<%BANNER%>

Voter Turnout

Permanent Link: http://ufdc.ufl.edu/UFE0045518/00001

Material Information

Title: Voter Turnout the Other Civil-Military Gap
Physical Description: 1 online resource (113 p.)
Language: english
Creator: Valentin, Sarah E
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2013

Subjects

Subjects / Keywords: absentee -- civil-military -- costs -- disenfranchisement -- participation -- politics -- resource -- turnout -- veterans -- voter
Political Science -- Dissertations, Academic -- UF
Genre: Political Science thesis, M.A.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The electoral chaos that erupted following the 2000 presidential contest and a decade of sustained warfare have ushered in a new era of research about the political attitudes and activity of military members.  Much of this work has focused on a) the potential disenfranchisement of service members or b) the so-called civil-military gap.  Very little emphasis has been placed on why those in uniform bother to vote at all, or the fact that they appear to do so at higher rates than their civilian counterparts.  This thesis seeks to fill a hole in the literature by placing recent survey data in a theoretical context.  Two sets of companion studies conducted by the NAES (2004) and Pew Research Center (2011) allow for direct comparisons between civilians, veterans and active duty members.  Using these data, it is possible to assess how well prevailing theories of electoral behavior account for elevated participation rates among service members, as well as whether specific military factors might be responsible.  Overall, the results suggest that there are substantive differences between civilians, veterans and active duty members, especially regarding the influence of socio-economic, demographic, and political identity variables.  However, no such discrepancies were detected for the effects of political resources such as interest, attention and efficacy.  Finally, service-related variables such as rank, combat experience, military responsiveness to needs, and draftee status do not appear to exert significant influence on the electoral participation of current or former members.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Sarah E Valentin.
Thesis: Thesis (M.A.)--University of Florida, 2013.
Local: Adviser: Wald, Kenneth D.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2013
System ID: UFE0045518:00001

Permanent Link: http://ufdc.ufl.edu/UFE0045518/00001

Material Information

Title: Voter Turnout the Other Civil-Military Gap
Physical Description: 1 online resource (113 p.)
Language: english
Creator: Valentin, Sarah E
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2013

Subjects

Subjects / Keywords: absentee -- civil-military -- costs -- disenfranchisement -- participation -- politics -- resource -- turnout -- veterans -- voter
Political Science -- Dissertations, Academic -- UF
Genre: Political Science thesis, M.A.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: The electoral chaos that erupted following the 2000 presidential contest and a decade of sustained warfare have ushered in a new era of research about the political attitudes and activity of military members.  Much of this work has focused on a) the potential disenfranchisement of service members or b) the so-called civil-military gap.  Very little emphasis has been placed on why those in uniform bother to vote at all, or the fact that they appear to do so at higher rates than their civilian counterparts.  This thesis seeks to fill a hole in the literature by placing recent survey data in a theoretical context.  Two sets of companion studies conducted by the NAES (2004) and Pew Research Center (2011) allow for direct comparisons between civilians, veterans and active duty members.  Using these data, it is possible to assess how well prevailing theories of electoral behavior account for elevated participation rates among service members, as well as whether specific military factors might be responsible.  Overall, the results suggest that there are substantive differences between civilians, veterans and active duty members, especially regarding the influence of socio-economic, demographic, and political identity variables.  However, no such discrepancies were detected for the effects of political resources such as interest, attention and efficacy.  Finally, service-related variables such as rank, combat experience, military responsiveness to needs, and draftee status do not appear to exert significant influence on the electoral participation of current or former members.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Sarah E Valentin.
Thesis: Thesis (M.A.)--University of Florida, 2013.
Local: Adviser: Wald, Kenneth D.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2013
System ID: UFE0045518:00001


This item has the following downloads:


Full Text

PAGE 1

1 VOTER TURNOUT: THE OTHER CIVIL MILITARY GAP By SARAH EGERTON VALENTIN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS UNIVERSITY OF FLORIDA 2013

PAGE 2

2 2013 Sarah Egerton Valentin

PAGE 3

3 To my husband, whose service inspires me and whose love sustains me

PAGE 4

4 ACKNOWLEDGMENTS I am deeply grateful to Dr. Ken Wald for shepherding me through this process for helping me develop the skills necessary to undertake this project and for serving on my committee. Finally, I want to thank my parents, who raised me to love learning and the Gators.

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 ABSTRACT ................................ ................................ ................................ ................... 10 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 12 2 LITERATURE REVIEW ................................ ................................ .......................... 15 Rational Choice Theory ................................ ................................ .......................... 15 Socio Economic Status Models ................................ ................................ .............. 16 Resource Theories ................................ ................................ ................................ 18 Mobilization Theories ................................ ................................ .............................. 23 Military Specific Literature ................................ ................................ ....................... 23 3 DATA AND METHODOLOGY ................................ ................................ ................ 26 NAES Surveys ................................ ................................ ................................ ........ 26 Dependent Variable ................................ ................................ .......................... 28 Independent Variables ................................ ................................ ..................... 28 Pew Surveys ................................ ................................ ................................ ........... 31 Dependent Variable ................................ ................................ .......................... 32 Independent Variables ................................ ................................ ..................... 32 4 RESULTS FROM 2004 NAES DATA ................................ ................................ ...... 35 NAES Merged Samples ................................ ................................ .......................... 35 NAES Panel Sample ................................ ................................ ............................... 43 NAES Active Duty Sample ................................ ................................ ...................... 48 5 RESULTS FROM 2011 PEW DATA ................................ ................................ ....... 62 Pew Merged Samples ................................ ................................ ............................. 62 Pew Veteran Sample ................................ ................................ .............................. 65 6 DISCUSSION AND CONCLUSION ................................ ................................ ........ 76 APPENDIX A VARIABLE CODING RULES ................................ ................................ .................. 85 D ependent Variables ................................ ................................ .............................. 85

PAGE 6

6 Socio Economic and Demographic Independent Variables ................................ .... 85 Political Independent Variables ................................ ................................ ............... 86 Military Specific Independent Variables ................................ ................................ .. 88 B DESCRIPTIVE STATISTICS FOR MODELS ................................ .......................... 90 LIST OF REFERENCES ................................ ................................ ............................. 105 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 113

PAGE 7

7 LIST OF TABLES Table page 4 1 Frequencies for Voter Turnout in 2000: Civilians, Veterans, and Active Duty Members (NAES) ................................ ................................ ............................... 51 4 2 One Way ANOVA Test of Means for Turnout and Demographics: Civilians, Veterans and Acti ve Duty (NAES) ................................ ................................ ...... 51 4 3 Effects of Demographic Variables on Voter Turnout: Civilians, Veterans, and Active Duty (NAES) ................................ ................................ ............................ 52 4 4 Effe cts of Attention to Political News on Voter Turnout: Civilians, Veterans, and Active Duty (NAES) ................................ ................................ ..................... 53 4 5 Effects of Interest in Campaign on Voter Turnout: Civilians, Veterans, and Active Duty ( NAES) ................................ ................................ ............................ 54 4 6 Effects of Trust in Government on Voter Turnout: Civilians, Veterans, and Active Duty (NAES) ................................ ................................ ............................ 55 4 7 Effects of Demo graphics on Voter Turnout: Civilians and Military (NAES) ......... 56 4 8 Effects of Political Interest and Internal Efficacy on Voter Turnout: Civilians and Military (NAES) ................................ ................................ ............................ 57 4 9 Effects of External Efficacy on Voter Turnout: Civilians and Military (NAES) ..... 58 4 10 Effects of Civic Duty on Voter Turnout: Civilians and Military ( NAES) ................ 59 4 11 Effects of Government Employment on Voter Turnout: Civilians and Veterans (NAES) ................................ ................................ ................................ ............... 60 4 12 Effects of Demographics o n Voter Turnout: Active Duty Military (NAES) ........... 61 4 13 Effects of Military Service on Voter Turnout: Active Duty Military (NAES) .......... 61 5 1 One Way ANOVA Test of Means for Registration and Demographics: Civilians and Veterans (Pew) ................................ ................................ .............. 70 5 2 Effects of Demographics on Voter Registration: Civilians and Veterans (Pew) .. 71 5 3 Effects of Patriotism on Voter Registration: Civilians and Veterans (Pew) ......... 72 5 4 Effects of Government Employment on Voter Registrat ion: Civilians and Veterans (PEW) ................................ ................................ ................................ .. 73 5 5 Effects of Demographics on Voter Registration: Veterans (Pew) ....................... 74

PAGE 8

8 5 6 Effects of Militar y Service on Voter Registration: Veterans (Pew) ...................... 74 5 7 Effects of Draft versus Volunteer Status on Voter Registration: Veterans (Pew) ................................ ................................ ................................ .................. 75 5 8 Effects of Military Response to Needs During Service on Registration: Veterans (Pew) ................................ ................................ ................................ ... 75 6 1 Significance of Variables: Number and Percentage of Models ........................... 83 6 2 Significance of Service Variables ................................ ................................ ....... 84 B 1 Descriptive Statistics for Demographics Model: Civilians, Veterans, and Active Duty (NAES) ................................ ................................ ............................ 90 B 2 Descriptive Statistics for Attention to Politics: Civilians, Veterans, and Active Duty (NAES) ................................ ................................ ................................ ....... 91 B 3 Descriptive Statistics for Political Interest Model: Civilian, Veterans, and Active Duty (NAES) ................................ ................................ ............................ 92 B 4 Descriptive Statistics for Trust in Government Model: Civilians, Veterans, and Active Duty (NAES) ................................ ................................ ............................ 93 B 5 Descriptive Statistics for Demographic Model: Civilians and Military (NAES) ..... 94 B 6 Descriptive Statistics for Political Interest and Internal Effica cy: Civilians and Military (NAES) ................................ ................................ ................................ ... 95 B 7 Descriptive Statistics for External Efficacy Model: Civilians and Military (NAES) ................................ ................................ ................................ ............... 96 B 8 Descriptive Statistics for Civic Duty Model: Civilians and Military (NAES) .......... 97 B 9 Descriptive Statistics for Government Employment: Civilians and Veterans (NAES) ................................ ................................ ................................ ............... 98 B 10 Descriptive Statistics for Demographic Model: Active Duty (NAES) ................... 99 B 11 Descriptive Statistics for Military Service Model: Active Duty (NAES) ................ 99 B 12 Descriptive Statistics for Demographic Model: Civilians and Veterans (PEW) 100 B 13 Descriptive Statistics for Patriotism Mod el: Civilians and Veterans (Pew) ........ 101 B 14 Descriptive Statistics for Government Employment Model: Civilians and Veterans (Pew) ................................ ................................ ................................ 102 B 15 Descriptive Statistics for Demographic Model: Veterans (Pew) ........................ 103

PAGE 9

9 B 16 Descriptive Statistics for Service Model: Veterans (Pew) ................................ 103 B 17 Descriptive Statistics for Draftee or Volunteer Model: Veterans (Pew) ............. 104 B 18 Descriptive Statistics for Military Met Needs During Service Model: Veterans (Pew) ................................ ................................ ................................ ................ 104

PAGE 10

10 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Arts VOTER TURNOUT: THE OTHER CIVIL MILITARY GAP By S arah Egerton Valentin May 2013 Chair: Kenneth D. Wald Major: Political Science The electoral chaos that erupted following the 2000 presidential contest and a decade of sustained warfare have ushered in a new era of research about the political attitudes and activity of military members. Much of this work has focused on a) the potential disenfranchisement of service members or b) the so called civil military gap. Very little emphasis has been placed on why those in uniform bother to vote at all, or the fact that they appear to do so at higher rates than their civilian counterparts. This thesis seeks to fill a hole in the literature by placing recent survey data in a theoretical context. Two sets of companion studies conducted by the NAES (2004) and Pew Research Center (2011) allow for direct comparisons between civilians, veterans and active duty members. Using these data, it is possible to assess how well prevailing theories of electoral behavior account for elevated participation rates among service members, as well as whether specific military factors might be responsible. Overall, the results suggest that there are substantive differences between civilians, veterans and active duty members, especially regarding the influence of socio economic, dem ographic, and political identity variables. However, no such discrepancies were detected for the effects of political resources such as interest, attention and efficacy.

PAGE 11

11 Finally, service related variables such as rank, combat experience, military respons iveness to needs, and draftee status do not appear to exert significant influence on the electoral participation of current or former members.

PAGE 12

12 CHAPTER 1 INTRODUCTION Scholars, journalists and politicians have d evoted increasing attention to uniformed vote rs in recent years, much of which has focused on two main areas. Concerns about military disenfranchisement are hardly a new phenomenon, but in the wake of the 2000 presidential debacle in Florida, they have risen to the fore of political discourse during every election cycle. News outlets, candidates, government entities, and non profit organizations routinely highlight the unique hardships service members face when voting and assess whether we are doing enough to alleviate them A few authors have ques tioned whether such efforts go too far (Mazur 2007, McNamara 2006 ), state and federal laws that govern the voting process especially by absentee. Accurate or not, the notion th at active duty voters encounter additional obstacles at the ballot box is pervasive and much discussed. The establishment of an all volunteer force (AVF) and a decade of sustained warfare have also generated renewed interest in the political attitudes and behavior of our men and women in uniform. Numerous studies have sought to validate or discredit the perception of a growing partisan/ideological gap between the military and civilians, often exploring its potential consequences for government, society and force effectiveness (Huntington 1957, Janowitz 1960, Szayna et al. 2007, Feaver & Kohn 2000 and 2001, Ricks 1997, Sarkesian & Conn or 2006, to name a few). Other scholars have attempted to measure and explain attitudinal differences between the two p opulations, specifically distinguishing between self selection and socialization effects

PAGE 13

13 (Jennings & Markus 1977, Schreiber 1979, Teigen 2007, Franke 2001, Wald & Feinstein 2012, Bachman et al. 2000 ). G iven the vast amount of research regarding how milit ary personnel vote ( both procedurally and as partisans), it is perplexing that there is so little about why they vote, especially considering they generally do so in greater percentages than civilians ( Betros 2001; Inbody 2009 ; U.S. Department of Commerce, Bureau of the Census 2000 2010 ) This thesis will attempt to fill a gap in the literature by examining military and civilian electoral participation in a theoretical context. Turnout was proportionally higher among service members in every presid ential general election between 1984 and 2004, sometimes by as muc h as fifteen percent (Inbody 2009), and there is evidence of a similar trend in the primaries from 2000 and 2004 (Teigen 2006). Despite the consistent and sometimes striking differences bet ween uniformed voters and civilians, many scholars simply mention it in passing. (Betros 2001), but fail to note that s imilar programs have had little impact on civilian turnout ra tes (Giammo and Brox 2010, Neeley & Richardson 2001, Fitzgerald 2005, Rabon 2006) I contend that understanding military participation, identifying which characteristics (if any) set this population apart, can make important contributions to a variety of fields Most obviously, it should be of great interest to the plethora of scholars investigating why turnout declined steadily after the 1960s and remains relatively low (Converse 1972, Boyd 1981, Abramson, Aldrich & Rohde 2010, Rosenstone & Hansen 1993, Miller & Shanks 1996). In addition to testing theories of political participation, there are implications for those regarding collective identity, personality, socialization behavior,

PAGE 14

14 and attitude formation. Candidates and campaigns that want to mobiliz e uniformed voters, and the elected officials who call them constituents, also have good reason to explore this phenomenon. Finally, it should be pertinent to military, legal and policy analysts, as well as students of democracy and civil military relatio ns. This thesis will endeavor to answer two primary research questions. First, how well do existing theories of voting behavior account for elevated participation rates can accoun t for the turnout differential ? Chapter 2 will provide an overview of the literature regarding voter turnout and military political participation. Chapter 3 discusses the data and methodology that were used for this study. Chapter 4 presents the regress ion results from the NAES data, while Chapter 5 does the same for the Pew surveys. Chapter 6 discusses my overall findings and conclusions.

PAGE 15

15 CHAPTER 2 LITERATURE REVIEW This chapter will provide an overview of existing literature regarding voter behavio r, as well as relevant findings pertaining to the political attitudes and participation of military members. Robert Lane defines the basic requirements for decide amo ng conflicting claims, sufficient relatedness to society for awareness of the election, capacity to implement emotional dispositions, an affirmative view of the century, ref lects much of the contemporary thinking about voting behavior. I will briefly summarize rational choice, socio economic, resource and mobilization theories and then introduce some of the most prominent studies of military personnel. Rational Choice Theo ry cost benefit analysis when deciding whether or not to vote. Costs are typically defined as the time and effort required to register, gather information, and cast a ballot, as well as the forfeiture of benefits that might be gained by doing something other than voting (Downs 1957, Riker & Ordeshook 1968, Wolfinger & Rosenstone 1980). A strict are linked directly to the outcome of the election (Downs 1957, Blais 2000, Meehl 1977). Since the probability that a single ballot will determine the electoral balance or the future of democracy is miniscule, it is generally irrational for a citizen to vote. And yet many of them do, including a disproportionate number of service members. Given the additional administrative obstacles they encounter (Wright 2008, Eversole 2010, Federal Voting

PAGE 16

16 Assistance Program 2011 ), this theory would seem to predict lo wer turnout among active duty members and little or no difference between veterans and civilians; neither is Shapiro 1994; Niemi 1976; Verba, Schlozman & Brady 199 5) and expansion (Blais 2000, Riker & Ordeshook 1968, Jackman 1993) of rational choice theory. Socio Economic Status Models Although some authors have suggested the costs of voting are minimal (N i e mi 1976, Aldrich 1993), scores of others have noted that the burdens are unequally distributed across socio economic lines. Age, income, education, race and gender are among the most consistent and widely accepted correlates of voting, which makes intuitive sense from a rational choice perspective. Conve rse once proclaimed that credit schooling with the cultivation of political interest and efficacy, as well as civic knowledge, skills and values (Rosenstone & Hansen 1993 Wolfinger & Rosenstone 1980, Verba et al. 1993). In addition to material wealth and the ability to meet opportunity costs, income often connotes status within and a connection to society (Lane 1959; Rosenstone & Hansen 1993; Walsh, Jennings & Stoker 200 4). Both education and income are positively correlated to voting and, more often than not, one another. They are al so associated with age and race, which considering the relationships described above, helps explain why youths and minorities are less lik ely to participate (Milbrath & Goel 1977, Calvo & Rosenstone 1989, Verba et al.1993,

PAGE 17

17 Converse & N i e mi 1971, Abramson, Aldrich & Rohde 2010.). 1 Finally, although earlier studies generally (and unsurprisingly) found that men were more likely to vote (Lane 1959, Campbell et al. 1960, Wolfinger & Rosenstone 1980), that trend has reversed as of late, with women turning out at higher rates than men (Baxter and Lansing 1983; Beckwith, 1986; Firebaugh and Chen 1995). There are significant differences on some o f these socio economic measures between active duty members, veterans, and the civilian population. For the sake of consistency, all demographic data reported here were drawn from government analyses and reflect conditions as of 2009 2010. As a group, ve terans tend to be older, have slightly higher income, and are more likely to be white and male than civilians ( National Center for Veterans Analysis and Statistics 2011a, U.S. Department of Commerce, Bureau of the Census 2010 ). They are also more likely t o have a high school diploma, degree ( National Center for Veterans Analysis and Statistics 2011b ). Those currently in uniform, by contrast, are typically younger t han their counterparts: 9.6% of civilians are between 18 and 24 years old, while 44.2% of active duty members fall in the same range ( ICF International 2010 U.S. Department of Commerce, Bureau of the Census 2010 ). Although a fairly large majority of the active duty force is white and male, the gaps have been shrinking in recent years as increasing numbers of women and minorities have joined the ranks. In terms of education, military members as a whole are more likely to obtain a high school diploma, 1 Although pre adolescent socialization and experiences are often deemed important contributors to political participation (Beck & Jennings 1982, Kam & Palmer 2008), they are beyond the scope of this analysis.

PAGE 18

18 but differences between officers and enlistees; the former are more than twice as likely to civi lians. Assessing the income levels of active duty members is extremely difficult, as noted by just about every study that undertakes such an analysis. In addition to base pay, service members receive a number of allowances and benefits, many of whic h are determined by geographical location, number of dependents, rank, years of service, duties performed and a variety of other factors that must be taken into account. Despite the complexity, recent government reports tend to agree that military pay com pares favorably with civilian (including federal) salaries ( U.S. Congress Congressional Budget Office 2011 and U.S. G overnment Accountability Office 2010 ). The discovery that socio economic factors are related to military turnout is almost inevitable, but the interesting question is whether or not these factors will exert the same influence as they do among civilians. If not, we must question whether SES theories are sufficient to the task at hand. As Verba et al. (1995) point out, these models are p redictive rather than explanatory; they can tell us who will vote but not why. Resource Theories This approach expands the rational choice model beyond SES factors by including the resources they provide as separate variables. In other words, the com mon correlates are broken down into their individual components to see how they reduce (or increase) the costs of voting. For example, Brady, Verba & Schlozman (1995), some of the scholars most closely associated with this theory, determined that educatio n was significantly related to voting primarily as a conduit for interest. This

PAGE 19

19 relationship would have been concealed in a standard SES model. That is also true of the results from an earlier study conducted by the same team (Verba, Schlozman, Brady & N ie 1993); racial disparities in participation largely evaporated after controlling for political resources such as civic skills and engagement. They concluded that social, providers, thereby mitigating the effects of socio economic disadvantages. Earlier works suggested that public service occupations and government employment can also foster political skills and participation (Lane 1959, Wolfinger & Rosenstone 1980). It is vocational school in the nation as well as a primary school for its disadvantaged how ever, that Leal (1999) dismissed this possibility after detecting differences in participation between Latino veterans who were dra fted and those who volunteered. as civic duty (Riker & Ordeshook 1968, Blais 2000, Campbell et al. 1960), social gratification (Verba, Schlozman & Brady 1995, Rosenstone & Hansen 1993) and, altruism (Fowler & Kam 2007). These rewards are derived solely from engaging in the act of voting, irrespective of the outcome. Some scholars have argued that such theories are tautological and distort or violate the tenets of rational choice theory (Blais 2000). If anything can be considered a benefit, what does the theory actually reveal? There are merits to t his criticism, but it encourages cautious interpretation rather than exclusion of the rewards above, which are fairly well established in the literature.

PAGE 20

20 Given the nature of their profession, it is possible that selflessness and a sense of citizen duty a re more prevalent among military members. Although I am unaware of any efforts to test this notion directly with regard to voting, scholars have suggested that those in uniform are more patriotic (Bachman et al. 2000, Teigen 2006) and less self oriented t han their civi lian counterparts (Franke 2001) 2 Franke also reports that 75% of suggesting an inclination toward service and duty. Several other political and social ch aracteristics can be credited with motivating and facilitating electoral participation. For example, marriage and community integration, as measured by home ownership, length of residence, and religious attendance, are believed to reduce information costs and provide social reinforcement (Stoker & Jennings 1995, Wolfinger & Rosenstone 1980; Miller & Shanks 1996; Abramson, Aldrich & Rohde 2010; Squire, Wolfinger & Glass 1987). Despite some debate about the magnitude of their independent effects, political trust, efficacy, and interest also appear to play a dual role, both decreasing costs and giving individuals a 1963, Brady, Verba & Schlozman 1995). Finally, there is wide spread agreement that strength of partisan identification operates in much the same manner and exerts a direct and powerful influence on the propensity to vote (Abramson, Aldrich & Rohde 2010, Campbell et al. 1960, Rosenstone & Hansen 1993, Verba et al 19 95). 2 iance to country (as opposed to the world), military service as the strongest indicator of good citizenship, the promotion of patriotism in citizen education, willingness to fight for country, and loyalty to country.

PAGE 21

21 A number of these factors could contribute to or detract from military voter participation. Veterans and active duty members are, on average, slightly more likely to be married than the general public ( ICF International 2010; National Center for Ve terans Analysis and Statistics 2011a; U.S. Department of Commerce, Bureau of the Census 2010 ). However, it should be noted that 39.1% of the current force reports never having been married. While this might be attributable to the relative youth of those now in uniform, Lundquist (2007) found that both the marriage and divorce rates among young enlistees tend to exceed those of their civilian peers. Veterans appear to be similar to the general public in their religious affiliations and attendance rates ( P rinceton Survey Research Associates International 2011 ), while Feaver and Kohn (2000, 2001) report that active duty officers were only slightly more religious than civilian elites. Finally, veterans are more likely to own their home than civilians ( Nation al Center for Veterans Analysis and Statistics 2011a). And while frequent relocation would seem to imply a nomadic and rootless existence for active duty members, Charles Moskos asserts that, s a social organization (2010: 23). As such, it might provide a sort of stability in motion. Although much of the politico military literature suggests only minor (or inconsistent) attitudinal differences between service members and civilians on a variety of issues (Szayna et al. 2007, Jennings & Markus 1977, Bachman et al. 2000, Schreiber 1979, Franke 2001), there are always exceptions to the rule. A few of them are particularly relevant to this paper. For example, some studies indicate that peop le who have donned the uniform exhibit greater political efficacy (Inbody 2009, Mettler 2005)

PAGE 22

22 Others, however, have detected no such difference (Schreiber 1979). On the other hand, the finding that service members are more likely to identify as Republicans and conservative is nearly unanimous. However, the degree to which this is true largely depends on the sample composition and is worthy of closer inspection. Early studies compared officers to civilian elites; they typically found that the former were far more likely to espouse conservative values (Huntington 1957, Janowitz 1960). More recently, Feaver & K ohn (2000, 2001) expanded this design to include the general public and obtained somewhat different results. They conclude that, although officers are far more conservative than civilian elites, they are actually less so than the general public. The part isan divide, however, is much larger; officers reported Republican affiliation by a margin of 8 to 1, while the other two groups were approximately evenly split. Inbody (2009) divided his sample to discriminate between officers, enlisted active duty, enl isted veterans, and civilians. All three military groups are less likely to identify themselves as Democrats than the general public. However, he found that neither sample of enlistees was any more likely than civilians to be Republican, and both were fa r more likely to report being Independent (or other). In each case, the tendencies were reversed for officers. He reports similar results regarding ideology, with enlistees In the aggregate, these studies suggest there is a Republican and conservative leaning within the armed forces, at least among the officer corps. However, they also underscore the fact that the uniformed community is not homogeneous; there are important d ifferences that must be taken into account.

PAGE 23

23 Mobilization Theories Rosenstone and Hansen (1993) posit that mobilization by political parties, campaigns, and other organizations is just as important as individual resources in determining who votes and who abstains. Because this theory is not empirically testable with the available data, I will merely offer a few thoughts for the reader to bear in mind. Active duty members are less likely to be contacted by political parties, campaigns, and non partisa n organizations for several reasons: 1) they are geographically dispersed around the globe 2) they change addresses frequently, and 3) they often live outside the state and district in which they are registered. However, it is also possible that the proli feration of government programs and media/public attention has made participation more salient and has activated a sense of collective identity or citizen duty. Finally, one could hypothesize that military members are indirectly mobilized by integration i n the community, while veterans are similarly affected by involvement in service related organizations or benefit programs, as well as their civilian social networks. Military Specific Literature Scholars have long been interested in the degr the AVF has reinvigorated efforts to discriminate between self selection and socialization effects. Collectively, the eviden ce suggests that the former is a much more powerful determinant of political attitudes and opinions among military members (Jennings & Markus 1977, Betros 2001, Teigen 2007, Schreiber 1979). Socialization effects, when they are detected at all, typically augment rather than alter pre existing

PAGE 24

24 inclinations (Bachman et al 2000, Franke 2001). However, it is entirely possible that socialization has a significant impact on political behavior but not attitudes or opinions. Unfortunately, there is a relative de arth of literature that specifically addresses political participation among those on active duty, presumably because of legal restrictions and a culture that encourages them to be apolitical (Hatch Act 1939, DOD Directive 1344.10 2008 Betros 2001, McNern ney 2006). Therefore, most studies have relied on samples of veterans, who are far more accessible and numerous than active duty members. As the previous discussion regarding demographics and partisanship makes clear, it is important to recognize that su ch data might conceal significant differences within the community. Nevertheless, the available results are instructive. Contrary to conventional wisdom, Bishin and Incantalupo (2008) found that uniformed voters are not more likely to support a candidat e who has military experience and do not appear to vote as a consistent or cohesive bloc. Teigen (2007) examined the 2004 presidential election and found that veterans did not differ significantly from civilians in terms of partisan allegiance, affect for candidates, or vote choice. It appears political choices, although the evidence regarding their participation is somewhat more mixed. Turnout has been disproportionat ely high among military members, especially in recent years (Teigan 2006, Inbody 2009, Betros 2001). Although Franzich (1982) reported that graduates of service academies were actually less likely to cast a vote than their civilian counterparts, his resul ts appear to be unusual. Teigen (2006) compared electoral participation over a 32 year period and determined that veterans

PAGE 25

25 consistently vote at moderately higher rates than civilians, except for those who served during the Vietnam era. Because participat ion was elevated for both WWII vets and those who joined an all volunteer force, he concludes that socialization effects are at least partially responsible for the intra cohort variations. With regard to race, Ellison (1992) found no difference in low le vel activities such as voting between African American vets and non vets, but he did detect a greater likelihood of engaging in high level activities, especially among those with combat experience. He also discovered that African Americans who served exhi bited lower levels of racial identification than their civilian counterparts, indicating that membership i n the armed forces could supplant such group identities. Overall, his results are similar to those obtained by Jennings and Markus (1976) regarding t he electoral participation of Vietnam era veterans. Leal (1999), on the other hand, reported that veteran (and specifically draftee) status had a greater positive association with voting and registration among Latinos who served than their white counterpa rts. It is difficult to draw general conclusions from these studies, many of which focus entirely on specific subpopulations within the military community or, alternately, ignore any distinctions between the groups that compose it. This study will attemp t to fill in some of the gaps by comparing civilian, veteran and active duty populations. When possible, I will also include dummy variables for officer/enlistee and volunteer versus draftee, as well as measures of length of service. By doing so, I hope to gain a clearer understanding of which military members are responsible for the higher turnout and why they participate.

PAGE 26

26 CHAPTER 3 DATA AND METHODOLOGY The data are drawn from two sets of companion surveys that permit direct comparisons between mili tary and civilian populations. Each has its own advantages and disadvantages regarding composition of the military samples and the questions asked. In combination, however, they allow a fairly detailed analysis of military voting behavior. All variables have sound theoretical justifications for their inclusion in the coded as missing values, and great care was taken to ensure that measures from the two sets of studies were as similar as possible. NAES Surveys In 2004, the National Annenberg Election Survey (NAES) randomly selected 656 active U.S. military households from the national rolling cross section for a special study. Interviews were conducted between September 22 and October 5, 2004. The sample contained 372 active duty members and 284 family members. Several of the questions asked are identical to those included in the 2004 General Election Panel survey conducted by NAES. A random sample of 8,664 adults (including 1537 veterans and 65 active duty members) was interviewed between July 15 and November 1, 2004; although the same respondents were re interviewed after the election, equivalent data is not available for the larger military sample. I have chosen to rely on the pre election panel, as opposed to the far more extensive national rolling cross section data, because the time frame, political context, and questionnaire content of the former are more compatible with the military survey.

PAGE 27

27 This study will use Ordinary Least Squares ( OLS) regression to estimate the effects of relevant independent variables on voting or registration to vote. Although probit or logit analysis is often favored for dichotomous dependent variables of this sort, OLS has proven rather robust to violations of linear regression assumptions, yielding similar results that are far easier to i nterpret and understand (Hellevi k 2009, Wald & Feinstein 2012, Cleary & Angel 1984, Noreen 1988). For large samples like those considered here, linear models are comparable t o other methods and therefore appropriate. Analysis of these data will be presented in three sections. The first combines and re codes the samples from both studies by introducing a dummy variable. Military family members and civilians were coded as 0, while service members (active and former) were coded as 1. Although the possibility that military life socializes dependents as well as members is fascinating, it is beyond the scope of this analysis because the Pew survey does not include an equivalent sample. An additional dummy codes veterans as 0 and current members as 1, allowing for comparisons between the two. The second section will compare the veteran and civilian samples from the panel study on a number of questions that were, unfortunately, e xcluded from the military survey. (Note: because the military sample only includes 65 active duty members, they will not be separated from veterans in this section.) Finally, I will analyze the active duty members from the military study to explore servi ce specific variables, such as rank and length of service. It is regrettable that direct comparisons between veterans and active duty members are not possible on these measures.

PAGE 28

28 Dependent Variable Because the 2004 election had not yet concluded when t he survey was taken and turnout in midterm elections is consistently lower, I will use self reports of voting (or abstention) in the 2000 general presidential election as my dependent variable. Despite the well known tendency for social desirability effec ts to inflate such reports (Bernstein, Chahda and Montjoy 2001; Silver, Anderson & Abramson 1986; Traugott & Katosh 1979; Karp & Brockington 2005) they provide the best available data for comparing the two populations. And although one could hypothesize t hat military members experience greater social pressure to vote, I am not aware of any evidence suggesting they are more or less inclined to overreport than civilians. Approximately 2.3% of the overall sample was ineligible to vote in 2000 due to age. I ndependent Variables Age is coded as an ordinal scale that includes five categories: 18 29, 30 39, 40 49, 50 64 and 65 or older. The education variable also has five levels: less than high school diploma, high school diploma or GED, some college (or techn ical/vocational variables are included for Hispanic (0=no, 1=yes), race (0=white, 1=non white or mixed race), and sex (0=male, 1=female). Income is measured on an ordinal scal e with the following categories: a) less than $25,000, b) $25 50,000, c) $50 75,000, d) $75 100,000, e) $100 150,000 and f) $150,000 or higher. Marital status has been re coded into a dummy variable so that 0=unmarried and 1=married; this is in keeping wi th the literature which suggests that currently being married exerts an influence on political participation (Stoker & Jennings

PAGE 29

29 1995). The ordinal scale for religious attendance includes seldom/never, a few times a year, once or twice a month, once a week or more than once a week. In addition to demographic and SES variables, this analysis will examine several political and attitudinal measures. Party ID has been recoded into a dummy with 0 indicating Independent, Other, or No Preference and 1 represen ting Republican or Democrat. I did this for two reasons. First, the literature suggests that Independents, leaners and those who do not identify with one of the major parties are generally less likely to vote (Abramson, Aldrich & Rohde 2010, Campbell et al. 1960, Rosenstone & Hansen 1993, Verba et al 1995). Although strength of party ID would provide a much better measure, no such question was asked in the military survey. Second, this s are more likely to be Independent than civilians or officers. Two measures of ideology are also included in the models. The first is a standard five point scale ranging from very conservative to very liberal. The second has been re coded as a measure of ideological strength, where 1=moderate, 2=liberal or conservative, and 3=very liberal or conservative. Two scales capture how much attention respondents had paid to political news on television or in the newspaper over the past week; both have four lev els and range from none to a great deal, with higher scores indicating more attention. Certain subsamples in the panel and all respondents in the military survey were asked about their interest in the campaign and how much of the time the federal governm ent can be trusted to do what is right. Responses regarding interest include some, not much and very; the trust scale consists of never, sometimes, most times and always. Unfortunately, the limited Ns for these variables require running separate

PAGE 30

30 regressi ons to reduce the standard errors. Any conclusions will therefore be tentative because it is impossible to see how the two interact and because there is only one measure for each trait. The panel survey included a number of additional politi cal questions that are particularly relevant to this paper. Among them is strength of party ID, which is coded as 0 for not strong and 1 for strong. Two additional interest variables measure how closely the campaign and politics in general are followed; both are coded into four part scales on which a higher value indicates greater interest or attention. I have also included a dummy variable for government employment, where 0=no and 1= yes. The survey also contained two measures of political efficacy and a variable for civic duty. The belief that politics is too complicated is measured on a 5 point scale that ranges from strongly disagree (low score) to strongly agree (high score). A similar item assesses how often candidates try to keep their promises, but a higher score in this case reflects greater trust. Respondents who indicated they vote because of candidates, issues or other reasons are coded as 0; those motivated by civic duty or a coded as missing values). Unfortunately, a number of these variables are also limited by small Ns and must be run separately. Consequently, results must be interpreted cautiously. The final section of this analysis will examine service speci fic variables among active duty members from the military cross section. Reasons for joining the armed forces were re coded so that 0 reflects personal benefits (such as education, job training, pay, new experiences, travel etc.) and 1 indicates service t o country or fighting terrorism. If the latter group is more likely to vote, it could suggest that self selection

PAGE 31

31 exerts greater influence on electoral participation than socialization. In order to assess have included questions about whether it is never (1) sometimes (2) or always (3) appropriate for m ilitary members to advocate for candidates. Because much of the literature finds differences according to rank, I have recoded pay grade so that 0=Enlist ed and 1=Officer. This variable could be classified in a variety of ways, but I have chosen a simple dummy for two reasons. First, the sample is not necessarily large or representative enough to permit narrower subdivisions. Second, the NAES and Pew sur veys include different categories; applying these broad designations makes comparisons between them possible. Finally, I have included years of service in the model because it is possible that socialization effects emerge only after an extended period. A lthough the survey provided different response options for the regular military and those serving in the National Guard or Reserves, they have been re coded to match: one= less than a year; two =1 2 years; three equals 2 3 years; four = 3 5 years; five =5 10 years and six = 10 or more years. Respondents who did not serve in that particular branch are coded as 0 to preserve a sufficient N for analysis. Pew Surveys In 2011, the Pew Research Center conducted companion surveys to investigate how the atti tudes of pre and post 9 11 veterans compare to one another, as well as those of civilians. Active duty members were specifically excluded from the military sample, which was composed of 1,853 veterans. Telephone and internet interviews were conducted bet ween July 28 and September 4, 2011. The civilian survey interviewed 2,003 members of the general public via telephone between September 1 and September 15, 2011. This sample included 256 veterans (12.8%) and 31 current active duty members (1.5%), who wil

PAGE 32

32 veteran sample. However, these respondents will be excluded from a separate analysis of the veteran survey because the civilian questionnaire did not include pertinent service related variables. Unfortunatel y, the active duty sample is small and likely unrepresentative, precluding detailed comparisons like those yielded by the NAES surveys. Dependent Variable This survey did not include any questions regarding voting behavior; therefore a dummy v ariable coded 0 for not registered to vote and 1 for registered will be used as an approximation. Although this is not a perfect measure, registration is a prerequisite for casting a ballot and is therefore a suitable substitute for the purposes of this p aper. Independent Variables Socio economic and demographic variables have been coded to match the NAES data. Party ID, ideology and government employment are also identical. The response sets for household income are slightly different, but the follo wing categories mirror the NAES as closely as possible: a) less than $20,000, b) $20 40,000, c) $40 75,000, d) $75 100,000, e) $100 150,000 and f) $150,000 or higher. The survey also includes a measure of whether the respondent is (1) less, (2) about as, or (3) more patriotic than most people in the country. Although the lack of questions regarding political behavior, interest and efficacy is disappointing, this survey does provide a replication of sorts for the SES and demographic regressions using NAES data. The military instrument contains a number of service related variables that appear throughout the literature. Dummies are included for draftee (0) or volunteer (1) status, combat experience (no=0, yes=1), and rank at time of discharge (0=enliste e, 1=officer, 5= still in National Guard/ Reserves and is coded as missing). To align with

PAGE 33

33 the NAES data, the scale for years in service includes the following categories: one= less than a year; two =1 2 years; three equals 2 3 years; four = 3 5 years; fi ve =5 10 years and six = 10 or more years. Measures of how useful military experience was for teaching cooperation, building self confidence, job/career preparation and personal growth were combined into a four point scale ranging from not at all useful t o very useful. A separate item regarding whether military service helped or hurt the respondent get ahead in life contains five categories: hurt a lot, hurt a little, made no difference, helped a little and helped a lot. Although not specifically politic al, these variables allow an indirect test of the possibility that the military fosters skills and efficacy. A dummy variable is included for service to country as a motivation for joining the armed forces, where 0=not important and 1= important. A third category, not applicable, identifies draftees and is coded as missing. Finally, Verba et al. (1993) have suggested that recipients of non means tested benefits (including veterans) tend to be more politically active than those who receive means tested be nefits. This raises the possibility that high turnout among veterans is partly motivated by their participation in such programs. Dummy variables are included for whether the respondents have received any benefits from the VA and whether the government h as given them all the help it should (0=no, 1= yes in both cases). Four point scales running from poor to excellent rate how well the VA is meeting needs of veterans and how well the military cared for the member and family while on active duty. One coul d also interpret these variables as partial (and indirect) measures of external political efficacy. For example, Mettler (2005) concluded that the GI Bill encouraged political participation among recipients because it demonstrated government responsivenes s to their needs.

PAGE 34

34 As I said before, each set of survey sets has its own advantages and drawbacks. The NAES data provide a relatively rare opportunity to compare current and former members to one another as well as civilians. They also contain a number of interesting and theoretically relevant political variables. Unfortunately, many of them were excluded from the active duty instrument and/or were limited to certain interview dates and subsamples. The Pew surveys, on the other hand, do not include any items directly related to political behavior (aside from party ID and ideology). However, they permit an exploration of which (if any) specific characteristics of military service are most correlated with electoral participation. Despite their limitatio ns, these data represent the best available resources for addressing the primary research questions of this thesis.

PAGE 35

35 CHAPTER 4 RESULTS FROM 2004 NAES DATA This chapter will present the results from three separate analyses of the 2004 NAES data. In each case, an initial model comprised of socio economic factors, as well as demographic and political identity variables, provides a baseline for comparisons between groups and across models: Voter Participa The tolerance levels for all of these variables, as well as tho se introduced in subsequent models, fall well above the 0.5 threshold for multicollinearity and the bi variate correlations do not exceed 0.7. The residuals for the dependent variable are normally distributed around a mean of zero for each regression. Co llectively, these tests confirm that the data satisfy many of the linear regression assumptions and verify that OLS is an appropriate method for these analyses. NAES Merged Samples The first NAES analysis combines the samples from the panel and military surveys, yielding a total N of almost eight thousand. Both veterans (1293) and active duty members (384) are fairly well represented in the sample, permitting separate analyses for each group. As noted earlier, there are several demographic differences b etween the three subsamples that might influence the interpretation of regression results. Therefore, it is important to assess whether these samples differ from what would be expected based on the population data presented above. The descriptive statist ics for these variables (as well as those introduced in subsequent models) can be

PAGE 36

36 found in Appendix B, and I will supplement this information with fre quencies (not reported here). The most notable disparities, at least for the purposes of this paper, are found on the dependent variable. The freque ncies are displayed in Table 4 1 Consistent with the literature, veterans are by far the highest t urnout group at approximately 90 %. They are followed by civilians, 8 3 % of whom participated, while those curre ntly in uniform lag behind by a full ten points. The lower rate among active duty members is not entirely surprising, given the administrative costs discussed earlier However, the size of the gap is quite striking and results from a one way ANOVA test i n Table 4 2 confirm that the differences between groups are statistically significant This suggests that veterans are primarily responsible for overrepresentation of the military at the polls. The independent variables reveal both similarities and di fferences between samples. Age and sex conform to expectations, with veterans being older and current service members being younger than civilians on average. The two military groups are overwhelmingly male, while women made up 65 percent of the civilian sample. Educational disparities are modest, however, and appear to be concentrated at the lower end of the scale; six percent of civilians report having less than a high school diploma, as opposed to approximately three percent of veterans and only one p ercent of active duty members. Income follows a similar pattern. Although veterans have a higher mean, less than a single percent of variation remains between the three groups among those earning more than $50,000 per year. Finally, more than eighty per cent of civilians and both types of military members are white. Hispanics represent a small and fairly similar portion of each sample (less than 7%), while non whites are slightly more

PAGE 37

37 prevalent among active duty members than the other two groups. Aside from the predictable differences for age and sex, the three subsets appear to be socio economically comparable overall. The demographic and political variables also suggest a mixture of shared and singular tendencies, as evidenced by the one way ANOVA results in Table 4 2. For example, the data reveal no substantial differences in terms of religious attendance, as a majority of each subsample attends only occasionally or not at all. On the other hand, active duty members and veterans are both more lik ely than civilians to be married, and those currently in uniform exceed their predecessors by an additional seven percent. There are also significant differences on the political identity variables. Although a majority of all three groups identifies with one of the major parties, civilians are more likely than respondents in either military set to label themselves as liberal or very liberal, in keeping with the literature. And while the frequencies regarding ideological strength are almost identical for civilians and veterans, active duty members are more inclined to be moderate and less likely to identify as very liberal or conservative by several points. Although they clearly diverge on certain individual items, the three groups are generally similar, and the disparities that emerge are consistent with the statistics and empirical data reported earlier. Table 4 3 presents the results for the base model, which was regressed separately for each of the subsamples to allow comparisons. To reca pitulate this model includes the core variables included in standard models of political participation. and related to voting in the predicted direction. The onl y exceptions, the two measures

PAGE 38

38 of ideology, produced results that might be obtained by simple chance for all three groups (and across all models). Age, education and income were all positive and ll, but there are somewhat notable differences in magnitude. For example, the unstandardized coefficient for age is nearly twice as large among active duty members as veterans, with civilians falling in between the two. Education, on the other hand, exer ts similar influence among civilians and current service members but a great deal less among veterans. In both cases, the measurement scales were identical and yet the size of the gaps exceeds the standard deviations, suggesting a substantive difference ( Miller 2005). Although the income coefficients also vary across groups, they do not meet this criterion. Finally, sex is negative for both military groups, indicating that uniformed females are less likely to vote than their male counterparts, but the re lationship is not statistically significant. These findings are fairly intuitive, given the relationships between them and the demographics of the sub samples, but others are not so easily explained. Coefficients for the racial/ethnic variables in Tab le 4 3 present something of a paradox. Being non white is negatively related to voting for both veterans and active (The p value of .076 for the veteran sample doe s satisfy less stringent requirements, however). Hispanic lineage, on the other hand, performs as expected (and quite strongly) among those currently in the service by depressing turnout, but is actually positive and insignificant for former members. His panics encounter additional barriers to political participation; controlling for citizenship and English proficiency might shed

PAGE 39

3 9 additional light on these results, but no such measures were available. 3 On the whole, however, it appears that military servi ce does mitigate some racial effects, if a bit inconsistently. Finally, religious attendance and party ID have positive and significant effects on the voting behavior of veterans, as predicted by the literature, but the p values for both relationships are well outside the acceptable range for those presently in uniform. Despite these variations across sub samples, the base model confirms the well documented fact that socio economic factors (specifically age, education and income) play a pivotal role i n determining who votes and who abstains. Model 2 in Table 4 4 introduces measures of how much attention respondents paid to political news on national television or in the newspaper during the previous week. Although the split sampling procedures and m ultiple questionnair e forms employed by NAES reduce the Ns somewhat for each group, there are still 4,350 civilians, 988 veterans and 267 active duty members included in the overall sample. a higher score indicating more attention paid. The descriptive statistics suggest only minor differences on these measures across groups; the frequencies (not reported here) indicate the same. However, the results of the regres sion presented in Table 4 4 reveal very different effects for each subsample. Both attention variables are positive veterans neither result differs from chance and increased attention to political news on 3 A citizenship question was include d in the panel questionnaire, but response rates were far too anemic for analysis. Opting to conduct the interview in Spanish could arguably be a proxy for proficiency, but it might also reflect simple preference rather than ability and was therefore excl uded.

PAGE 40

40 TV actually has a negative coefficient. Among current service members, both effects conditions for significance. It appears that at tention to political news via these media does not have as much influence among military members as it does among civilians. However, the addition of these variables does alter some of the socio economic effects from the base model. Interestingly, the c hanges are least dramatic among civilians, for whom nearly all of the variables remain significant, even though some of the coefficients decline slightly. The only exception is marital status, which drops out as a statistically important predictor for civ ilians and veterans alike and retains this negligible status among active duty members. Similarly, the p value for religious attendance rises above the .05 level for veterans (p=.068) and is still insignificant for those currently in uniform. Income is a lso no longer significant for veterans and, while the p value increases to .063 for active duty members, it remains extremely low for civilians. Finally party ID, one of the most heralded predictors of electoral participation, continues to be significant for veterans and civilians but has no such statistical association for those on active duty. The variable was coded as a dummy, with a one indicating affiliation with either major party. It seems that uniformed independents and those who identify with o ther parties are not any less likely to vote than their Republican regarding the middle of the road partisan and ideological preferences of current and former enlistees, as wel l as their propensity to participate.

PAGE 41

41 Model 3, presented in Table 4 5, returns to the base model and expands it by adding a specific measure of interest in the ongoing 2004 presidential campaign. Ideally, this variable would be run in the same model as the attention items just discussed, but the limited number of respondents who were asked all three questions precludes this course of action. A combined regression (not reported here) resulted in higher standard errors for all variables, indicating less reliable results. Even without including the attention variables, the Ns for this model are smaller than the base, but sample sizes of 2,315 civilians, 446 veterans, and 341 active duty members are more than adequate. The descriptives suggest that veter ans are the most likely to be very interested in the campaign, followed by civilians, while active duty members bring up the rear. Once again, however, the regression results in Table 4 5 seem to defy logical and theoretical predictions regarding the effe cts. Veterans might indeed be more interested, but it does not appear to influence whether or not they participate at the polls, since the p value is well outside the accepted parameters. At the same time, there is a positive and significant association for civilians and active duty members. In fact, interest surpasses assertion that the latter functions primarily as a conduit for the former. The coefficient for those currently in service is almost one and a half times larger than for civilians, suggesting that interest is a more important resource for them when all other factors are held constant. However, these results must be interpreted cautiously for two reasons. First, a single item of interest, especially one that is limited to the campaign, is not necessarily a reliable indicator. Second, the dependent variable measures voting

PAGE 42

42 behavior in 2000, four years prior to the campaign about which respondents are being asked. These limitations notwithstanding, the interest variable exerts an indirect influence on the socio economic factors in the model, most notably rendering income insignificant for all three subsamples. For civilians, being female is no longer statistically associated with higher turnout, but all of the other base variables retain their significance, despite diminished magnitude (except for marital status, which actually increased). Among veterans, party ID, religious attendance and rac e now exceed the acceptable range for p values. Interestingly marital status, which was not significant in the base, now satisfies this requirement and has a much larger coefficient. For the most part, there were only modest changes in the active duty re sults. Income was the lone item that crossed out of (or into) significance, and although the age coefficient declined and education increased, neither shift was very large. The Hispanic variable, on the other hand, exhibits much greater magnitude in this model than the base and continues to be by far the most powerful predictor for this group. Model 4 in Table 4 6 tests the hypothesis that greater trust in the federal government might contribute to higher turnout rates among military members. Although th is model too is limited by the inclusion of a single measure and dramatically reduced Ns, it at least allows for a preliminary investigation. The sample contains 1 396 civilians, 220 veterans, and 326 active duty members. Respondents were asked how much of the time they trust the government to do the right thing; a higher score on this scale indicates greater trust. The means suggest, and the frequencies verify, that active duty members do in fact exhibit higher levels of trust, while the other two group s are

PAGE 43

43 rather similar in their distribution. Considering their integration in and dependence on the government, it is not terribly surprising that current service members are more trusting of it. However, one would not expect this quality to be negatively and significantly associated with voting, which is exactly what the regression results in Table 4 6 indicate. The coefficients for civilians and veterans, on the other hand, are positive and the p values are quite high. Bi variate correlations between t rust and voting (not shown here) confirm these somewhat baffling findings. It is possible that the limitations of the model are at least partially responsible; more extensive measures of trust and larger active duty samples might yield more predictable re sults. Or it could be that explain why positive evaluations diminish turnout. NAES Panel Sample This section will examine a number of pertinent political questions th at were excluded from the active duty survey. The analysis is therefore limited to the sixty five current service members, 1,193 veterans and 5,601 civilians who were part of the panel sample. As a result, it is no longer possible to regress the two mili tary groups separately and uniformed respondents have been combined with veterans. The descriptives for the base model, which now includes a measure for strength of party ID, are located in the appendix. The new variable is coded as a dummy, with one ind icating a strong partisan affiliation. The mean for military members is 0.72, while the civilian average a slightly lower 0.67. Interestingly, the means for Party ID are nearly identical but reversed. The rest of the figures are quite similar to those o btained for the merged sample and will therefore not be discussed in great detail here.

PAGE 44

44 For the most part, the regression results presented in Table 4 7 are also comparable to the initial base model. Among civilians, all of the variables continue to be significant with only minor fluctuations in magnitude. (Again, ideology is the exception.) The only notable difference for military members occurs on the race variable, which shifted from borderline to clearly significant (p=.004) and increased in magnit ude from .043 to .070. As before, being female, married and Hispanic are all insignificant, and the coefficient for the latter remains positive. The new strength of party ID measure performs quite well for both groups; the p values are extremely sma ll and the coefficients are identical at .056. Model 2, located in Table 4 8, includes two measures of political interest and one for internal efficacy. The sample is comprised of 1,860 civilians and 445 military members. Although the reduced Ns are di sappointing, running these three items separately did not improve the sample sizes; nor did it substantively alter the results or reduce the standard errors. The interest variables indicate how closely respondents are following the campaign and how often they follow politics in general. Both are measured on a four tier scale that increases from not at all to a great deal. Although these two items are positively and significantly related to one another and to voting, the correlations do not exceed accepta ble limits. A five point efficacy scale measures indicating greater agreement with that statement. The descriptives presented in the Appendix suggest that military members are more inclined to follow political matters and less likely to believe they are too complicated. Both of these trends are evident in the frequencies as well. Table 4 8

PAGE 45

45 displays the regression results for this model. While the coefficient for t he efficacy measure is negative, as expected, it also fails to achieve significance for both groups. Holding a low level of what scholars call internal political efficacy does not appear to prevent these individuals from voting. The two interest variable s, on the other hand, are politics are logically and theoretically linked to political participation, so these findings are not exactly shocking. However, it is intere sting to note that the magnitude of both variables is greater among military members; the coefficient for following politics is twice as large as that for civilians. This contrasts with the attention to political news and interest in campaign variables, w hich were not significant for the veterans who make up a vast majority of this military sample. It could be that the previous measures were too specific and narrow; or perhaps question wording and order contributed to the discrepancies. The two samples also differ in terms of the consequences that result from adding these items to the equation. Among civilians, income and being female are no longer significant predictors of voting behavior. Race, which has been consistently and highly significa nt in every model up to this point, now has a borderline p value of 0.73 and the coefficient dropped by half to .042. For military members, race and education are no longer even remotely significant (both p values approach or exceed .500), while income a ctually increases its positive and powerful influence. Because income and education are so closely related, it is difficult to determine why they behave differently for the two groups, or if it even matters. Party ID and strength of attachment both perfo rm as

PAGE 46

46 expected among civilians, while the latter is insignificant f or military members and the former has an increased p value of .077. Model 3 in Table 4 9 adds a measure of external political efficacy to the base model. Respondents were asked how much of the time candidates try to keep their promises and the five point scale ranges from never (low) to always (high). Again, this would ideally have been included with the model above or combined with the trust government variable, but the reduced Ns of 6 01 civilians and 136 military members require a separate analysis. The descriptives reveal means that are almost identical for average answer. The frequencies (not reported) show that the distributions are fairly similar for higher and lower scores in both groups. Table 4 9 displays the results of the regression. The coefficient for this variable is positive but insignificant across the board, which is not surprisi ng for a single item that is quite narrow in focus. Despite its poor performance, adding this item does drop sex and marital status as significant predictors for civilians. Among military members, the p values for religious attendance, being Hispanic, an d education no longer fall within an acceptable range. The rest of the independent variables show minor variations in the coefficients, but do not differ substantially from the base model. Returning to the base, Model 4 adds a dummy variable indicatin g whether civic were equal are coded as one; those who replied they do not vote were exclud ed from the analysis.) While the mean for civilians falls just above the midpoint at .54, military

PAGE 47

47 members have an average of .62. The frequencies, which reveal an 8% discrepancy between the two groups, also suggest that more service members are driven b y a sense of duty. But there is yet another disconnect between the answers given and the effects of the variable, as evidenced by the results in Table 4 10. Although the coefficient for civic duty was positive for both groups, it only attained significan ce among civilians (p=.025). The magnitude is fairly large at .054 and renders race, marital status and party ID insignificant. For military members, the only significant predictors were age, being married, and the party ID measures. However, it should be noted that there are once again substantially reduced Ns of 595 civilians and 138 military. It is difficult to be certain that these respondents are representative of the 7,000 who were not asked this question. The final model for the panel survey da ta tests whether government employment facilitates or encourages political participation, as suggested by Lane (1959) and Wolfinger & Rosenstone (1980). A simple dummy variable was coded 1 for government employees. The sixty five active duty respondents, who obviously meet this criterion, were excluded to avoid skewing the results. A separate analysis (not reported here) indicated their inclusion had just that effect, as discussed below. There are thus 3,057 civilians in the regression, approximately 26 % of whom work for the government. A slightly higher 31% of the 470 veterans are so employed. The results, which appear in Table 4 11, show that government work is positively associated with voting for both rising that the coefficient is twice as large for former service members (.056), since they have prior as well as current experience. However, this relationship could simply reflect self selection effects and does not

PAGE 48

48 necessarily imply causation. The coe fficient was also positive in the regression that included active duty members but it did not attain significance. (Obviously civilian results were unaffected.) Having included this item in the model, marital status no longer even approaches significanc e for civilians and the p value for being female increases to .056. All of the other base variables continue to exert significant influence. For veterans, education and strength of party ID drop out, while party ID clings to the last remnants of signific ance with a p value of .094. Income, being non white, and religious attendance all continue to be significant predictors of voting, although the latter now only satisfies the more lenient criterion (p=.062). NAES Active Duty Sample This section examines the internal differences among active duty members, rather than comparing them with civilians or veterans. The demographic characteristics of the 321 active duty members from the military cross section were discussed above and will not be repeated here. However, a separate descriptives table is provided in the appendix. Results from the base model are presented in Table 4 12 and closely resemble those obtained in previous sections. There is, however, one noteworthy exception: for the first time, income is positive but insignificant right from the start. This is even more striking considering the coefficients for education and age are nearly identical to those from the merged base model. In that instance, income had a p value of 0.46 for active duty me mbers while those for the other two groups hovered around zero. There appears to be a pattern, and yet the complete absence of significance is surprising. The other variables behave more predictably; race, sex, and all of the demographic and political va riables are insignificant. Age, education and being

PAGE 49

49 Hispanic are all influential as expected and the coefficient for the latter remains extremely high at .232. Model 2 in Table 4 13 introduces several service related variables and includes 300 responde nts. A dummy for rank is coded so that 0 denotes an enlistee and 1 an officer, with just over a quarter of respondents falling into the latter category. A similar variable measures whether service members joined the military to obtain personal benefits s uch as education and job training (0) or to serve the country or fight terrorism (1). A majority (73%) of subjects selected the former. Two questions regarding whether it is appropriate for a military member to advocate for local or presidential candidat es were combined into a three point scale, with a high score indicating either behavior is appropriate. A mean score of 1.71 suggests that the average service member believes it is acceptable some of the time or not at all. Finally, separate variables we re included for years in the regular military or the National Guard/Reserves, with members who did not serve in that branch coded as zero. Table 4 13 reveals that none of the military variables were significantly related to voting, except for years in se rvice. Both of these items were positive and significant, although the p value for the National Guard/Reserves measure is much lower (.008 versus .072) and the coefficient is one and a half times larger (0.50). These results are fairly compelling since t he tolerance levels were well below the .500 cutoff point that hints at multicollinearity and the model controls for age, making this a strict test. A separate regression that excluded years of service (not reported here) confirms the poor predictive powe rs of the other military variables.

PAGE 50

50 Adding these service related variables increases the p values for age and education from almost zero to .056 and .070, respectively, and reduces the magnitude of both by quite a bit. The impact is reversed and even mo re dramatic for the Hispanic variable, which now has a coefficient of .333. Finally, religious attendance, which was insignificant in the base model, becomes borderline significant with a p value of .070. It is disappointing, but not surprising, that t he military variables performed so poorly. Only rank had even a bi variate correlation with voting that attained significance; and a close association with age and income meant it was unlikely to exert much independent influence. The reason for joining v ariable was included as a secondary measure of civic duty, which was also insignificant. And the question about advocating for candidates was an intriguing but hardly all encompassing attempt to capture the apolitical culture that is so prominent in the l iterature. After all, there is a big difference replication of some, but not all, of these variables and are the subject of the next chapter.

PAGE 51

51 Table 4 1 Frequencies for Voter Turnout in 2000: Civilians, Veterans and Active Duty Members (NAES) Frequency Percent Civilian Did not Vote 1112 15.1 Voted 6101 82.9 Total 7213 98.0 Veteran Did not Vote 112 7.6 Voted 1321 89.7 Total 1433 97.4 Active Duty Did not Vote 107 24.5 Voted 318 72.8 Total 425 97.3 Table 4 2 One Way ANOVA Test of Means for Turnout and Demographics: Civilians, Veterans and Active Duty (NAES) Sum of Squares df Mean Square F Sig. Voted in 2000 General Between Groups 11.826 2 5.91 3 47.708 .000 Within Groups 1123.875 9068 .124 Total 1135.701 9070 Religious Attendance Between Groups 2.308 2 1.154 .643 .526 Within Groups 16564.503 9229 1.795 Total 16566.811 9231 Marital Status Between Groups 15.936 2 7.968 33.752 .000 Within Groups 2183.507 9249 .236 Total 2199.443 9251 Party ID Between Groups 2.039 2 1.019 4.605 .010 Within Groups 2001.031 9040 .221 Total 2003.070 9042 Ideology Between Groups 97.115 2 48.558 47.55 1 .000 Within Groups 9241.562 9050 1.021 Total 9338.678 9052 Ideology Strength Between Groups 7.622 2 3.811 8.017 .000 Within Groups 4302.150 9050 .475 Total 4309.772 9052

PAGE 52

52 Table 4 3 Effects of Demographic Variables o n Voter T urnout: Civilians, Veterans and Active Duty (NAES) Coe f. Std. Error Std. Coe f. Sig Civilian a (Constant) .278 .026 .000 Age Group .071 .003 .252 .000 Education .053 .004 .185 .000 Income .015 .003 .065 .000 Hispanic .154 .018 .103 .000 Ra ce .068 .013 .062 .000 Sex .027 .009 .038 .001 Religious Attendance .024 .003 .092 .000 Marital Status .027 .009 .037 .003 Party ID .064 .009 .086 .000 Ideology Strength .000 .006 .001 .953 Ideology .001 .004 .002 .866 Veteran b (Co nstant) .578 .055 .000 Age Group .044 .007 .165 .000 Education .024 .007 .104 .000 Income .016 .006 .088 .004 Hispanic .028 .038 .020 .461 Race .043 .024 .048 .076 Sex .033 .024 .037 .183 Religious Attendance .012 .006 .061 .033 Marita l Status .006 .016 .010 .726 Party ID .057 .015 .103 .000 Ideology Strength .002 .011 .006 .851 Ideology .011 .009 .040 .192 Active Duty c (Constant) .050 .171 .769 Age Group .089 .022 .201 .000 Education .076 .022 .188 .001 Income .038 .019 .113 .046 Hispanic .213 .085 .118 .012 Race .088 .055 .075 .110 Sex .069 .054 .060 .203 Religious Attendance .021 .017 .061 .220 Marital Status .044 .048 .045 .357 Party ID .011 .044 .012 .797 Ideology Strength .030 .039 .0 45 .448 Ideology .016 .033 .031 .613 a. N=6,308; R Square=.172; b. N=1,293; R Square=.083; c. N=384; R Square=.212; Estimated by OLS

PAGE 53

53 Table 4 4 Effects of Attention to Political News o n Voter Turnout: Civilians, Veterans, and Active Duty (NAES) Co e f. Std. Error Std. Coe f. Sig Civilian a (Constant) .329 .033 .000 Age Group .063 .004 .244 .000 Education .038 .004 .143 .000 Income .013 .003 .062 .000 Hispanic .114 .022 .074 .000 Race .071 .015 .070 .000 Sex .027 .010 .042 .004 Rel igious Attendance .023 .004 .097 .000 Marital Status .012 .010 .019 .229 Party ID .051 .010 .076 .000 Ideology Strength 8.954E 005 .007 .000 .989 Ideology .004 .005 .012 .424 Attention to Pol. News TV .011 .005 .032 .038 Attention to Pol. N ews Paper .015 .005 .043 .005 Veteran b (Constant) .693 .057 .000 Age Group .031 .008 .136 .000 Education .015 .007 .076 .025 Income .007 .005 .049 .164 Hispanic .007 .040 .006 .855 Race .012 .024 .016 .624 Sex .018 .025 .023 .46 5 Religious Attendance .010 .006 .060 .068 Marital Status .002 .016 .005 .889 Party ID .037 .015 .080 .013 Ideology Strength .002 .011 .006 .851 Ideology .006 .008 .025 .484 Attention to Pol. News TV .005 .009 .021 .544 Attention to P ol. News Paper .011 .008 .048 .162 Active Duty c (Constant) .176 .219 .423 Age Group .059 .026 .142 .026 Education .067 .027 .167 .014 Income .042 .023 .128 .063 Hispanic .027 .106 .015 .798 Race .048 .070 .040 .492 Sex .101 .066 .090 .126 Religious Attendance .027 .020 .084 .168 Marital Status .017 .058 .018 .770 Party ID .059 .052 .069 .257 Ideology Strength .001 .045 .002 .977 Ideology .043 .037 .087 .243 Attention to Pol. News TV .029 .031 .060 .341 Atten tion to Pol. News Paper .047 .026 .112 .076 a. N=4,350; R Square=.146; b. N=988; R Square=.051; c. N=267; R Square=.174; Estimated by OLS.

PAGE 54

54 Table 4 5 Effects of Interest in Campaign o n Voter Turnout: Civilians, Veterans, and Active Duty (NAES) Coe f. Std. Error Std. Coe f. Sig Civilian a (Constant) .279 .047 .000 Age Group .062 .005 .224 .000 Education .046 .006 .162 .000 Income .006 .005 .025 .273 Hispanic .144 .029 .098 .000 Race .061 .021 .057 .003 Sex .018 .014 .026 .193 Religio us Attendance .017 .005 .067 .001 Marital Status .034 .015 .048 .023 Party ID .062 .014 .085 .000 Ideology Strength .012 .010 .024 .230 Ideology .006 .007 .018 .379 Interested in Campaign .058 .010 .110 .000 Veteran b (Constant) .536 .096 .000 Age Group .043 .013 .154 .001 Education .024 .012 .100 .046 Income .003 .009 .017 .748 Hispanic .041 .076 .025 .589 Race .001 .041 .001 .977 Sex .036 .044 .038 .414 Religious Attendance .012 .010 .060 .233 Marital Status .072 029 .127 .014 Party ID .042 .026 .076 .109 Ideology Strength .002 .019 .006 .907 Ideology .019 .014 .068 .186 Interested in Campaign .027 .019 .070 .144 Active Duty c (Constant) .069 .191 .718 Age Group .091 .024 .204 .000 Educati on .068 .025 .165 .006 Income .028 .021 .081 .179 Hispanic .244 .089 .139 .007 Race .067 .059 .057 .256 Sex .062 .058 .054 .285 Religious Attendance .020 .018 .057 .283 Marital Status .042 .052 .042 .422 Party ID .017 .047 .019 .717 Ideology Strength .021 .042 .032 .607 Ideology .009 .034 .018 .784 Interested in Campaign .083 .034 .123 .016 a. N= 2315; R Square=.160; b. N=446; R Square=.101; c. N=341; R Square=.208; Estimated by OLS.

PAGE 55

55 Table 4 6 Effects of Trust in Government o n Voter Turnout: Civilians, Veterans, and Active Duty (NAES) Coe f. Std. Error Std. Coe f. Sig Civilian a (Constant) .376 .067 .000 Age Group .081 .007 .286 .000 Education .042 .008 .142 .000 Income .011 .007 .044 .129 Hispanic .194 .039 .126 000 Race .061 .027 .057 .022 Sex .014 .018 .019 .442 Religious Attendance .018 .007 .069 .012 Marital Status .023 .019 .032 .236 Party ID .058 .019 .076 .002 Ideology .013 .009 .037 .186 Ideology Strength .006 .013 .012 .631 Trust Gover nment .009 .015 .016 .526 Veteran b (Constant) .341 .152 .027 Age Group .078 .021 .255 .000 Education .020 .018 .086 .250 Income .023 .015 .120 .139 Hispanic .203 .132 .107 .125 Race .096 .068 .094 .159 Sex .032 .058 .038 .578 Religious Attendance .011 .015 .054 .470 Marital Status .019 .046 .031 .682 Party ID .052 .041 .089 .199 Ideology .014 .024 .046 .555 Ideology Strength .001 .031 .001 .985 Trust Government .017 .030 .038 .578 Active Duty c (Constant) .272 .205 .187 Age Group .073 .024 .166 .003 Education .079 .025 .194 .001 Income .036 .021 .106 .085 Hispanic .192 .092 .108 .038 Race .087 .060 .075 .148 Sex .070 .058 .062 .228 Religious Attendance .031 .018 .091 .091 Marital Stat us .049 .053 .049 .364 Party ID .007 .047 .007 .889 Ideology .006 .034 .011 .870 Ideology Strength .017 .042 .026 .690 Trust Government .077 .035 .115 .028 a N=1,396; R Square=.172; b. N=220; R Square=.133; c. N=326; R Square=.215; Estimated by OLS.

PAGE 56

56 Table 4 7 Effects of Demographics on Voter Turnout: Civilian s and Military (NAES) Coe f. Std. Error Std. Coe f. Sig. Civilian a (Constant) .268 .028 .000 Age Group .066 .003 .242 .000 Education .052 .004 .182 .000 Income .014 .003 .064 .000 Hispanic .159 .018 .108 .000 Race .085 .014 .078 .000 Sex .025 .009 .035 .005 Religious Attendance .023 .003 .092 .000 Marital Status .030 .009 .043 .001 Party ID .063 .010 .082 .000 Party ID Strength .056 .009 .077 .000 Ideology .003 .004 .009 .522 Ideology Strength .005 .006 .010 .423 Military b (Constant) .493 .056 .000 Age Group .052 .007 .204 .000 Education .026 .007 .107 .000 Income .019 .006 .103 .001 Hispanic .031 .037 .023 .403 Race .070 .024 .079 .004 Sex .039 .025 .044 .110 Religious Attendance .012 .006 .056 .048 Marital Status .009 .017 .016 .578 Party ID .059 .016 .101 .000 Party ID Strength .056 .017 .093 .001 Ideology .010 .009 .036 .239 Ideology Strength .005 .012 .014 .647 a. N=5,601; R Square =.177; b. N=1,258; R Square=.111; E stimated by OLS

PAGE 57

57 Table 4 8 Effects of Political Interest and Internal Efficacy on Voter Turnout: Civilian s and Military (NAES) Coe f. Std. Error Std. Coe f. Sig. Civilian a (Constant ) .123 .058 .033 Age Group .055 .006 .211 .000 Education .050 .007 .181 .000 Income .007 .006 .034 .183 Hispanic .138 .030 .100 .000 Race .042 .024 .039 .073 Sex .020 .015 .029 .188 Religious Attendance .026 .006 .107 .000 Marital Stat us .040 .015 .061 .009 Party ID .073 .016 .099 .000 Party ID Strength .040 .015 .057 .010 Ideology .009 .007 .030 .202 Ideology Strength .004 .010 .008 .698 Following Campaign Closely .022 .011 .051 .050 Follow Politics .031 .011 .078 .005 Po litics Too Complicated .001 .005 .006 .793 Military b (Constant) .382 .107 .000 Age Group .026 .011 .110 .018 Education .003 .011 .014 .774 Income .028 .009 .152 .003 Hispanic .011 .049 .010 .825 Race .027 .037 .032 .473 Sex .053 .039 .063 .176 Religious Attendance .013 .009 .069 .141 Marital Status .019 .026 .035 .460 Party ID .043 .024 .079 .077 Party ID Strength .037 .026 .065 .155 Ideology .015 .014 .056 .276 Ideology Strength .030 .019 .080 .107 Following Campaign Closely .037 .018 .104 .040 Follow Politics .075 .019 .213 .000 Politics Too Complicated .007 .008 .042 .379 a. N=1,860; R Square=.192; b. N=445; R Square=.194; Estimated by OLS.

PAGE 58

58 Table 4 9 Effects of External Efficacy on Voter Tur nout: Civilians and Military (NAES) Coe f Std. Error Std. Coe f. Sig Civilian a (Constant) .206 .102 .045 Age Group .067 .010 .246 .000 Education .049 .011 .184 .000 Income .021 .010 .095 .039 Hispanic .149 .056 .103 .008 Race .059 .040 057 .138 Sex .056 .026 .084 .033 Religious Attendance .020 .010 .083 .044 Marital Status .036 .029 .054 .208 Party ID .057 .029 .077 .048 Party ID Strength .053 .028 .073 .057 Ideology .009 .014 .028 .505 Ideology Strength .008 .019 .017 .661 Candidates Keep Promises .025 .017 .056 .147 Military b (Constant) .492 .204 .017 Age Group .069 .021 .297 .001 Education .004 .019 .020 .821 Income .012 .016 .072 .452 Hispanic .040 .082 .043 .624 Race .104 .066 .139 .119 Sex .020 .065 .027 .757 Religious Attendance .011 .017 .054 .524 Marital Status .035 .043 .071 .417 Party ID .101 .041 .204 .014 Party ID Strength .072 .043 .142 .095 Ideology .030 .021 .125 .167 Ideology Strength .006 .033 .016 .866 Candi dates Keep Promises .028 .026 .089 .279 a. N=600; R Square=.171; b. N=136; R Square=.242; Estimated by OLS.

PAGE 59

59 Table 4 10 Effects of Civic Duty on Voter Turnout: Civilians and Military (NAES) Coe f. Std. Error Std. Coe f. Sig Civilian a (Co nstant) .324 .086 .000 Age Group .061 .010 .239 .000 Education .039 .011 .150 .000 Income .018 .010 .090 .054 Hispanic .122 .057 .084 .032 Race .061 .039 .062 .112 Sex .056 .026 .087 .030 Religious Attendance .021 .010 .089 .037 Marita l Status .029 .028 .046 .290 Party ID .046 .028 .065 .101 Party ID Strength .058 .027 .085 .031 Ideology .010 .013 .032 .455 Ideology Strength .008 .018 .017 .669 Vote Because of Civic Duty .054 .024 .087 .025 Military b (Constant) .6 03 .176 .001 Age Group .065 .020 .283 .002 Education .002 .019 .011 .897 Income .012 .016 .070 .451 Hispanic .043 .082 .046 .597 Race .103 .064 .141 .111 Sex .027 .064 .036 .678 Religious Attendance .011 .016 .055 .514 Marital Status .0 33 .043 .066 .452 Party ID .103 .040 .209 .012 Party ID Strength .075 .042 .149 .077 Ideology .032 .021 .137 .137 Ideology Strength .008 .032 .023 .809 Vote Because of Civic Duty .009 .040 .019 .819 a. N=595; R Square=.146; b. N=138; R Squ are=.235; Estimated by OLS

PAGE 60

60 Table 4 11 Effects of Government Employment on Voter Turnout: Civilians and Veterans (NAES) Coef. Std. Error Std. Coef. Sig. Civilian a (Constant) .266 .038 .000 Age Group .069 .005 .216 .000 Education .051 006 .166 .000 Income .016 .005 .064 .001 Hispanic .172 .024 .120 .000 Race .103 .018 .094 .000 Sex .023 .012 .032 .056 Religious Attendance .028 .005 .104 .000 Marital Status .006 .013 .008 .665 Party ID .066 .013 .084 .000 Party ID Str ength .069 .012 .093 .000 Ideology .002 .006 .006 .748 Ideology Strength .008 .009 .016 .357 Government Employee .028 .013 .035 .037 Veteran b (Constant) .611 .096 .000 Age Group .046 .014 .158 .001 Education .017 .013 .067 .170 Inco me .022 .010 .112 .032 Hispanic .018 .058 .014 .751 Race .085 .040 .097 .034 Sex .050 .041 .056 .224 Religious Attendance .019 .010 .088 .062 Marital Status .029 .030 .048 .333 Party ID .046 .028 .077 .094 Party ID Strength .040 .027 .068 .141 Ideology .021 .016 .071 .174 Ideology Strength .017 .020 .041 .418 Government Employee .056 .027 .094 .041 a. N=3,057; R Square=.187; b. N=470; R Square=.113; Estimated by OLS.

PAGE 61

61 Table 4 12 Effects of Demographics on Voter T urnout: Active Duty Military (NAES) Coe f. Std. Error Std. Coe f. Sig Military (Constant) .084 .187 .653 Age Group .087 .025 .196 .000 Education .078 .025 .188 .002 Income .032 .021 .092 .143 Hispanic .232 .092 .131 .012 Race .074 .060 .06 4 .222 Sex .072 .059 .064 .225 Religious Attendance .028 .019 .081 .134 Marital Status .030 .055 .030 .580 Party ID .001 .048 .001 .987 Ideology .009 .035 .018 .792 Ideology Strength .021 .042 .032 .616 N= 321; R Square =.198; Estimated b y OLS Table 4 13 Effects of Military Service on Vo ter Turnout: Active Duty Military (NAES) N=300; R Square=.241; Estimated by OLS Coe f. Std. Error Std. Coe f. Sig Active Duty (Constant) .035 .207 .867 Age Group .054 .028 .121 .056 Education .053 .029 .128 .070 Income .018 .022 .054 .410 Hispanic .333 .101 .177 .001 Race .081 .061 .072 .184 Sex .077 .061 .069 .206 Religious Attendance .034 .019 .100 .070 Marital Status .022 .056 .022 .691 Party ID .049 .049 .055 .311 Ideology .004 .035 .008 .910 Ideology Streng th .007 .043 .010 .877 Rank .067 .064 .073 .299 Reason for Joining .018 .047 .021 .696 Appropriate for Member to Advocate for a Candidate .037 .024 .081 .134 Years in G uard or Reserves .050 .019 .313 .008 Years in Regular Mil itary .032 .018 .2 00 .072

PAGE 62

62 CHAPTER 5 RESULTS FROM 2011 PEW DATA This chapter will present results from the 2011 data collected by the Pew Research Center. Although the primary intent of both surveys was to assess generational ch ange and attitudinal differences regarding military matters, the instruments permit limited replication of the regressions just discussed. The veteran survey also includes additional service variables that appear in the literature, such as draft status an d combat experience. In all of the regressions that follow, the dependent variable is registration to vote and all independent variables have been coded to match the NAES values as closely as possible (see coding table in Appendix 2 for details). Pew Merged Samples Combining the samples from the two surveys produced 1,295 civilians and 1,932 military members for inclusion in the base model. Although the latter group is composed overwhelmingly of veterans, it also contains thirty two active duty me mbers who obviously could not be regressed separately. The descriptive statistics, located in Appendix 1, suggest these samples are generally similar to their NAES counterparts. A vast majority of both groups report being registered to vote, although the 9 1 % of service members wh o do so exceeds civilians by ten points. Table 5 1 presents the results from a one way ANOVA test, which confirms that the differences are statistically significant. These numbers are so high as to arouse a bit of suspicion that social desirability effects are at work, but I am unaware of any reason to suppose they operate differently across groups and the data do not permit an empirical test of this possibility. In keeping with official statistics, a greater percentage of civi lians fall on the lower end of the educational scale, although they are more likely to have a four year degree.

PAGE 63

63 Military members, by contrast, are more likely to have attended graduate or professional school. Overall, however, the differences are fairly small. Similarly, the gaps in income are minimal for those earning more than $20,000 a year, but there are twice as many civilians (21.2%) who fall below that point. Hispanics are slightly more prevalent in the general public sample, as are non whites by nearly ten points. Females compose over half of civilians (61%) but only twelve percent of service members. As before, the latter are far more likely to be married and, although the differences on the other demographic and political variables are fairly minor, they are all significant with the exception of party ID. The results from the base model are presented in Table 5 2. Most of the variables are once again significant for civilians, but there are a few noteworthy differences. The p values for ra ce and religious attendance exceed acceptable limits for the first time in a base model, and those for income (.030) and marital status (.070) are quite a bit higher than before. By contrast, the effects of Hispanic heritage and affiliating with a major p arty continue to exert significant influence in opposite (and expected) directions. For military members, the only variables that even approached significance were age, marital status, and interestingly enough, ideology. The latter squeaks by with a p va lue of .099, and the negative coefficient suggests that liberal service members are less likely to vote than conservatives. However, the consistently poor performance of the two ideology measures and the razor thin margin of significance raise questions a bout whether the results have any substantive value. Finally, it is interesting to note that the coefficients for race and Hispanic heritage are both positive, if insignificant.

PAGE 64

64 The second model in Table 5 3 incorporates a measure of patriotism, an el usive concept that can nevertheless be loosely interpreted as a distant relative of civic duty. Respondents were asked whether they are less (1), about as (2), or more (3) patriotic than most of the country. The regression includes the 1,272 civilians an d 1,895 military members who provided valid answers. The frequencies for this measure are rather interesting, as less than 10% of either group chose the neutral ground. Instead, a somewhat striking 53.7% of civilians reported they are less patriotic than most, as opposed to the 60.8% of service members who claimed to be more so. The results of the regression are displayed in Table 5 3. Although patriotism was positively associated with registration for both groups, it was only significant for civilians ( p=.023). It seems that military members are more likely to possess (or at least believe they have) this quality and yet less likely to be influenced by it. Adding this item to the equation had relatively little impact on the other independent variables for either sample. Among civilians, although there were minor fluctuations in coefficients and p values, the same predictors from the base model remain significant. For military members, the results are nearly identical except that party ID and, far mor e predictably, ideology are no longer statistically associated with registration. The final model for the merged samples, presented in Table 5 4, attempts to replicate the NAES results regarding the effects of government employment. A dummy variable is coded 1 for those who satisfy this criterion and 0 for those who are otherwise employed. Of the 809 civilians in the sample, approximately 20% work for the government, as opposed to 30% of the 930 veterans in the sample. (For the sake of

PAGE 65

65 consistency, act ive duty members were excluded from this analysis. However, a separate regression shows that, in this case, their inclusion made no difference.) Table 5 4 displays the results of the analysis, which contrast sharply with those obtained using the NAES dat a. This time, government employment was positively related to registration but quite insignificant for both groups, with p values above .400. These differences are fairly dramatic and do not appear to be the product of sample size; although there were th ree times as many civilians in the NAES regression, the veteran sample was reduced by half compared to the Pew analysis. The mere existence of these contradictory findings makes it unwise to draw firm conclusions about the effects of this variable for eit her civilians or military members based on these data. There were relatively few changes to the other independent variables. Among civilians, education and income are now significant only at the 0.1 level (p values = .087 and .086 respectively), and marit al status no longer satisfies even this criterion. For veterans, age continues to be extremely influential, while the positive coefficients for being non white or Hispanic take on significance for the first time. Both p values just miss the 0.5 cutoff po int and the magnitudes of .111 for Hispanics and .077 for non whites are rather impressive. This could simply reflect the reduced Ns in the model, but the positive (if insignificant) relationship for Hispanics was also noted in the NAES study and the base model above. Pew Veteran Sample Three hundred thirty military members, including those on active duty, were interviewed as part of the general public survey. Because they were not asked any of the service related questions, they have been excluded leaving 1,602 veterans for this analysis. The results of the base model regression appear in Table 5 5. Age,

PAGE 66

66 education, religious attendance, marital status and party ID are all extremely significant, while income has a borderline p value of .059. The Hispanic variable on c e again has a positive coefficient and the association is negative for non whites, but neither effect is significant. Overall, these results are fairly consistent with those discussed above. Model 2 in Table 5 6 introduces a num ber of service related variables that can be loosely classified as military demographics, experience and benefits. The first category consists of rank, years of service, combat experience and motivation for joining the armed forces. The measures for rank and years of service are identical to those in the NAES regression, while dummy variables are included for the other two. Respondents with combat experience were coded (1), as were those who said serving the country or fighting terrorism were important r easons they joined. There are two military experience variables. The first is a five point scale measuring whether service hurt or helped respondents get ahead in life, with a higher score indicating the latter. The second measure combines four questio ns about how useful military experience was for teaching cooperation, building self confidence, job/career preparation and personal growth. This four point scale ranges from not at all useful (1) to very useful (4). Three types of military benefits are included in this model, the first two of which are dummies. Respondents were asked whether the government has given veterans all the help that it should and whether they have received VA benefits. In both cases, a e indicating how well the VA is meeting

PAGE 67

67 The results are displayed in Table 5 6 and reveal that overall the service specific variables once again perform poorly. They also have relatively little impact on the other independent variables; although education and income drop out as significant predictors, the rest of the base variables retain their influence. None of the military demographics were significant, including years of service, and all of them h ave negative coefficients. For combat and years of service, this direction is confirmed by significant bi variate correlations with registration. Previous studies have yielded conflicting results regarding the effects of combat on political participation and in some ways a negative relationship makes intuitive sense. The effects of length of service, however, are a bit more puzzling. In addition to being positive and significant, the NAES data showed a difference between years served in one of the fou r major branches and time in the National Guard or Reserves. The Pew survey allowed respondents to list as many branches as apply, a practice I heartily endorse, but it did not separate out length of service for each. It is also possible that the differe nt findings reflect sample composition, as the NAES survey was limited to active duty members. If that is the case, recency of service might explain the results obtained among veterans. The military experience variables yielded mixed results, as the sca le for usefulness of service did not attain significance. This was not all that surprising since only one of the original items, that for building confidence, had a significant zero order correlation with registration (p=.021). However, the measure for w hether military experience helped respondents get ahead in life performed quite well (p=.001) and has a greater magnitude (.036) than age in this model (.034). The two experience

PAGE 68

68 measures are highly correlated (.443), so it is interesting that one should outshine the other so dramatically. It might be that the usefulness measures were too specific or excluded relevant response options, while the question about getting ahead allowed respondents to decide exactly what was being asked. Of course, such vagu e question wording also leaves researchers to wonder what the response actually means. Nevertheless, it seems that military service does provide resources that are positively associated with political participation. Both receipt of VA benefits and respondent evaluations of that department were insignificant and, while the coefficients were negative, they were also minute ( .008 and .006, respectively). It is possible that those who need VA assistance are more likely to be disadvantaged, disabl ed or infirm and therefore incur higher administrative costs, but this cannot be confirmed with the available data. The measure for whether respondents are receiving sufficient help from the government, on the other hand, was borderline significant (p=.08 0) and the coefficient of .034 is identical to that for age. Once again, however, the question is so ambiguous as to provide little information about exactly hy pothesize that this variable provides an indirect measure of political efficacy or simply reflects self interest, but it is difficult to say for sure without additional information. Separate regressions were run for two additional variables that, ideally, would have been included in the model above. Doing so, however, would have reduced the Ns by several hundred and increased standard errors quite a bit. Model 3 in Table 5 7 therefore returns to the base model and adds a dummy for status as a draftee or v olunteer. Just over a quarter of the sample (26.3%) falls into the former category and

PAGE 69

69 is coded as 0; the vast majority of the 840 respondents who volunteered are coded as 1. Both the bi variate correlation and regression coefficient are insignificant (p =.464 and .398, respectively), although the first value is positive and the second negative. It is unclear why volunteering for service would be negatively associated with registration or why sex attains borderline significance (p=.065) in this model, des pite the fact that females represent only three percent of the sample. The rest of the regression results in Table 5 7, however, are entirely consistent with the base model, indicating the introduction of this additional variable had relatively little ove rall impact. The final model in Table 5 8 expands the base to include a measure of how well duty. The mean for the 907 veterans included in the regression is 2.96 on a scale that ranges from poor (1) to excellent (4). Although the bi variate correlation is positive and significant (p=.043), the regression results presented in Table 5 8 are similar to those obtained for the other military benefits. Recency effects coul d be partially responsible for the extremely large p value (.834) and miniscule magnitude (.002), but the sample included a large number of veterans who served after September 11, 2001 (ten years prior to the survey). It might also be that such evaluation s are simply no longer relevant to former service members. Either way, the rest of the independent variables are identical to the base model, except for sex, which is once again significant (p=.007). Comparisons between these findings and those from the NAES survey will be discussed in the next chapter.

PAGE 70

70 Table 5 1 One Way ANOVA Test of Means for Registration and Demographics: Civilians and Veterans (Pew) Sum of Squares df Mean Square F Sig. Registered to Vote Between Groups 13.835 1 13.835 71.960 000 Within Groups 738.455 3841 .192 Total 752.289 3842 Religious Attendance Between Groups 8.206 1 8.206 3.919 .048 Within Groups 7997.082 3819 2.094 Total 8005.288 3820 Marital Status Between Groups 38.964 1 38.964 169. 247 .000 Within Groups 883.110 3836 .230 Total 922.074 3837 Party ID Between Groups .360 1 .360 .834 .361 Within Groups 1620.646 3756 .431 Total 1621.006 3757 Ideology Between Groups 33.470 1 33.470 26.700 .000 Within Groups 4648.219 3708 1.254 Total 4681.689 3709 Ideology Strength Between Groups 18.600 1 18.600 21.330 .000 Within Groups 3233.510 3708 .872 Total 3252.110 3709

PAGE 71

71 Table 5 2 Effects of Demographics on Voter Registration: Civilian s and Veterans (Pew) Coe f. Std. Error Std. Coe f. Sig Civilian a (Constant) .548 .057 .000 Age Group .001 .001 .028 .273 Education .033 .009 .103 .000 Income .014 .008 .054 .089 Hispanic .285 .035 .216 .000 Race .052 .024 .058 .031 Sex .102 .021 .129 .000 Religious Attendance .012 .007 .044 .109 Marital Status .067 .022 .087 .003 Party ID .094 .021 .119 .000 Ideology .009 .011 .025 .368 Ideology Strength .001 .015 .002 .942 Military b (Constant) .613 .042 .000 Age Group .010 .003 .090 .000 Education .029 .006 .112 .000 Income .005 .005 .025 .337 Hispanic .001 .027 .001 .979 Race .032 .018 .040 .085 Sex .009 .020 .010 .671 Religious Attendance .017 .005 .084 .000 Marital Status .089 .015 .146 .000 P arty ID .067 .013 .115 .000 Ideology .007 .008 .023 .363 Ideology Strength .013 .010 .032 .198 a. N=1,309; R square = .153; b. N=1,881; R square = .091; Estimated by OLS.

PAGE 72

72 Table 5 3 Effects of Patriotism on Voter Registration: Civilians and Vetera ns (Pew) Coe f. Std. Error Std. Coe f. Sig Civilian a (Constant) .383 .063 .000 Age Group .045 .008 .163 .000 Education .034 .009 .106 .000 Income .016 .008 .065 .038 Hispanic .272 .035 .207 .000 Race .024 .024 .027 .319 Sex .080 .021 .1 02 .000 Religious Attendance .009 .007 .035 .201 Marital Status .039 .022 .051 .080 Party ID .081 .021 .103 .000 Ideology .006 .011 .016 .565 Ideology Strength .002 .015 .003 .914 Patriotic .025 .011 .060 .023 Veteran b (Constant) 634 .068 .000 Age Group .053 .009 .143 .000 Education .007 .011 .015 .560 Income .009 .009 .025 .339 Hispanic .056 .048 .027 .241 Race .026 .033 .019 .418 Sex .054 .036 .036 .133 Religious Attendance .009 .008 .026 .268 Marital Status .069 .026 .065 .008 Party ID .005 .015 .007 .753 Ideology .015 .010 .035 .152 Ideology Strength .000 .011 .001 .973 Patriotic .002 .011 .005 .204 a. N=1,272; R Square=.182; b. N=1895; R Square=.033; Estimated by OLS.

PAGE 73

73 Table 5 4 Effects of G overnment Employment on Voter Registration: Civilians and Vet eran s (PEW) Coef. Std. Error Std. Coef. Sig Civilian a (Constant) .502 .078 .000 Age Group .040 .011 .132 .000 Education .021 .012 .064 .087 Income .017 .010 .068 .086 Hispanic .28 9 .044 .222 .000 Race .029 .032 .031 .366 Sex .076 .026 .098 .004 Religious Attendance .010 .010 .037 .295 Marital Status .022 .028 .029 .432 Party ID .099 .026 .126 .000 Ideology .007 .014 .018 .600 Ideology Strength .011 .019 .019 .5 75 Government Employee .026 .033 .027 .422 Veteran b (Constant) .697 .074 .000 Age Group .045 .012 .129 .000 Education .025 .015 .060 .108 Income .001 .012 .004 .914 Hispanic .111 .057 .065 .051 Race .077 .040 .066 .056 Sex .023 .04 3 .018 .593 Religious Attendance .001 .011 .002 .954 Marital Status .016 .035 .017 .644 Party ID .002 .019 .003 .931 Ideology .015 .013 .041 .262 Ideology Strength .001 .014 .003 .923 Government Employee .002 .025 .003 .931 a. N=809; R S quare=.167; b. N=930; R Square=.033; Estimated by OLS.

PAGE 74

74 Table 5 5 Effects of Demographics on Voter Registration: Veterans (Pew) Coef. Std. Error Std. Coef. Sig. (Constant) .527 .048 .000 Age Group .038 .006 .173 .000 Education .020 .007 .075 .00 5 Income .011 .006 .052 .059 Hispanic .022 .030 .018 .471 Race .011 .020 .013 .596 Sex .029 .023 .031 .215 Religious Attendance .014 .005 .069 .006 Marital Status .096 .016 .153 .000 Party ID .065 .014 .110 .000 Ideology .007 .009 .023 .408 Id eology Strength .007 .011 .016 .554 N=1602; R Square = .114; Estimated by OLS. Table 5 6 Effects of Military Service on Voter Registration: Veterans (Pew) Coe f. Std. Error Std. Coef. Sig (Constant) .419 .090 .000 Age Group .034 .007 .149 .000 Edu cation .015 .009 .054 .107 Income .006 .007 .029 .379 Hispanic .034 .035 .027 .334 Race .012 .024 .015 .610 Sex .041 .027 .045 .128 Religious Attendance .017 .006 .083 .005 Marital Status .110 .020 .167 .000 Party ID .071 .017 .116 .000 Ideology .009 .011 .027 .409 Ideology Strength .008 .014 .018 .571 Rank .017 .026 .020 .531 Years of Service .003 .004 .021 .462 Combat .019 .018 .032 .284 Reason for Joining Serve Country .010 .035 .008 .770 Service Helped Get Ahead .036 .011 .10 7 .001 Received Gov. Help as Veteran .034 .020 .054 .080 Received VA Benefits .008 .018 .014 .636 Veterans Affairs Evaluation .006 .011 .016 .607 Service Useful .000 .005 .001 .966 N=1,167; R Square = .147; Estimated by OLS.

PAGE 75

75 Table 5 7 Effects of Draft versus Volunteer Status on Voter Registration: Veterans (Pew) Coef Std. Error Std. Coef Sig (Constant) .524 .091 .000 Age Group .039 .016 .081 .018 Education .020 .007 .107 .004 Income .012 .006 .073 .063 Hispanic .017 .039 .015 .668 Race .007 .026 .009 .790 Sex .087 .047 .062 .065 Religious Attendance .016 .005 .107 .002 Marital Status .097 .017 .199 .000 Party ID .043 .015 .095 .005 Ideology .000 .009 .002 .969 Ideology Strength .005 .012 .016 .682 Draftee or Volunteer .014 .017 .028 .398 N=840; R square=.112; Estimated by OLS. Table 5 8 Effects of Military Response to Needs During Service on Registration: Veterans (Pew) Coe f. Std. Error Std. Coe f. Sig (Constant) .449 .076 .000 Age Group .043 .008 .179 .000 Educati on .021 .009 .080 .027 Income .018 .008 .089 .018 Hispanic .067 .042 .051 .112 Race .025 .025 .032 .327 Sex .076 .028 .090 .007 Religious Attendance .012 .007 .061 .069 Marital Status .098 .023 .144 .000 Party ID .068 .019 .117 .000 Ideology .00 9 .012 .030 .440 Ideology Strength .014 .016 .033 .378 Military Met Needs on Active Duty .002 .011 .007 .834 N=907; R Square=.125; Estimated by OLS.

PAGE 76

76 CHAPTER 6 DISCUSSION AND CONCLUSION This thesis has explored a number of possible explanations for t he higher turnout rate among military members. The large samples in the NAES and Pew survey sets provided a relatively rare opportunity to compare veterans and active duty members to one another, as well as the general public. Although there are limitati ons to these data, as discussed below, my results suggest both similarities and differences across all three groups. Table 6 1 details the number and percentage of mo dels in which each variable some of the most interesting findings involve items that were not statistically influential, including most of the service related variables. T he samples that combined former and current military members were overwhelming ly comprised of veterans, but the fact that active duty members were included warrants a separate category. One of the primary research questions for this thesis was how well existing theories explain military turnout. Although the data do not permit an empirical test relating to rational choice, the active duty members in the NAES sample were less likely to report having voted than either civilians or veterans. There is a g ood bit of literature suggesting that higher administrative costs and institutional factors are at least partially responsible for their lower participation rates (Wright 2008; Eversole 2010; Alvarez, Hall & Roberts 2007; Inbody 2009). The fact that the f ederal government has implemented assistance programs, revised voting laws, and sued a number of states for noncompliance lends credence to this assertion. According to Smith (2010), reports of higher turnout among this group are inflated because the Fede ral Voter Assistance Program (FVAP) tallies attempts to vote rather than counted ballots. If this is the case,

PAGE 77

77 then active duty members might not be abstaining after all. Either way, it is extremely difficult to obtain an accurate total of active duty vo ters because states and government agencies use a variety of sampling and reporting procedures. Socio economic theories perform fairly well for the civilian samples: age, education, income, and being Hispanic or female are significantly related to voti ng or registration in at least 75% of the models. Age exerts this statistical influence in all models for all groups, but the other SES variables yield mixed results for the military samples. For example, income is significant in only half of the models for both veterans and active duty members, while education is a far more consistent predictor for the latter (who are similar to civilians). Hispanics were significantly less likely to vote than whites in 83% of the active duty models, but the relationshi p is positive and rarely significant among veterans. Being non white, on the other hand, did not have a p value below 0.1 in a single model for those currently in uniform. This is fairly striking considering race was significant in almost 60% of the civi lian regressions, with veterans and the combined samples falling in between. It is unclear why military service should mitigate the effects of being non white but not Hispanic; past literature has found the reverse to be true, at least with regard to voti ng (Leal 1999, Ellison1992). This might indicate that military socialization (if any) is indeed temporary. Finally, females were more likely to participate in 75% of ci vilian models, but sex was never significant for active duty members and rarely so for veterans. Apparently women in uniform are no more or less likely to vote than their far more numerous male counterparts. At least one of the military groups is fairly distinctive on every socio economic variable except

PAGE 78

78 for age, suggesting that SES theories do not provide sufficient explanation for military participation. The rest of the variables can be considered resources that theoretically reduce the costs or inc rease the benefits of voting. In terms of demographic and political identity variables, veterans resemble civilians much more closely than their successors. Religious attendance was significant in twice as many models for the general public and veterans (70 75%) as for those on active duty (33%). And even though current service members are more likely to be married, this status did not attain significance in a single regression; it was a statistical predictor in half of the models for the other two group s. Similarly, the party ID variable (coded 1 for Republican or Democrat) behaved predictably for civilians and veterans a majority of the time but was insignificant in every active duty analysis. Unfortunately, this sample was not asked about strength of party ID, which was significant in 75% of the models for the combined military samples, as opposed to 100% for civilians. With one anomalous exception, both of the ideological scales were insignificant across the board. Frankly, I can offer no explanati on since they are common measures and have only been re coded to reverse direction. Overall, however, the results suggest that the effects of demographic variables are less powerful and consistent for those currently in uniform. The results pertaining t o political resources are conflicting. Attention to political news and interest in the campaign were positive and significant for civilians and current service members but apparently had little impact on veterans. At the same time, measures of how closel y respondents followed politics and the campaign were significant for the combined military sample, which was primarily composed of those

PAGE 79

79 exact same veterans. It is difficult to account for this discrepancy other than to note that single measures which ca nnot be combined into a comprehensive scale are notoriously unreliable. Given the vast literature attesting to the power of interest and attention and the overall pattern of my results, it seems safe to conclude that these are important political resource s for military members and the general public alike. The political efficacy variables, by contrast, were insignificant across all groups. (The exception for trust in government among active duty members was previously discussed). Their lack of influenc e is not terribly surprising since such concepts are difficult to capture, especially with a single item. Even if there was a significant difference in the efficacy of each group, it is unlikely the available measures would reflect it. Conflicting result s from the two studies make it difficult to assess the impact of government employment; however, the results were either significant or not for both civilians and veterans, suggesting little difference between them. Finally, despite frequencies that favor service members and my own expectations, the patriotism and civic duty variables were significant only for civilians. Overall, the results suggest that there are substantive differences between civilians, veterans and active duty members, especially reg arding the influence of SES, demographic, and political identity variables. However, no such discrepancies were detected for the effects of political resources such as interest, attention and efficacy. If the usual suspects are, at best, equally powerful for military members, what accounts for their higher rate of turnout? To test whether service in the armed forces is responsible, I regressed a number of military specific variables for active duty members from the NAES survey and veterans from the Pew s tudy. The results were

PAGE 80

80 inconsistent across the two studies. There is, however, one interesting finding: those who reported that service helped them get ahead in life were s tatistically more likely to be registered to vote. Although the question was infuriatingly vague, it does suggest that military service can foster political skills and participation. Veterans who said the government had given them all the help it should were also more likely to vote, but this seems to reflect external efficacy, i.e. constituent service as opposed to military service. There are significant limitations to both the NAES and Pew surveys. The NAES data measured voting behavior four years pr ior to the survey and, because self reports cannot be verified, there is likely a good bit of measurement error. The Pew study measured current registration rather than voting. This was an acceptable approximation for the purposes of this paper, but more recent and precise measures would have been preferable. In addition to the drawbacks of using single measures discussed above, split sampling and reduced response rates often required separate regressions. Such analyses might conceal relationships or in teractions with variables from another model. In other words, the results paint a partial picture. Several of the most interesting questions were limited to relatively few respondents or were excluded from the military instruments altogether. Additional ly, about half of the veterans in the Pew sample had served (and separated) since September 11, 2001. Although such oversampling was perfectly appropriate and probably necessary for their purposes, it might have skewed the results obtained in my analyses. Finally, uniformed voters serving overseas were excluded from both survey sets, a reasonable practice considering the logistics of

PAGE 81

81 obtaining a representative sample here in the states. However, a discussion of the costs of military voting should probabl y include those who are most likely to pay them. Future research could shed additional light on military turnout, first and foremost, by conducting a study specifically dedicated to that topic. The data used here were collected almost as an afterthoug ht; the primary purpose of the surveys was to assess political attitudes and opinions regarding the campaign and the wars in Iraq and Afghanistan. There is scant literature regarding whether or not military members actually act on these preferences or why they choose to do so. Expanding the sample composition to include service members who are stationed overseas would allow for an empirical test of the conventional wisdom that uniformed voters are disenfranchised. Government agencies attempt to track whe ther or not they vote and if they make use of available programs but seldom address why they want to cast a ballot in the first place. Open ended follow ups regarding how military service helped respondents get ahead in life could help design better quest ions and response options. In addition to measuring rank, it might be useful to include questions about specialty or type of service in the armed forces. Additionally, this study suggests there are a number of differences between former and current milit ary members that warrant a more thorough investigation, and measuring recency of service could help distinguish between self selection and socialization effects. Military members could be placed in a broader theoretical context if background information w ere collected, a practice that might also reveal whether military provision of resources or socialization is greater among the disadvantaged. Finally, future efforts to compare civilians and the military should ask comparable and thorough questions of eac h sample.

PAGE 82

82 Ultimately, this study found that veterans do indeed register and vote at higher rates than civilians, while the opposite is true of active duty members. References to guish between former and current service members. Why these two groups differ from one another, as well as civilians, is not yet clear. However, the distinctions are both impressive and consistent and warrant further exploration.

PAGE 83

83 Table 6 1 Signif icance of Variables: Number and Percentage of Models Civilians Veterans Active Duty Former & Current Military Num. Pct. Num. Pct. Num. Pct. Num. Pct. Age 12 100% 10 100% 6 100% 6 100% Education 12 100% 6 60% 6 100% 1 17% Income 9 75% 5 50% 3 50% 2 33% Hispanic 12 100% 1 10% 5 83% 0 0% Race (Non White) 7 58% 3 30% 0 0% 1 17% Sex (Female) 9 75% 2 20% 0 0% 0 0% Religious Attendance 9 75% 7 70% 2 33% 1 17% Married 6 50% 5 50% 0 0% 2 33% Party ID (Major Party) 11 92% 7 70% 0 0% 4 66% Par ty ID Strength 5 100% 0 0% 0 N/A 3 75% Ideology (Very Liberal) 0 0% 0 0% 0 0% 1 17% Ideology Strength 0 0% 0 0% 0 0% 0 0% Attention to Pol. News TV 1 100% 0 0% 0 0% 0 N/A Attention to Pol. News Paper 1 100% 0 0% 1 100% 0 N/A Interested in Camp 1 1 00% 0 0% 1 100% 0 N/A Trust in Government 0 0% 0 0% 1 100% 0 N/A Following Camp 1 100% 0 N/A 0 N/A 1 100% Follow Politics 1 100% 0 N/A 0 N/A 1 100% Politics Too Complicated 0 0% 0 N/A 0 N/A 0 0% Candidates Try to Keep Promises 0 0% 0 N/A 0 N/A 0 0% C ivic Duty 1 100% 0 N/A 0 N/A 0 0% Gov. Employment 1 50% 1 50% 0 N/A 0 N/A Patriotism 1 100% 0 N/A 0 N/A 0 0%

PAGE 84

84 Table 6 2 Significance of Service Variables Significant Active Duty Members Rank No Years of Service Yes Reason for Joining To Serve Country/Fight Terrorism No Appropriate to Advocate for Candidates No Veterans Rank No Years of Service No Reason for Joining To Serve Country/Fight Terrorism No Combat Experience No Service Useful No Service He lped Get Ahead Yes Received Government Help as a Veteran Yes Received Veterans Affairs Benefits No Veterans Affairs Evaluation No Drafted No Military Met Needs No

PAGE 85

85 APPENDIX A VARIABLE CODING RULES This appendix lists the coding rules for each variab le included in the analyses. Independent variables have been subdivided into socio economic factors and demographics, political identity and attitudes, and military specific variables. Within each section, items appear in alphabetical order. Unless othe rwise noted, the scales and measures used in my regressions are identical for both studies. Dependent Variables Voted in the 2000 General Election (NAES) and currently registered to vote (Pew) were both coded as dummies: 0=no; 1= yes. Socio Economic a nd De mographic Independent Variables Age is coded as an ordinal scale that includes five categories: one =18 29; two= 30 39; three= 40 49; four= 50 64; and five = 65 or older. Education also has five levels: 1= less than high school diploma; 2=high schoo l and 5= graduate or professional school. Hispanic heritage is denoted by a dummy variable: 0=no, 1=yes. The two surveys differ slightly, as the Pew question includes Spanish as well as Latino descent, but are otherwise identical. Income is measured differently in the two studies, but the ordinal scales have been aligned as closely as possible. For the NAES surveys, income includes the following categories: one= les s than $25,000; two =$25 50,000; three= $50 75,000, four= $75 100,000; five =$100 150,000 and six= $150,000 or higher. The Pew scale

PAGE 86

86 also has six levels, which include: one=less than $20,000; two= $20 40,000; three= $40 75,000; four= $75 100,000; five= $1 00 150,000 and six =$150,000 or higher Marital status has been re coded into a dummy variable so that 0=unmarried and 1=married. Race is also captured by a dummy, with 0 indicating white and 1 non white or mixed race. The two surveys provided different response categories, but are thus coded to match. Religious attendance is measured on an ordinal scale that includes: 1=seldom or never; 2= a few times a year; 3= once or twice a month; 4= once a week, or 5=more than once a week. Political Independent Variables Attention to political news during the past week (via cable/network television or newspaper) was measured on four tier scales: none at all (1), not much (2), some (3) and a great deal (4). Candidates try to keep promises is a reflection of how often respondents believe candidates try to follow through once elected: 1=never, 2=rarely, 3=sometimes, 4=most times, 5= always. Civic duty vote because of candidates, issues or other rea sons are coded as 0; those motivated by re coded as missing values). Following campaign closely is also captured by an ordinal scale with four levels: 1=not at all, 2 = not too, 3=somewhat, 4=very.

PAGE 87

87 Follow politics is a more general assessment of interest, with possible options including: 1=hardly at all, 2=now and then, 3=sometimes, and 4=most times. Government employment is measured on a simple dummy variable: 0= not employed by government, 1=employed by government. The Pew question asks whether respondents are employed by government or the military, while the NAES excluded the latter. Ideology is measured on a standard five point scale ranging from very conservati ve (1) to very liberal (5), with self reported moderates coded as a three. Ideological strength falls on a three point scale: 1=moderate, 2=liberal or conservative, and 3=very liberal or conservative. Interest in the campaign includes three possible res ponse options: 1=not much, 2=somewhat, 3 = very. Party ID has been recoded into a dummy with 0 indicating Independent, Other, or No Preference and 1 representing Republican or Democrat. Party ID strength was measured only in the NAES panel survey. It i s coded so Patriotism is a subjective measure of whether respondents believe they are (1) less, (2) about as, or (3) more patriotic than most people in the country. Politics are too complicated measures whether respondents believe people like themselves can understand political matters. A standard 5 point agree/disagree scale is used, with 1 representing strong disagreement with that statement and 5 denoting strong agreement.

PAGE 88

88 Tr ust in government measures how much of the time respondents trust the federal government to do the right thing. There are four possible responses on this ordinal scale: 1= never, 2= sometimes, 3= most times and 4=always. Military Specific Independent Va riables Appropriate for military member to advocate for local or presidential candidates is coded as follows: 1=never inappropriate, 2=appropriate sometimes (i.e. for one or the other), and 3= always appropriate. Combat experience is coded so that 0 reflec ts a lack of experience and 1 identifies those who did engage in combat. Draft indicates whether respondents were selected for duty (0) or volunteered for service in the armed forces (1). Government help as veteran assesses whether respondents believ e the government has provided them all the assistance it should. It is a simple dummy coded as 0=no and 1=yes. Military responsiveness assesses how well veterans believe the military responded to their needs and those of their families while on active dut y. Possible responses included: 1=poor, 2=only fair, 3=good, 4= excellent. Rank is coded as a dummy, with 0 denoting enlistee and 1 indicating officer. In measures the current rank of active duty members. Service to country/fighting terrorism as reasons for joining the military are coded as 1, while personal benefits such as education and career training are coded as 0.

PAGE 89

89 Service helped get ahead in life is measured o n a 5 point scale that includes: 1=hurt a lot, 2= hurt a little, 3=made no difference, 4=helped a little or 5=helped a lot. Those who indicated it was too soon to tell were re coded as missing. Service useful measures whether respondents believe militar y experience helped them develop self confidence, personal growth, cooperation and job skills. The original four scales were combined into one that has four possible categories: 1=not at all useful, 2=no t too useful, 3=somewhat useful and 4=very useful. VA benefits those who have not (0). VA evaluation notes how well that department has cared for veterans and can take the following values: 1=poor, 2=only fair, 3=good, 4=excellent. Years of service fall on a six point scale that i ncludes: one= less than a year ; two =1 2 years; three equals 2 3 years; four = 3 5 years; five =5 10 years and six = 10 or more years. The NAES survey includes separate measures for regular military versus Natio nal/Guard Reserves; respondents who did not serve in that particular branch are coded as 0.

PAGE 90

90 APPENDIX B DESCRIPTIVE STATISTICS FOR M ODELS Table B 1. Descriptive Statistics for Demographi cs Model: Civilians, Veterans and Active Duty (NAES) Min. Max. Mea n Std. Dev. N Civilian Voted in 2000 General 0 1 .86 .348 6308 Age Group 1 5 3.21 1.237 6308 Education 1 5 3.38 1.201 6308 Income 1 6 2.93 1.501 6308 Hispanic 0 1 .06 .232 6308 Race 0 1 .11 .314 6308 Sex 0 1 .64 .479 6308 Religious Attendanc e 1 5 2.91 1.337 6308 Marital Status 0 1 .60 .490 6308 Party ID 0 1 .68 .466 6308 Ideology Strength 1 3 1.78 .689 6308 Ideology 1 5 2.87 1.030 6308 Veteran Voted in 2000 General 0 1 .92 .265 1293 Age Group 1 5 4.02 .990 1293 Education 1 5 3.40 1.126 1293 Income 1 6 3.03 1.452 1293 Hispanic 0 1 .04 .191 1293 Race 0 1 .10 .299 1293 Sex 0 1 .10 .298 1293 Religious Attendance 1 5 2.89 1.328 1293 Marital Status 0 1 .67 .470 1293 Party ID 0 1 .65 .478 1293 Ideology Strength 1 3 1.76 696 1293 Ideology 1 5 2.61 .954 1293 Active Duty Voted in 2000 General 0 1 .76 .427 384 Age Group 1 5 2.13 .964 384 Education 1 5 3.49 1.055 384 Income 1 6 3.00 1.259 384 Hispanic 0 1 .06 .238 384 Race 0 1 .16 .368 384 Sex 0 1 .17 .373 384 Religious Attendance 1 5 2.91 1.258 384 Marital Status 0 1 .74 .437 384 Party ID 0 1 .66 .473 384 Ideology Strength 1 3 1.65 .652 384 Ideology 1 5 2.55 .809 384

PAGE 91

91 Table B 2 Descriptive Statistics for Attention to Politics: Civilians, Veterans a nd Active Duty (NAES) Min. Max. Mean Std. Dev. N Civilian Voted in 2000 General 0 1 .89 .312 4350 Age Group 1 5 3.33 1.214 4350 Education 1 5 3.46 1.175 4350 Income 1 6 3.05 1.514 4350 Hispanic 0 1 .04 .203 4350 Race 0 1 .11 .309 4350 Sex 0 1 .65 .478 4350 Religious Attendance 1 5 2.93 1.312 4350 Marital Status 0 1 .62 .486 4350 Party ID 0 1 .69 .462 4350 Ideology Strength 1 3 1.76 .682 4350 Ideology 1 5 2.87 1.010 4350 Attention to Pol. News 1 4 3.05 .873 4350 Attenti on to Pol. News Paper 1 4 2.92 .906 4350 Veteran Voted in 2000 General 0 1 .95 .219 988 Age Group 1 5 4.08 .955 988 Education 1 5 3.47 1.128 988 Income 1 6 3.12 1.467 988 Hispanic 0 1 .03 .174 988 Race 0 1 .10 .294 988 Sex 0 1 .09 .281 988 Religious Attendance 1 5 2.90 1.306 988 Marital Status 0 1 .69 .465 988 Party ID 0 1 .65 .478 988 Ideology Strength 1 5 1.73 .694 988 Ideology 1 3 2.65 .946 988 Attention to Pol. News 1 4 3.14 .868 988 Attention to Pol. News Paper 1 4 2.99 .935 988 Active Duty Voted in 2000 General 0 1 .79 .408 267 Age Group 1 5 2.17 .980 267 Education 1 5 3.54 1.019 267 Income 1 6 3.10 1.238 267 Hispanic 0 1 .06 .231 267 Race 0 1 .13 .342 267 Sex 0 1 .16 .365 267 Religious Attendanc e 1 5 2.98 1.264 267 Marital Status 0 1 .76 .430 267 Party ID 0 1 .64 .482 267 Ideology Strength 1 5 1.63 .650 267 Ideology 1 3 2.61 .813 267 Attention to Pol. News 1 4 3.14 .833 267 Attention to Pol. News Paper 1 4 2.73 .974 267

PAGE 92

92 Table B 3. Descriptive Statistics for Political Interest Model: Civilian, Veterans and Active Duty (NAES) Min. Max. Mean Std. Dev. N Civilian Voted in 2000 General 0 1 .87 .341 2315 Age Group 1 5 3.18 1.240 2315 Education 1 5 3.37 1.213 2315 Inc ome 1 6 2.92 1.505 2315 Hispanic 0 1 .06 .231 2315 Race 0 1 .12 .322 2315 Sex 0 1 .66 .475 2315 Religious Attendance 1 5 2.91 1.321 2315 Marital Status 0 1 .61 .487 2315 Party ID 0 1 .68 .467 2315 Ideology Strength 1 3 1.76 .675 2315 Ideolo gy 1 5 2.88 1.013 2315 Interested in Campaign 1 3 2.55 .649 2315 Veteran Voted in 2000 General 0 1 .92 .266 446 Age Group 1 5 4.05 .948 446 Education 1 5 3.39 1.128 446 Income 1 6 3.04 1.494 446 Hispanic 0 1 .03 .162 446 Race 0 1 .11 .307 446 Sex 0 1 .09 .279 446 Religious Attendance 1 5 2.93 1.342 446 Marital Status 0 1 .67 .471 446 Party ID 0 1 .64 .481 446 Ideology Strength 1 3 1.74 .698 446 Ideology 1 5 2.65 .958 446 Interested in Campaign 1 3 2.59 .674 446 Active Duty Voted in 2000 General 0 1 .76 .430 341 Age Group 1 5 2.13 .965 341 Education 1 5 3.48 1.033 341 Income 1 6 2.98 1.250 341 Hispanic 0 1 .06 .246 341 Race 0 1 .16 .363 341 Sex 0 1 .16 .371 341 Religious Attendance 1 5 2.94 1.246 341 Marital Status 0 1 .75 .432 341 Party ID 0 1 .66 .475 341 Ideology Strength 1 3 1.65 .650 341 Ideology 1 5 2.57 .814 341 Interested in Campaign 1 3 2.51 .635 341

PAGE 93

93 Table B 4. Descriptive Statistics for Trust in Government Model: Civilians, Veterans and Active Du ty (NAES ) Min. Max. Mean Std. Dev. N Civilian Voted in 2000 General 0 1 .85 .353 1396 Age Group 1 5 3.16 1.244 1396 Education 1 5 3.35 1.195 1396 Income 1 6 2.91 1.466 1396 Hispanic 0 1 .06 .228 1396 Race 0 1 .12 .329 1396 Sex 0 1 .65 .478 13 96 Religious Attendance 1 5 2.88 1.349 1396 Marital Status 0 1 .62 .487 1396 Party ID 0 1 .68 .467 1396 Ideology 1 5 2.85 1.034 1396 Ideology Strength 1 3 1.78 .698 1396 Trust Government 1 4 2.21 .600 1396 Veteran Voted in 2000 General 0 1 .91 .282 220 Age Group 1 5 4.00 .917 220 Education 1 5 3.53 1.195 220 Income 1 6 3.09 1.470 220 Hispanic 0 1 .02 .149 220 Race 0 1 .08 .275 220 Sex 0 1 .13 .334 220 Religious Attendance 1 5 2.81 1.380 220 Marital Status 0 1 .70 .457 220 Part y ID 0 1 .64 .482 220 Ideology 1 5 2.62 .911 220 Ideology Strength 1 3 1.71 .687 220 Trust Government 1 4 2.26 .636 220 Active Duty Voted in 2000 General 0 1 .76 .427 326 Age Group 1 5 2.14 .968 326 Education 1 5 3.48 1.046 326 Income 1 6 2.98 1.243 326 Hispanic 0 1 .06 .240 326 Race 0 1 .16 .367 326 Sex 0 1 .17 .378 326 Religious Attendance 1 5 2.95 1.253 326 Marital Status 0 1 .76 .429 326 Party ID 0 1 .66 .476 326 Ideology 1 5 2.56 .823 326 Ideology Strength 1 3 1.66 .654 326 Trust Government 1 4 2.46 .640 326

PAGE 94

94 Table B 5. Descriptive Statistics for Demographic Model: Civilian s and Military (NAES) Min. Max. Mean Std. Dev. N Civilian Voted in 2000 General 0 1 .87 .340 5601 Age Group 1 5 3.23 1.238 5601 Education 1 5 3. 39 1.200 5601 Income 1 6 2.94 1.503 5601 Hispanic 0 1 .06 .231 5601 Race 0 1 .11 .312 5601 Sex 0 1 .65 .477 5601 Religious Attendance 1 5 2.92 1.333 5601 Marital Status 0 1 .60 .490 5601 Party ID 0 1 .73 .443 5601 Party ID Strength 0 1 .67 .469 5601 Ideology 1 5 2.87 1.035 5601 Ideology Strength 1 3 1.78 .686 5601 Military Active or Vet eran Voted in 2000 General 0 1 .92 .273 1258 Age Group 1 5 3.95 1.070 1258 Education 1 5 3.41 1.133 1258 Income 1 6 3.05 1.463 1258 Hispanic 0 1 .04 .197 1258 Race 0 1 .10 .306 1258 Sex 0 1 .10 .303 1258 Religious Attendance 1 5 2.91 1.328 1258 Marital Status 0 1 .68 .468 1258 Party ID 0 1 .68 .465 1258 Party ID Strength 0 1 .72 .450 1258 Ideology 1 5 2.61 .953 1258 Ideology Stren gth 1 3 1.76 .693 1258

PAGE 95

95 Table B 6. Descriptive Statistics for Political Interest and Internal Efficacy: Civilian s and Military (NAES) Min. Max. Mean Std. Dev. N Civilian Voted in 2000 General 0 1 .88 .326 1860 Age Group 1 5 3.28 1.241 1860 Educatio n 1 5 3.38 1.182 1860 Income 1 6 2.94 1.492 1860 Hispanic 0 1 .06 .235 1860 Race 0 1 .10 .297 1860 Sex 0 1 .66 .475 1860 Religious Attendance 1 5 2.95 1.329 1860 Marital Status 0 1 .58 .493 1860 Party ID 0 1 .73 .442 1860 Party ID Strength 0 1 .68 .468 1860 Ideology 1 5 2.84 1.044 1860 Ideology Strength 1 3 1.80 .688 1860 Following Campaign Closely 1 4 3.31 .770 1860 Follow Politics 1 4 3.24 .829 1860 Politics Too Complicated 1 5 2.81 1.547 1860 Military Voted in 2000 General 0 1 .93 .259 445 Age Group 1 5 3.89 1.101 445 Education 1 5 3.38 1.112 445 Income 1 6 3.03 1.409 445 Hispanic 0 1 .06 .231 445 Race 0 1 .10 .305 445 Sex 0 1 .11 .308 445 Religious Attendance 1 5 2.85 1.341 445 Marital Status 0 1 .65 .477 445 Party ID 0 1 .67 .470 445 Party ID Strength 0 1 .71 .455 445 Ideology 1 5 2.62 .953 445 Ideology Strength 1 3 1.76 .685 445 Following Campaign Closely 1 4 3.45 .732 445 Follow Politics 1 4 3.49 .734 445 Politics Too Complicated 1 5 2.60 1.528 4 45

PAGE 96

96 Table B 7. Descriptive Statistics for External Efficacy Model: Civilians and Military (NAES) Min. Max. Mean Std. Dev. N Civilian Voted in 2000 General 0 1 .88 .325 600 Age Group 1 5 3.19 1.203 600 Education 1 5 3.40 1.210 600 Income 1 6 3.01 1.505 600 Hispanic 0 1 .05 .225 600 Race 0 1 .11 .311 600 Sex 0 1 .63 .484 600 Religious Attendance 1 5 2.90 1.330 600 Marital Status 0 1 .63 .484 600 Party ID 0 1 .75 .435 600 Party ID Strength 0 1 .71 .453 600 Ideology 1 5 2.84 1.004 600 Ideology Strength 1 3 1.75 .681 600 Candidates Try to Keep Promises 1 5 3.24 .718 600 Military Voted in 2000 General 0 1 .94 .236 136 Age Group 1 5 4.05 1.021 136 Education 1 5 3.51 1.095 136 Income 1 6 2.89 1.381 136 Hispanic 0 1 .07 .250 136 Race 0 1 .11 .314 136 Sex 0 1 .11 .314 136 Religious Attendance 1 5 2.93 1.209 136 Marital Status 0 1 .66 .475 136 Party ID 0 1 .66 .475 136 Party ID Strength 0 1 .68 .467 136 Ideology 1 5 2.66 .990 136 Ideology Strength 1 3 1.79 .679 136 Candidates Try to Keep Promises 1 5 3.22 .737 136

PAGE 97

97 Table B 8. Descriptive Statistics for Civic Duty Model: Civilians and Military (NAES) Min. Max. Mean Std. Dev. N Civilian Voted in 2000 General 0 1 .89 .308 595 Age Group 1 5 3.22 1.199 595 Educa tion 1 5 3.42 1.190 595 Income 1 6 3.02 1.507 595 Hispanic 0 1 .05 .212 595 Race 0 1 .11 .310 595 Sex 0 1 .63 .483 595 Religious Attendance 1 5 2.91 1.332 595 Marital Status 0 1 .62 .485 595 Party ID 0 1 .75 .436 595 Party ID Strength 0 1 72 .449 595 Ideology 1 5 2.83 1.009 595 Ideology Strength 1 3 1.76 .686 595 Vote Because of Civic Duty 0 1 .54 .498 595 Military Voted in 2000 General 0 1 .94 .235 138 Age Group 1 5 4.06 1.016 138 Education 1 5 3.50 1.089 138 Income 1 6 2.92 1 .394 138 Hispanic 0 1 .07 .248 138 Race 0 1 .12 .321 138 Sex 0 1 .11 .312 138 Religious Attendance 1 5 2.95 1.216 138 Marital Status 0 1 .67 .473 138 Party ID 0 1 .67 .473 138 Party ID Strength 0 1 .68 .468 138 Ideology 1 5 2.64 1.003 138 Ideology Strength 1 3 1.81 .689 138 Vote Because of Civic Duty 0 1 .62 .488 138

PAGE 98

98 Table B 9. Descriptive Statistics for Government Employment: Civilians and Veterans (NAES) Min. Max. Mean Std. Dev. N Civilian Voted in 2000 General 0 1 .86 .351 3057 Age Group 1 5 2.85 1.094 3057 Education 1 5 3.55 1.152 3057 Income 1 6 3.12 1.427 3057 Hispanic 0 1 .06 .246 3057 Race 0 1 .12 .320 3057 Sex 0 1 .59 .491 3057 Religious Attendance 1 5 2.82 1.308 3057 Marital Status 0 1 .60 .489 3057 Party ID 0 1 .72 .447 3057 Party ID Strength 0 1 .66 .475 3057 Ideology 1 5 2.95 1.035 3057 Ideology Strength 1 3 1.78 .676 3057 Government Employee 0 1 .26 .441 3057 Veteran Voted in 2000 General 0 1 .91 .279 470 Age Group 1 5 3.46 .958 470 Educati on 1 5 3.48 1.080 470 Income 1 6 3.41 1.391 470 Hispanic 0 1 .05 .216 470 Race 0 1 .11 .319 470 Sex 0 1 .11 .317 470 Religious Attendance 1 5 2.77 1.295 470 Marital Status 0 1 .69 .462 470 Party ID 0 1 .69 .464 470 Party ID Strength 0 1 .66 .474 470 Ideology 1 5 2.65 .923 470 Ideology Strength 1 3 1.71 .686 470 Government Employee 0 1 .31 .464 470

PAGE 99

99 Table B 10. Descriptive Statistics for Demogra phic Model: Active Duty (NAES) Min. Max. Mean Std. Dev. N Voted in 2000 General 0 1 .76 428 321 Age Group 1 5 2.15 .967 321 Education 1 5 3.48 1.037 321 Income 1 6 3.00 1.244 321 Hispanic 0 1 .06 .242 321 Race 0 1 .16 .369 321 Sex 0 1 .17 .377 321 Religious Attendance 1 5 2.94 1.250 321 Marital Status 0 1 .76 .426 321 Party ID 0 1 .6 6 .475 321 Ideology 1 5 2.57 .823 321 Ideology Strength 1 3 1.66 .653 321 Table B 11. Descriptive Statistics for Military Service Model: Active Duty (NAES) Min. Max. Mean Std. Dev. N Voted in 2000 General 0 1 .77 .424 300 Age Group 1 5 2.13 .940 300 Education 1 5 3.49 1.033 300 Income 1 6 2.99 1.242 300 Hispanic 0 1 .05 .225 300 Race 0 1 .17 .376 300 Sex 0 1 .17 .379 300 Religious Attendance 1 5 2.92 1.258 300 Marital Status 0 1 .77 .424 300 Party ID 0 1 .65 .478 300 Ideology 1 5 2.58 .812 3 00 Ideology Strength 1 3 1.65 .639 300 Rank 0 1 .31 .462 300 Reason for Joining 0 1 .38 .485 300 Appropriate to Advocate Candidate 1 3 1.71 .938 300 Years in G uard or Reserves 0 6 2.65 2.653 300 Years in Reg ular Mil itary 0 6 2.24 2.627 300

PAGE 100

100 Table B 12 Descriptive Statistics for Demographic Model: Civilians and Veterans (PEW) Min. Max. Mean Std. Dev. N Civilian Registered to Vote 0 1 .82 .387 1295 Age Group 1 5 3.15 1.387 1295 Education 1 5 3.14 1.194 1295 Income 1 6 3.00 1.525 1295 Hispan ic 0 1 .09 .293 1295 Race 0 1 .24 .430 1295 Sex 0 1 .61 .489 1295 Religious Attendance 1 5 2.80 1.436 1295 Marital Status 0 1 .49 .500 1295 Party ID 0 1 .61 .488 1295 Ideology 1 5 2.84 1.003 1295 Ideology Strength 1 3 1.75 .682 1295 Military Registered to Vote 0 1 .91 .286 1880 Age Group 1 5 3.76 1.297 1880 Education 1 5 3.26 1.104 1880 Income 1 6 3.20 1.396 1880 Hispanic 0 1 .06 .241 1880 Race 0 1 .15 .360 1880 Sex 0 1 .12 .324 1880 Religious Attendance 1 5 2.71 1.439 1880 Ma rital Status 0 1 .68 .466 1880 Party ID 0 1 .59 .491 1880 Ideology 1 5 2.53 .937 1880 Ideology Strength 1 3 1.80 .682 1880

PAGE 101

101 Table B 13. Descriptive Statistics for Patriotism Model: Civilians and Veterans (Pew) Civilian or Military Min. Max. Mean S td. Dev. N Civilian Registered to Vote 0 1 .82 .385 1272 Age Group 1 5 3.15 1.389 1272 Education 1 5 3.14 1.190 1272 Income 1 6 3.01 1.527 1272 Hispanic 0 1 .09 .292 1272 Race 0 1 .24 .429 1272 Sex 0 1 .61 .489 1272 Religious Attendance 1 5 2.78 1.432 1272 Marital Status 0 1 .50 .500 1272 Party ID 0 1 .61 .488 1272 Ideology 1 5 2.84 1.003 1272 Ideology Strength 1 3 1.75 .682 1272 Patriotic 1 3 1.81 .932 1272 Military Registered to Vote 0 1 .91 .286 1856 Age Group 1 5 3.76 1.297 1 856 Education 1 5 3.27 1.101 1856 Income 1 6 3.20 1.389 1856 Hispanic 0 1 .06 .239 1856 Race 0 1 .15 .359 1856 Sex 0 1 .12 .325 1856 Religious Attendance 1 5 2.71 1.439 1856 Marital Status 0 1 .68 .466 1856 Party ID 0 1 .59 .492 1856 Ideo logy 1 5 2.52 .934 1856 Ideology Strength 1 3 1.80 .680 1856 Patriotic 1 3 2.25 .956 1856

PAGE 102

102 Table B 14. Descriptive Statistics for Government Empl oyment Model: Civilians and Veterans (Pew) Min. Max. Mean Std. Dev. N Civilian Registered to Vote 0 1 82 .385 809 Age Group 1 5 2.91 1.273 809 Education 1 5 3.31 1.181 809 Income 1 6 3.37 1.516 809 Hispanic 0 1 .10 .295 809 Race 0 1 .22 .415 809 Sex 0 1 .54 .499 809 Religious Attendance 1 5 2.75 1.404 809 Marital Status 0 1 .55 .498 809 P arty ID 0 1 .61 .488 809 Ideology 1 5 2.86 .992 809 Ideology Strength 1 3 1.74 .674 809 Government Employee 0 1 .20 .400 809 Military Registered to Vote 0 1 .91 .286 916 Age Group 1 5 3.25 1.232 916 Education 1 5 3.43 1.060 916 Income 1 6 3.65 1.373 916 Hispanic 0 1 .07 .257 916 Race 0 1 .16 .370 916 Sex 0 1 .13 .339 916 Religious Attendance 1 5 2.71 1.417 916 Marital Status 0 1 .72 .447 916 Party ID 0 1 .60 .490 916 Ideology 1 5 2.51 .912 916 Ideology Strength 1 3 1.78 .684 916 Government Employee 0 1 .29 .454 916

PAGE 103

103 Table B 15. Descriptive Statistics for Demographic Model: Veterans (Pew) Min. Max. Mean Std. Dev. N Registered to Vote 0 1 .91 .290 1602 Age Group 1 5 3.74 1.306 1602 Education 1 5 3.25 1.101 1602 Income 1 6 3.18 1.392 1602 Hispanic 0 1 .06 .236 1602 Race 0 1 .15 .361 1602 Sex 0 1 .11 .313 1602 Religious Attendance 1 5 2.67 1.445 1602 Marital Status 0 1 .69 .463 1602 Party ID 0 1 .59 .492 1602 Ideology 1 5 2.52 .925 1602 Ideology Strength 1 3 1.79 .681 1602 Table B 16. Descriptive Statistics for Service Model: Veterans (Pew) Min. Max. Mean Std. Dev. N Registered to Vote 0 1 .90 .303 1167 Age Group 1 5 3.56 1.321 1167 Education 1 5 3.25 1.072 1167 Income 1 6 3.21 1.391 1167 Hispanic 0 1 .07 .247 1167 Race 0 1 .17 .374 1167 Sex 0 1 .13 .337 1167 Religious Attendance 1 5 2.68 1.442 1167 Marital Status 0 1 .69 .461 1167 Party ID 0 1 .58 .493 1167 Ideology 1 5 2.51 .916 1167 Ideology Strength 1 3 1.78 .681 1167 Rank 0 1 .17 .372 1167 Years of Service 1 7 4.20 1.999 1167 Combat 0 1 .52 .500 1167 Serving Country Important Reason for Joining 0 1 .94 .244 1167 Service Helped Get Ahead 1 7 4.50 .904 1167 Service Useful 4 36 14.38 2.050 1167 Received Gov. Help as Veteran 0 1 .65 .478 1167 Rece ived VA Benefits 0 1 .61 .488 1167 VA Evaluation 1 4 2.68 .846 1167

PAGE 104

104 Table B 17 Descriptive Statistics for Draftee or Volunteer Model: Veterans ( Pew ) Min. Max. Mean Std. Dev. N Registered to Vote 0 1 .95 .218 840 Age Group 1 5 4.70 .456 840 Educa tion 1 5 3.26 1.169 840 Income 1 6 3.05 1.346 840 Hispanic 0 1 .04 .192 840 Race 0 1 .10 .297 840 Sex 0 1 .03 .156 840 Religious Attendance 1 5 2.77 1.466 840 Marital Status 0 1 .72 .447 840 Party ID 0 1 .61 .489 840 Ideology 1 5 2.47 .950 840 Ide ology Strength 1 3 1.85 .683 840 Draftee or Volunteer 0 1 .74 .439 840 Table B 18 Descriptive Statistics for Military Met Needs During Service Model: Veterans ( Pew ) Min. Max. Mean Std. Dev. N Registered to Vote 0 1 .91 .288 907 Age Group 1 5 3.70 1. 189 907 Education 1 5 3.35 1.099 907 Income 1 6 3.38 1.392 907 Hispanic 0 1 .05 .220 907 Race 0 1 .17 .374 907 Sex 0 1 .14 .344 907 Religious Attendance 1 5 2.79 1.441 907 Marital Status 0 1 .77 .423 907 Party ID 1 5 .58 .493 907 Ideology 1 3 2.49 .910 907 Ideology Strength 1 4 1.79 .688 907 Military Met Needs on Active Duty 0 1 2.96 .860 907

PAGE 105

105 LIST OF REFERENCES Abramson, Paul R., John H. Aldrich and David W. Rhode. 2010. Change and Continuity in the 2008 Elections. Washington: CQ Press. Aldr Science 37 (February): 246 278. Almond, Gabriel A. and Sidney Verba. 1963. The Civic Culture: Political Attitudes and Democracy in Five Nations Princeton: Princeton University Press. Fordham Urban Law Journal 34 (3): 935 997. Alvarez, R. Michael, Thad E. Hall, and Betsy Electoral Studies 27 (4): 673 683. Annenberg Public Policy Center. 2006. 2004 National Annenberg Election Survey: General Election Panel Codebook. Philadelphia: University of Pennsylvania. 2006. 2004 National Annenberg Election Survey: Military Cross Section Codebook. Philadelphia: University of Pennsylvania. Bachman, Jerald G., Peter Freedman 76 1997: Self Armed Forces & Society 26 (4): 561 585. Baxter, Sandra and Marjorie Lansing. 1983 Women and Politics: The Visible Majority. Ann Arbor: University of Michigan Press. Beck, Paul Allen and M. Kent Jennings. 198 The American Political Science Review 76 (March): 94 108. Beckwith, Karen. 1986. American Women and Political Participation: The Impacts of Work, Generation, and Feminism. New York: Greenwood. Bernstein, Robert, Anita Chadha and Robert Montjoy. 2001. "Overreporting Voting: Why It Happens and Why It Matters." Public Opinion Quarterly 65 (Spring): 22 44. Armed Forces & Society 27 (Summer) : 501 523 Bishin, Benjamin G., and Matthew B. Incantalupo. 2008. From Bullets to Ballots? The Role of Veterans in Contemporary Elections. University of California, Riverside. Typescript.

PAGE 106

106 Blais, Andre. 2000. To Vote or Not to Vote?: The Merits and Limits of Rationa l Choice Theory. Pittsburgh: University of Pittsburgh Press. Boyd, Richard W. 1981. American Politics Quarterly 9 (April): 133 159. Brady, Henry E., Sidney Verba and Kay Lehman Schlozman. 1995. Beyond SES: A The American Political Science Review 89 (June): 271 294. C SPAN. 2012. Veterans Data from the American Community Survey and Dece nnial Censuses. Washington, DC : United States Census Bureau. Calvo, Maria Antonio and Steven J. Rosenstone 1989. Hispanic Political Participation San Antonio: Southwest Voter Research Institute. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. The American Voter. Chicago: University of Ch icago Press. Journal of Health and Social Behavior 25 (September): 334 348. http://www.lpcohn.squarespace.com (October 11, 2012). voting Among Young Adults in Political Parties and Political Behavior eds. William J. Crotty, Donald A. Freeman, and Douglas S. Gatlin. Boston: Allyn & Bacon. The Human Meaning of Social Change eds. Angus Campbell and Phillip E. Converse. New York: Russell Sage Foundation. Human Events, May. Defense Manpower Data Center. 2009. 2008 Post Election Voting Survey of Uniformed Service Members: Tabulations of Responses: Active Duty Military. Arlington, VA: U.S. Department of Defense. Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper. Participation among Black Adult Males 73 (2): 360 378.

PAGE 107

107 Evers Families United Military Voter Protection Project and AMVETS Legal Clinic at the Chapman University School of Law. www.mvpproject.org/MVPProject_study_download.pdf.(July 2 7, 2012). The National Interest 61 ( Fall ): 29 37. 2001. Soldiers and Civilians: The Civil Military Gap and American National Security. Camb ridge: MIT Press. Federal Voting Assistance Program. 2011. 2010 Post Election Survey Report to Congress. Alexandria, VA: U.S. Department of Defense. Women: The Enduring Effect of The American Journal of Sociology 100 (January): 972 996. American Politi cs Research 33 (November): 842 867. Journal of Politics 69 (August): 813 827. arison of Attitudes and Journal of Political and Military Sociology 29 (Summer): 92 119. Franzich, S. E. 1982 Citizens in Uniform: Political Participation among Military and Civilian Samples. Journa l of Political and Military Sociology 1 0 (1): 15 28. Political Research Quarterly 63 (2): 295 303. Green, Donald P. and Ian Shap iro. 1994. Pathologies of Rational Choice Theory: A Critique of Applications in Political Science. New Haven: Yale University Press. Quality & Quantity 4 3 (January): 59 74. Huntington, Samuel P. 1957 The Soldier and the State; the Theory and Politics of Civil Military Relations. Cambridge: Belknap Press of Harvard University Press. ICF International. 2010. Demographics 2010: Profile of the Military Commun ity Office of the Deputy Under Secretary of Defense (Military Community and Family Policy). Washington, DC : U. S. Department of Defense.

PAGE 108

108 Republicans? Political Party and Ideolog ical Preferences of American Enlisted Inspector General U.S. Department of Defense. 2012. Assessment of the Federal Voting Assistance Program Office Implementation of the Military and Overseas Voter E mpowerment Act. Alexandria, VA: U.S. Department of Defense. American Journal of Political Science 37 (February): 279 290. Jacobson, Car Group Marriage and United Journal of Political and Military Sociology 31 (Summer): 1 22. Janowitz, Morris. 1960. The Professional Soldier Glencoe, IL: Free Press. Jennings, M. Kent and Greg The Social Psychology of Military Service, eds. Nancy L. Goldman and David R. Segal. Beverly Hills: Sage Publications. American Political Science Review 71 (March): 131 147. Acta Politica 39 (December): 342 379. Kam, Cindy D. and Carl L. P almer. 2008. Journal of Politics 3 (July): 612 631. Validity: A Comparative Analysis of Ov erreporting Voter Turnout in Five The Journal of Politics 67 (August): 825 840. Lane, Robert Edwards. 1959. Political Life: Why People Get Involved in Politics. Glencoe, Ill.: Free Press. Leal, David L. 1999 rvice and Latino Political Political Behavior 21 (2): 153 74. Journal of Political and Military Sociology 35 ( Winter): 199 217. Journal of Political and Military Sociology 36 (Winter): 167 188.

PAGE 109

109 American Journal of Sociology 103 (March): 1235 1266. Voting and Civil Election Law Journal: Rules, Politics, and Policy 4 (June ): 105 131. Network World 23 (37): 104. Journal of Political and Military Sociology 34 (Winter): 281 288. Away Vote Argumen American Political Science Review 71 (March): 11 30. Mettler, Suzanne. 2005 Soldiers to Citizens: The G.I. Bill and the Making of the Greatest Generation. New York: Oxford University Press. Milbrath, Lester W. and M. Goel. 1977. Political Participati on: How and Why Do People Get Involved in Politics? 2 nd ed. Boston: Rand McNally College Publishing. Miller, Jane E. 2005. The Chicago Guide to Writing about Multivariate Analysis. Chicago: University of Chicago Press. Miller, Warren E. and J. Merrill Shan ks. 1996. The New American Voter. Camb ridge: Harvard University Press. Armed Forces and Society 29 (Spring): 373 391. The Military and American History Reviews in American History 1 (March): 53 59. Mosko s, Charles C. Jr. [1977] 2010 Volunteer Military: Calling, Profession, or Parameters 40 ( Winter ): 23 31. National Center for Veterans Analysis and Statistics. 2011 Educational Attainment of Veterans: 2000 to 2009. Washington, DC : U. S. Department of Veterans Affairs. 2011. Profile of Veterans: 2009 Data from the American Community Survey. Washington, DC : U. S. Department of Veterans Affairs. Neeley, Grant W. Social Science Journal 38 (July): 381 92.

PAGE 110

110 Public Choice 27 (Fall) : 115 119. rical Comparison of Probit and OLS Regression Journal of Accounting Research 26 (Spring): 119 133. Princeton Survey Research Associates International. 2011. Generational Change/Veterans GP Survey Washington: Pew Social Trends. Rabon, Jo University of Florida. The Atlantic Monthly, July, 66 78. Riker, William H., and Peter C. Ordesh ook. 1968. A Theory of the Calculus of Voting. American Political Science Review 62 (March): 25 42. Rosenstone, Steven J. and John Mark Hansen. 1993. Mobilization, Participation and Democracy in America. New York: MacMillian. Ryan, Garry D. & Timothy K. Ne nninger, eds. 1987. Soldiers and Civilians: The U.S. Army and the American People. Washington: National Archives and Records Administration. Sarkesian, Sam C. and Robert E. Connor, Jr. 2006. The US Military Profession into the Twenty First Century: War, P eace and Politics. 2nd ed. New York: Routledge. Schlozman, Kay Lehman, Sidney Ve rba, and Henry E. Brady. 2012. The Unheavenly Chorus: Unequal Political Voice and the Broken Promise of American Democracy Princeton: Princeton University Press. Schreiber, E. M. 1979 Social Forces 57 (3): 824 39. American Political Scien ce Review 80 (June): 613 624. OVF Research Newsletter 2 (May): 4 6. Journal o f Political and Military Sociology 36 (Summer): 131 145. Cycle Transitions and Political American Political Science Review 89 (June) 421 433.

PAGE 111

111 Squire, Peverill, Raymond E. American Political Science Review 81 (March): 45 66. Szayna, Thomas S., Kevin F. McCarthy, Jerry M. Sollinger, Linda J. Demaine, Jefferson P. Marquis and Brett Steele. 2007. The Civil Military Gap in the United States: Does It Exist, Why, and Does It Matter? Santa Monica: RAND Corporation. Political Research Quarterly 59 (4): 60 1 607. Armed Forces & Society 33 (April): 414 437. of Public Opinion Quarterly 43 (Autumn): 359 377. U.S. Congress. Congressional Budget Office. 2011. Analysis of Federal Civilian and Military Compensation Washington, DC : CBO. U.S. Congress. Senate. 1939. The Hatch Act 76th Cong. S. Rep t. 76 1. U.S. Department of Comm erce. Bureau of the Census. 2000 Current Population Survey Washington, D.C.: Department of Commerce. 2002 Current Population Survey Washington, D.C.: Department of Commerce. 2004 Current Population Survey Was hington, D.C.: Department of Commerce. 2006 Current Population Survey Washington, D.C.: Department of Commerce. 2008 Current Population Survey Washington, D.C.: Department of Commerce. 2010. Current Population Survey Washington, D.C.: Department of Commerce. U.S. Department of Defense. 2008. DOD Directive 1344.10: Political Activities by Members of the Armed Forces. Washington: U.S. Department of Defense. U.S. Election Assistance Commission. 2009. Uniformed and Oversea s Citizens Absentee Voting Act. Washington: U.S. Election Assistance Commission. U.S. Government Accountability Office. 2010. Military Personnel: Military and Civilian Pay Comparisons Present Challenges and Are One of Many Tools in Assessing Compensation Washington, DC : Government Accountability Office. Verba, Sydney, Kay Lehman Schlozman, Henry E. Brady and Norman H. Nie. 1993. British Journal of Political Science 23: 453 497

PAGE 112

112 American Political Science Review 87 (June ): 303 318. Verba, Sidney, Kay Lehman Schlozman, and Henry E. Brady. 1995. Voice and Equality: Civic Voluntarism in American Politics. Cambridge: H arvard University Press. Class Identifica British Journal of Political Science 34 (July): 469 495. Wolfinger, Raymond E. and Steven J. Rosenstone. 1980. Who Votes? New Haven: Yale University Press. Voting in America. Vol. 1 of Praeger Perspectives, ed. Morgan E. Felchner. Westport, CT: Praeger.

PAGE 113

113 BIOGRAPHICAL SKETCH Sarah Valentin earned her b d egree in history with a minor in p sychology from the College of Cha rleston in 1999. She received her Master of Arts in political science from the University of Florida in the spring of 2013. As a proud Navy wife, she has also live d and travel ed extensively overseas.