University of Southern California Dissertations and Theses

The geography of voter power in the U.S. electoral college from 1900-2012

(USC Thesis Other)

The geography of voter power in the U.S. electoral college from 1900-2012

PDF

Download

Open document

Flip pages

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content THE GEOGRAPHY OF VOTER POWER IN THE U.S. ELECTORAL COLLEGE FROM 1900-2012 by Luke Wenschhof A Thesis Presented to the FACULTY OF THE USC GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree MASTER OF SCIENCE (GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY) May 2014 Copyright 2014 Luke Wenschhof ACKNOWLEDGEMENTS I have to thank Dr. Ren Vasiliev, Dr. David Aagesen, and Dr. Darrell Norris of the Geography department at the State University of New York at Geneseo for starting me on my path in Geography and Geographic Information Systems and for keeping me on course while at Geneseo. I want to thank Paul Schenkel. Paul gave me the chance that I needed in 2002 and was my first boss in the GIS world. I have never worked for someone who can do everything in the place better than those under him the way Paul can. He talks the talk and walks the walk, as they say. I also want to thank Dr. Karen Kemp for inspiring me to pursue voter power as a thesis topic and pointing me towards Dr. Vos. This endeavor could not have been successful without his wide-ranging knowledge and background, creativity, and perseverance. Finally, I need to thank my father. I have always tried to live by his ethic that honesty and hard work will take you a long way in life and he was right, it has. Without those values, and his funding of course, none of this would have ever been possible. Thanks Dave! ii TABLE OF CONTENTS Acknowledgements ii List of Tables v List of Figures vi List of Abbreviations viii Abstract ix Chapter One: Introduction 1 1.1 The Importance of the Electoral College 1 1.2 Controversy and Debate around the Electoral College 2 1.3 Research Questions 4 Chapter Two: Related Work 6 2.1 Debate That Formed the Electoral College 6 2.2 Contemporary Debate Surrounding the Electoral College 8 2.3 Voter Power 9 2.4 Opposing Arguments on the Applicability of Voter Power Measures 12 2.5 Political Geography, Geographic Information Systems, and Voter Power 14 Chapter Three: Methodology 16 3.1 Measuring Voter Power 16 3.2 Data Sources and Units of Analysis 19 3.3 Statistical Analyses of the Rural, Agricultural, and Total Population Map Series 23 3.4 Ordinal Classification 24 3.5 Geographic Information Systems, Cartography, and Map Symbology 25 Chapter Four: Results 29 4.1 Statistical Analyses of the Rural, Agricultural, and Total Population Map Series Results 30 4.2 The Rural Population Map Series 37 4.3 The Agricultural Labor Map Series 41 4.4 The Total Population Map Series 45 4.5 Jim Crow, the Great Migration, and Voter Power 49 4.6 Counterfactual Analysis of Voter Power within the Electoral College and a National Popular Vote 57 Chapter Five: Discussion and Conclusions 63 5.1 Contemporary Support for the Electoral College and Voter Power Analysis 63 5.2 Review of Findings 64 5.3 Future Research in Assessing Voter Power at the County Level 68 5.4 Final Thoughts 68 References 71 iii Appendices 74 Appendix A: POPULATION VARIABLE/VOTER POWER SCATTERPLOT SERIES 74 Appendix B: RURAL POPULATION/VOTER POWER MAP SERIES 103 Appendix C: AGRICULTURAL LABOR POPULATION/VOTER POWER MAP SERIES 132 Appendix D: TOTAL POPULATION/VOTER POWER MAP SERIES 161 Appendix E: VOTER POWER RANKINGS 190 Appendix F: GREAT MIGRATION/JIM CROW SCATTERPLOT SERIES 219 Appendix G: COUNTERFACTUAL ANALYSES VOTER POWER MAP SERIES 226 iv LIST OF TABLES Table 1: SPSS output of the Pearson’s chi-square tests for independence 30 Table 2: SPSS Pearson’s product-moment correlation coefficient output 32 v LIST OF FIGURES Figure 1: Tabular Vermont national election data from website Atlas of U.S. Presidential Elections 20 Figure 2: Typical scatterplot of voter power and percentage of each state’s population classified as rural 34 Figure 3: Typical scatterplot of voter power and percentage of each state’s population engaged in agricultural labor 34 Figure 4: Scatterplot of voter power and total population during the election of 1900 35 Figure 5: Scatterplot of voter power and total population during the election of 1964 36 Figure 6: Scatterplot of voter power and total population during the election of 2012 36 Figure 7: Map showing voter power during the election cycle of 1900 overlain on the percentage of each state’s population classified by the United States Census Bureau as rural 38 Figure 8: Map showing voter power during the election cycle of 2012 overlain on the percentage of each state’s population classified by the United States Census Bureau as rural 40 Figure 9: Map showing voter power during the election cycle of 1900 overlain on the percentage of each state’s population engaged in agricultural labor as determined by the United States Census Bureau 42 Figure 10: Map showing voter power during the election cycle of 2012 overlain on the percentage of each state’s population engaged in agricultural labor as determined by the United States Census Bureau 44 Figure 11: Map showing voter power during the election cycle of 1900 overlain on each state’s population as determined by the United States Census Bureau 46 Figure 12: Map showing voter power during the election cycle of 2012 overlain on each state’s population as determined by the United States Census Bureau 47 Figure 13: Voter turnout and voter power in the continental United States throughout the 20th Century 50 Figure 14: Black population change in the United States during The Great Migration 53 Figure 15: Black population and voter power in the continental United States across the 20th Century 56 Figure 16: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1900 58 Figure 17: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1912 58 Figure 18: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1932 59 Figure 19: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1952 59 Figure 20: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1964 60 vi Figure 21: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1972 60 Figure 22: Depiction of states whose voters historically have a higher voter power measure when voting within the Electoral College than in a national popular vote 66 Figure 23: Depiction of states whose voters have a higher voter power measure in contemporary elections when voting within the Electoral College than in a national popular vote 68 vii LIST OF ABBREVIATIONS CSA Confederate States of America EC Electoral College GIS Geographic Information Systems GW George Washington University INDVP Individual Voter Power IRV Instant Runoff Voting MAUP Modifiable Areal Unit Problem NPV National Popular Vote SVP State Voter Power viii ABSTRACT The Electoral College (EC) has occasioned controversy at several points in its history, most recently in 2000 when George W. Bush was elected without winning the popular vote. One principal historical and contemporary argument in favor of the EC is that it performs a balancing function to lift the power of rural and less populous states. Using Geographic Information Systems (GIS) and the measure of voter power as formulated by Banzhaf (1968), this study puts this argument to an empirical test. It finds that the EC has not functioned to balance the electoral power of voters in urban and populous states with those in rural and less populous states throughout the 20th Century and into the 21st Century. Counterintuitively, by late in the 20th Century it actually enhances the electoral power of the largest and most heavily urbanized states. One partial exception to this finding is that the EC did significantly enhance the power of voters in the South in the decades before the Great Migration took place and civil rights legislation ensured equal voting rights. Analyses in this study uncover the voting rules within the EC that are behind these variations in voter power. The analyses and findings in this study leave a foundation for further study at the county scale that may aid in validating the results here. ix CHAPTER ONE: INTRODUCTION The Electoral College has been controversial since it was born of a compromise at the Constitutional Convention of 1787 (Bowman 2011). In 2000, George W. Bush was elected President without winning the popular vote; in 2004, while winning the popular vote, he would not have been elected, if a few counties in Ohio had gone Democratic. Outcomes such as these in the national elections periodically rekindle debates over the legitimacy of the Electoral College, which ultimately decides the outcome of presidential elections in the United States. Although there are many historical and contemporary arguments for and against the Electoral College, a principal argument is rooted in geography. Advocates argue that without it urbanized and populous states would overwhelm the influence of rural and less populous states in the national elections. This study uses the measure of voter power to focus on how the Electoral College has or has not balanced the power of voters in urban versus rural and populous versus less populous states across a dynamic part of its history, from the start of the 20 th Century up until the election of 2012. 1.1 The Importance of the Electoral College As the 21 st Century opens, the United States is the third most populous nation on Earth and the fourth largest in terms of land area. It also possesses both the world’s largest military (Stockholm International Peace Research Institute 2014) and the world’s largest economy (World Bank 2014). Militarily, politically, technologically, and culturally the United States exerts enormous influence. It is crucial that the people’s choice be reflected in the outcome of the national election. There have been four instances throughout the history of the United States in which the popular vote has been usurped by the calculus of the Electoral College. In the election of 1824, John Adams received 38,149 fewer votes than his opponent Andrew Jackson. This happened twice more in the 19 th Century. In 1876, Rutherford B. Hayes received 254,235 fewer votes than Samuel J. Tilden, and in 1888 Benjamin Harrison received 90,596 fewer votes than Grover Cleveland in 1888. The final instance 1 occurred at the start of the 21 st Century. In the 2000 election 450,520 more voters cast their ballots for Al Gore than did for George W. Bush (Office of the Federal Register n.d.). In these four elections the Electoral College awarded the presidency of the United States to John Adams, Rutherford B. Hayes, Benjamin Harrison, and George W. Bush though each received fewer votes from the citizens of the United States than their opponent did. The historical significance of events in foreign policy that followed George W. Bush’s election in 2000 underscore the importance of the national elections in the United States and cannot be understated: the attacks of September 11 th , 2001 on New York City and Washington, D.C.; the signing into law of The Patriot Act (Department of Justice n.d.); the invasions and occupations of Afghanistan and Iraq; the detentions and abuses at the military prisons in both Guantanamo Bay, Cuba and Abu Ghraib, Iraq (White and Mintz 2004); and the expansion of the extraordinary rendition program in which individuals suspected of terrorism are abducted and transferred for interrogation to countries suspected of torture (American Civil Liberties Union 2005), a practice that continued long after President George W. Bush left office (Savage 2009). 1.2 Controversy and Debate around the Electoral College The Electoral College has undergone various changes since its inception in 1787 and with nearly seven hundred amendment proposals no other aspect of the Constitution has been the subject of as much controversy and debate (Slonim 1986). At first electors were not obliged to vote according to the results of the popular vote in their respective states and, though not required by law, it was originally expected that electors would remain politically independent in order to objectively debate and elect the President and Vice President. George Washington was elected unanimously in the first two elections with the Electoral College working as the Founding Fathers envisioned (Gregg 2011). The advent of political parties, which were not in existence at the time of the Constitutional Convention, soon subverted this though as parties 2 found ways to put electors in place who would reliably vote along partisan lines (Bowman 2011). They did just this in 1800 and a tie of electoral votes resulted. As specified in the Constitution, the House of Representatives took up the matter, and after thirty-six rounds of voting a President was selected. The 12 th Amendment was added which put safeguards in place that Congress hoped would stave off a repeat of this outcome (Cornell University Law School n.d.). Though not required by law, as the years went by electors began casting their votes according to the results of the popular vote in their states and by 1836 South Carolina was the only state to not follow this convention, coming on board only after the American Civil War (Bowman 2011). Interestingly, even by 2012 during the contest between Barack Obama and Mitt Romney only twenty- seven states legally required their electors to vote for a specific candidate (Office of the Federal Register n.d.). And though it has only rarely happened, the electors in the other twenty-four states are free to vote for whomever they choose or to abstain. The other major change occurred in 1961 when the 23 rd Amendment gave the residents of the District of Columbia, which is not a state, three electoral votes (Bowman 2011). There was renewed controversy after the election of 2000 with calls for reform becoming louder. Reform proposals are varied and include the elimination of the Electoral College altogether in favor of a direct nationwide popular vote, instant runoff voting (IRV), and proportional voting. In IRV if no candidate receives at least 50 percent of voters’ first choice on the ballot, the candidate in last place drops out and the results are recalculated with this process repeated until a candidate gets to 50% or more (McKenna 2008). Proportional voting requires that each state’s electoral votes be cast relative to the results of the popular vote within their respective states (Bowman 2011). As all previous attempts at reform of the U.S. Constitution have been defeated, a movement termed the National Popular Vote (NPV) has come about that leaves the Electoral College in place but circumvents it. Under the NPV, each state pledges to 3 cast their electoral votes not according to the results of the popular election within their borders but according to the results of the popular vote nationwide (National Popular Vote n.d.). In the past, it was common to employ indirect and/or weighted elections such as the Electoral College to appoint executives. However, the United States is one of only a handful of democracies in the modern era still using this system (Shelley 2002). 1.3 Research Questions From the very beginning of our nation’s founding in 1787 the issue of representation was paramount. The smaller, less populous states favored measures in which all of the states, irrespective of population, would have the same number of representatives or votes. The larger, more populous states favored weighted measures in which the number of representatives or votes increased with population. The Electoral College is a compromise between these two groups that attempts to ensure that voters from the more populous states will not overwhelm those from the less populous states, while at the same time not grossly over-representing the voter power of citizens of less powerful states. The goal of this proposed study is to examine voter power, a measure of an individual vote’s influence in the outcome of a weighted election, state by state under the Electoral College in the United States from William McKinley’s election victory in 1900 as America continued its transition from a rural, agrarian society to an urbanized, industrial one through to Barack Obama’s re-election in 2012. In doing so, the study tests the empirical underpinnings of the argument that the Electoral College protects less populous or rural states. The main research questions are: • Has the Electoral College kept the voter power of citizens in more populous and urbanized states from overwhelming that of the less populous and rural states? • How has this balance changed as the political landscape of the country has changed over more than one century? 4 • Are there states that wield more or less voter power than others in particular elections and what may be behind these disparities? • How does voter power of voters in particular states in elections across the study period compare with the expected voter power of citizens in these states under a “one person, one vote” scheme? To answer these questions, the study engages in an extensive data analysis and creates maps using GIS software. The study includes two main empirical tests. First, there is a factual historical analysis of voter power relative to the rurality of the states (i.e., where rurality indicates the degree of the rural versus the urban characteristics summed at the state level). Second, there is a counterfactual historical analysis of voter power in the individual states under the one person, one vote standard. The study finds that the Electoral College has never performed the anticipated balancing function and that there is no relationship between rurality and voter power. It also finds that there is no relationship between rurality and whether or not a given state's voters have more or less power in the Electoral College than they would have had under a hypothetical national popular election according to the voter power measure for both types of elections. Counterintuitively, towards the end of the 20th Century and the start of the 21st Century, the Electoral College actually enhances the voter power of voters in large states. 5 CHAPTER TWO: RELATED WORK The Electoral College was crafted at the Constitutional Convention in the late 18 th Century as a compromise between the large (more populous) and small (less populous) states. The debate over this aspect of the Constitution of the United States has continued into contemporary times. Some argue that it is a fair way of electing Presidents and should remain as the founding fathers of the nation envisioned it (Gregg 2011), some say it needs to be reformed (McKenna 2008), and still others call for its elimination (National Popular Vote n.d.). Political scientists, mathematicians, and other researchers have devised a measure of voter power, which gives a means of assessing whether or not the Electoral College really is a balancing factor in the national elections. Though this measure has its critics, it is highly useful for historical analysis of the large time scale employed in this study. This chapter addresses criticisms of voter power measures and explains how these measures can be employed to test the voter power of voters in different states in actual election cycles. This use of the voter power measure builds on the existing body of work picking up where other researchers have left off. 2.1 Debate That Formed the Electoral College As the Constitution of the United States was crafted at the Constitutional Convention in 1787 a point of contention from the beginning which disrupted the Convention nearly to a point of impasse concerned large (more populous) states, small (less populous) states, and the issue of representation. As the Convention progressed, one of the debates that ensued related to representation and the way in which the President and Vice President should be chosen with a principal concern being that through sheer numbers of votes, voters in populous states could overwhelm any choice the less populous states would make (Slonim 1986). Out of this debate came a proposal known as the Virginia Plan, which would have a popularly elected legislature from each state, with a number of representatives in proportion to each state’s population, appoint the executives through a majority vote (Slonim 1986). A second proposal, known as 6 the New Jersey Plan, also proposed that a popularly elected legislature from each state appoint the executives through a majority vote with the difference being that each state was to have one representative with one vote each (Slonim 1986). The small states came out against the Virginia Plan as they feared that through sheer numbers they would be overwhelmed by the large states. As each state irrespective of size would have the same number of votes, the large states came out against the New Jersey Plan as they feared that the smaller states would exert undue influence. Neither proposal was adopted as many at the Convention felt that under either of these arrangements partisanship would be the order of the day, with the President and Vice President catering to those within these bodies with the power to elect them (Bowman 2011). A third proposal, which was also rejected, was a direct, national popular vote to appoint the President and Vice President. Communication in the late 18 th Century was difficult. So much so that a candidate well known in one part of the country would be virtually unknown elsewhere. It was felt that in a national popular vote, states would vote only for “native sons” from within their borders friendly to their own issues and familiar to their populace. Representatives from the smaller states feared that in a direct, popular vote their smaller populations would be overwhelmed by voters from the more populous states, their candidates would never be heard, and their issues never addressed (Slonim 1986). This debate was representative of an issue that overshadowed the entire convention: the smaller states fear that the populous states would overwhelm their interests and the populous states fear that the smaller states would exert undue influence through voting measures where each state had the same number of votes, as in the Senate (Slonim 1986). The issue went to committee on August 31, 1787 and a proposal was crafted in which each state would appoint a number of individuals equivalent to that state’s Senators and Representatives in Congress who were to each cast two votes for president, with the requirement that one be from outside that elector’s own state (Leip 2008). The votes would then be tallied and the candidate with the most 7 votes elected President and the runner up Vice President. These individuals were known as electors and the proposal, though not so termed at the time, was the compromise that became the Electoral College (Bowman 2011). 2.2 Contemporary Debate Surrounding the Electoral College In the contemporary debate over the legitimacy of the Electoral College, both those on the side of reform and those on the side of maintaining the Electoral College posit that it increases the influence that less populous states have on the national election. However, those on the side of keeping the Electoral College say that it is a balancing influence while those for its elimination or reform say that it increases the power of a voter from the less populous states’ influence in the national election to such a level that it violates an underlying conception of political equality, “one person, one vote” (FindLaw 2013). Gary L. Gregg writing in support of the Electoral College in The American Conservative, a magazine founded in 2002 by conservative luminaries Pat Buchanan, Scott McConnell, and Taki Theodoracopulos, states that a simplification of the electoral process in the national election in the United States to a popular vote would “render small states irrelevant, and enthrone urban areas as undisputed kingmakers” (Gregg 2011, 34). He continues saying that the path to the presidency would become “one where smaller states and rural areas could be ignored with impunity” and that “an aspirant might be able to win the presidency by campaigning only in major metropolitan areas” (Gregg 2011, 35). Charles D. Snelling, himself a Republican Party activist and one-time Elector, also wrote in support of the Electoral College in the November 18, 2000 edition of The Morning Call, a paper from the eastern part of Pennsylvania, stating that the Electoral College keeps densely populated areas from dominating elections in the United States (Snelling 2000). As Gregg does in The American Conservative Snelling addresses the elimination of the Electoral College in favor of a popular vote. He asks: “Why 8 would small states, who benefit from this arrangement to this day, want to do this?” (Snelling 2000, A37). Neal Peirce and Lawrence Longley criticized the Electoral College for the two electoral votes that each state receives corresponding to the two Senators representing each state regardless of size. Known as the “Federalism bonus” they claim this distorts the popular vote to the point that voters are unequally represented in the national election due to the disproportionate influence the bonus gives to small states (Longley and Peirce 1981). Randall Adkins and Kent Kirwan, writing neither in support of nor against the Electoral College, observe that in the 2000 presidential election in California, 551,112 residents are represented by each of the state’s fifty-four electors, while in Wyoming three electors represent only 151,196 people each (Adkins and Kirwan 2002). 2.3 Voter Power One way of quantifying a voter’s influence is through measures of voter power. Measurements of voter power are all “based on the…proposition that the purpose of any voting system is to allow each voting member some chance, however small, to affect the decisions that must be made” (Banzhaf 1968). At its most basic level, voter power is defined as the likelihood that a single voter or group of voters, known as a coalition, can sway the results and make up the deciding vote or votes in an election (Gelman and Tuerlinckx 2002). However, not all voting systems are created equal. Each is influenced by the way in which the votes within them are weighted and aggregated into a single outcome (Gelman and Tuerlinckx 2002). In the example of the national elections in the United States, an individual’s vote is weighted through the electoral votes allocated to the state he or she lives in. And there are actually two aggregates and outcomes. The popular vote is aggregated by state boundaries and each state’s electoral votes by the United States itself. Voter power must be calculated with this in mind: an individual’s vote must be decisive in the state in which he or she lives and their state’s electoral votes must be decisive within the Electoral College (Gelman, Katz, and Tuerlinckx 2002). 9 Voter power measures began to be explored by mathematicians and social scientists near the middle of the 20 th Century and were subsequently quoted in various legal settings (Banzhaf 1968). The early, major players were Lionel Penrose and his 1946 The Elementary Statistics of Majority Voting and Lloyd Shapley and Martin Shubik’s A Method for Evaluating the Distribution of Power in a Committee System published in 1954. The seminal study is John Banzhaf’s 1965 article, Weighted Voting Doesn't Work: A Mathematical Analysis (Gelman, Katz, and Tuerlinckx 2002). Penrose, known mainly for his work in genetics but also a notable mathematician, wrote that the mathematics involved around the counting of votes had been little studied and noted in his article that he hoped his work would encourage further exploration of the subject (Penrose 1946). Lloyd Shapley and Martin Shubik, in their article A Method for Evaluating the Distribution of Power in a Committee System, discuss their thoughts on how voting is affected as structural changes are made within legislative bodies and committees saying that “the effect of a revision usually cannot be gauged in advance except in the roughest terms” and that “it can easily happen that the mathematical structure of a voting system conceals a bias in power distribution unsuspected and unintended by the authors of the revision” (Shapley and Shubik 1954, 787). In this article, they offer initial ideas in calculating voter power before structural changes are made in voting coalitions, such as legislatures, so that the fairest means of representation are produced. Similar to the Penrose article, Shapley and Shubik produced what they considered to be an initial offering and hoped to encourage others to explore the topic further. The theorist most often associated with the measure of voter power is John F. Banzhaf III. A professor of law at George Washington University (GW) in Washington, D.C., Banzhaf began his career in the practice of public interest law and now teaches it in the course “Legal Activism” at GW (The George Washington University n.d.). His initial review of weighted voting in 1965, Weighted Voting Doesn't 10 Work: A Mathematical Analysis, looks at two real world examples of weighted voting and why he believes weighted voting systems cannot fairly represent the individuals voting within them. The first example came from the New Jersey Senate in which a resolution was passed to implement a weighted voting plan. Similar to the Electoral College, the plan allocated a number of votes to each of its twenty-one members in proportion to the population of the county that each represented. The plan was never implemented because it was found to not be in compliance with New Jersey state law (Banzhaf 1965). Banzhaf (1965) calculated the voter power of each representative had the plan been implemented by analyzing the two million combinations of legislative votes possible in the New Jersey Senate. He found that the representative of the most populous county, Essex with nineteen votes, could cast the deciding vote over 477,000 times while those from the least populous, Sussex and Cape May with one vote each, only a little over 22,000 times (Banzhaf 1965). A second example of Banzhaf analysis comes from an actual voting system at the Nassau County, New York Board of Supervisors. The representatives of the most populous municipalities in Nassau County receive thirty-one votes each while those from the least populous receive two each. Mathematical analysis of voter power similar to that done in New Jersey reveals that when the Nassau County Board of Supervisors meets the only representatives with the power to effect the outcomes of voting within that body are those from the county’s three most populous municipalities, and thereby the largest number of votes. If any two of the three work in coalition, they are able to pass or defeat any measure and there is nothing the representatives of the other municipalities are able to do about it (Banzhaf 1965). According to the results of these analyses, the smaller counties in New Jersey would have been, and the smaller municipalities in Nassau County are, essentially disenfranchised. Banzhaf’s second article on voter power, 1968’s One Man, 3.312 Votes: A Mathematical Analysis of the Electoral College relates directly to this study. In the mid-1960’s, the Congress of the United States was considering replacing the Electoral College with one of two alternatives. The first was a 11 proportional voting scheme in which each state’s electoral votes would be divided in proportion to the statewide popular vote. The second was a district plan in which the winner of the popular vote in each congressional district would be awarded that district’s electoral vote (Banzhaf 1968). Banzhaf finds through the mathematical analysis of voter power that neither of these alternatives nor the Electoral College in its present form produces balanced voting. Banzhaf’s calculation of voter power is a multi-stage process. This process looks first at the number of possible voting combinations in which an individual, by changing his or her vote, can affect the outcome of their state’s popular vote and thereby which candidate receives their state’s electoral votes (Bahnzaf 1968). Next it considers the number of possible voting combinations in which a state, by changing how it casts its electoral votes, can affect the outcome of the national election. It is this formula that is used throughout this study to calculate voter power at the state level. 2.4 Opposing Arguments on the Applicability of Voter Power Measures The main thrust of the argument that critics of voter power take is that the measures’ assumption of random voting is false (Albert 2003). Geoffrey Garrett and George Tsebelis argue that voter power measures “reduce actors’ preferences and institutional rules of the game to mere probability distributions” and ignore the facts that voters are strategic and that ideological groups vote alike (Garrett and Tsebelis 1999, 305). One of the rules of the game that Garrett and Tsebelis claim voter power indices ignore is that some policy proposals are selected partly on the basis that the outcome of votes on them are guaranteed (Garrett and Tsebelis 1999). Powers (2009) makes a similar argument. Though a proponent of the Banzhaf measure, Powers advocated for a modification of the measure to account for partisanship and the fact that in the United States within the Electoral College some states typically vote for a certain candidate, or party for that matter, irrespective of the candidate (Powers 2009). 12 Andrew Gelman, Jonathan Katz, and Joseph Bafumi offer an additional argument that the measures are a priori measures relying only on the mathematical rules of a voting system and not on past or anticipated future patterns of voting within the system (Gelman, Katz, and Bafumi 2004). Their criticism and those of others, though interesting observations of realities in the political realm, are acknowledged and refuted by proponents of voter power measures principally by reminding critics that neither voting systems nor calculations of voter power were intended to account for the idiosyncrasies of human behavior and electoral politics. Indeed, in a study like this one with a historical length that spans more than a century, the anticipated ideological patterns shift significantly. This would make adding nuance to the measure of voter power on the basis of such patterns tricky to say the least. In regard to voting systems themselves, strong advocates of electoral democracy put individual equality ahead of federalism, arguing that any country’s constitution should create voter power as equal as possible amongst all of its citizens (Felsenthal and Machover 2003). Thus, calculations of voter power should not be crafted solely to ascribe principles of fairness in present ideological arrangements, but instead help to prescribe how best equality should be put into practice for the long term (Felsenthal and Machover 2003). They also acknowledge the a priori aspect of measures of voter power and that they must ignore the individual characteristics of voters along with the candidates and bills that are being contested in the election in question (Felsenthal and Machover 2003). As mathematical formulae have difficulty accounting for intangibles that do not readily translate into statistics, Banzhaf himself admits that calculations of voter power do “not reflect the actual ability of any given voter to affect the outcome of a particular election” (Banzhaf 1968, 810). These intangibles include elements as complex as the influence that political parties have on specific regions and individual voter’s feelings and political leanings to those as simple as voter intimidation, weather, or election result forecasting (Banzhaf 1968). What calculations of voter power do represent however, are 13 an average of a voter's effectiveness in an election by exposing the inequalities that may be present in a voting system due to the rules governing it (Banzhaf 1968). 2.5 Political Geography, Geographic Information Systems, and Voter Power Disciplines such as political science, mathematics, and statistics are well represented in the discussion and debate on voter power and the Electoral College. However, political geography and geographic information science and technology are under-utilized in discussions of voter power and the Electoral College. Much of political geography’s contemporary contributions to the discussion of the Electoral College relate to examinations of campaign spending or the distribution of voting for candidates and parties. Typical of these is a series of articles written for the journal Political Geography after the 2000 presidential election. This series maps the popular vote by county and the electoral votes in that election by state to see where each candidate garnered the most support (Archer 2002). Fred Shelley writing in the same series gives a brief history of the Electoral College along with his thoughts on how its vote aggregation directed campaign spending leading up to the 2000 election, observing that both the Republican and Democratic campaign teams kept their candidates and money in states not known to reliably vote for one party or the other. He notes that both Bush and Gore both made at least six campaign stops in Florida but spent no time in New York, a safely Democratic state, or Texas which votes reliably Republican (Shelley 2002). In addition to observations of voting and campaign spending, political geographers will often discuss the future of the Electoral College in terms of the effects of reform proposals as well as the changing electorate in the United States. In a counterfactual analysis, Ron Johnston, David Rossiter, and Charles Pattie examined the 2000 and 2004 presidential elections in the United States to determine if reform of the Electoral College would have changed the outcomes. The format they looked at, actually used in Maine and Nebraska, gives one-fifth of a state’s electoral votes to the winner of the popular vote 14 within the state and the remainder to the candidate who wins in each of the state’s Congressional districts in favor of the “winner take all” format used in most states which awards all of that state’s electoral votes to the winner of the popular vote within the state (Johnston, Rossiter, and Pattie 2006). In a discussion of the changing electorate, Burmila (2009) uses projections of the 2010 and 2020 censuses along with historic voting behavior, finding that the Republican Party can expect gains in electoral votes due to the changing demographics within the electorate through both immigration as well as migration within the country. Warf examines the effect that the Electoral College has on voting in the national election, embracing the concept of voter power. While at Florida State University he and Cynthia Waddell contributed to the 2002 series of articles in Political Geography mentioned above. Their article examines the 2000 election focusing on the state of Florida and the way in which the Electoral College enables small groups of voters to affect the outcome of the national election (Warf and Waddell 2002). The authors, though focusing on one state in one election, employ John Banzhaf’s voter power measure detailed in his 1968 article One man, 3.312 votes: A mathematical analysis of the electoral college to compare the influence that voters in Florida had in the election with voters in the rest of the nation as a whole (Warf and Waddell 2002). Warf (2009) gives a brief overview of the Electoral College, covers some of the debate around it, defines voter power as it functions within the Electoral College, and introduces his own calculation of voter power. Warf (2009) presents a high level cartographic overview of presidential elections from 1960 to 2004 along with a brief but detailed look at the 2000 and 2004 election cycles. He also introduces methods from various researchers on determining the voter power of individuals based on those individual’s demographic characteristics. 15 CHAPTER THREE: METHODOLOGY The main goal of this study is to determine whether or not weighted voting as historically practiced in the U.S. Electoral College balances voting between the populous and less populous states and to describe what areas (if any) have held higher or lower than expected voter power in U.S. Presidential elections. To accomplish this requires a multi-stage process that is driven by calculations of voter power. Although there are many methods of calculating voter power (described in Chapter 2 above), this study uses the Banzhaf voter power measure with a small modification so that the measure reflects actual voter turnout. The analyses begin by calculating individual voter power within the Electoral College starting with the 1900 election throughout the 20 th Century and into the 21 st Century, ending with the 2012 election. Using statistical analysis and GIS software, voter power within each time period for each state is compared to the percentages of each state’s population classified as rural by the United States Census Bureau, the percentages of each state’s population determined by the United States Census Bureau to be engaged in agricultural labor, and each state’s total population. 3.1 Measuring Voter Power Voting in the national election in the United States is a multi-stage process. In the first stage, individuals vote within their respective states and in the second stage, the states vote within the Electoral College. Banzhaf’s calculation of voter power is a multi-stage process as well, combining the two probabilities to determine the probability of a voter changing the outcome of the national election. Individual Voter Power (INDVP), the number of possible voting combinations in which an individual by changing his or her vote can affect the outcome of their state’s popular vote and thereby which candidate receives their state’s electoral votes, determines the first probability (Banzhaf 1968). The second probability, State Voter Power (SVP), is determined by considering the number of possible voting combinations in which that voter’s state, by changing how it casts its electoral votes, can affect the outcome of the national election. The final calculation multiplies these two probabilities to determine 16 Voter Power (VP), the probability of a voter changing the outcome of the national election in the United States, with the following formula: VP = 𝐼𝑁𝐷 𝑉𝑃 ∗ 𝑆𝑉𝑃 (1) To calculate INDVP, the first probability, Banzhaf used combinatorial analysis through an expression that implements Stirling’s Approximation, used in calculating the factorials of large numbers (Banzhaf 1968). As it is derived from voting within a one person, one vote majority system the voting power of a voter is the probability that their vote is the one that tips the election to one candidate or measure being voted on (Miller 2013). The laws of statistics tell us that this first probability can be closely approximated though the expression: INDVP = �2/ 𝜋𝑛 (2) where n is the number of individuals voting within the election, indicating that, voter power, the probability of a vote changing the outcome of the election, is not inversely proportional to the number of voters, but to the square root of the number of voters (Miller 2013). It is this formula that is used throughout this study to calculate INDVP, the first probability value needed to calculate Banzhaf voter power. It is also this formula that is used in the counterfactual assessment of voter power in this paper that looks at an individual’s voter power in a national popular vote in place of voting within the Electoral College. Banzhaf’s original assessment of voter power uses each state’s population in calculating an individual’s voter power within his or her state. This study attempts to improve on this by using instead the total number of actual votes in each election. Votes that are thrown out and not counted or eligible voters who fail to turn out are not factored into this number. Although each state’s electoral votes are calculated based on its population not every resident in every state votes in the national election. Those in the population that do not vote should not be factored into calculations of voter power as their numbers will tend to dilute the results. In a very simple example that demonstrates this, consider a 17 measure being voted on by a group of ten individuals in a simple one vote per person, majority rules voting scheme. If all ten participate the measure will pass with a majority of six. The above formula tells us that each voter’s power is 0.252310. However, if two out of the ten abstain, then the majority needed for the measure to pass decreases to five while the participating voters’ power increases to 0.282090. To calculate SVP, the second probability, Banzhaf (1968) was assisted by researchers at The Rand Corporation using some of the most powerful computers available at the time to implement the Monte Carlo technique. The Monte Carlo technique is a method of computing the probability of a certain event, in this case a state’s electoral votes changing the result of the national election, occurring in large data sets in which possible outcomes are sampled some number of times and occurrences of the certain event counted. The true probability of the event occurring is then determined by dividing the number of times the certain event is counted by the number of samples taken. For accurate results, all of the possible events need to be tested for and a large number of samples need to be taken (The University of North Carolina 2003). Similar to Banzhaf (1968), this study receives assistance in calculating SVP and also takes advantage of a powerful computer. Not one at The Rand Corporation but via a calculator available online hosted at The University of Warwick which allows for calculation of Banzhaf voting power that members within voting bodies implementing weighted voting possess (The University of Warwick n.d.). In the context of the Electoral College, members are the states themselves. The calculator is able to count all possible voting outcomes equally and derives each of the member’s Banzhaf voter power by dividing the number of times each member is the swing vote, the vote that tips the election to one measure, or political party in this case. The calculator then divides that number by the number of possible outcomes among the other members (The University of Warwick n.d.). 18 Mathematics tells us that the probability of two independent events occurring together can be determined by multiplying the independent probability of each event (The University of North Carolina 2003). VP then, as the probability that an individual’s vote is decisive in the national election, is calculated with formula (1) above and is the product of INDVP, the probability that an individual’s vote is decisive within his or her state’s popular election, and SVP, the probability that the state that individual resides in is decisive in voting within the Electoral College. This product is the Banzhaf voter power of an individual voter in the national election in the United States voting within the Electoral College and is the number that is used as the basis for all analyses within this study. In Banzhaf’s calculation of voter power he combines a voter’s effectiveness within his or her state with the effectiveness of that voter’s state within the Electoral College to determine a voter’s power in voting in the national presidential elections. Using the first calculation from his voter power measure to measures a voter’s effectiveness in the simple majority, one person one vote system in each voter’s state, it is possible to measure voter power in a national popular vote. This study uses this calculation, but in a simple majority, one person one vote system in the nation as a whole, in a counterfactual analysis to examine how voter power when voting within the Electoral College compares with that in one of the most popular contemporary Electoral College reform proposals, a national popular vote. 3.2 Data Sources and Units of Analysis Calculations of voter power in this paper are based on the past work of mathematicians and political scientists. Election data come from the private website the Atlas of U.S. Presidential Elections (Leip 2012). States are required to collect, tabulate, and report their own election results and these data are available through their various state election departments, election boards, and voting divisions. The Atlas of U.S. Presidential Elections, which is thoroughly referenced throughout, has collected these data sets making election results from the first presidential election held in 1789 to the 19 most recent in 2012 available for personal, public, and private use (Leip 2012). Outside of the number of electoral votes a state possesses, the only data necessary to calculate voter power within this paper include simply the number of votes for each political party. The Atlas of U.S. Presidential Elections provides these in tables throughout the time period examined within this paper. A data set typical of those available on the site can be seen below in Figure 1. Figure 1: Tabular Vermont national election result data from website Atlas of U.S. Presidential Elections The United States Census Bureau began assessing urban versus rural populations for the entire nation beginning in 1900. At the start of the 20 th Century, a transition from a rural to an urban nation was already underway with the shift from an agrarian to an industrial society that started in the 19 th Century (Morris 1996). The Census Bureau, for instance, assessed 20% of the population of Florida as urban in 1900 while in 2010 over 91% of its citizens were classified as urban dwellers (United States Census Bureau 2013). As a thorough and reliable assessment of America’s population, it is the Census Bureau’s determination of who is an urban dweller and who is a rural dweller that is used in this paper for analyses of voter power. The criteria the Census Bureau uses to make this determination have changed over the years. Beginning with the 1900 census, an individual was determined to be an urban dweller if they lived 20 within an incorporated place, such as a town or city, with a population of more than 2,500. It was realized though, that this definition left out many densely populated areas for the simple reason that they were not incorporated. In an attempt to overcome this, prior to 1950 certain areas, especially minor civil divisions, were classified as “urban under special rules” with their residents then falling into the urban dweller classification (United States Census Bureau 2013). Despite the “urban under special rules” classification, many densely built up areas were still being excluded from the urban category. As a result, starting with the census of 1950 the bureau began defining urban as any place with 2,500 or more residents, whether they were incorporated or not (United States Census Bureau 2013). Places, as defined by the Census Bureau, are densely populated areas including those with their own government, known as incorporated places; those without their own government, referred to as census designated places; and consolidated cities which are municipal governments that have merged the county and city or cities within that county into one jurisdiction (United States Census Bureau 2013). The definition of a rural dweller has remained the same throughout the history of the bureau’s assessment and is simply someone who does not live in an area designated as urban. While the Census Bureau’s assessment of urban and rural populations is a convenient number for analysis of large numbers of voters, the historical shift in the criteria behind this classification clouds it as a straightforward measure of rurality. The Census Bureau also makes available additional demographic measures including number of persons employed in agricultural labor. A majority of agricultural labor and the industries the Census Bureau places under that classification typically take place within rural areas. Thus, in order to further validate the assessment of voter power and rural population, this study also makes use of the occupational data as a measure of rurality. Unfortunately, as with the measure of rural places, the occupational measure is also somewhat clouded by historical changes in the census classification. Throughout the study period, the Census 21 Bureau has collected counts of agricultural laborers, but it has not always classified agricultural labor or reported its findings in the same way. For example, in the 1900, 1910, and 1920 censuses, the Census Bureau reported on employed persons ten years and older. However, once child labor laws began to be enacted from 1930 forward, the census reported on adult employed persons only. In addition, there are certain industries that are not always included under the classification of agricultural pursuits and are broken out separately. In order to standardize across all of the years, this study modifies the census definition of agricultural labor to include a consistent set of census occupational categories related to rural places. The United States Census is a decennial survey reported at the beginning of each decade. To accommodate for elections that take place during off census years this study uses a strategy of linear interpolation which bases rural population, agricultural labor, and total population counts on differences between the censuses preceding and following a given election. For example, to calculate these demographics for the 1904 election year, the counts for each from the 1900 census were subtracted from the counts for each from the census that took place in 1910. To correspond to the 1904 election year these numbers were then divided by four and added to the numbers from the 1900 census counts. A total of twenty-two of the twenty-nine elections in this study fall on off census years and the process was repeated for each of these election cycles. The 2012 election presents a particularly interesting dilemma as there is no 2020 census data at the time of the writing of this paper. To interpolate the demographic data for 2012, the 2000 census numbers were subtracted from the 2010 counts, which were then divided by ten to get the yearly increase from the start of the decade until the end. These numbers were then multiplied by two and added to the numbers reported in the 2010 census. 22 3.3 Statistical Analyses of the Rural, Agricultural, and Total Population Map Series In order to look for relationships that may be present in the data, a series of statistical analyses were performed. ArcGIS was used to create simple scatterplots and the analysis software SPSS Statistics was used to run Pearson’s chi-square test for independence (chi-square test) and Pearson’s product- moment correlation coefficient (Pearson’s correlation) on each election cycle from 1900 to 2012. Chi- square tests for independence reveal whether or not two phenomena are related and if further statistical exploration should be undertaken (Laerd Statistics 2013a), Pearson’s correlation measures the strength and direction of the relationship between two variables (Laerd Statistics 2013b). Both SPSS tests in this study output P-values which report the probability of the results occurring in the data set being tested, if the null hypothesis is true. A P-value of .05 or less indicates that there is a significant relationship and that one can be 95% certain that the relationship is not due to chance (Laerd Statistics 2013b.). For accurate chi-square test results, data sets that include frequencies of less than five must use the Fisher’s exact test, which this study implements. Typically used on data sets with small sample sizes, Fisher’s exact tests produce exact measures of P-values and not approximations as other chi-square tests do. In addition to P-values, Pearson’s correlations output a value known as the Pearson correlation value that indicates the strength of the relationship. The closer this value is to -1 or 1 the stronger the relationship. A Pearson correlation value represented by a positive value indicates a positive relationship in which the dependent variable goes up as the explanatory variable does the same. The opposite is true as well in a positive relationship with the dependent variable decreasing as the explanatory variable decreases. A negative Pearson correlation value indicates that as the explanatory variable increases in value the dependent variable will also decrease. Or as the explanatory variable goes down in value the dependent variable will go up. 23 3.4 Ordinal Classification ArcGIS was used to create the data sets consumed by SPSS for the chi-square and Pearson’s correlation tests in this study. Voter power, the percentages of each state’s population classified as rural by the United States Census Bureau, the percentages of each state’s population determined by the United States Census Bureau to be engaged in agricultural labor, and each state’s total population were all ranked into five classes. Classes with the highest values were given a value of one, classes with the group of second highest values a two, and so on down to the group with the lowest values. The results of these rankings were then used to run chi-square tests for independence and Pearson’s product- moment correlation coefficient comparing voter power with each population variable separately. While ArcGIS presents multiple classification methods including Jenks (natural breaks), equal interval, and quantiles, after examining the data and actually running it through each classification method, it was determined that of these three options the quantiles classification was the most appropriate method of grouping each variable in this study for the statistical analyses in SPSS. Key to this study is visualizing the changing electorate and the measures of voter power it possesses across the 20 th Century and into the 21 st through various map series. For consistency, the same classification method used in the statistical analyses is also used in symbolizing the data in these map series. Quintiles (i.e., quantiles separated into five ordinal classes) was chosen because it allows for representation of a state’s rank both within and across election cycles. The variables of rural population, agricultural labor, and total population change dramatically in absolute values across the time period this study covers. To be meaningful, the statistics used in this study must capture the relative position of the states to one another in any given election cycle and across the historical sweep. The Jenks (natural breaks) classification groups data into classes with the most similar values so that the differences between the classes are the greatest (Environmental Systems Research Institute 2014a). When Jenks was used to classify the data sets in this study though, it was seen that a majority 24 of the values fell into three or four classes with only one or two values falling into the classes at the high or low ends of the distribution. Not only do groupings such as these skew the results of any statistical analyses done in SPSS, they may also mislead viewers of any maps basing their symbology on it. Important to this study is to fairly visualize voter power and its effects on voting within the Electoral College across the entire time period this study examines. The ranges of values can be so specific between data sets when classifying with Jenks that it is often difficult to compare maps based on these different data sets symbolized with Jenks. Equal interval takes the entire range of attribute values and divides them into groupings with equal ranges (Environmental Systems Research Institute 2014a) while quantiles breaks data into classes with the same number of features or observations in each (Environmental Systems Research Institute 2014a). Both produced fairly similar groupings of data avoiding the clumping seen when using Jenks. It was seen though, that by partitioning the data into ordinal classes consisting of more equal numbers quantiles better display the data sets evenly over the long time period this study covers. This allows for the visualization of the historical map series of each population variable and voter power so important to this study. It is worth noting that the data in this study were tested with all three of the classification schemes above and the central conclusions about key relationships between variables did not change. 3.5 Geographic Information Systems, Cartography, and Map Symbology Unlike earlier studies, this study features a deep historical engagement with political geography by considering how the urban or rural character of various states may be associated with voter power in the Electoral College. Producing a factual and a counterfactual analysis of voting in the national presidential elections in the United States, this study builds on past work by examining elections throughout an expansive period of time. It takes as its basis Banzhaf’s metric of voter power: the theoretical ability of an individual to affect who is elected President and Vice President every four years in the United States. 25 Key to this is GIS and its ability to organize and facilitate the visualization of data in order to tell a story through a series of maps. This carefully crafted series of maps reveals change and stability across elections cycles when it might otherwise be hidden in individual elections. This study maps voter power and three population variables over a period of more than a century, covering a total of twenty-nine election cycles. Three map sets are produced that are viewable as a historical series. These allow the viewer to watch not only the electorate change as well as the cultural and historical contexts in which each election takes place but also the migration of voter power from state to state and region to region by simply flipping quickly through each map series. The software package ArcGIS was key to this study due to its ability to quickly produce and reproduce complex map layouts. Ultimately, the historical map series is put to use in the creation of complex maps displaying up to six inset maps in one clear graphic. These maps illustrate both the historical narrative and counterfactual analysis by focusing on the key transitional points in the political and cultural climate over the course of the study period. GIS contributes to the study by allowing for efficient placement on clear maps of complex measures of the historic ability of a voter to effect the national elections in the United States (voter power) in relation to the changing electorate and nation in which these elections take place. These maps bring these historical facts into a contentious political debate that is ongoing even in contemporary times. In fact, both those in favor of and opposed to reform of the Electoral College argue with assumptions in direct opposition to the findings reported and visualized here. The data necessary for these historical findings can be pulled from graphs and charts, but GIS and its ability to clearly translate the facts and figures contained within them is a key element of this work and ultimately key to communicating the facts to those engaged in this debate. In order to not distort the story visually, it is also important to adhere to sound principles of cartography and in the selection of map symbology. This study needs to be able display multiple data 26 sets measured and reported at the fairly small cartographic scale of U.S. state boundaries (Montello 2001). These include voter power calculations as well as the three population variables this study looks at in relation to voter power. Often times mapping related to political topics and issues employ choropleth maps for their ability to display varying quantities of data, and this study does so as well. Choropleth maps typically employ a single color scheme to represent the distribution of values, where the best practice is to use darker shades to represent higher values with decreasing shades representing lower and lower values (Harvard University Graduate School of Design n.d.). In mapping throughout this study, all three population variables are shown using choropleth symbology. To visualize the comparison of the population variables with voter power, for the purposes of this study another data set measured at the scale of U.S. states, it was decided to use proportional symbols, a technique in which the size of each symbol increases with the values, overlain on the population variable choropleth maps (Environmental Systems Research Institute 2014b). This allows for the clear viewing of both data sets simultaneously. Cartographers symbolizing data sets with proportional symbols run the risk that viewers of the map will have difficulty distinguishing differences between values, particularly if there are many values (Environmental Systems Research Institute 2014b). This study avoids this through the careful selection of the classification scheme of quantiles and five classes, making it intuitive for the map reader to discern the range of values within the voter power data sets. Aggregated data fall victim to two traps: the modifiable areal unit problem (MAUP) and ecological fallacies (Harvard University Graduate School of Design n.d.). The two are closely related in that the means by which the data is aggregated will embellish certain aspects of the data while other aspects are minimized (i.e., the MAUP), while the impressions taken by viewers of maps based on the aggregated data may cause misleading generalizations across spatial scales (i.e., ecological fallacies) 27 (GISC 2013). In this study, these issues may arise in terms of the visualization of rurality, which is aggregated at the level of the individual states and should not be seen as depicting conditions at other scales. Nevertheless, the relevant analytical point with regards to the argument about Electoral College explored here is conditions aggregated at the level of the each state in the U.S. 28 CHAPTER FOUR: RESULTS The overarching story this study’s analysis tells is consistent throughout the 20 th Century: rural states do not have greater than expected voter power in spite of their slightly higher (proportional) representation in the Electoral College. However, prior to the enactment and enforcement of voting rights laws, voters in the Southern (ex-confederate) rural states enjoyed much greater voter power than would otherwise be expected. Also, by the end of the study period, extensive urbanization gives populous states greater voter power than less populous states in spite of their slightly lower (proportional) representation in the Electoral College. In fact, quite counterintuitively, the data show that by the end of the 20 th Century voters in the most populous states enjoy more power than they would under a one person, one vote scheme. The counterfactual analysis is key to revealing these changes in the voter power of citizens in various states. This is true whether the degree of a state’s rurality is measured either as determined by the United States Census Bureau or as the percent of a state’s population employed in agricultural pursuits. Analyses in this study do reveal a relationship between voter power and voter turnout in each state. The following section looks at each of these in the 1900 election cycle when: percentages of population classified by the United States Census Bureau as rural in the United States were highest in the time period covered by this study; percentages of population determined by the Census Bureau to be engaged in agricultural labor were highest; and total population counts in the decennial censuses taken by the Census Bureau were lowest. For comparison and in an effort to contrast contemporary with historic elections, the 2012 election cycle is examined: when the percentage of the population in the United States classified as rural was it its lowest in the time period covered by the study; the percentages of population engaged in agricultural labor were lowest; and total population the highest. This section next focuses on voter power in the Southern United States throughout the 20 th Century and how it was influenced by Jim Crow laws, the Great Migration, and the Civil Rights Movement in the 29 United States. The discussion focuses on the election cycles directly after the turn of the decade following the decennial census when electoral votes are reallocated. It wraps up with a counterfactual examination of voter power within a national popular vote in comparison with that of voting within the Electoral College again within the historical context of Jim Crow, the Great Migration, and the Civil Rights Movement. 4.1 Statistical Analyses of the Rural, Agricultural, and Population Map Series Results The results of the chi-square tests reveal inconsistent relationships between all three population variables and the voter power individuals from each state possess. Over the course of the twenty-nine election cycles in this study, the relationship between voter power/percentage of population engaged in agricultural labor and voter power/percentage of population classed as rural is significant only sporadically and infrequently. Total population though, was significant in every election cycle such that the null hypothesis can be rejected and that there is a relationship, between it and voter power. As population is the determinant of the number of electoral votes each state receives, voter power and total population are not independent variables and as such, this result is expected and not significant. These inconsistent relationships undermine the assertion that the Electoral College balances voting between urbanized, populous states and the rural, less populous states. The results of the chi-square analyses can be seen below in Table 1 (values highlighted in green are P-values low enough to indicate a relationship). Table 1: SPSS output of the Pearson’s chi-square tests for independence Election Cycle Agricultural Labor / Voter Power P-value Rural Population / Voter Power P-value Total Population / Voter Power P-value 1900 0.381 0.006 0.001 1904 0.023 0.030 0.001 1908 0.043 0.567 0.002 1912 0.280 0.590 < .0005 1916 0.009 0.840 0.005 1920 0.393 0.248 0.001 30 1924 0.756 0.230 0.001 1928 0.532 0.125 < .0005 1932 0.415 0.117 < .0005 1936 0.288 0.126 < .0005 1940 0.124 0.293 < .0005 1944 0.746 0.661 0.002 1948 0.844 0.014 < .0005 1952 0.518 0.483 < .0005 1956 0.901 0.441 < .0005 1960 0.419 0.170 < .0005 1964 0.327 0.288 < .0005 1968 0.019 0.501 < .0005 1972 0.004 0.743 < .0005 1976 0.085 0.506 < .0005 1980 0.056 0.560 < .0005 1984 0.145 0.411 < .0005 1988 0.016 0.316 < .0005 1992 0.027 0.963 < .0005 1996 0.047 0.153 < .0005 2000 0.069 0.266 < .0005 2004 0.009 0.132 < .0005 2008 0.125 0.145 < .0005 2012 0.349 0.013 < .0005 It is also possible to compare data for all elections covered in this study simultaneously. Similar to the chi-square tests, the Pearson correlations produced inconsistent findings when comparing all of the agricultural labor, census rural population, and state population variables to voter power values across all of the election cycles in the study. Seen below in Table 2 is test output from the Pearson correlation tests and again, output from variables which reported P-values low enough to indicate a relationship are highlighted in green. Total population in election cycles towards the end of the 20 th Century were the only data sets with Pearson correlation values that indicated relationships of any strength. But again, as population at the decennial census is the determinant of the number of electoral 31 votes each state receives, it is not independent of voter power and as such, the Pearson’s product results are not surprising. The agricultural labor and rural population variables do begin to show a relationship to voter power during this same time period. Although weak, this relationship is negative and was strengthening during the time period in which total population was seeing a strengthening positive relationship to voter power. Table 2: SPSS Pearson’s product-moment correlation coefficient output Election Cycle Agricultural Labor / Voter Power Pearson Correlation Agricultural Labor / Voter Power P- value Rural Population / Voter Power Pearson Correlation Rural Population / Voter Power P-value Total Population / Voter Power Pearson Correlation Total Population / Voter Power P-value 1900 0.257 0.088 0.123 0.421 0.604 < 0.0005 1904 0.317 0.034 0.152 0.319 0.667 < 0.0005 1908 0.287 0.054 0.148 0.326 0.681 < 0.0005 1912 0.226 0.122 0.112 0.450 0.694 < 0.0005 1916 0.198 0.176 0.083 0.575 0.631 < 0.0005 1920 0.185 0.208 0.133 0.369 0.635 < 0.0005 1924 0.146 0.323 0.030 0.841 0.641 < 0.0005 1928 0.217 0.139 0.061 0.681 0.632 < 0.0005 1932 0.089 0.547 (0.044) 0.766 0.734 < 0.0005 1936 0.110 0.459 0.017 0.910 0.718 < 0.0005 1940 0.195 0.184 0.039 0.793 0.714 < 0.0005 1944 0.083 0.581 (0.012) 0.937 0.656 < 0.0005 1948 0.056 0.707 0.015 0.918 0.715 < 0.0005 1952 (0.057) 0.702 (0.067) 0.651 0.770 < 0.0005 1956 (0.113) 0.443 (0.024) 0.872 0.774 < 0.0005 1960 (0.089) 0.538 0.050 0.733 0.693 < 0.0005 1964 (0.211) 0.137 (0.096) 0.503 0.704 < 0.0005 1968 (0.143) 0.315 (0.077) 0.591 0.692 < 0.0005 1972 (0.215) 0.130 (0.234) 0.099 0.797 < 0.0005 1976 (0.310) 0.027 (0.147) 0.304 0.757 < 0.0005 1980 (0.323) 0.021 (0.174) 0.222 0.812 < 0.0005 1984 (0.414) 0.003 (0.302) 0.031 0.867 < 0.0005 1988 (0.389) 0.005 (0.298) 0.034 0.869 < 0.0005 32 1992 (0.375) 0.007 (0.226) 0.110 0.842 < 0.0005 1996 (0.459) 0.001 (0.356) 0.010 0.854 < 0.0005 2000 (0.446) 0.001 (0.392) 0.004 0.844 < 0.0005 2004 (0.486) < 0.0005 (0.457) 0.001 0.831 < 0.0005 2008 (0.346) 0.013 (0.400) 0.004 0.866 < 0.0005 2012 (0.344) 0.013 (0.422) 0.002 0.825 < 0.0005 Scatterplots for elections at key historical intervals also help to depict the SPSS results from above: no real relationship between rural population or percentage of the population engaged in agricultural labor in each state and the voter power of individuals from those states; a weak relationship between total population and voter power throughout the first half of the 20 th Century and into the 1960’s; and a strengthening relationship between total population and voter power from the early 1970’s through the early 21 st Century. Figures 2 and 3 below illustrate the relationships that are consistent between both rurality and agricultural labor and voter power that are consistent throughout the time period this study covers. 33 Figure 2: Typical scatterplot of voter power and percentage of each state’s population classified as rural Figure 3: Typical scatterplot of voter power and percentage of each state’s population engaged in agricultural labor 34 Figure 4 shows the relationship typical of each state’s total population and voter power from the first election cycle of the 20 th Century up until the 1960’s. Figures 5 and 6 show the progressive strengthening of the relationship in the second half of the 20 th Century and into the early part of the 21 st Century. Scatterplots for each population variable from each election cycle in the study can be seen in Appendix A. Figure 4: Scatterplot of voter power and total population during the election of 1900 35 Figure 5: Scatterplot of voter power and total population during the election of 1964 Figure 6: Scatterplot of voter power and total population during the election of 2012 36 4.2 The Rural Population Map Series The rural population map series, all of which are available in Appendix B, shows the results of the voter power analyses overlain on each state’s rural population for each election cycle this study examines: 1900 through 2012. Using the quantiles classification, the maps present the percentage of a state’s population classified as rural by the United States Census Bureau divided into five classes. Voter power was also divided into five classes using quantiles classification as well. The maps symbolize the most rural states with a dark green fill while progressively less rural states use lighter and lighter shades of green down to the least rural which use white. The maps symbolize voter power using graduated symbols. The largest circles are seen on states with the highest measures of voter power for the election cycle shown on the map down to the smallest circles, which are seen on states with the lowest measures of voter power. The rural map series reveals that voter power values vary widely among states from the most rural to the least. In the map that shows the election cycle of 1900, seen below in Figure 7, there are states such as South Carolina that fell into the group of most rural states but also had the highest measure of voter power in 1900 with 4.7 times that of Colorado, the state with the lowest measure of voter power. New York, one of only four states that fell into the category of least rural states in 1900, also fell into the highest voter power category during the election cycle of 1900 with a voter power 3.7 times that of Colorado. 37 Figure 7: Map showing voter power during the election cycle of 1900 overlain on the percentage of each state’s population classified by the United States Census Bureau as rural 38 Also observed in this map series are states like Florida and Nevada. Both of these states begin the 20 th Century as very rural states with both falling into the category of most rural states with 80% and 83% of their respective populations classified as rural in 1900. By the 2012 election cycle, seen below in Figure 8, both fall into the least rural group of states but Florida’s measure of voter power is 1.5 times that of Nevada’s. In the 1900 election cycle the reverse is true. Nevada’s voter power is 1.7 times that of Florida’s though they are both rural states. This illustrates that urban and rural character seem to be little associated with power of a given states’ voters in the Electoral College. 39 Figure 8: Map showing voter power during the election cycle of 2012 overlain on the percentage of each state’s population classified by the United States Census Bureau as rural 40 4.3 The Agricultural Labor Population Map Series The agricultural labor map series, all of which are available in Appendix C, shows the results of the voter power analyses overlain on each state’s agricultural labor population for each election cycle this study examines: 1900 through 2012. Using quantiles classification, the maps present the percentage of each state’s population classified as engaged in agricultural labor pursuits by the United States Census Bureau divided into five classes. Voter power was again divided into five classes using quantiles classification. The maps symbolize states with the highest percentages agricultural laborers with a dark brown fill while those with progressively lower numbers of agricultural workers use lighter and lighter shades of brown down to those with the fewest which use white. The maps symbolize voter power similarly to the rural series using graduated symbols with the largest circles seen on states with the highest measures of voter power down to the smallest circles on states with the lowest measures of voter power. Though the percentage of workers engaged in agricultural labor dropped significantly from the election cycle in 1900 to the one in 2012, similar to the rural map series, voter power values vary widely among the states throughout the five classifications of percentages of agricultural laborers. It can be seen in Figure 9 below that South Carolina, with the highest measure of voter power in 1900, fell into the category of states with the highest percentages of agricultural laborers. North Dakota, which also fell into the category of states with the highest percentages of agricultural laborers, had a voter power 3.86 times lower than South Carolina’s. 41 Figure 9: Map showing voter power during the election cycle of 1900 overlain on the percentage of each state’s population engaged in agricultural labor as determined by the United States Census Bureau 42 The map and data in Figure 10 shows that by 2012 both North Dakota and South Carolina remained in the category of states with the highest percentage of agricultural laborers. While North Dakota remained in the group of states with the lowest measures of voter power South Carolina’s voters went from possessing the highest voter power in the country in 1900 to a measure 4.5 times lower than those living in California, whose voters wielded the highest voter power in the 2012 election cycle. 43 Figure 10: Map showing voter power during the election cycle of 2012 overlain on the percentage of each state’s population engaged in agricultural labor as determined by the United States Census Bureau 44 4.4 The Total Population Map Series The total population map series, all of which are available in Appendix D, shows the results of the voter power analyses overlain on each state’s total population for each election cycle this study examines: 1900 through 2012. Using quantiles classification the maps present each state’s population divided into five classes. As in the rural and agricultural map series, voter power was also divided into five classes using quantiles classification. The maps symbolize the grouping of states with the highest population counts with a dark gray fill while progressively less populous states use lighter and lighter shades of gray down to the least populous which use white. As above, the maps symbolize voter power using graduated symbols with the largest circles seen on states with the highest measures of voter power down to the smallest circles on states with the lowest measures of voter power. Early in the 20 th Century there were states in the lowest population classifications that also possessed some of the highest voter power measures in the nation. In 1900, seen below in Figure 11, Nevada with only 42,335 residents fell into this classification because it was also in the grouping of states with voters with the highest measure of voter power. By the 2012 election cycle, shown in Figure 12, Nevada still had a comparatively low number of residents that put them into the group of states with the second lowest population but the state’s voter power in the national election had dropped significantly. In 2012, California’s voters exhibited almost five times the voter power of those in Nevada. Mississippi, with one of the highest measures of voter power in 1900 and a population count that put them into the classification of states with the second lowest number of residents, in 2012 again found themselves in the group of states with the second lowest number of residents but now with a voter power nearly 5.5 times less than that of California. 45 Figure 11: Map showing voter power during the election cycle of 1900 overlain on each state’s population as determined by the United States Census Bureau 46 Figure 12: Map showing voter power during the election cycle of 2012 overlain on each state’s population as determined by the United States Census Bureau 47 The exploration of voter power shows minimal relationships between rurality, urbanity, and population, but the exploration has revealed a relationship between how many voters participate in an election and voter power. Neither Nevada nor Virginia saw a significant enough increase in population in the census taken in 1900 for either state to receive any additional electoral votes in the election of 1904, but both states saw the voter power of individuals in their states go in very different directions due to the number of voters participating in elections within them. Nevada had only 10,196 votes counted in the election of 1900 and with the minimum number of electoral votes of three, voters in the state had a voter power only 1.34 times less than that of South Carolina with the highest. In the same year, Virginia had twelve electoral votes in an election in which the state had 264,208 voters and a voter power 1.7 times less than that of South Carolina’s. In the election of 1908 though, the number of voters in Nevada more than doubled while the number in Virginia decreased by 127,143. Voter turnout alone dropped the voter power of voters in Nevada to more than 2 times less than that of Mississippi’s, the state with the highest in the 1908 election cycle, and raised those in Virginia so that it was only 1.19 times less than that of those in Mississippi. Delegates at the Constitutional Convention in 1787 were concerned that urbanized and more populous states would electorally overwhelm the more rural and less populous states through sheer numbers. Population does factor into voter power but not as the delegates had envisioned. In the census of 1900, the population of the state of Texas increased enough for the state to be awarded three more electoral votes for the election in 1904 than in 1900. In the election of 1900 Texas, with fifteen electoral votes, had 423,706 votes counted and the ninth highest voter power: 1.72 times less than that of South Carolina with the highest voter power in the nation. But in the next election in 1904 Texas had 189,698 fewer voters than it did in 1900. The increase in Texas’ electoral votes combined with the drop in voter turnout immediately jumped the voter power of Texas voters to fourth highest in the nation. 48 4.5 Jim Crow, the Great Migration, and Voter Power As seen in the maps produced for this study, voter power in national presidential elections in the United States can be placed into a broader historical context. The experience of black Americans throughout the 20 th Century illustrates what can affect an individual’s voter power within the Electoral College. Reconstruction was the process of incorporating the states of the Confederacy and the society within them back into the United States after the American Civil War (Morris 1996). One key demand of the victorious Northern states was that the states write new constitutions and that the newly freed slaves be allowed to vote. This and other aspects of Reconstruction were enforced through a series of laws as well as the military occupation of the Southern states. However, the old South soon found ways of nullifying the requirements of Reconstruction though through both legal and extra-legal means. Legally and as required, beginning in 1890 Southern state governments did rewrite their constitutions or amend their existing ones but also included a series of laws within them that kept almost all black citizens, and poor whites as well, from voting. Extra-legally a concerted campaign took place in which groups such as the Ku Klux Klan used intimidation and outright violence to keep black citizens from voting. The map series in Figure 13 is an examination of voter turnout in relation to voter power beginning with the election of 1900 and shows not only the effect these laws had in the South, but also that these effects have persisted throughout much of the 20 th Century. Historic data on registered voters is not readily available and as a result this map series calculates voter turnout by the total number votes as a percentage of the population. This is by no means an exact measure of voter turnout, but as each state’s electoral votes are determined by each state’s population including those who cannot vote, this map gives a fairly even picture nationally of voter participation in the national elections. The story the data tells nationally is clear. 49 Figure 13: Voter turnout and voter power in the continental United States throughout the 20 th Century. In this map states in red have the HIGHEST percentage of their population turning out to vote. Voter turnout decreases as the colors move from red to orange and finally to yellow in states that have the LOWEST percentage of their population voting. Voter power measures for voters in each state are represented using proportional symbols with voter power increasing as the symbol size increases. Both data sets are classified using quantiles. States symbolized with the crosshatch pattern were territories and had not been given their statehood yet at the time of the election the map depicts. For reference to the historical narrative of this paper, the blue border around the Southern states represents the borders of the ex-Confederate States of America; states within it were part of the CSA and are ex-slave states. 50 The examples of Nevada, Virginia, and Texas in the elections early in the 20 th Century show how voter power is driven by electoral votes and voter turnout. The states of the old confederacy used Jim Crow laws, intimidation, and violence to suppress the ability of black citizens to vote driving down voter turnout thereby increasing the voter power of those who could vote, mostly whites, to such a level that these states consistently held the highest levels of voter power in the nation throughout the first half of 20 th Century. This is readily apparent by examining the election data sets year by year. In 1900, South Carolina had a population of over 1.3 million giving the state nine electoral votes but with only 50,812 votes counted in the presidential election in the state that year voters in South Carolina had the highest voter power in the nation. In the next election cycle though, the state dropped to three on the list as Mississippi took over the top spot and Louisiana the second. After the census in 1900, electoral votes were reapportioned with both Mississippi and Louisiana receiving an additional electoral vote each for the election of 1904. Though the population in both of these states increased between the elections of 1900 and 1904 voter turnout dropped. In 1900, 59,055 and 67,096 votes were counted in Mississippi and Louisiana respectively. Yet, in 1904 less than 58,000 voted in Mississippi and in Louisiana less than 54,000. This low voter turnout combined with the extra electoral vote each received for the election gave voters in those two states the first and second highest measures of voter power in nation. This story is told over and over again in the Southern states throughout the first half of the 20 th Century and can be seen in Appendix E which ranks each state by their voter power and contains each state’s electoral votes, population, votes counted in the election, and voter power values throughout the entire study period. In response to the poor economic and social conditions that existed for them in the Southern United States, beginning early in the 20 th Century until around 1970 black Americans began moving from the South to the Northeast, Midwest, and Western states mostly to large industrialized cities in an 51 historic movement known as the “Great Migration”. Figure 14 below shows the percentage which each state’s black population changed during this time period. 52 Figure 14: Black population change in the United States during The Great Migration 53 As each state’s population determines the number of electoral votes it possesses, whether or not slaves would be counted in each state’s population became part of the debate over representation at the Constitutional Convention in 1787. Though slaves could not vote, to the slave states’ advantage it was ultimately decided slaves would count as three fifths of a person in census population totals increasing the number of electoral votes each of these states possessed. The historical record shows that even after the slaves were freed and given the right to vote, their vote was suppressed, particularly as reconstruction failed and race relations reached a nadir in the early decades of the 20 th Century (Morris 1996). Election data would seem to support this and ex-slave states continued to take electoral advantage of the black citizens within their boundaries. After the American Civil War, former slaves still resident in the South further increased the voter power of individuals voting within states from the now defunct Confederacy as they would now be counted as one person in the census increased from the three fifths they were counted as prior to the American Civil War. The Great Migration and the social advances that black Americans saw through the civil rights movement in America changed the advantage ex-slave states had. These states saw a reduction in their electoral votes as black citizens moved out and their population numbers decreased in the decennial census while the states they moved to saw population increases and a corresponding increase in electoral votes. Laws passed during the civil rights movement made it increasingly difficult to suppress the vote of black Americans. Those that remained in the South after the Great Migration were then able to vote thereby increasing voter turnout and lowering these states’ voter power values. Figure 15 below illustrates the effect the Great Migration had on voter power not only in the states black Americans left but also the states they moved to. California is particularly illustrative of this. From 1900 to 1972 the state saw an over 13,000% increase in its black population and the number of electoral votes it possessed increased from nine to forty-five. During this time period, its voters went from being ranked twenty-first in the nation in voter power during the election of 1900 to first during 54 the election that took place in 1972 and apart from the election of 1980 when it was ranked second, California’s voters have retained the highest voter power in the nation since. While part of this population, electoral vote, and voter power increase can be attributed to immigration of all ethnicities to California, the Great Migration undoubtedly contributed to a large part of this particularly during World War II and after as black Americans flocked to the west coast to work in the burgeoning defense industry there. This map calculates black population for off census years using the same linear interpolation formula used to calculate total population in off census years discussed above in section 3.2. 55 Figure 15: Black population and voter power in the continental United States across the 20 th Century. In this map states in red have the HIGHEST total counts of black Americans within their populations. Black population decreases as the colors move from red to orange and finally to yellow in states that have the LOWEST total counts of black Americans. Voter power measures for voters in each state are represented with proportional symbols with voter power increasing as the symbol size increases. Both data sets are classified using quantiles. States symbolized with the crosshatch pattern were territories and had not been given their statehood yet at the time of the election the map depicts. For reference to the historical narrative of this paper, the blue border around the Southern states represents the borders of the ex-Confederate States of America; states within it were part of the CSA and are ex-slave states. 56 4.6 Counterfactual Analysis of Voter Power within the Electoral College and a National Popular Vote One of the most common discussions around reforming the Electoral College concerns replacing it with a national popular vote. A national popular vote would not be a weighted election, as voting within the Electoral College is, but rather a true one person, one vote scheme in which every voter has the same voter power. Using the first calculation from the Banzhaf voter power calculation which measures a voter’s effectiveness in a simple majority, one person one vote system it is possible to measure voter power in a national popular vote. Figures 16 through 21 show simple scatterplot analyses of voter power within the Electoral College from the time Jim Crow laws were enacted, throughout the Great Migration, and finally until after the Civil Rights movement and voting rights laws were put in place. These graphics use a red line to show where voter power would be in a national popular vote illustrating the difference between the two voting schemes. Red dots represent ex-slave states that had seceded from the United States to join the Confederate States of America (CSA) in the middle of the 19 th Century. Green dots represent states that had never been part of the CSA. 57 Figure 16: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1900 Figure 17: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1912 58 Figure 18: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1932 Figure 19: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1952 59 Figure 20: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1964 Figure 21: Counterfactual scatterplot analyses of voter power within the Electoral College and a national popular vote in the election of 1972 60 It is apparent from even the few scatterplots above that through the legal and extra-legal means by which the votes of black citizens, and poor whites, were suppressed many of the ex-slave states were able to increase their voter power through the first half of the 20 th Century. These states’ voter power was not only greater than what most other states in the nation possessed, but also higher than what it would have been had the election taken place within a national popular vote and not the Electoral College. The ex-slave states began losing voter power as many millions of their black citizens left during the Great Migration taking electoral votes with them. These states also lost voter power as they lost the ability to suppress the vote of the black citizens who remained. Scatterplots displaying the same data as those above are available from 1900 to 1972 throughout the Jim Crow era, the Great Migration, and the Civil Rights movement in Appendix F. The scatterplots above contrast voter power within the Electoral College voting scheme and voter power in a national popular vote with total population during the Jim Crow era, the Great Migration, and the Civil Rights movement. Scatterplot series contrasting the two voting schemes with each population variable including percentage of the population classified as rural, percentage of the population engaged in agricultural labor, and total population for each election cycle covered by this study are available in Appendix A. The story they tell is that through voting within the Electoral College, the voter power of almost all voters in the United States is below what it would be if the national elections were a national popular vote, irrespective of these voters’ classification as rural or urban dwellers. However, towards the end of the 20 th Century and into the 21 st Century a new trend begins to be seen. Through sheer population numbers, and thereby electoral votes, in combination with voter turnout relative to the number of electoral votes, California, Texas, and New York are the only states whose voter power measures in voting within the Electoral College are consistently above what they would be in a national popular vote. This has been the case since the election of 1960 when the last of 61 the Southern states dropped out of this class of states. This is in sharp contrast to the argument that those on either side of the debate over the preservation of the Electoral College make: that it increases the influence that less populous and rural states have in voting for the President and Vice President in the United States. 62 CHAPTER FIVE: DISCUSSION AND CONCLUSIONS There are many observations that can be made related to the historical contexts in which every election cycle this study examines takes place. And though voter power can be placed within each election’s historical context it is important to recognize that, independent of what may have been going on in the country at the time, the analyses of voter power in this study expose that historically not every person’s vote has the same chance of affecting outcomes of the national presidential elections. The analyses in this study expose that the Electoral College has not balanced voting between urban and rural states or more populous and less populous states and that its rules allowed the Southern voters undue influence on the national election through voter suppression. Analysis results indicate that, until the end of the 20 th Century when the two or three states with both the highest total population counts and number of electoral votes begin to overwhelm the rest of the nation, the geographic character of the state, whether that be a rural state, an urbanized state, a populous state, or a less populous state, has little to no effect on their ability to influence the national election in the United States. 5.1 Contemporary Support for the Electoral College and Voter Power Analyses This goal of this study is to evaluate a principal argument for the founding and preservation of the Electoral College: that without it urbanized and populous states would wield greater influence in the national elections than rural and less populous states. Supporters of the Electoral College argue that it has a balancing effect on the national election in the United States and that without it Presidential and Vice Presidential candidates would need to only cater to the urbanized areas and more populous states to be elected. Those who call for its reform claim that it raises the influence that small states have on the national elections to such a level that voters from these states far surpass the “one person, one vote” maxim (FindLaw 2013). Empirical analysis in this study demonstrates that at least since 1900, in practice the Electoral College has never performed the balancing function envisioned by its advocates. 63 Instead, the influence of voters in various states has been magnified beyond the “one person, one vote” standard for a variety of other reasons. A quick look at where Barack Obama and Mitt Romney, the candidates from the two major parties in the election of 2012, spent most of their campaign time in the last six months of the election it is apparent that candidates already cater to voters in particular states. Though Romney spread his stops around the country a little more than Obama, only Ohio, Florida, Virginia, California, New York, Colorado, and Iowa saw double digit numbers of visits from both candidates (The Washington Post 2012). California and New York, both of whose electoral votes reliably go to the Democrats, were primarily visited for the money that can be raised there by both candidates. The other five states on that list were battleground states in the 2012 election meaning it was not a sure thing which way these states were going to vote and thereby which candidate would receive these states’ electoral votes. Due to their low numbers of electoral votes and the fact that they are known to vote solidly Republican or Democratic, fourteen states saw neither of these candidates during this period of the election campaign (The Washington Post 2012). It may be the winner takes all distribution of electoral votes that drives candidates to states where the campaigns believe they have the best chance of tipping the electoral balance. 5.2 Review of Findings Voter power, as formulated by John Banzhaf and the basis for the analyses in this study, is a multi-step calculation that combines the number of possible voting combinations in which an individual can affect the outcome of their state’s popular vote, and thereby which candidate receives their state’s electoral votes, with the number of possible voting combinations his or her state by changing how it casts its electoral votes, can affect the outcome of the national election. The statistical analysis package SPSS was used to look for relationships between voter power and percentage of each state’s population classified as rural, the percentage of each state’s population engaged in agricultural labor, and total 64 population. The Pearson’s chi-square test for independence and Pearson’s product-moment correlation coefficient analyses both indicated the same inconsistent relationship throughout all of the elections this paper studies in the 20 th Century and the first three of the 21 st : voter power and total population begin to show a defined relationship during elections past 1960 and into the 21 st Century, but as the voter power and population variables are not independent of each other this is not unexpected and may not be a significant observation. Simple scatterplot analyses report findings similar to those of the SPSS results and counterfactual analyses also show that most states’ voter power has historically fallen below what it would be had the elections in question been held within a national popular vote rather than the Electoral College. Going further, this counterfactual analysis exposes that the only time the voter power of voters from more rural states or those from less populous states was increased occurred when the Southern states disenfranchised the black and poor white population within their borders. These ex- slave states from the by then dissolved Confederate States of America were able to take advantage of low voter turnout and the electoral votes their population gave them to boost the voter power of their electorate in national elections to levels above what every voter in the nation would have in a national popular vote. By contemporary times and the elections towards the second half of the 20 th Century and into the 21 st , the South had experienced a reduction in their black population while the West, Midwest, and Northeast saw an increase through the Great Migration. With this loss in population, many of these states also saw a reduction in their electoral votes and a corresponding drop in voter power. In addition to these, the Civil Rights Movement and the voter rights laws that it put in place made it increasingly difficult to suppress the vote of those black citizens who had remained in the South through the Great Migration. These combined to increase the number of votes counted in each election further dropping these states’ measures of voter power, Figure 22 illustrates this phenomenon. 65 Figure 22: Depiction of states whose voters historically have a higher voter power measure when voting within the Electoral College than in a national popular vote. In this map voters in red states had HIGHER voter power measures voting within the Electoral College than they would had the election been a national popular vote and voters in green states had LOWER voter power measures voting within the Electoral College than they would had the election been a national popular vote. States symbolized with the crosshatch pattern were territories and had not been given their statehood yet at the time of the election the map depicts. For reference to the historical narrative of this paper, the blue border around the Southern states represents the borders of the ex-Confederate States of America; states within it were part of the CSA and are ex-slave states. 66 At the end of the time period this study covers, seen below in Figure 23, the only states with voter power levels that surpass that of what it would be in a national popular vote are also the states with the two or three highest population totals and thereby the highest numbers of electoral votes as well. Only one of these is a state from the ex-Confederate States of America, Texas. It appears that with the reapportionment of electoral votes occurring only once every ten years at the decennial census that the Electoral College may not be able to keep up with the pace of population change in the contemporary United States of America. Maps comparing voter power within the Electoral College with voter power within a national popular vote are available in Appendix G for each election cycle this study covers. 67 Figure 23: Depiction of states whose voters have a higher voter power measure in contemporary elections when voting within the Electoral College than in a national popular vote. In this map voters in red states had HIGHER voter power measures voting within the Electoral College than they would had the election been a national popular vote and voters in green states had LOWER voter power measures voting within the Electoral College than they would had the election been a national popular vote. For reference to the historical narrative, the blue border around the Southern states represents the borders of the ex-Confederate States of America; states within it were part of the CSA and are ex-slave states. 68 5.3 Future Research in Assessing Voter Power at the County Level Findings in this study are all performed at the state level using the Banzhaf measure of voter power. However, assessing voter power and rurality at the county scale may also be a valid exploration of the relationship between voter power in the Electoral College and the character of places across the U.S. A study of this nature would be feasible from the election of 1960 forward as election results at the county level are available starting with the election of 1960 (Leip 2012). A highly detailed urban/rural classification scheme is available from the United States Department of Agriculture that breaks the United States Census Bureau urban classification into three smaller groups based on population (United States Department of Agriculture 2013). The USDA scheme breaks the Census Bureau’s rural classification into six smaller delineations based on degree of urbanization and their adjacency to metropolitan areas (United States Department of Agriculture 2013). A means of calculating Banzhaf voter power at the county level needs to be formulated for an exploration at the county level such as this. Banzhaf factors in each state’s electoral votes into his calculation of voter power within the Electoral College. A measure of voter power at the county level would need to factor in the county in question’s share of the electoral votes possessed by the state that county is within. This could be accomplished with an equation that determines a county’s electoral votes in proportion to that county’s population in the state. For example, consider a fictional state with a population of 10,000 and ten electoral votes in the Electoral College. A county within that state that has a population of 4,500 would be awarded 4.5 electoral votes while one with 2,000 residents two electoral votes, and so on. 5.4 Final Thoughts Militarily, economically, politically, and culturally the United States has become increasingly powerful over the time period this study covers. It is essential that the voting system electing individuals to the offices within the Executive branch of the United States government be as free of 69 inequality as possible. The Electoral College was created at the founding of the United States with the goal of balancing voting between large, urbanized (more populous) states and smaller, rural (less populous) states. Analyses in this study have revealed though that voting within the Electoral College does not balance voting in the 21 st Century and never did as it was envisioned throughout the 20 th Century. Thus it appears that that on this basis, critics of the Electoral College are correct in calling for its reform. 70 References Adkins, R. and K. Kirwan. 2002. What Role Does the "Federalism Bonus" Play in Presidential Selection? Publius 32 (4): 71-90. Albert, M. 2003. The Voting Power Approach: Measurement without Theory. European Union Politics 4 (3): 351-366. Archer, J. C. 2002. The Geography of an Interminable Election: Bush v. Gore, 2000. Political Geography 21: 71-77. Banzhaf, J. F. 1965. Weighted Voting Doesn't Work: A Mathematical Analysis. Rutgers Law Review 19 (2): 317–343. ———. 1968. One Man, 3.312 Votes: A Mathematical Analysis of the Electoral College. Villanova Law Review 13: 304-332. Bowman, M. 2011. Electoral College. In: Kazin, M., R. Edwards, and A. Rothman, eds. 2011. The Concise Princeton Encyclopedia of American Political History. Princeton University Press: Princeton, pp. 204-207. Burmila, E. 2009. The Electoral College after Census 2010 and 2020: The Political Impact of Population Growth and Redistribution. Perspectives on Politics 7 (4): 837-847. Cornell University Law School. n.d. 12 th Amendment|Constitution |US Law|LII/Legal Information Institute. http://www.law.cornell.edu/constitution/amendmentxii (last accessed 2 June 2014). Department of Justice. n.d. What is the USA Patriot Web. http://www.justice.gov/archive/ll/highlights.htm (last accessed 27 April 2014). Environmental Systems Research Institute. 2014a. ArcGIS Help 10.2 – Classifying numerical fields for graduated symbology. http://resources.arcgis.com/en/help/main/10.2/index.html#//00s50000001r000000 (last accessed 31 March 2014). Environmental Systems Research Institute. 2014b. ArcGIS Help 10.2 – About symbolizing layers to represent quantity. http://resources.arcgis.com/en/help/main/10.2/index.html#//00s500000034000000 (last accessed 2 June 2014). FindLaw. 2013. FindLaw | Cases and Codes. http://caselaw.lp.findlaw.com/scripts/getcase.pl?court=US&vol=372&invol=368 (last accessed 2 November 2013). Garrett, G. and G. Tsebelis. 1999. Why Resist the Temptation to Apply Power Indices to the European Union? Journal of Theoretical Politics 11 (3): 291-308. Gelman, A., J. N. Katz, and F. Tuerlinckx. 2002. The Mathematics and Statistics of Voting Power. Statistical Science 17 (4): 420-435. 71 Gelman, A., J. N. Katz, and J. Bafumi. 2004. Standard Voting Power Indexes Do Not Work: An Empirical Analysis. British Journal of Political Science 34: 657-674. The George Washington University. n.d. GW Law – Faculty Directory. http://www.law.gwu.edu/faculty/profile.aspx?id=1759 (last accessed 27 July 2013). GISC. 2013. Geographic Data Assumptions: MAUP and Ecological Fallacies|GISC. http://giscollective.org/geographic-data-assumptions-maup-and-ecological-fallacies/ (last accessed 2 June 2014). Gregg, G. 2011. Unpopular Vote. The American Conservative December 2011: 33-35. Harvard University Graduate School of Design. n.d. GIS Manual: Mapping with Quantitative Data. http://www.gsd.harvard.edu/gis/manual/normalize/ (last accessed 2 June 2014). Johnston, R., D. Rossiter, and C. Pattie. 2006. Changing the scale and changing the result: Evaluating the impact of an electoral reform on the 2000 and 2004 US Presidential elections. Political Geography 27: 557-569. Laerd Statistics. 2013a. Chi-Square Test for Association using SPSS – Procedure, assumptions, and reporting the output. https://statistics.laerd.com/spss-tutorials/chi-square-test-for-association- using-spss-statistics.php#procedure (last accessed 31 March 2014). Laerd Statistics. 2013b. Pearson’s Product - Moment Correlation in SPSS – Procedure, assumptions, and output using a relevant example. https://statistics.laerd.com/spss-tutorials/pearsons-product- moment-correlation-using-spss-statistics.php (last accessed 31 March 2014). Leip, D. 2008. Atlas of U.S. Presidential Elections: The Electoral College – Origin and History. http://uselectionatlas.org/INFORMATION/INFORMATION/electcollege_history.php (last accessed 14 August 2013). Leip, D. 2012. Atlas of U.S. Presidential Elections. http://uselectionatlas.org (last accessed 10 May 2014). Longley, L. and N. Peirce. 1981. The People's President: The Electoral College in American History and the Direct Popular Alternative. Yale University Press: New Haven, pp. 111-113. McKenna, P. 2008. Vote of no confidence. New Scientist April 12, 2008: 30-33. Miller, Nicolas R. 2013. A Priori Voting Power and the U.S. Electoral College. In Power, Voting, and Voter Power – 30 Years After, eds. Holler, M.J. and H. Nurmi, 411-442. Berlin Heidelberg: Springer- Verlag. Montello, D. R. Scale in Geography. In: Baltes, P. B. and N. J. Smelser, eds. 2001. International Encyclopedia of the Social and Behavioral Sciences. Pergamon Press: Oxford, pp. 13501-13504. Morris, Richard B. 1996. Encyclopedia of American History: Seventh Edition. New York, NY: Harper Collins. National Popular Vote. n.d. National Popular Vote -- Electoral college reform by direct election of the President. http://www.nationalpopularvote.com/pages/explanation.php (last accessed 8 May 2014). 72 Office of the Federal Register. n.d. U.S. Electoral College: Historical Election Results. http://www.archives.gov/federal-register/electoral-college/historical.html (last accessed 30 June 2013). Penrose, L.S. 1946. The elementary statistics of majority voting. Journal of the Royal Statistical Society 109 (1): 53-57. Powers, V. 2009. A Note on Banzhaf Power in the Electoral College in the 2008 U.S. Presidential Election [pdf]. Available at: <http://www.mathcs.emory.edu/~vicki/preprint/power09note.pdf> (last accessed 14 September 2013). Shapley, L.S. and M. Shubik. 1954. A Method for Evaluating the Distribution of Power in a Committee System. American Political Science Review 48: 787–792. Shelley, F. 2002. The Electoral College and the election of 2000. Political Geography 21: 79-83. Snelling, C. 2000. Electoral College was part of Founders’ compromise. The Morning Call 18 November: A37. Stockholm International Peace Research Institute. April 14, 2014. Trends in World Military Expenditure, 2013 [pdf]. Available at: <http://books.sipri.org/product_info?c_product_id=476> (last accessed 7 May 2014). United States Census Bureau. 2013. Urban and Rural Classification – Geography – U.S. Census Bureau. http://www.census.gov/geo/reference/urban-rural.html (last accessed 27 July 2013). United States Department of Agriculture. 2013. USDA ERS – Rural-Urban Continuum Codes. http://www.ers.usda.gov/data-products/rural-urban-continuum-codes/.aspx (last accessed on 11 May 2014). The University of North Carolina. April 14, 2003. Banzhaf Power Index. http://www.cs.unc.edu/~livingst/Banzhaf/ (last accessed 3 April 2014). The University of Warwick. n.d. Algorithms for Computing Power Indices. http://homepages.warwick.ac.uk/~ecaae/ (last accessed 14 September 2013). Warf, B. 2009. The U.S. Electoral College and Spatial Biases in Voter Power. Annals of the Association of American Geographers 99 (1): 184–204. Warf, B. and C. Waddell. 2002. Florida in the 2000 presidential election: historical precedents and contemporary landscapes. Political Geography 21: 85-90. The Washington Post. 2012. 2012 Presidential campaign stops: See where Obama and Romney are going. http://www.washingtonpost.com/wp-srv/special/politics/2012-presidential-campaign- visits/ (last accessed 28 April 2014). World Bank. 2014. GDP (current US$) | Data | Table. http://data.worldbank.org/indicator/NY.GDP.MKTP.CD/ (last accessed 7 May 2014). 73 APPENDIX A: POPULATION VARIABLE/VOTER POWER SCATTERPLOT SERIES 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 APPENDIX B: RURAL POPULATION/VOTER POWER MAP SERIES 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 APPENDIX C: AGRICULTURAL LABOR POPULATION/VOTER POWER MAP SERIES 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 APPENDIX D: TOTAL POPULATION/VOTER POWER MAP SERIES 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 APPENDIX E: VOTER POWER RANKINGS * next to the state name indicates a state whose Banzhaf Voter Power was higher in the corresponding election cycle voting within the Electoral College than it would be had the election been a national popular vote 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 APPENDIX F: GREAT MIGRATION/JIM CROW SCATTERPLOT SERIES 219 220 221 222 223 224 225 APPENDIX G: COUNTERFACTUAL ANALYSES VOTER POWER MAP SERIES In these maps voters in red states had HIGHER voter power measures voting within the Electoral College than they would had the election been a national popular vote and voters in green states had LOWER voter power measures voting within the Electoral College than they would had the election been a national popular vote. States symbolized with the crosshatch pattern were territories and had not been given their statehood yet at the time of the election the map depicts. Also, for reference to the historical narrative of this study the blue border around the Southern states represents the borders of the ex-Confederate States of America; states within it were part of the CSA and are ex-slave states. 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240

Abstract (if available)

Abstract The Electoral College (EC) has occasioned controversy at several points in its history, most recently in 2000 when George W. Bush was elected without winning the popular vote. One principal historical and contemporary argument in favor of the EC is that it performs a balancing function to lift the power of rural and less populous states. Using Geographic Information Systems (GIS) and the measure of voter power as formulated by Banzhaf (1968), this study puts this argument to an empirical test. It finds that the EC has not functioned to balance the electoral power of voters in urban and populous states with those in rural and less populous states throughout the 20th Century and into the 21st Century. Counterintuitively, by late in the 20th Century it actually enhances the electoral power of the largest and most heavily urbanized states. One partial exception to this finding is that the EC did significantly enhance the power of voters in the South in the decades before the Great Migration took place and civil rights legislation ensured equal voting rights. Analyses in this study uncover the voting rules within the EC that are behind these variations in voter power. The analyses and findings in this study leave a foundation for further study at the county scale that may aid in validating the results here.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Investigating electoral college reform: geography's impact on elections, and how maps influence our perception of election outcomes

PDF

Wake County District Overlay: an online electoral data visualization application

PDF

Is the likelihood of waterfowl presence greater on conserved lands?

PDF

Evaluating transit and driving disaggregated commutes through GTFS in ArcGIS

PDF

Modeling geopolitics in Tikal through least cost paths

PDF

Modeling nitrate contamination of groundwater in Mountain Home, Idaho using the DRASTIC method

PDF

Measuring seasonal variation in food access: a case study of Everett, Washington

PDF

Determining the utility of GIS in border disputes case study: Sudan and South Sudan

PDF

Finding food deserts: a study of food access measures in the Phoenix-Mesa urban area

PDF

Drawing better lines: comparing commissions to legislatures on compactness and coterminosity

PDF

Using Maxent to model the distribution of prehistoric agricultural features in a portion of the Hōkūli‘a subdivision in Kona, Hawai‘i

PDF

Closed landfills to solar energy power plants: estimating the solar potential of closed landfills in California

PDF

Finding environmental opportunities for early sea crossings: an agent-based model of Middle to Late Pleistocene Mediterranean coastal migration

PDF

Mapping firing ranges as social capital generators in Houston, Texas

PDF

Evaluating spatial changes in the rate of insurgency‐violence in Central Africa: the Lord's Resistance Army 2008-2012

PDF

Exploring urban change using historical maps: the industrialization of Long Island City (LIC), New York

PDF

The role of GIS in asset management: integration at the Otay Water District

PDF

Using pattern oriented modeling to design and validate spatial models: a case study in agent-based modeling

PDF

Water Rights Permit System (WRPS): a GIS‐based tool for the Umpqua Drainage Basin

PDF

Redefining urban food systems to identify optimal rooftop community garden locations: a site suitability analysis in Seattle, Washington

Asset Metadata

Creator Wenschhof, Luke (author)

Core Title The geography of voter power in the U.S. electoral college from 1900-2012

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 08/08/2014

Defense Date 05/29/2014

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Civil rights,electoral college,Geographic Information Science and Technology,geographic information systems,GIS,GIST,Great Migration,Jim Crow,John Banzhaf,OAI-PMH Harvest,presidential election,voter power

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Vos, Robert O. (committee chair), Kemp, Karen K. (committee member), Warshawsky, Daniel N. (committee member)

Creator Email lukewenschhof@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-458688

Unique identifier UC11287519

Identifier etd-WenschhofL-2793.pdf (filename),usctheses-c3-458688 (legacy record id)

Legacy Identifier etd-WenschhofL-2793.pdf

Dmrecord 458688

Document Type Thesis

Format application/pdf (imt)

Rights Wenschhof, Luke

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA