reSponse is +2. The sign of the scale score gives the directionality of the response. If the net deviation score for a county is high, there must have been a high rate of individual ticket- splitting and/or selective nonvvoting. But if the score is low, it provides little information about the rate of these individual acts, because acts Of ticket-splitting and/or selective non—voting may have occurred at either low or high rates but in different directions, cancelling each other out. There would be little effect on the BSE deviation statistics. 71 When there is high net deviation voting, then, using the regression model to predict future net devia- tion voting is making inferences about individual voting behavior from aggregate statistics. The hypothesis linking education to net deviation voting, for example, states that net deviation voting increases as the median years of school completed increases. This infers that the voter who has more education will contribute more to the net deviation voting in the county by responding to the differential attractive- ness of the candidates by either Splitting his ticket or selectively non—voting more frequently than the voter who has less education. Blalock has develOped an argument to justify making inferences from one level to another—~in the present case, from the county to the individual level. He argues that ". . . in Shifting from one unit of analysis to another we are very likely to affect the manner in which outside and possible disturbing influences are operating on the dependent and independent variables under consideration."15 If we are able to-control for these nuisance variables, the correlation between the dependent variable, Y, and the independent variable, X, should increase, since the 15Hubert M. Blalock, Causal Inferences in Non— experimental Research (Chapel Hill: UniverSity of North Carolina Press, 1964), p. 98. 72 magnitude of the correlation coefficient is just an indicator of how well Y can be predicted from X, and its numerical value depends on the amount of variation in the independent variables X and the degree to which the nuisance variables have been controlled. The slope of the regression curve, byx’ Should not change, except for sampling errors.l6 Blalock argues that these variables may be controlled by grouping units of data. He analyzes several ways of grouping, three of which are applicable to the present study: 1) grouping randomly, which does not change the value of the correlation coefficient, rx , or Y the values of byx and bxy; 2) grouping so as to maximize the variation in the independent variable, X, which increases the values of rxy and bxy’ but does not change the value of byx; and 3) grouping by proximity, which increases rXy and bxy’ but does not appreciably change b . yx Grouping by proximity is a combination of grOUping randomly and grouping by the independent variable, X. The larger the groupings, the more heterogeneous and more nearly random is the grouping.17 Grouping by proximity is the method which fits 16Ibid., p. 100. l7Ibld., p. 111. 73 the county level data used in this study. By using the county as the unit of grouping, the individuals in the county are grouped both geographically and politically. The individuals in a county are in closer geographical proximity to each other than to individuals in other counties, and, Since they share the same county political unit, they are in closer political proximity to each other than to individuals in other county political units. Blalock shows that grouping by proximity makes the multiple correlation coefficient (R) unreliable as a measure of the corresponding individual characteristics, i.e., R for counties is not a good estimate of R from individual data. A high correlation coefficient for aggregate variables doesn't necessarily eXplain voting at the individual level. However, as Blalock also shows, byx has not been changed appreciably by the grouping process, so the regression slopes of the dependent variable on the independent variables can be used to test the hypotheses about individual behavior because the value of the regression Slope will be approximately the same for a population of individuals as it will be for a pOpulation of groups of individuals (counties).18 18Ibid., p. 112. 74 The solution of the ecological problem for this study thus becomes simple: the predictive power Of the model for the collective variable, net deviation voting by counties, is measured by the multiple correlation coefficient, R. Inferences about individual behavior can be made on the basis of the regression slope of the dependent variable on each independent variable, byx' High values of the multiple correlation coefficient R for these equations may mean only that the grouping of individuals by counties has obscured the effects of individual variables which vary substantially within counties but not between counties. Application of the RegressiOn Model To determine the relationship between the vari- ables for each pair of offices over all counties in a Single presidential election, a regression equation was calculated for each pair of offices in a single presi- dential election year, using the coefficient of deter— mination to show the prOportion of the variance in net deviation voting over counties that can be estimated from the concomitant variation in the independent variables in the equation. The coefficient of determination R2 is the best measure of the Fgoodness" of fit of the model for pre- dictive purposes because it specifies the per cent of the 75 variability in the dependent variable that is associated with the independent variables in the equation. It is the square of the multiple correlation coefficient. The statistic used in this study, R2, is the unbiased estimate of the coefficient of determination corrected for degrees of freedom. To determine the relationship between the vari— ables for each pair of Offices over all counties in all presidential elections, the variables in the regression equations for each pair of Offices in each of the presi— dential election years were combined, and the coefficient Of determination of the resulting regression equation, after a deletion process was used to eliminate variables which were not significant in the overall equation, was used to Show the prOportion of the variance in net devia- tion voting over all offices over time which can be estimated from the concomitant variation in the indepen— dent variables. To determine the relationship between the vari- ables for each particular pair of offices over time, for the state as a whole and for individual counties, a regression equation for each pair of offices was calculated using all the variables which were significant in the regression equations for that particular pair of Offices in each of the six presidential election years, and the deletion process was used to eliminate the variables 76 which were not Significant in the overall regression equation. The resulting coefficient of determination showed the prOportion of the variance for a particular pair of Offices over time which can be estimated from the concomitant variation in the independent variables. Summary In this chapter the regression model has been developed, the independent variables in the model have been defined, methods of obtaining their values were explained, hypotheses linking them to the dependent vari- able have been stated, and the theoretical justifications of the hypotheses have been presented. The method of regression analysis, its application in this study, and the statistics used in the model and for interpreting and testing the results of the application of the model to Montana's voting patterns have been described. Inferences made about individual behavior from aggregate statistics in this study were justified. Chapter 5 TESTING THE HYPOTHESES Introduction In this chapter the model is applied to determine the relationships between net deviation voting and the independent variables for each pair of offices over all counties in a single election. The validity of the model is tested, and correlation coefficients and regression coefficients are used to test the ten hypotheses in the model. The hypothesis that net deviation voting is inversely related to unionization is tested and is not confirmed. Three reasons for the failure of the regression model to Show a consistent relationship between net devia- tion voting and the independent variables, and a new approach is introduced for use in the next chapter. Testing_the Regression Model The regression model was first applied to deters mine the relationship between the variables for each pair of offices over all counties in a single presidential election. To do this, a regression equation was calculated for each Of the pairs of offices in each of the six 77 78 presidential election years using all the variables linked by hypotheses to the model. The computer program used was a stepwise deletion program which calculated an initial least squares equation and then tested each independent variable to determine if any variable should be deleted from the equation. The F test was used as the criterion for deletion. If the amount of variation in NDV a variable eXplained was not significant at the .05 level, it was deleted from the equation, and a new least squares equation was calculated. The deletion process continued until all variables remain— ing in the equation were significant at the .05 level. The product of the least squares equation was the coefficient of determination corrected for degrees of freedom (R2), which shows the prOportion of the variance in net deviation voting over counties that can be esti— mated from the concomitant variation in the independent variables in the equation. Table 10 shows the R2 values for each of the three pairs of offices in each election year. The R2 values are frequently very low. In seven of the sixteen equations less than twenty per cent of the variation in net deviation voting was Shown to be associ- ated with the independent variables. The median value is about .23, with the values widely distributed between .03 and .44, showing no consistent pattern. 79 Table 10. R2 Values Year R2 Sig. 1944a NDl .1866 .05 ND2 -- -— ND3 .3114 .001 1948b NDl .3721 .001 ND2 .4375 .001 ND3 .0269 .10 1952C NDl .1039 .10 ND2 .3860 .001 ND3 .2088 .01 1956d NDl .3080 .01 ND2 -- —- ND3 .4045 .001 1960e NDl .0633 .10 ND2 .3015 .001 ND3 .2577 .05 1964f NDl .0732 .10 ND2 .1817 .01 ND3 .1196 .05 Note: NDl is President and Governor ND2 is U. S. Senator and Governor ND3 is U. S. Congressman and Governor 80 Table 11 shows the variables remaining in each of the regression equations. None of the variables was significantly related to net deviation voting in all the election years. Eight of the variables (IPC, Per Cent Urban, Mobility, Turnout (l), (3) and (4), Population, and POpulation Density) were deleted from twelve or more of the sixteen regression equations; five of the variables (Country of Origin (Scandinavia), Turnout (2), Per Cent Farm Employed, Education, and Age) were deleted from more than half of the sixteen regression equations; and only three (Non—white, Country of Origin (British Islands), and Foreign Born) were deleted from less than half of the sixteen regression equations. The inconsistency of the relationship between net deviation voting and the independent variables is further shown by the values of the byx in Table 13. Every vari- able had both positive and negative values for byx’ indicating that in some elections net deviation voting was higher in counties with high value of the independent variable, and in other elections, lower. The best fitting model has a different set of Variables in each of the sixteen equations indicating that different variables are associated with net deviation voting in different pairs of offices in different election years. Hence, the model cannot be a predictive model 81 Table 11. Number of Regression Equations in Which Variables Were Retained, by Year Variable: 1944 1948 1952 1956 1960 1964 Total IPC Turnout (1) Turnout (2) Turnout (3) Turnout (4) Age Education H FJiA xlco o c: o <3 0 F4 o c: H F4 o H +4 N F414 Faro H o 63 o c><3 H‘t‘ o N +4 N FJIO H‘CD H o FJCD o tho o k4 .5 ext» a..n O\J> w POpulation POpulation Density Urban Farm Employed Mobility Country of Origin (Scandi- navia) 0 2 2 l 2 0 7 Country of Origin (British Islands) 1 3 0 0 2 2 Foreign Born 1 Non—White l 2 l l 2 2 o +4 o c: c: o c: o +4 H F4 o o be H F4 o to o LO id Hard 0 N O\LO p 82 because the best fitting equation for an election cannot be calculated until the election takes place. It must be concluded that this method of applying the regression model was not successful in showing a substantial amount of net deviation from straight party ticket voting to be clearly associated with the independent variables in the model. Testing the Hypotheses in the Model The ten hypotheses which made up the regression model stated the relationship between net deviation voting and each of the independent variables in the model. Two kinds of statistical measures were used to test the validity of the hypotheses: 1. the Pearson product—moment correlation coefficient, r; and 2. the regression coefficient, byx’ representing the units increment in the dependent variable, net deviation voting, associated with a unit increment in an independent variable with the relationship with the remaining independent variables partialed out. The Correlation Coefficients The simple correlation coefficients between net deviation voting and the independent variables were 83 inconsistent and uniformly low. Each variable had sixteen correlation coefficients, one with each net deviation score for each pair of offices (three) in each election year (six). Table 12 Shows the highest and lowest correla- tion coefficients for each of the independent variables. In no case were correlation coefficients high enough to support a conclusion that a significant and consistent correlation existed between net deviation from straight party ticket voting and any of the independent Variables. Many relationships appear to be predominantly in a direction Opposite to that hypothesized. The Regression Coefficients In the discussion of the ecological correlation problem in Chapter 4, it was concluded that byx' the slope of the regression curve of the dependent variable on each independent variable was the best measure of the hypotheses linking individual Variables to the aggregate variable, net deviation voting. In using the regression coefficient byx to Show the amount of change in the dependent variable that can be associated with a given change in the independent vari— able, it must be remembered by byx is a function not only of the strength of the relationship between the variables but also of their variances. If there is considerable variance in the independent variable relative to the 84 Table 12. Range of Coefficients of Correlation between Net Deviation Voting and the Independent Variables High V V Low IPC -.28 —.08 Turnout -.26 -.03 Age .59 .10 Education .44 -.08 Population .30 .02 Farm Employed -.47 .14 Country Of Origin (Scandinavia) —.41 .05 Country of Origin (British Islands) —.48 —.05 Foreign Born —.45 —.07 Non—white .18 .07 85 dependent variable, the value of byx will be relatively small, indicating that it takes a large change in the independent variable to produce a moderate change in the dependent variable. The numerical values of the byx's thus depend on the size of the units measured.l Table 13 shows the values Of the regression coefficients between each variable and the three different net deviation scores for each of the six presidential election years. Hypothesis 1: The amount of net deviation voting in a county increases as inter— party competition in the county increases. Both positive and negative values were present. This inconsistency of the IPC coefficients means the relationship between IPC and net deviation voting varied considerably from one kind of net deviation to another and from one election year to another. The hypothesis states a consistent relationship, so it is not sustained by these data. Hypothesis 2: Net deviation decreases as the median age of the county increases. lHubert M. Blalock, Social Statistics. Table 13. Regression 1944: IPC Turnout (2) Age Education Population POpulation Density Urban Farm Employed Mobility Scandanavian Immigrants British Immigrants Foreign Born Non-white 1948: IPC Turnout (2) Age Education Population Population Density Urban Farm Employed Mobility Scandanavian Immigrants British Immigrants Foreign Born Non—white 86 Coefficients (b ) NDl: .402 - .151 - .051 .792 - .002 ---a -—-a .006 ---a .012 -2.l79 .276 .133 .326 .048 .688 .686 .001 ---a ---a .016 ---a .207 -l.887 .808 .073 YX ND2: .309 .009 .488 .433 .001 —--a ———a - .014 ———a l .827 —l.230 .114 .001 ND3: .230 - .093 .659 - .956 - .004 ---a ---a - .163 —--a .136 —2.575 - .213 .018 - .001 - .042 .047 -2.961 — .001 —--a -—-a - .204 ---a .345 1.022 - .879 - .361 Table 13 (continued) 1952: IPC Turnout (2) Age Education POpulation POpulation Density Urban Farm Employed Mobility Scandanavian Immigrants British Immigrants Foreign Born Non-white 1956: IPC Turnout (2) Age Education Population Population Density Urban Farm Employed Mobility Scandanavian Immigrants British Immigrants Foreign Born Non-white 87 NDl: .127 .148 .602 .909 .001 ---a -—-a .015 CIA—Ha .066 -l.439 .223 .208 .208 .005 .291 1.179 .001 —--a —--a .089 —--a .507 -3.l47 .363 .065 ND2: .054 .105 .733 .457 .001 ---a -—-a .079 ---a .012 .217 .005 .187 ND3: .132 .404 .769 .712 .001 ---a ---a .130 —--a .457 —1.971 .309 .108 .962 .029 .257 .793 .001 ---a ---a .063 ---a -1.045 .698 .093 Table 13 (continued) 1960: IPC Turnout (2) Age Education POpulation Population Density Urban Farm Employed Mobility Scandanavian Immigrants British Immigrants Foreign Born Non—white 1964: IPC Turnout (2) Age Education POpulation Population Density Urban Farm Employed Mobility Scandanavian Immigrants British Immigrants Foreign Born Non—white 88 NDl: .431 .043 .033 .223 .001 ---a —--a .004 -—-a .124 .601 .332 .025 .052 .108 .129 .613 .001 ---a --—a .085 --—a .512 .697 .095 .106 aDeleted from regression equation ND2: - .471 - .051 .053 — .312 .001 ---a ---a - .030 ——-a -l.167 -4.351 '- 0209 .108 .127 .041 .304 .777 .001 ---a --—a .041 ---a .298 —3.995 .257 .031 ND3: - .109 .069 .005 - .343 - .001 ---a ---a ~ .028 ---a — .313 —2.855 - .322 .044 .139 .091 .182 .310 .001 ---a ---a .025 ---a - .412 -1.083 - .651 — .022 on first run. 89 Both positive and negative values are present. The inconsistency of the relationship between this vari- able and net deviation does not support the hypothesis. Hypothesis 2 is not sustained by these data. Hypothesis 3: Net deviation voting increases as the median years of school completed in the county increases. Both positive and negative values are present. There were a few coefficients high enough to support the hypothesis, but the number of low or negative values present makes the overall relationship between this vari— able and net deviation voting too inconsistent to support the hypothesis. Hypothesis 3 is not sustained by these data. Hypothesis 4: Net deviation voting is more frequent in rural counties, less frequent in more urban counties. Two of the alternative measures of the rural nature of the counties, pOpulation density and per cent urban, were deleted from twelve of the sixteen regression equations on the first run, indicating that they were not significantly related to net deviation voting. The coefficients for the population variable were inconsistent with both positive and negative values present. 90 The same is true of the farm employed variable. None of the four measures of the rural nature of the county supported the hypothesis. Hypothesis 4 is not sustained by these data. Hypothesis 5: Net deviation voting is higher in counties with high mobility, lower in counties with low mobility. This variable was deleted from fourteen of the sixteen regression equations on the first run, indicating that there was no Significant relationship between mobility and net deviation voting. Hypothesis 5 is not sustained by these data. Hypothesis 6: Counties with high concentrations of Scandinavian immigrants have more net deviation voting than counties with few or no Scandinavian immigrants. Both positive and negative values are present. The consistent high positive coefficients required by this hypothesis are not present. Hypothesis 6 is not sustained by these data. Hypothesis 7: Net deviation voting decreases as the per cent of immigrants from the 91 British Islands in the county increases. There were twelve equations in which the high negative coefficients required for this hypothesis were present making this the only variable which shows any consistent relationship to net deviation voting. This variable is multidimensional, however, and, as noted above, is associated with the factors of high voter turn- out, urban residence, ease Of assimilation into a politi- cal party culture, and membership in labor unions which encourage party loyalty. These factors tend to orient the immigrants from Great Britain toward the party—~in this case, the Democratic party—~rather than toward the candidate. However, it is very difficult to interpret this modest consistency, embedded as it is in equations with diverse other variables with inconsistent associations with net deviation voting. However, the two low negative coefficients and two high positive coefficients made this variable too inconsistent to support the hypothesis. Hypothesis 7 is not sustained by these data. Hypothesis 8: Net deviation voting increases as the per cent of foreign born in the county increases. 92 The coefficients for this variable were consis- tently low and had both positive and negative values. Hypothesis 8 is not sustained by these data. Hypothesis 9: Net deviation voting increases as the per cent of non—whites in the county increases. The coefficients for this variable were consis— tently low with both positive and negative values. Hypothesis 9 is not sustained by these data. The Union Influence Hypgthesis Net deviation voting is inversely related to unionization. The political demands made upon the political system by union voters differ from those made by voters in non-union occupations. Their demands are also similar at the local, state, or federal levels of government, and they do not usually demand conservative legislation at one level of government while demanding liberal legisla— tion at a different level. Non—union voters such as businessmen, farmers, and ranchers Often do make different demands on different levels of government. Unlike union voters, who behave as if their demands can be satisfied best by one party at all levels of government, non—union voters may have to Split 93 their tickets in order to get what they want from different levels of government. This hypothesis is a test of Montana's adherence to the general pattern of union voters predominantly voting straight party tickets in support of the Democratic party, and farmers Splitting their tickets most often of any occupation group. This pattern has been found by Lipset and Campbell2 in national surveys and are individ— ual behavior patterns. They should show up in aggregate data analysis of net deviation voting. Montana is classified as a strong union state. Alexander Heard lists Montana as one of the seventeen states in which "the financial resources and potential voting strength of unions are concentrated . . . where the percentage of union members among persons of voting age equalled or exceeded the national average and in which the percentage of non—agricultural employees was thirty or more."3 The occupation groups most prevalent in Montana and examples of some of the demands they make on the different levels of government are as follows: 2Seymour M. Lipset, Political Man (New York: Doubleday and Company, Inc., 1959), p. 233; Campbell, gr 21" The American Voter, p. 213. 3Alexander Heard, The Costs of Democracy (Chapel Hill: University of North Carolina Press, 1960), p. 175. 94 Strong Union Occupations: the mining and lumber unions are the largest and make the most demands on govern- ment, and their demands are satisfied by the Democratic party at all levels of government. Protection of the right of collective bargaining, the right to strike, freedom from harrassment by management, and mine safety laws are supported more strongly by the Democrats in the state legislature. The Democratic party's support of these and other union demands at the federal level of government is well-known. Therefore, we eXpect union members to vote straight party tickets, and net deviation voting should be lower in counties in which the dominant occupa— tions are highly unionized, and higher in counties which have few or no unionized occupations. Non-union occupations: the dominant non—union occupations are farming, ranching, and business. The demands made on the political system by farmers and ranch— ers are so Similar that they may be considered together. Many of the farmers, in fact, do some ranching in addition to their farming, while many of the ranchers so some farming in conjunction with their ranching. Table 14 shows the counties grouped by dominant occupations and their rank in net deviation voting. The variation in rank order of net deviation is too great, and the range from third to fiftieth is too large to indicate any Significant correlation between the kind of 95 Table 14. Counties in which Dominant Occupations are Highly Unionized, Ranked by Net Deviation Voting County: Major Economic Interest: Net Deviation Rank: Missoula Manufacturing 3 Sanders Lumber 18 Granite Mining 20 Mineral Lumber 30 Lincoln Lumber 40 Deer Lodge Mining 46 Musselshell Mining 47 Cascade Manufacturing 48 Silver Bow Mining 49 Flathead Lumber 50 96 dominant occupation in the county and the amount of net deviation voting which occurs. The hypothesis is not con— firmed. A New Approach If net deviation voting cannot be shown to be consistently associated with the political, socio—economic, and demographic variables included in the regression model tested in this study, what does explain the high amount of net deviation voting which occurs? A new approach is needed, and this approach must start with the reasons the regression model did not work. The first reason for the failure of the regression model is the fact that the counties are highly variable from one election to the next in amount of net deviation voting, while the independent variables are all highly stable, some actually constant because they are census figures that hold for two elections. The amount and consistency of net deviation voting varies considerably in the counties, not only between counties but within counties over time and between different pairs of Offices. A county may rank sixth in the amount of net deviation voting between the Offices of President and Governor and fifty-fifth in the amount of net deviation voting between the offices of Congressman and Governor in one election year, and rank thirty—seventh on both pairs of offices in 97 another election year, as Rosebud County did in 1944 and 1956. Or, a county may rank third in the amount Of net deviation voting between the Offices of President and Governor in one election year and rank fiftieth for the same offices in the next election year, as Daniels County did in 1948 and 1952. No county was consistent in its ranking for all pairs of offices in a single election year or for a Single pair of offices for all years. The significance of these extreme variations in net deviation voting is that they sharply define a problem in building a regression model of net deviation Voting with variables whose values are obtained from aggregate statistics: if there are extreme variations in the dependent variable between the aggregate units, the independent variables in the model must show enough varia— tion between the aggregate units to explain the variation in the dependent variable. If a county has a low amount of net deviation voting between the offices of President and Governor and a high amount of net deviation voting between the offices of U. S. Senator and Governor and the same amount of inter-party competition, it is difficult to determine whether the inter-party competition is related to high or low amounts of the dependent variable. Inter—party competition is a stable variable, with one value for all elections in this study. 98 Given the extreme variation in net deviation voting and the lack of variation in the independent vari- ables in the present study, it is not surprising that the regression model was not successful in showing a consistent amount of net deviation voting to be associated with variables whose values were obtained from relatively stable aggregate statistics. The second reason for the failure of the regres- Sion model is that there are variable conditions in each election which the counties respond to differently from one election to the next, thus producing the inconsistent relationships between net deviation voting and the social, economic and political characteristics of the counties. The differential attractiveness between candidates is the result of factors other than party membership which change from election to election and which cause gains or losses in the amount of electoral support the candidates for a particular office receive. 'An analysis of each of the elections in this study Shows that the most common of these factors are incumbency, the record of the candidate, the candidates' relationship to the party leaders at all levels of party organization, the candidates' appeal to normal non-voters and to marginal members of the Opposition party, and the relative amount of Opposition each receives from the other party. 99 The third reason for the failure of the regression model is the possibility of the relationships with differ- ential attractiveness overwhelming any ticket-splitting relationships which might exist. Ticket—Splitting may occur and sometimes be associated with the independent variables, but at the aggregate level, these effects are overwhelmed by the shifting responses of the counties to these features. The effect of incumbency, for example, shifts from one election to another. Differential attractiveness increases as the number Of factors which separate the candidates of the same party increases. There will be more differential attractiveness between an incumbent candidate and a candidate of the same party who is not the incumbent than there will be between two incumbents from the same party, other factors being equal. There will be less differential attractiveness between an incum- bent candidate who haS very little opposition and an incumbent candidate from the same party who has consider— able Opposition. In order to avoid these shortcomings, a case study approach is used in the next chapter to show that the patterns of changing partisan control of state and national political Offices which result in high rates of net devia— tion voting correspond to the historical patterns of changing corporate partisan activity in Montana. In this 100 case study, the amount of net deviation voting in Montana is compared with that in other states, and the changing patterns of net deviation voting in different periods of Montana history are compared to different corporate strategies in these periods. Summary In this chapter the regression model was applied to determine the relationships between net deviation voting and the independent variables for each pair of offices over all counties in a single presidential elec— tion. Correlation coefficients and regression coefficients were used to test the ten hypotheses in the model. It was concluded that the model was not a valid instrument for determining the association between net deviation voting and the independent variables. None of the hypotheses was sustained by the data. The hypothesis that net deviation voting is inversely related to unionization was tested and rejected. Three reasons were given for the failure of the regression model to Show a consistent relationship between net deviation voting and the inde— pendent Variables, and a new approach was introduced for use in the next chapter. Chapter 6 CORPORATE POLITICS IN MONTANA Introduction In this chapter, historical patterns of corporate political activity are identified and compared with shifts over time in the extent to which control over state and national offices, respectively, has been split between the two major political parties. It is concluded that the ways in which Montana corporations have used their power to influence state politics are closely associated with changing patterns of party control over elections for state and national Offices in Montana. A New Approach The analysis of this chapter has a somewhat different focus than that of earlier chapters. Rather than analyzing variation in the net deviation statistic, which is derived from vote percentages regardless of win-loss outcome, the approach taken in this chapter examines the partisan outcomes of elections -- the way in which party victories are distributed over state and national offices. And rather than attempting to explain 101 102 intercounty differences, the analysis of this chapter assesses shifts in party control over an eighty-year period of time, for the state as a whole. A principal motivation for studying net deviation voting in the first place was its connection to the strength of party identification, particularly as this variable determines homogeneous party control of elections for state and national offices within a state. Net deviation voting is an appropriate aggregate statistic of individual behavior in an election, tapping the extent to which a body of voters splits its votes for state and national level offices between the two political parties. It may be, however, that much net deviation voting can be explained more economically not as the result of the characteristics of individual voters and their interaction with elements of unique campaigns, but rather as a byprOduct of the efforts of long-standing and powerful political and economic forces --- in this case, a few dominant industrial corporations --- to control what offices they could at the level of government most salient to them. The corporations, obviously, are not interested in achieving high levels of net deviation voting. Their goal is to win elections, to control at least those offices most salient to their interests. In the case of Montana, it is state Offices that the corporations are most concerned to control, for reasons which will 103 be apparent. The necessity for the corporations to concentrate on state level offices is the result of the competitive, almost evenly split partisan politics in the state. If a state's politics are one-party dominant against a strong corporate interest, the corporate interest may not be effective and may withdraw from political activity. On the other hand, if the state's politics are one-party dominant in favor of the corporate interest, there may be no need to settle for control of state offices when national offices may be controlled as well. Only where the electoral politics of the state are highly competitive, or where, as in Montana, the party Opposed to the corporate interests has a slight advantage, will the corporate interests find it necessary and fruitful to con- centrate on control of state level offices. When the corporate interests have the amount of both economic and political power that they have in Montana, it rarely makes much difference which party is in power at the state level because, as will be shown below, both Democratic and Republican parties are controllable by the corporate interests. The corporations prefer the Repub- lican party, however, because it is more sympathetic to their interests. If the state leans toward the Democratic party, as Montana now does in elections for national offices, corporations may nevertheless be successful in achieving 104 Republican control of the state in many elections. In this event, party control of state offices and of elections for national offices in the state will frequently be split between the parties as a byproduct of corporate political influence. That is the argument of this chapter. The extent to which party control of state offices and national elections is split between the parties may be measured directly. The important federal elections are for President, Senator, and two Congressmen. The most important points of control over state government are the governor- ship and the majorities of the two houses of the state legislature. Now consider each possible pair of offices (or majorities) made up of one federal Office and one state office. In a presidential election year, for example, there will be twelve such pairs: four federal offices paired with each of three state offices or majorities. For each of these pairs, either one party will carry both offices, or control will be split between the parties. The proportion split of all possible such pairs is here termed "Split control". It measures the extent to which control of state offices or majorities and elections to federal office in the state is split between the parties. The measure may be obtained for a single election or, more reliably, accumulated over a 105 period of time. Incidence of Split Control over State and National Offices AS documented in Tables 1-3 and Figures 1-4 of Chapter 2, Montana has a high incidence of split control relative to the other states as a group, especially for the period Since 1940. As is Shown below, however, Montana has not always voted so as to split control of state and national elections between the parties. In some periods of her history, the incidence of split control has been much lower, and sub- stantial shifts in the rate of split control appear to arise as byproducts of shifts in corporate support of the parties. In order to establish the connection between corporate strategies and the incidence of split control, it is necessary to give detailed attention to the political activities of Montana's major economic structures, the politically powerful mining and power companies, over the state's history. A major portion of this chapter is given over to an account of their political efforts and of the evidence of their substantial success in controlling the state offices most salient to their political interests. The last section of the chapter lays out the evidence for the relationship between corporate strategy and the Split of party control between the elections for state and federal office. 106 Corporate Influence in Montana Politics The corporations which exert the most influence on Montana politics are the Anaconda Copper Company and the Montana Power Company. The railroads, particularly the Northern Pacific, have occasionally exerted some influence, but usually in collaboration with the Anaconda Company. The extent to which the Anaconda Company has influenced Montana politics has become something of a classic example of a single interest dominating a state. Zeigler uses Montana to illustrate the pattern of a "single dominant interest" in his classification of interest group activity into four distinct patterns.2 In discussing this example, he makes these remarks: In the turbulent frontier atmosphere of the West, Anaconda played the classic role of economic royalist. Its forays into the electoral process were frequent and extravagant. Indeed, politicians such as Burton K. Wheeler became legendary as courageous foes of the Company in much the same manner that Senator Borah built a political career upon the crusade against the trusts. Indeed, much of Montana's political history seems to reflect a basic division of the population: either you were for the Company, or (in the case of the unions and Farmer's Union) against it. . . . The conflict in the state seemed to be structured around the Company rather than the political parties.3 2Harmon Zeigler, "Interest Groups in the States," in Herbert Jacob and Kenneth N. Vines, eds., Politics in the American States (Boston: Little, Brown and Company, 1965), p. 117. 31bid., pp. 119-20. 107 He goes on to note that the Company's dominance is not complete because its candidates have been defeated from time to time.4 Those candidates who were supported by the company but did not win election usually were, like Senator Wheeler, candidates for federal rather than state office. Senator Wheeler, in fact, was strongly supported by the company in the primary election of 1946 but was defeated.5 The office of U. S. Senator had become a focus for anti—company voting, thus polarizing the voters in both parties on the company issue and resulting in a high level of net deviation voting between the offices of U. 8. Senator and Governor.6 Senator Wheeler also recognized the role of the company in Montana politics. In his autobiography, his opposition to the company is the major focus of his discussion of the early years of his career. In the last legislature before pOpular election of U. S. Senators went into effect, for example, he writes: "In the 1911— 1912 legislature the Democrats controlled the House, the Republicans controlled the Senate, and the Company 4Ibid., p. 120. 5K. Ross Toole, Montana: An Uncommon Land (Norman, Okla.: University of Oklahoma Press, 1959), p. 224. 6 Ibid., p. 226. 108 controlled the leaders of both."7 Professor Thomas Payne, describing the extent of Anaconda's influence, writes: Its strength rests not only in its wealth and resources, but also in its elaborate network of relationships with key citizens, banks, legal firms, and business organizations throughout the state. Rare is that unit of local government—-county, city, or school district—~that does not have among its Official family an associate, in some capacity, of the Anaconda Company. And Anaconda men are invariably among the leaders of the power structure in every large community in the state.8 The Anaconda Company and the Montana Power Company, although separate corporations on paper, function politi- cally as a single unit. In addition to sharing the traditional economic interests and concern with govern- mental regulation Of big business corporations, their mutuality of interests also includes a common corporate origin, both having been part of the same holding company, the Amalgamated Copper Mining Company; partial common ownership, Montana Power Company having begun its Opera- tions as a subsidiary of the Anaconda Company, with the same financial institutions controlling large blocks of 7Burton K. Wheeler and Paul Healy, Yankee From the West (Garden City, N. Y.: Doubleday and Company, I962), p. 84.' 8Thomas Payne, "Under the Copper Dome: Politics in Montana," in Frank H. Jonas, ed., Western Politics (Salt Lake City, Utah: University of Utah Press, 1961), pp. 197-8. 109 stock in each company;9 and mutual concern for each other's economic interests resulting from an interdependent economic relationship in which the Montana Power Company provides all the gas and electric power used by the Anaconda Company and the Anaconda Company uses over seventy per cent of the gas and electric power sold by the Montana Power Company.10 This interdependent economic relationship between the mining and power interests of the state began in the 1890's when, as Montana Power Company reports: Without the mines of Butte, the Great Falls Water Power and Townsite Company (a subsidiary of the Anaconda Company until it was consolidated with other subsidiaries in 1912 to form the Montana Power Company) might have starved to death before it found a cash customer. Without cheap and abundant water power readily available, the enormous COpper deposits of the Butte hill might have remained indefinitely merely a potential, and not a develOped, natural resource—- a resource that to date has yielded more than tyg and one—half billion dollars of mineral wealth. An additional relationship is described: And now the Great Falls of the Missouri, together with a dozen other plants located along this and other streams, are producing the current 9Ferdinand Lundberg, America's Sixty Families (New York: The Citadel Press, 1937), p. 241. 10Richard G. Sheridan, Politics is Thicker Than Water (Bozeman, Montana: Montana State Extension Service, I968T, p. 243. 11 Ibid., p. 242. 110 that mines the copper, that makes the wire, that creates and distributes additional electrical energy, thus completing a gesture of productive reciprocity between the two foremost industries of the state.1 Another reciprocal relationship involving these "two foremost industries" resulted in an Interstate Commerce Commission condemnation in 1928. The Chicago, Milwaukee and Saint Paul Railroad went bankrupt after the eXpenditure of $180 million for electrification of its western lines. The electrification program, started to provide better transportation for the products of the Butte mines, got its copper supplies from the Anaconda Copper Mining Company and its electrical power from the Montana Power Company, then a subsidiary of the Anaconda Company. John D. Ryan, President of the Anaconda Com— pany, sat on the board of directors of the Milwaukee railroad. William D. Rockefeller, a director of the Anaconda Company, was a major stockholder in the Milwaukee railroad. The electric power contracts of the railroad with the Montana Power Company were condemned by the Interstate Commerce Commission because they obligated the railroad to pay for power it did not use.13 Among the Specific political interests these two corporations share are: government regulation of lzIbid., p. 242. 13 pp. 241-2. Ferdinand Lundberg, America's Sixty Families, lll corporate activity; property tax assessments; securities regulation; and pollution control. Each corporation has interests peculiar to its activity. The Anaconda Company, for example, is interested in mine safety laws, laws regulating claim disputes, and wage and hour laws for miners. The Montana Power Company is interested in water rights legislation, laws regulating the methods of electric and gas power transmission, the regulation of Rural Electrification Administration COOperatives, and govern— ment Sponsored power projects. While these interests are of more concern to one company than the other, each sup— ports the other's position on these issues. The corporations attempt to disguise their unified political front. Sensitive to public concern over cor- porate influence in politics, the corporations wage public "battles" to prove that they have conflicting interests and do not work together to control the economy of the state. An example of this tactic is the "conflict" over the rates which the Montana Power Company can charge its customers. These rates are fixed by the Public Service Commission. Increases are constantly sought by the Power Company while the Anaconda Company, a consumer of power, Opposes these increases. This "battle" is waged as Often through the newspapers of the state as before the Public Service Commission. The conflict is not as real as it may seem, however. The strength of the Anaconda Company's 112 opposition is determined by the relative impact an increase will have on the two corporations. If the existing economic conditions, such as the current market price of copper, eXpansion or reduction of mining activities, etc., are such that a rate increase will result in more profit for the Montana Power Company than the reduction in profits it will cost the Anaconda Company, the Anaconda Company will offer only token resistance. If the rate increase will result in less profit for the Montana Power Company than the reduction in profits it will cost the Anaconda Company, the Anaconda Company will strongly Oppose the increase and, invariably, prevail. Since both corporations are controlled by the same financial insti— tutions, such a reciprocity is easily arranged. Also, since the Public Service Commission is strongly influenced by both corporations, it does not often hinder the corporations' plans. In most cases the Montana Power Company gets its increase and a kickback, in the form of discounts for volume consumption of power, is given the Anaconda Company. The smaller users of power, including homeowners, thus are the ones who pay the increased rates. Publicly, however, the companies appear to be at war with each other. 113 The Corporate Influence on Split Control Montana votes primarily Democratic. In all state- wide elections between 1890-1968, the Democratic party has won forty-four while losing twenty-three to the Repub- licans. Included in the Democratic victories are ten presidential elections, twenty-one U. S. Senate elections, and thirteen elections for Governor. The Republican victories include nine presidential elections, six U. S. Senate elections, and eight elections for Governor. In elections for the U. 5. House of Representatives for the same period, the Democrats have won forty-two while losing twenty-eight to the Republicans. The elections for seats in the state legislature have been won more often by Republicans, although the margin has been small. In thirty-nine elections to the state house of representatives, the Republicans have won control of twenty sessions, with the Democrats winning nineteen. In thirty-nine elections to the state senate, the Republicans have won control of twenty-three sessions, with the Democrats winning sixteen. Montana corporations support candidates, regardless of party, who will protect corporation interests. Policies and regulations of the state government threaten corpora- tion interests more than do policies and regulations Of the federal government, at least on day-to-day Operations. 114 Corporate political actions are concerned most with pro- tecting their interests from adverse action by the state. National elections are influenced by national issues, over which the corporations have little influence and which pose few threats to corporation interests. As a general rule, national issues are not injected into contests for state offices. Montana is a good weathervane of national voting trends in its support of national candidates, but this voting behavior seems to have little effect on the outcome of state elections.l4 State elections are more subject to their influ- ence than are national elections: Powerful economic groups often have their greatest impact upon the governments of the states in which their principal resources are based. Thus the Anaconda COpper Company was almost synonymous with state government in Montana until the federal government during the New Deal became a formidable competitor.15 There are practical reasons for the corporations to stay out of federal elections, eSpecially those for Congressmen. The most success they could hope for would be the election of two Senators and two Congressmen who, in a Congress of 535 members, would have little influence 14Thomas Payne, "Under the Copper Dome: Politics in Montana," p. 204. 15Joseph A. Schlesinger, "The Politics of the Executive," in Herbert Jacob and Kenneth N. Vines, eds., Politics in the American States (Boston: Little, Brown and Company, 1965), p. 213. 115 over national legislation. At the state level, they could elect a Governor and conceivably every member of the legislature. A majority of either house, with or without the Governor, would be sufficient to block any undesirable legislation. Another practical reason for not expending resources on federal elections is to avoid out—of—state competition. Within the state, the only organized competi- tion comes from other state-wide organizations, none of which is as powerful as either of the corporations. A U. S. Senator or Congressman can acquire support from organizations who have nationwide interests and want his vote on federal legislation which they favor, and who will be willing to provide financial and organizational support for his re—election. This support can be of sufficient magnitude as to discourage the corporations from contesting his re-election, especially if he can do them little harm or little good. The support can be strong enough to discourage the corporations even if the Senator or Congress- man can do them harm, as happened in Senator Lee Metcalf's 1966 re-election campaign. Senator Metcalf, in his book Overcharge, had attacked the investor-owned utilities of the nation, paying particular attention to the Montana Power Company. The Montana Power Company, after beginning an effort to Oppose Senator Metcalf, decided against such an effort When out—of—state organizations pledged over 116 $200,000 to support his-campaign.l6 Conservative politicians, usually Republicans in Montana,17 are more sympathetic to the interests of the corporations than are liberal politicians. The corpora— tions want as little attention from the state government as possible, i. e., conservative policies, since the state threatens regulatory attention. A majority of the state's voters are Democratic, so the corporate interests concentrate their efforts in such a way that their interests can be protected in state politics without the necessity of getting a majority of the votes. One way this has been done in the past was by supporting Republican candidates from farming and ranching counties to the state senate.18 Until the state legislative districts were reapportioned in 1965, a majority of the state senate could be elected from counties containing only sixteen per cent of the state's population. As a result of corporation support, the Republican party, already strong in the eastern counties because of the support they got from the large—scale farming and 6Sources on Senator Lee Metcalf's staff. l7Thomas Payne, "Under the Copper Dome: Politics in Montana," p. 195. 8Jerre C. Murphy, The Comical History of Montana (San Diego, Calif.: E. L. Scofield, 1912), p. I45. 117 ranching interests, managed to control the state senate from 1894 until 1956, with three exceptions: between 1898 and 1904, when the issue of free silver caused the "silver Republicans" to fuse with the Democrats and elect all Democratic candidates except one U. S. Congressman, control the state senate for three sessions, and control the state house of representatives for two sessions;19 in 1912, when Wilson carried a majority of Democrats into the state senate for one term; and between 1936 and 1940, when F. D. Roosevelt's popularity finally reached the state senate and a majority of Democrats was elected for two sessions. Control of the state senate has been in the hands of the Democrats since 1956, but the Republicans have maintained a veto on state legislation by electing a Republican Governor during that period. In the seventy—five years of Montana's state— hood, from 1889 to 1964, the Republicans have controlled at least one house of the state legislature or the Governor for a total of sixty—five years.20 19Merrill G. Burlingame, Political Party Activities in Mgntang, unpublished manuscript, pp. 22—4. 20In the most recent elections, 1964—1970, they have controlled the House in each election, making their total years of veto power over state legislation seventy- one of eighty-one years. 118 Montana voters respond well to the political needs of the state's largest non-government employer. They generally select liberal Democratic politicians to repre- sent them in Congress while preferring conservative Republican politicians in the state legislature. "This pattern of marching both ways at once is an illustration of the proposition that differing factors, indeed . . . differing constitutiences are involved in the selection of national as contrasted with state representation."21 This ambivalence is not irrational. The voters rather than opposing corporation interests may understand the economic and political advantages of being ambivalent. As Professor Thomas Payne notes: The same configuration of interests which apparently receives benefit from political liberalism in national politics may sense the advantages of political conservatism in state affairs. In a state with sparse population and an economy based mainly on the production of raw materials rather than finished products, a policy of restricting state expenditure and seeking supplementary spending from Washington may well coincide with the interests of a substantial majority of the population.22 The success of the corporations in maintaining this "configuration of interests" by persuading the voters 21Thomas Payne, "Under the Copper Dome: Politics in Montana," p. 199. 221bid., p. 205. 119 that it exists, through both public relations and political activity, determines the amount of split control that occurs. The more successful the corpo- rations, the more control will be split as a result of the voters supporting different parties at each level of government. Five Periods of Corporate PartiSan Activity The partisan activity of the Anaconda Company, later joined by its subsidiary, the Montana Power Company, can be divided into five distinct periods: 1. Copper Goes Democratic. The period between 1890-1900, characterized by strong corporation support of the Democratic party at all levels of government and very little split control. 2. Copper Goes Republican. The period between 1902-1906, characterized by strong corporation support of the Republican party at all levels of government and increased but still relatively low incidence of split control. 3. Corporate Diversification Leads to Political Diversification. The period between 1908-1930, charac- terized by corporate support and an effort to control both parties at all levels of government, with a subsequent rise in split control. 4. Riding With the Tide--Corporate Support of the Democratic party. The period between 1932-1938, 120 characterized by strong corporation support of the Democratic party at all levels of government with a decline in split control. 5. Modern Corporation Politics: The Path of Least Resistance--Support of the Current Majority at Each Level of Government. The period between 1940-1968, characterized by strong corporation support of both parties, but more support of the Republican party at the state level and more support of the Democratic party at the national level of government and a sharp rise in the amount of split control. Period One, 1890-1900--Copper Goes Democratic. Copper interests in this period included not only the Anaconda Company, formed late in the period by its owner, Marcus Daly, but the United Verde Copper Company, owned by W. A. Clark. Both these men supported the Democratic party, the only really virile political force in the Territory. It was strong not only because of the southern element in the population, but also because the industrial leaders had chosen the strongest vehicle available for their purposes.23 The political battles of the well-publicized "War of the Copper Kings" took place within the structure of 23K. Ross Toole, An Uncommon Land, pp. 173-4. 121 the Democratic party. The first protagonists in this eighteen-year battle were Marcus Daly and William A. Clark, the two leading copper owners. These two men were two of the "Big Four" of the Democratic party until 1888, when Daly and Clark had a falling out as Daly quietly supported Thomas H. Carter against Clark in a race for territorial delegate to Congress.24 This started a twelve—year feud which affected the balance of the political structure for more than a decade. Among other Spectacular events, it included a state-wide battle over the selection of a site for the state capital, won by Clark after an election in which both sides engaged in large-scale vote-buying. When Clark attempted to win election to a seat in the U. S. Senate by buying votes in the state legislature, he was blocked by Daly when he convinced the Committee on Privileges and Elections of the Senate not to seat Clark because of the Open bribery employed by Clark's supporters. A second attempt by Clark to get to the U. S. Senate succeeded when he supported enough candidates to the state legislature to send him there. The first period«of corporation partisan activity ended when Standard Oil bought the Anaconda Company from Daly in 1899 and incorporated it at Trenton, N. J. as 24Ibid., p. 178. 122 the Amalgamated COpper Company. Its officers were Henry H. Rogers, William G. Rockefeller, and Albert C. Burrage, all of Standard Oil, and James Stillman, President of the New York City Bank. Standard Oil had begun preparations for the purchase by persuading the Montana legislature, through the Anaconda lobby headed by Anaconda attorney E. D. Matts, to pass House Bill Number 132, which provided for the legality of stock transfers without minority consent.25 Period Two, 1902-1906. COpper Goes Republican. This period began with the alliance of W. A. Clark with F. Augustus Heinze, a young mining engineer who had entered the arena of copper politics in 1889. Heinze was described even by his enemies as having charm, wit, courage, and a peculiar power over the miners of Butte, while at the same time they asserted his unmitigated 26 The Clark-Heinze alliance to Oppose Amal— rascality. gamated was a short—lived one, however. After Clark succeeded in getting elected to the U. S. Senate by the legislature in 1900, with substantial help from Heinze, he sold most Of his prOperty in Butte to Amalgamated. This left Heinze, with little money and only one good 251bid., pp. 164—6. 26Ibid., p. 197. 123 mine, to do battle with Amalgamated. For six years he stood alone in front of the great juggernaut of Standard Oil and stOpped them.27 This third and final episode in the COpper wars hastened the develOpment of two-party politics in the state. The Republican party became more powerful, encouraged and supported by Amalgamated partly because Heinze had strong support in the Democratic party and partly because the eastern owners Of Amalgamated, headed by Standard Oil, were influential in the national Repub- lican party and wanted to increase their influence by delivering the state to the party in Presidential elec- tions and by adding Republicans to Congress, especially to the U. 5. Senate. During the years of the Heinze— Amalgamated fight, 1901-1907, the Republicans managed to sweep almost all elected offices, both state and federal, carrying the state for Taft in 1904; winning the races for the U. 8. House of Representatives in 1902, 1904, 1906, and 1908; electing a U. S. Senator in 1906; winning control of the state house of representatives in 1902, then winning control of both houses of the legislature for the first time in 1904, and repeating this victory in 1906. The only state—wide Office they did not win in this period 27Ibid., p. 197. 124 was the office of Governor, won by the incumbent Democrat Joseph K. Toole in 1904. The Heinze-Amalgamated fight was waged primarily in the political arena, each side using the tactics most appropriate to its political resources. Heinze, with little money and less influence on state politics, fought his major battles in the two state district courts in Butte where his immense popularity with the miners enabled him to select judges, elected by pOpular vote, sympathetic to his point Of View. That point of view, eXpressed in judicial terms, was usually that when Heinze took a case to the district court, he was right, regard— less Of the evidence or points of law. He was successful in obtaining injunctions against Amalgamated by arguing under the "apex law" that the veins of copper in the Amalgamated mines actually apexed on his claim and were rightfully his. While the mines were thus closed down, he would tunnel into them from his mine and take out as much ore as he could until a higher court dismissed the injunctions. In one case alone, he took out almost one million dollars worth of ore.28 Both sides in the fight began to broaden the base of their political activity. Amalgamated sought to gain 28C. B. Glasscock, The War of the COpper Kings (New York: Grosset and Dunlap, 1935), p. 276. 125 influence with state and county officials, particularly with state legislators, and with judges, sheriffs, county commissioners, and assessors. It developed a state-wide political network which Heinze could not compete with because of limited resources. Amalgamated at this time began to put together a state-wide newspaper network.29 In 1901 Heinze made a tactical error. One Of his employees who owned one—fifteenth of one per cent Of the stock of the Boston and Montana mine, owned by Amalgamated, obtained an injunction restraining the Boston and Montana from paying dividends on ninety-eight per cent of its stock under the rule of common law that one corporation could not own, vote or control shares of stock in another corporation, as Amalgamated did. Over a period Of two years over two million five hundred thousand dollars in dividends piled up in the Boston and Montana treasury. When the injunction was made permanent in 1903, Amalgamated ordered all mines, mills and smelters of their subsidiary companies shut down at once. Twenty thousand men, constituting approximately four—fifths Of the wage earners in the state, were thrown out of work, paralyzing the economy of the state.30 Amalgamated then demanded a 29K. Ross Toole, An Uncommon Land, pp. 205—6. 30 pp. 275-7. C. B. Glasscock, The War of the COpper Kings, 126 special session of the legislature be called to pass a "Fair Trials Bill" allowing a party to a civil suit to get a change of venue if it considered the judge corrupt or prejudiced. The mines would remain closed until such a law was passed. The Special session was called and Amalgamated's bill was quickly passed.31 What the corporation wanted was not in itself a Special interest bill; it was a necessary step in the develOpment Of a fair judicial system. Its actions to get a fair bill, however, were brutal. In the words of one historian: But the fact remained that a combination of cor- porations had, through the threatened starvation of one hundred thousand persons, forced the unwilling governor of a sovereign state to call a legislature to enact laws for its benefit . . . A combination Of capital, for the first time in the history of the United States, had openly dictated to a state in which it had invested money and grown vastly rich, and under the laws of which it was Operating.32 With the passage of this bill, Heinze's political power was taken from him. Amalgamated had won the war. The copper trust controlled ninety per cent of the press in the state; it controlled a majority of the legislature; and it had proved its power to control a majority of the 31K. Ross Toole, An Uncommon Land, p. 208; C. B. Glasscock, The War Of the COpper Kings, p. 278. 32 p. 288. C. B. Glasscock, The War of the COpper Kings, 127 voters——if it had to starve them into submission.33 By 1904 it was evident to party leaders that success was improbable for any ambitious man without the consent, if not the support, of the corporation management.34 Heinze stayed in Montana for three years, but the "War of the COpper Kings" was over. In 1906 Amalgamated paid him $10,000,000 for his holdings, and lawsuits tying up over $200,000,000 of Amalgamated mines were dismissed. His prOperties were worth much less, so the price repre- sented his nuisance value.35 Heinze then went to New York and promptly lost everything on the stock market when William Rockefeller and Henry Rogers, in a vindictive stock manipulation of his United COpper Company, wiped him out. Wall Street's punishment of Heinze precipitated the panic of 1907.36 Period Three, 1908-1930. Corporate Diversification Leads to Political Diversification. After Heinze's departure, Amalgamated moved to establish its influence in the Democratic party as well as the Republican party, and the third cycle of its partisan activity began. 33Ibid., p. 290. 34Jerre C. Murphy, The Comical History of Montana, p. 134. 35K. Ross Toole, An Uncommon Land, p. 209. 36 C. B. Glasscock, The War Of the COpper Kings, pp. 303-4. 128 The method of Operation was first, to exert influence over the selection of candidates by both parties, and second, to contribute money, organizational support, and the support Of the corporation press to the election of candidates acceptable to the corporation. An example of this bipartisan activity at the beginning of this period is provided in the election of 1908. The corporation supported Theodore Roosevelt, Republican, for President, and Joseph K. Toole, Democrat, for Governor. ". . . Company support . . . would easily account for the election of both so far as the state of Montana was concerned, regardless of the relative strength or weakness of the defeated candidates."37 The party division of corporation officers also indicated this bipartisan activity. The party division of Amalgamated officers in 1912 was as follows, according to a contemporary observer: Nominally, President John D. Ryan is a partisan Democrat. So was Mr. John G. Morony when managing director. So is Mr. "Con" F. Kelly who succeeded him. Mr. E. P. Mathewson, general manager of the great smelters and almost everything else in Anaconda, is the same kind of a Republican. So is Mr. C. H. Goodale, who has charge Of Great Falls smelters and some mining Operations in Butte. Mr. John Gillie, general superintendent of the company, Democrat; Mr. L. 0. Evans, head of the legal department, Republican; Mr. Harry Gallwey, manager of the combine railway company, Democrat; Mr. 37 pp. 137-8. Jerre C. Murphy, The Comical History of Montana, 129 Charles Schwartz, company tax commissioner, Republican; Mr. John R. Toole head of the great lumber department, Democrat.3é He goes on to add: But all of the gentlemen named, and almost all of the thousands unnamed, actively participate in political primaries, in conventions, at elections, or as lobbyists; regardless of party politics, or promises, or interests; and with as strict devotion to the wishes of their superiors and the lawless purposes of the combine as they do in their respec- tive positions in the industrial organization Of the company.39 ‘ Influence over the nomination process was exer- cised by having company officers, especially attorneys, become leaders in each party. Rather than try to main— tain control of state politics through the Republican party, a difficult proposition due to the increasing strength of the unions and the Progressive movement, the company worked to get acceptable candidates elected from either party. If they could not manage this, they attempted to convert the candidates who did get elected to their political viewpoint. They were successful in doing this, not only in the state legislature but in the U. S. Senate as well. Every Senator elected either by the legislature or by the peOple between 1904 and 1936—- Republicans Carter and Dixon, Democrats Myers, Walsh, and Wheeler--were either elected with the support of the 38Ibid., p. 131. 391bid., p. 132. 130 company, as were Carter, Dixon and Myers, or were con- verted, if not to the company's vieWpoint at least to a neutral position, and supported by the company for re-election, as were Walsh and Wheeler. Senator Thomas Murray, elected in the Roosevelt landslide of 1936, was the first U. 8. Senator from Montana who was not indebted to the company at any time during his Senate career. Because of increasing prejudice against corporate influence in politics, a new tactic was introduced in 1908: ". . . corporate agencies were shrewdly employed to utilize the pOpularity of capable men and to applaud patriotism and public Spirit whenever it could be exercised without injury to combine interests or Obstruc— tion to combine plans."4O Control Of the state's press helped make this tactic easy to practice. Influence over nominations was public knowledge. A contemporary newspaper editor noted that ". . . men of good public character and reputation and without corrupt purpose have been known to visit Butte to learn from headquarters if their candidacy would be satisfactory to the combine managers prior to announcing it to the 41 public." Company support was not always consistent. When two company candidates opposed each other in the 4011311, p. 134. 41Ibid., p. 135. 131 race for Governor in 1908, company contributions were made to the campaign funds of each. This was not irrational from the company's point of view, however. Republican State Senator Edward Donlan's chances of success in his campaign for Governor were reduced when the company used its influence in the Democratic party to get Democrats to support the Republican candidate for President (Taft) in exchange for Republican support for the Democratic candidate for Governor, Edwin L. Norris. The company benefitted three ways: its choice for President, Taft, carried the state; its choice for Governor, Norris, carried the state; and State Senator Donlan, grateful for company support in the campaign, continued to support the company from his position as leader of the state senate in the legislative session.42 By 1910 the company had slightly altered the form Of its organization under the pressure of laws promoted by the trust-busting activities of Theodore Roosevelt and changed its name to the Anaconda COpper Mining Company. It had been given new influence by legislation secured in 1909, and continued to enlarge its Sphere of influence: "The Anaconda controlled the hill (Butte), the smelters (in Anaconda), the plants at Great Falls, the timber resources of the state, a large part of the water—power 42Ibid., pp. 135—6. 132 rights, the leading newspapers, the great company store, Hennessey's, and the political organizations."43 The relationship between the company and Republi— can Senator Joseph Dixon illustrates how the company exercised its influence. Dixon was elected to the U. S. House of Representatives in 1902, re-elected in 1904, and elected to the U. S. Senate in 1907. He was supported by Amalgamated, as was every candidate who was opposed to or Opposed by Heinze. Mr. Dixon rewarded this support by favoring the Amalgamated Company against Heinze in all matters in which both were interested.44 However, in 1909 Dixon vigorously supported the Interstate Commerce Commission law prohibiting the railroads from charging discriminatory rates to Montana Shippers, although the copper interests controlled the railroads. "Because of this activity both Democratic and Republican party leaders and newspapers aided in conSigning him to political ."45 President Taft lost confidence in oblivion. . . Senator Dixon's judgment, and the Senator's recommenda— tions on patronage were ignored. The Democratic press 43C. B. Glasscock, The War of the COpper Kings, p. 307. 4Jerre C. Murphy, The Comical History of Montana, p. 139. 45 M. G. Burlingame, Political Party Activities in Montana, p. 26. 133 commenced serious war against him, and the Republican press, subject to the same corporate influences as the Democratic press, commenced to ignore or misrepresent him. After Dixon announced his support of Mr. Roosevelt in 1912, and was made manager of Roosevelt's nationwide campaign, the company announced its support of Taft. A delegation favorable to Taft was sent to the convention in Chicago with the declared purpose of "getting Joe Dixon."46 Dixon, running on the Progressive ticket, was defeated by Thomas J. Walsh in 1912, the first year in which U. S. Senators were elected directly by the peOple. Walsh, though not a company man, was considered the lesser of two evils. Dixon was to have one last hurrah. He ran for Governor in 1920 and was supported once again by the company, but only because the company had been unable to Split the vote in either party primary and was faced with a choice between the Republican, Dixon, and the Democrat, Burton K. Wheeler-~"both liberal, both dangerous, "47 and both unbuyable. This time the company chose him as the lesser of two evils, since "Bolshevik Burt" 46Jerre C. Murphy, A Comical History of Montana, pp. 139-40. 47 K. Ross Toole, An Uncommon Land, p. 222. 134 Wheeler was hated by the company for his refusal to prosecute, as U. S. District Attorney, radical labor agitators who were trying to oust the company union. The company strongly disliked Dixon, but they detested and feared Wheeler. J. Bruce Kremer, an Anaconda attorney and a Democratic national committeeman, worked for Dixon against his own party's candidate.48 . Dixon was elected Governor but could not accom— plish his goal of revision of mining taxation laws because the company Opposed him at every turn. They turned their newspapers over to the process of his excoriation, leaving him no medium for expression; the state press was controlled, and there were no radios. A Democratic legislature hostile to Dixon was elected in 1922 with company support.49 When Dixon ran for re-election in 1924, the company attacked him in one of the state's most virulent campaigns. They warned laborers of mine shutdowns and depression if Dixon won; they said his tax prOposals were confiscatory and would not only destroy old business but would prevent new business from coming into the state. They defeated him soundly, but in the heat of the campaign overlooked Initiative Number 28, prepared by 48Ibid., p. 222. 491bid., p. 223. 135 Dixon, which provided for a graduated tax on mine produc- tion. It passed, and Dixon departed the political scene with a victory for the state of Montana which resulted in its tax revenue from the mining industry increasing from $13,000 in 1924 to $300,000 in 1925.50 The relationships between the company and Senators Walsh and Wheeler illustrate what happens when anti— company Officials become more tolerant in their attitudes toward the company and begin to get company support. Senator Walsh, before going to the U. S. Senate, had been an attorney in Helena who had successfully prose— cuted actions for individuals against corporations. He had refused an annual retainer from the Anaconda Company. In his first race for the Senate in 1910 against the incumbent Senator Carter, the company Opposed him and, taking advantage of the malapportionment of the state legislature favoring the less populated ranching and farming counties, managed to elect enough legislators to prevent the election of Walsh. It was a close election, however, with Walsh's Democratic party having a majority of fifty-three to fortyénine Republicans on the joint ballot. Walsh was defeated when ten Democrats, including eight company-elected representatives from Silver Bow county, refused to Sign a caucus call or an agreement to 5°1bid., p. 226. 136 be governed by its actions. Among the representatives refusing to Sign the call was Representative Roy S. Alley, secretary and confidential man for Anaconda President John D. Ryan, and state Senator Harry Gallwey, for years a mine manager for the company and a trusted lobbyist at Helena before he was elected to the senate. Henry L. Myers, Democrat, was elected as a compromise with the , 51 company s approval. Walsh was elected in the first public election of U. S. Senators in 1912, as noted above, when the company supported him as the lesser of two evils. When he ran for re-election in 1918, the company put pressure on him to split with his protege, Wheeler, whom he had gotten appointed as U. S. District Attorney and who was causing the company so much trouble. Walsh gave in and asked Wheeler to resign. The company then supported him and he was re—elected. His law partner wrote him after his victory: The election returns Show most conclusively that the company did all that it possibly could do to bring about your election, and without the financial assistance that was given by Con Kelly (later chairman of Anaconda's board) our situation would be critical in meeting the expense account that was contracted.52 51 pp. 144-9. 52 Jerre C. Murphy, The Comical History of Montana, K. Ross Toole, An Uncommon Land, p. 220., 137 In the Senate Walsh became more interested in national affairs, as Montana Senators invariably do, almost to the exclusion of state affairs. After breaking the "Teapot Dome" case, along with Wheeler, he achieved national fame and "walked the tightrOpe of senatorial liberalism and local conservatism with consummate skill."53 He remained in the Senate until 1932, without Opposition by the company, when he resigned to take the post of U. S. Attorney General in Roosevelt's cabinet. He died, however, before the new administration was sworn in. Senator Wheeler, the firebrand anti-company U. S. District Attorney, was an equally perceptive pragmatist. After his defeat in the race for Governor in 1920, he took two years to simmer and to contemplate Walsh's strategic political withdrawal. He was a better politician when he ran for the U. S. Senate in 1922, and realized that all he needed to reach the Senate was, if not the neutrality, no more than mild Opposition from the company. Wheeler approached the company for support, with Senator Walsh as intermediary. He got it and won by a large margin. He ran on the Progressive ticket as a candidate for Vice— President with LaFollette in 1924, following the pattern of Montana Senators in emphasizing national affairs. He 53Ibid., p. 215. 138 was a liberal until 1937, when he fell out with Roosevelt because of his Opposition to Roosevelt's court—packing plan. He attacked the New Deal, became one Of the most vocal Of isolationists, and by 1946 was an avowed con- servative. He was defeated in the Democratic primary that year in spite of strong company support.54 One of the most important reasons for his defeat was the opposition of Senator Murray, the first U. S. Senator from Montana who was not associated in any way with the Anaconda Company. Wheeler also lost support in the Democratic party because of his friendship with the two leading Republicans in the state, J. Wellington Rankin, state Republican chairman, and Governor Sam T. Ford, whose nomination for Governor in 1940 had been made by Senator Wheeler in the Republican convention. He had campaigned for Ford while running for re—election to the Senate in 1940. Progressivism in Montana The Progressive movement in Montana was repre- sented to a great extent by Senators Dixon, Walsh and Wheeler. In the judgment of historian K. Ross Toole, . . . progressivism in Montana lasted from 1910 to 1924, and the most notable thing about it, in the end, was that, save for sound and fury and one small tax reform, it got no place. The forces of conservatism were overwhelming. On no single 54Ibid., p. 225. 139 occasion did any spirit of rebellion infect the legislature.55 Historian Richard B. Roeder takes issue with this conclusion. In extending the period of the movement to include the years 1900 to 1910, he argues that the passage of legislation dealing with direct primaries, the commission form of government, extension of the initiative and referendum to cities and towns, non- partisan nomination of judges, woman suffrage, a prefer- ential vote for U. S. Senators, attempts to define and regulate trusts, to regulate railroads and other public utilities, and to outlaw railroad passes for public officials suggest that Montana "responded to the same progressive impulses as the rest Of the nation."56 Toole, however, argues in rebuttal that the Anaconda Company fought the direct primary, but thereafter thoroughly controlled the results of the primary; that it did not fight the creation of the Public Service Commis— sion as a regulatory agency because it knew it could control the Commission and has controlled it in fact; that it did not fight the commission form of city govern- ment for the same reason; and that Anaconda still elected 55Ibid., p. 226. 56Richard B. Roeder, "Sound and Fury . . . and One Small Tax Reform," Montana, the Magazine of Western History, Autumn, 1970, pp.“18—26. 140 and controlled judges and other Officials in spite of non—partisan nominations. Other reforms did not concern them: the eight—hour day was already in effect, and they were not interested in compulsory school attendance, child labor, or juvenile courts. As Toole argues, a reform isn't a reform if it doesn't work.57 The reforms enacted in Montana could not be effective against the company's pervasive and overwhelming power. The company protected its interests the way it always had: by taking whatever action was necessary to keep the political system from threatening its economic well-being. Period 4, 1932—1938. Riding With The Tide—— Corporate Support of the Democratic Party. The defeat on the initiative on tax revision in 1924 convinced the company that its wisest course lay in making more moderate demands and building a better public image. During the 1920's the company press was vitriolic in its treatment of its enemies; by the early 1930's the company papers became less venomous and ignored them by not mentioning them or their activities.58 57K. Ross Toole, "Rebuttal: When Is A Reform A Reform?" Montana, the Magazine of Western History, Autumn, 1970, p. 27. 58 Harmon Zeigler, "Interest Groups in the States," p. 121. 141 The Democratic wave of the depression years of the 1930's resulted in the company switching most of its support to the Democrats, who were winning most of the Democratic races in the state government as well as those for offices in the federal government. The company began to support the Democrats at the national level as early as 1924, when President John D. Ryan contributed $5,000 to the national party.59 He followed this with a $2,000 contribution in 1928 to help pay the deficit left over from 1924, and with a $27,000 contribution to Alfred E. Smith's campaign.60 Ppriod 5, 1940—Present. Modern Corporate Politics: The Path of Least Resistance-~Support of the Current Majority at Each Level of Government. During the war years of 1941—1945 the Republicans regained control of the state senate, so the company had little to fear from adverse state legislation. Their interests could best be protected at the federal level by Democratic officeholders who have, with the exceptions of 1946 in the Senate and 1946 and 1952 in the House of Representatives, controlled Congress since 1932. In the 1946 election Montana retained its representation in the majority party in the 59Ferdinand Lundberg, America's Sixty Families, p. 172. 6OIbid., pp. 179-81. 142 Senate by sending to it the only Republican to win a Senate seat in Montana. The Republicans also regained control of the Second Congressional district for the first time since 1932 and kept it until 1956. In the modern period, there is no evidence to indicate that the corporations have lost their potency; only their methods of operation have changed. The Montana Power Company has become more active in state politics, and the Anaconda Company has become less visible. By the late 1940's the company newspapers had again changed their policies and, rather than simply ignoring its enemies, actually began to print Speeches hostile to the company. Editorials took on a more neutral cast.61 In 1959 the company disposed of its neWSpapers which, at that time, were owned and maintained through a subsidiary, the Fairmont Corporation. Its control had been extensive. According to an Observer in 1937: "The Anaconda COpper Mining company . . . is one of the great neWSpaper publishers of the country. It owns nearly every neWSpaper in Montana that has an Associated Press franchise."62 The seven company papers had 55.4 per cent 61 p. 121. 62 Harmon Zeigler, "Interest Groups in the States," Ferdinand Lundberg, America's Sixty Families, p. 272. 143 of the total daily circulation in Montana, and 59.4 per cent of the total Sunday circulation.63 In the late 1950's the company began to be con— cerned with its public image, as reflected in its press policies, and tended to work quietly in protecting its economic interests, confining itself to blocking adverse legislation and reducing its efforts to influencing the electoral process.64 These efforts were not in vain. A contemporary observer in 1959 noted: Today the influence of the Anaconda Company in the state legislature is unSpectacular but very great. It has been a long time since the company showed the mailed fist. But no informed person denies its influence or the fact that the basic use to which it is put is to maintain the status quo—~to keep taxes down, not to rock the boat.65 The Anaconda Company is known for the excellence of its lobbying activities, and every measure put before the legislature is followed with keen scrutiny by Anaconda's attorneys. Anaconda has used its power mostly in a negative way to prevent undesirable government 63Thomas Payne, "Under The COpper Dome: Politics in Montana," p. 190. 64 p. 121. 65 Harmon Zeigler, "Interest Groups in the States," K. Ross Toole, An Uncommon Land, pp. 244—5. 144 legislation, and in this effort it has been joined by the Montana Power Company.66 The Montana Power Company has become more powerful and better organized as a lobby, partially to divert public attention from the Anaconda Company, and partially because it has developed a state—wide political organiza- tion built into its company organization. Its Officers, regional directors, and even its clerks and linemen are public relations agents for the company. Most executives are skilled in the traditional public relations tactics: buying drinks at the country club, sponsoring Little League baseball teams, speaking to high school and college classes on the importance of the company to the community, holding office in community service organizations, and diverse other efforts. The message is always the same: Montana Power Company's purpose is to serve the public, and it can do that best if there is no interference from the government. Much of the political activity of the Power Company is conducted through its legal branch. One of its more effective tactics is to practice the principle "if you can't lick'em, let'em join you." An example of this tactic is the practice of hiring bright young county 66Thomas Payne, "Under the COpper Dome: Politics in Montana," pp. 197-8. 145 attorneys (or state legislators, Legislative Council or Public Service Commission lawyers and accountants) from both political parties who become popular because of their Opposition to the company. Company salaries are much more attractive than state government salaries. Old political pros in the state maintain that the best way to get a job with Montana Power Company is to Oppose them politically: the more successful the Opposition, the more attractive the job offer from the company. Few politicians can resist the temptation of a soft job with a large salary. Attempts to gain favor with incumbent officials by means less than ethical, if not illegal, are not unknown. In 1970 one issue of the Missoulian neWSpaper carried two stories about investigations into Montana Power Company land deals. In one deal, the company had sold choice land to the wife of a Republican state Supreme Court Justice at one—thirtieth of its market value. In the other deal, the company had sold choice lakefront prOperty to the Democratic Governor's firm at a fraction of its original value. Both officials denied that they had shown any "official favoritism" to the company.67 The corporations' activities are primarily con- cerned with keeping the state legislature conservative and 67Missou1a Missoulian, August 20, 1970, p. 18. 146 keeping Montana's delegation to Washington in the majority party in Congress. The company's support of Democratic Senator Wheeler until he lost in 1946 primary and their support of Republican Zales Ecton in the general election of that year maintained their consistency in supporting the winner in one of the state's Senate seats from 1922 to 1952. When Mike Mansfield defeated Ecton in 1952, the company did not support him, but did not Oppose him, and have not Opposed him seriously since then. The success of company-supported Senators was not due to company support completely, but also due to the fact that Montana Senators invariably support the programs in Washington that Montanans want supported. It is a happy coincidence for the company that those programs are also the programs that are acceptable to the company. The programs of the federal government relating to the tariff, metals policy, and price support for basic crOps critically affect the economy and politics of Montana. Stockpiling of critical materials such as zinc and COpper by the federal government affects the state's producers of these materials, and price supports for farm products, especially wheat, are important to a large segment of the state's voters. Politicians who are not sympathetic to these concerns do not last long in 147 Montana.68 The Anaconda Company is vitally concerned with copper and zinc stockpiling and tariff policy, and is not opposed to farm supports. It does not find it difficult to support Democratic Congressmen who are dedicated to supporting the policies which satisfy the large majority of voters in the state. In the state legislature the company's interests have been protected by a generally conservative legisla- ture. Regardless of the partisan alignment in either or both houses, a conservative orientation is normally characteristic. Rarely are prOposals broader in SCOpe than the local or sectional interests of legislators brought up for consideration.69 Things were so favorable in state government that in 1961, after the legislature passed a law allowing state corporations to issue stock Options to its executives and the Governor had signed it, the Montana Power Company moved its headquarters from New Jersey and incorporated under the laws of Montana forty-nine years after it had been formed in 1912. This put it outside the jurisdiction of the Federal Power Commission since at the same time a subsidiary corporation, the Montana—Dakota Utilities 68Thomas Payne, "Under the COpper Dome: Politics in Montana," p. 204. 691bid., p. 200. 148 Company, was created to handle its interstate operations. The same bill had been passed by the legislature in 1951 but had been vetoed by a Democratic Governor. The Relationship Between Split Control and Corporate Strategy The incidence of split control has shifted as the corporations have altered their strategy to meet changing circumstances and opportunities. In the early periods of the state's history, 1890-1907, which corresponds to the first two phases of corporate political activity described above, the corporations supported first one political party, then the other. After the Republican victories of 1892 and 1894, we may see the effects of corporate support Of the Democratic party from 1896 to 1900 in Table 15 (next page), which gives the party of the winning candidate for selected offices and the party with state legislative majorities, 1890-1968. With corporate support, Democrats virtually swept these offices between 1896 and 1900. AS described in detail above, the corporations shifed their support to the Republican party at all levels during the period 1901-1907. In elections held in this period, Republicans won nine out of eleven of the offices and majorities recorded in Table 15. If we apply the measure of split control to these election outcomes, we find the incidence Of Split control to be moderately low in both of these periods. In the period 1890-1900, 22.6 per cent of the relevant 149 Table 15. Partisan Results of Montana Elections, 1890-1968 Year 'Pres. Sen. Gov. Cong. District State legislative U.S. (1) (2) majorities House Senate 1890 D R D 1892 R R R R T D 1894 R R R R 1896 D D R (Silver) D R 1898 D D D D 1900 D D D D D D 1902 R R D 1904 R R D R R R 1906 R R R R 1908 R D R D R 1910 D R D R 1912 D D D D D D D 1914 D D D R 1916 D D D D R D R 1918 D D R R R 1920 R R R R R R 1922 D D R R R 1924 R D D D R R R 1926 D R R R 1928 R D D D R R R 1930 D D R R R 1932 D D D D D R 1934 D D D D T 1936 D D D D D D D 1938 R D D D 1940 D D R R D D R 1942 D D D D R 1944 D R D D R R 1946 R D R R R 1948 D D D D R D T 1950 D R R T 1952 R D R D R R R 1954 D D R D R 1956 R R D D D D 1958 D D D D D 1960 R D R D D R D 1962 D R R D 1964 D D R D R D D 1966 D D R R D 1968 R D D D R D R - Republican; D _ Democrat; T — tied' Source: Ellis Waldron, An Atlas of Montana Politics Since 1864 (Missoula, Montana: Montana State University Press, 1958). pp. 176—8, 196—200, 220—2, 240—2, 246, 264—6, 270, 286—8, 306—8, 326—8, 356—8, 384-6. Richard Scammon, America Votes, 18 (Wash., D.C.: Governmental Affairs Institute, 1970), 227—9. 150 100 90 80 7O 6O 50 40 3O 2O 10 1890- 1902- 1908- 1932- 1940- 1900 1906 1930 1938 1968 Figure 5: Incidence of Split Control in Montana Elections, 1890-1968. Source: Table 16. Table 16: Incidence of Split Control in Montana Elections, 1890-1968. Periods of Corporate Incidence of Split Congrol in Pairs Support of State and Federal Offices Fraction Per Cent Democratic party, 1890-1900 7/31 22.6 Republican party, 1902-1906 4/15 26.6 Both parties, 1908-1930 39/93 41.9 Democratic landslide, 1932-1938 5/28 17.9 Republican party at state level, 1940-1968 58/119 48.7 Source: Table 15 Note: The method of calculating split control is descri— bed in the text, at the beginning of Chapter 6. 151 state-national Office-pairs were split between the parties; for the years 1901-1907, 26.6 per cent involved party control split between state and national offices. (Figures for these and later periods are recorded in Table 16 and Figure 5, page 150.) Between 1908 and 1930, however, corporate strategy shifted to support candidates of BEER parties at all levels of government and, we may assume, to exert substan- tial influence in both parties. The legislature tended to go Republican in both houses. U. S. Senate seats were held exclusively by Democrats. Other offices were shared. With corporate support not exclusively behind either party, party control over state and national Offices was frequently split: the incidence of split control for elections between 1908-1930 was 41.9 per cent. The Democratic landslide of the 1930's greatly reduced the incidence of Split control to 17.9 per cent. Perhaps it also suggested to the corporations some new limitations on their ability to control the Democratic party; and surely that party's ideological shift must have made it seem an unattractive and potentially un- reliable agent. Whatever the precise causes, a distinc- tively modern pattern appears for the first time in the election of 1940, in which Republicans captured the gover- norship and one house of the legislature, but Democrats won three out of four elections to national offices. With variations, this pattern most clearly typifies the period 152 since l940--apparently the result of the concentration of corporate effort to elect Republicans at the state level in a predominantly Democratic state. The effects of these shifts of circumstances and strategy on split control is seen clearly in Table 16 and Figure 5. In the modern period, the incidence of split control rises to the high level of 48.7 per cent. The result of the strong impact of corporate strategy is to make Montana's current levels of net deviation voting and split control higher than those of most states. Summary The flexibility of corporate partisan support and the configuration of interests in Montana which permits that flexibility leads to the seemingly ambivalent voting by Montana voters, that puts conservative Republicans in state Offices and liberal Democrats in federal offices. This may be a strategy to maximize benefits from the federal govern- ment by sending liberals to Congress while at the same time restricting spending and regulation by state government by eknting conservatives to state offices. Net deviation voting and split control are the result. The corporations encourage this thinking. It allows them to support the Democratic party nationally under the pretext of trying to get as much federal support for the state as possible, which incidentally reduces the state's need to tax the corporations. The more successful 153 the corporations are in convincing the voters that such a configuration of interests exists, the higher the incidence Of net deviation voting. In the mOdern period, then, net deviation voting and split control may be considered indicators of the success of the corporations in Montana in persuading the voters that their well-being is enhanced by actions that enhance the well-being of the corporations. The highly competitive nature of electoral politics in the state, with the Democrats holding a slight advantage, is a contributory factor in the success of the corporate interests in pursuing a strategy of splitting their efforts between levels of government. By allowing Democrats to dominate elections for national Offices, they avoid a direct confrontation with opposing forces, predominantly union oriented, in the Democratic party. The corporate interests in Montana are so powerful, however, that it doesn't always matter which party controls state government, as the corporations Often manage to get what they want in the way of substantive legislative enact- ments, in spite of the differing symbolic positions of the parties. They usually prefer to work within the structure of the Republican party, at least in recent years, because they exhibit business orientations more acceptable to them, and because the possibility of an anticorporation movement taking over control of the Republican party is slight. Chapter 7 SUGGESTIONS FOR FURTHER RESEARCH Three approaches were used in this study in an attempt to determine the variables in Montana elections which influence voters to deviate from straight party ticket voting. The first approach was to attempt to develOp a model of net deviation voting by regressing net deviation statistics for Montana counties on fifteen political, socio—economic, and demographic variables. This attempt was not successful. It.was concluded that no model could be develOped using aggregate data which would consistently eXplain a Significant amount of net deviation voting. The second approach was an attempt to determine if the degree of unionization of the dominant occupation in a county was related to the amount of net deviation voting in the county. It was concluded that there was no apparent relationship between the two variables. The third approach was a historical case study in which the political history of the state was examined to determine what forces have influenced voting behavior in the state over a long period of time. It was concluded 154 155 that in a highly competitive state such as Montana, where the partisan division of the vote is almost evenly split, political activity by corporations to protect their interests by supporting conservative Republicans for state offices and liberal Democrats for national office results in the state having a high level Of net deviation voting. While the historical case study method was successful in explaining net deviation voting in Montana, the results cannot be generalized to other states. The unique circumstances which obtain in Montana, and the particular kinds of political activity which result Offer no explanations for states with other circumstances. An understanding of net deviation voting in a broader context than that of a single state requires a different approach. Survey research methods offer two possibilities for fruitful research on net deviation voting. While survey research methods cannot deal adequately with historical develOpmentS over a long period of time, they can serve to identify relationships which exist at a particular point in time and explain net deviation voting as it occurs in contemporary elections. A comparison of varying relationships between net deviation voting and explanatory variables under different electoral circum— stances would increase Our knowledge about the net devia— tion phenomenon. 156 The first method could be a study of the relation- ship between Split-ticket voting and the voters' percep- tions of the functions of the different levels Of political organization. If it could be shown that voters who split their tickets have different perceptions of the function of political party organizations than do those who do not split their tickets, and if there is a common element in their perceptions, then a relationship could be shown to exist between ticket-splitting and the common element shared by all ticket-splitters. If, for example, ticket- splitters perceive the functions of the state party organization to have little or no relation to the function of the national party organization and non-ticket-splitters see the two party organizations to be strongly related, split-ticket voting could be shown to be related to the differing perceptions of party functions held by the voters. If these differing perceptions have partisan or ideological meanings, it is possible that ticket-splitting could be related to the voters' perceptions of which party can best Perform the different functions at each level of government. It is possible for a voter to be a Democrat at one level of government and a Republican at another level of government without having any ideological con- flict. Such a voter would be expected to Split his ticket regularly. 157 The second method could be to determine if split- ticket voting is related to inconsistent partisan attach- ments and strong interest group influence. If inconsistent partisan attachments as discussed above could be shown to vary as ticket-splitting varies and a relationship could be shown to exist between inconsistent partisan attach— ments and strong interest group influence, the possibility of a strong relationship between Split-ticket voting and strong interest group influence (or activity) could be investigated by survey research methods. It is possible that inconsistent partisan attachments are more prevalent in those states where there is one dominant interest group whose actions reinforce inconsistent partisan attachments. In a state like Montana, for example, the actions of the corporations in emphasizing the importance of state politics and ignoring the national politics could reinforce inconsistent partisan attachments. In those states where two or more interest groups compete politically, and especially where one interest group is oriented toward state politics and the other is oriented toward national politics, inconsistent partisan attachments would occur less frequently. If such patterns do exist, it would be possible to relate the rate of split-ticket voting across the states to the number and influence of the interest groups in the states. 158 These examples do not exhaust the possibilities that survey research methods Offer, but serve to indicate one direction in which research may be fruitful. Methods other than survey techniques may also be used. Methods using aggregate statistics should be used with caution, since the problems encountered may be of sufficient magnitude as to make the effort impractical, if not entirely useless. The best method for investigating a phenomenon such as net deviation voting is one which uses the individual voter as the unit of analysis. BIBLIOGRAPHY BIBLIOGRAPHY 1. Books Bates, J. Leonard. Senator Walsh of Montana, 1918—1924: A Liberal Under Pressure, Unpublished Doctor's dissertation, University of North Carolina, 1952. Bennett, C. A. and N. C. Franklin. Statistical Analysis in Chemistry and the Chemical Industry. New York: John Wiley & Sons, Inc., 1954. Blalock, Hubert M., Jr. Causal Inference in NoneXperi- mental Research. Chapel Hill: University of North Carolina Press, 1964. . Social Statistics. New York: McGraw—Hill Book Company, Inc., 1960. and Ann B. Blalock. Methodology in Social Research. New York: McGraw-Hill Bock Company, 1968. Bogue, Donald J. and Dorothy L. Harris. Comparative Population and Urban Research via Multiple Regression and Covariance Analysis. Chicago: Scripps Foundation for ResearCh in POpulation Studies and Population Research and Training Center, University of Chicago, #8, 1954. Burlingame, Merril G. The Montana Frontier. Helena, Montana: State PubliShing Co., 1942. . Political Party Activity in Montana. Unpublished manuscript, 1964. and K. Ross Toole, eds. A History of Montana. 3 vols. New York: The Lewis Publishing Co., 1957. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. The American Voter. New York: John Wiley & Sons, Inc., 1964. Connolly, Christopher P. The Devil Learns to Vote. New York: Covici Friede, 1938. 159 160 Cummings, Milton C., Jr. Congressmen and the Electorate. Elections for the U. S. House and the President, 1920—I964. New York: The Free Press, 1966. Draper, N. R. and H. Smith. Applied Regression Analysis. New York: John Wiley & Sons, Inc., 1966. Duncan, Otis D., Ray P. Cuzzort, and Beverly Duncan. Statistical Geography. Glencoe, Ill.: The Free Press, 1961. Epstein, Leon D. Politics in Wisconsin. Madison: The University Of Wisconsin Press, 1958. Ezekiel, Mordecai and Karl A. Fox. Methods Of Correla- tion and Regression Analysis. 3rd ediEiOn. New York: JOhn Wiley & Sons, Inc., 1959. Glasscock, C. B. The War of the Copper Kings. New York: Grosset & Dunlap, 1935. Hamilton, James M. From Wilderness to Statehood: A History of Montana. Ed. by Merrill G. Burlingame. Portland, Ore.: Binfords and Mort, 1957. Hays, William L. Statistics for Psychologists. New York: Holt, Rinehart, and Winston, 1963. Heard, Alexander. The Costs of Democracy. Chapel Hill: University of North Carolina Press, 1960. Herbert Jacob and Kenneth N. Vines, eds. Politics in the American States. Boston: Little, Brown & Co.,il965. Howard, Joseph K. Montana: High, Wide and Handsome. New Haven: Yale University Press, 1943} Hutchens, John K. One Man's Montana: An Informal Portrait of a State. Philadelphia: J. P. Lippincott Co., 1964. ' Jonas, Frank H. Western Politics. Salt Lake City: University of Utah Press, 1961. Key, V. 0., Jr. The Responsible Electorate. With the Assistance of Milton C. Cummings, Jr. Cambridge, Mass.: The Belknap Press of Harvard University, 1966. Kirk, Roger E. EXperimental Design: Procedures for the Behavioral Sciences. Belmont, Calif.: BrookS7Cole Publishing Company, 1969. 161 Lindsay, John. The Amazing Experiences of a Judge. Philadelphia: Dorrance and Co., 1939. Lundberg, Ferdinand. America's Sixty Families. New York: The Citadel Press, 1937. Metcalf, Lee and Vic Reinemer. Overcharge. New York: David McKay Co., Inc., 1967. Murphy, Jerre C. The Comical History of Montana. San Diego: E. L. Scofield, 1912. Pool, Ithiel de Sola, Robert P. Abelson, and Samuel POpkin. Candidate Issues and Strategies: A Computer Simulation of the 1960 and 1964 Presidential Elections. Cambridge, Mass.: The M. I. T. Press, 1965. Raymer, Robert G. Montana: The Land and the Pegple. 3 vols. Chicago: The Lewis Publishing Co., 1930. Renne, Roland R. The Government and Administration of Montana. New York: Thomas Y. Crowell Co., 1958. Sheridan, Richard G. Politics Is Thicker Than Water. Bozeman, Mont.: Montana State Extension Service, 1968. Shinn, Charles H. Mining Camps: A Study in American Frontier Government. New York: Harper & Row, 1965. Smothers, Frank, ed. The Book of The States, 1954-1955. Chicago: Council of State Governments, 1954} Stuart, Granville. Forty Years on the Frontier. Cleveland: Arthur A. CIark Co., 1925. Toole, K. Ross. Montana: An Uncommon Land. Norman, Okla.: University of Oklahoma Press, 1959. Waldron, Ellis. An Atlas of Montana Politics Since 1864. Missoula, Mont.: Montana State University Press, 1958. Webb, Walter Prescott. The Great Plains. Boston: Houghton Mifflin, 1936. Wheeler, Burton K. and Paul Healy. Yankee From the West. Garden City, N. Y.: Doubleday & Co., 1962. 162 2. Articles Adamany, David. "The Size of Place Analysis Reconsidered," Western Political Quarterly, 16 (1964). Alford, Robert P. "The Role of Social Class in American Voting Behavior," Western Political Quarterly, 15 (1963). Althoff, P. and S. Patterson. "Political Activism in a Rural County," Midwest Journal of Politics, 1 (February, 1966). Baggaley, A. R. "Patterns of Voting Change in Wisconsin Counties," Western Political Quarterly, 12 (1959). Beyle, Thad L. "Contested Elections and Voter Turn-Out in a Local Community: A Problem in Spurious Correla— tion," American Political Science Review, 59 (1965). Browning, Rufus. "Computer Programs as Theories of Political Processes," Journal of Politics, 24 (1962). Burnham, Walter D. "American Voting Behavior and the 1964 Elections," Midwest Journal of Political Science (February, 1968). . "The Changing Shape of the American Political Un1verse," American Political Science Review, 59, NO. 1 (March, 1965). Campbell, Angus and Warren E. Miller. "The Motivational Basis of Straight and Split-Ticket Voting," American Political Science Review, 59 (1957). Cox, Edward F. "The Measurement of Party Strength," Western Political Quarterly, 13 (1960). Cutright, Philip and Peter H. Rossi. "Grass Roots Politicians and the Vote,“ American Sociological Review, 23, #2 (April, 1958). Dawson, R. E. and J. A. Robinson. "Interparty Competition, Economic Variables and Welfare Policies in American States," Journal of Politics, 25 (1963). Duncan, 0. D. and B. Davis. "Alternative to Ecological Correlation," American Sociological Review, 18 (1953). Eldersveld, Samuel. "Theory and Method in Voting Behavior Research," Journal of Politics, 13 (1951). 163 Epstein, Leon D. "Size of Place and the Division of the Two-Party Vote in Wisconsin," Western Political Quarterly, 9 (1956). Eulau, Heinz. "Class Identification and Projections in Voting Behavior," Western Political Quarterly, 8 (1955). Glaser, William A. "Intention and Voting Turnout," American Political Science Review, 52 (1958). Gold, David. "Some Problems in Generalizing Aggregate Associations," American Behavioral Scientist (December, 1964). Golembiewski, R. T. "A Taxonomic Approach to State Political Party Strength," Western Political Quarterly, 11 (1958). Goodman, L. A. "Ecological Regression and Behavior of Individuals," American Sociological Review, 18 (1953). Goodman, Leo A. "Some Alternatives to Ecological Correlations," American Journal of Sociology, 64 (1968). Hacker, Andrew. "Does a Divisive Primary Harm a Candi- date's Election Chances?" American Political Science Review, 59 (1965). Hofferbert, R. I. "Classification of American State Party Systems," Journal of Politics, 26 (1964). . "Ecological Development and Policy Change in the American States," Midwest Journal of Politics, 4 (November, 1966). Howard, Joseph K. "The Decline and Fall of Burton K. Wheeler," Harper's Magazine (March, 1947). . "The Montana Twins in Trouble," Harper's Magazine (September, 1944). Jacob, Herbert. "State Political Systems," in H. Jacob and K. Vines, Politics in the American States (Boston: Little, Brown & Co., 1965). Jones, Charles O. "Inter-Party Competition for Congres- sional Seats," Western Political Quarterly, 18 (September, 1965). 164 Kenney, David and William Butts. "Variables Associated with Population Change and Negro Voter Registration in Mississippi," Public Affairs Bulletin, I (January- February, 1968). Knapp, Henry R. "William Andrews Clark," Cosmopolitan, 34 (February, 1903). Masters, Nicholas and Deil Wright. "Trends and Variations in the Two-Party Vote: The Case of Michigan," American Political Science Review, 52 (1958). Moore, Joan W. "Social Deprivation and Advantage as Sources of Political Values," Western Political Quarterly, 15 (june, 1962). Neuberger, Richard L. "Wheeler Of Montana," Harper's Magazine (May, 1940). Ogden, Daniel M., Jr. “A Voting Behavior Approach to Split—Ticket Voting in 1952," Western Political Quarterly, 11 (1958). Payne, Thomas. "Under the COpper Dome: Politics in Montana," in Frank H. Jonas, ed., Western Politics (Salt Lake City, Utah: University of Utah Press, 1961). Pitchell, Robert J. "The Electoral System and Voting Behavior: The Case of California's Cross-Filing," Western Political Quarterly, 12 (1959). Press, Charles. "Voting Statistics and Presidential Coattails," American Political Science Review, 52 (1958). Reutten, Richard T. "Anaconda Journalism: End of an Era," Journalism Quarterly, 37 (1960). Robinson, W. S. "Ecological Correlation and Behavior of Individuals," American Sociological Review, 15 (1950). Roeder, Richard B. "Sound and Fury . . . and One Small Tax Reform," Montana, the Magazine Of Western History (Autumn, 1970). Schlesinger, Joseph A. "The Politics of The Executive," in Herbert Jacob and Kenneth N. Vines, eds. Politics in The American States (Boston, Mass.: Little, Brown and Company, 1965). 165 Schlitz, John. "Montana's Captive Press," Montana Opinion (1956). Simmons, James W. "Voting Behavior and Socio—Economic Characteristics: The Middlesex East Federal Election, 1965," Canadian Journal of Economics & Political Science (August, 1967). Standing, William H. and James A. Robinson. "Inter- Party Competition and Primary Contesting: The Case of Indiana," American Political Science Review, 52 (1958). Toole, K. Ross. "Rebuttal: When is A Reform A Reform?" Montana, the Magazine of Western History (Autumn, ‘1970). Walker, Jack. "Ballot Forms and Voter Fatigue: An Analysis of the Office Block and Party Column Ballote," Midwest Journal of Politics, 4 (November, 1966). ”'11111111111111(Immujwm‘s