This item is only available as the following downloads:
EVALUATING THE IMPACT OF DIRECT DEMOCRACY ON VOTER TURNOUT
DAVID L. SCHECTER
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
To M. Margaret Conway,
The best teacher I have ever had
This dissertation could not have been completed without the help of my
committee. M. Margaret Conway was always the perfect chairperson, mentor and editor.
Her ideas and suggestions were instrumental in putting this work together in the early
stages. Her help and constant encouragement were instrumental in getting it done. David
Hedge, Richard Scher, Wayne Francis and Ronald Formisano provided tremendous,
thoughtful feedback at important stages of the research and their support of the project
was paramount to its success. I owe a great debt to each of these scholars for showing
me the importance of good teaching, good writing and reasoned analysis. I also owe
them for their friendship and personal support during my years in the graduate program.
This dissertation also could not have been completed without the help of a
number of colleagues in the department. I am grateful for the friendship of Elizabeth
Oldmixon, Josh Gordon, Marian Currinder, Scott Richards, Maureen Tartaglione and
TABLE OF CONTENTS
ACKN OW LEDGM ENTS..................................................................... .. iii
AB STRA CT............ ... ........ .............................. ........... ........................ vi
1 RATIONALE FOR THE RESEARCH..................................................1
Introduction and Background .............................................. .......... 1
Statement of the Problem: The Gap in the Research.................................2
Summary of Research Questions.......... ...........................................
Significance.................. ..................... ..........7
O operational D efinitions....................... .........................................
Delimitations............. ........ .................................. 10
2 THE DIRECT DEMOCRACY MOVEMENT IN THE STATES..................12
History of the Movement: The Progressive Era........................................12
Discussion ........................ .............. ............ ...... .......... 20
Researching Direct Democracy: Review of the Literature and
Em pirical M odels........................................ ... ........ .... ...........22
Bivariate Approaches ....................... ..................................22
Theoretical Support for Direct Democracy's Impact on Voter Turnout....24
Anecdotal Support........................................ ....... ...... 25
The Rational Choice Perspective..........................................26
3 POLITICAL PARTICIPATION AND VOTER TURNOUT: REVIEW OF THE
LITERATURE AND EMPIRICAL MODELS........................................ 29
M odels of V oting.............................. ........ .............. .................. 29
The Socioeconomic Status and Political Psychology Approaches...............30
The Mobilization Approach: The Role of Parties, Groups and
Institutional Structures............... .......................................... 38
Additional Review of the Rational Choice Perspective.............................43
Discussion.................. ... ........ .. ..... .... .................. .... .... ..... .. 44
The Boyd Hypotheses......................................................................47
Additional Review of the Role of Voter Attitudes........... ........................48
4 THE SYNTHESIS...........................................................................51
Bringing the Two Research Areas Together.........................................51
The Overarching Question: The Impact of Direct Democracy
on V other Turnout.................................... ..... ............. ........... 53
Additional Research Questions.........................................................56
Predicted Socioeconomic Impacts........................................... 56
Predicted Psychological / Attitudinal Impacts............................. 56
Predicted Structural / Election-Specific Impacts.............................57
Predicted Mobilization Impacts................................................59
5 DATA, METHODS AND PROCEDURES.............................................. 61
Sum m ary........................................................................... ....... 61
The Individual-Level Approach: Data, Discussion of the Variables,
Advantages and Disadvantages................................................... 62
The State as the Unit of Analysis: Data, Discussion of the Variables,
Advantages and Disadvantages .................................................... 67
6 ANALYSIS AND IMPLICATION OF FINDINGS....................................74
The Individual-Level Approach: Analysis of the Findings.........................75
The State-Level Approach: Analysis of the Findings..............................82
Implications and Discussion......................................................... 90
7 TOWARD A TYPOLOGY OF DIRECT DEMOCRACY ISSUES....................94
A STATES WITH THE CITIZEN INITIATIVE .........................................109
B NUMBER OF INITIATIVES ON THE BALLOT NATIONWIDE, 1900-1999....110
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
EVALUATING THE IMPACT OF DIRECT DEMOCRACY ON VOTER TURNOUT
David L. Schecter
Chairperson: M. Margaret Conway
Major Department: Political Science
This dissertation examines the development of the direct democracy
movement and its impact on the most important act of political participation, the act of
voting. Specifically, the research looks at whether states that offer initiatives and
referendums to their citizenry have higher levels of voter turnout than those states that do
not allow for direct democracy options, controlling for a number of important variables
not previously considered in other research.
This dissertation builds several strong, multivariate models of the impact of direct
democracy on turnout by combining measures of direct democracy in the states with
important socioeconomic, structural and attitudinal variables found in other subsets of the
political participation literature. Data sets using the American National Election Studies
survey results are analyzed, in addition to new data sets built to analyze state-level data.
Contrary to some earlier research in the discipline, the findings here support the
contention that direct democracy can lead to high levels of voter turnout. Specifically, in
an individual-level analysis logistic regression is used to study all state ballots
(presidential and congressional election years) between 1976 and 1996. In a second data
set, in which the state becomes the unit of analysis, ordinary least squares is used and
further evidence is provided that shows direct democracy has a positive and significant
impact on voter turnout. The findings suggest that direct democracy makes an important
contribution to understanding the variation in voter turnout levels.
The dissertation ends by looking more closely at certain elections in a number of
states to study the impacts of particular kinds of initiatives and referendums on turnout.
Differences in political participation levels are gauged for elections where highly salient
(or controversial) ballot questions are presented to the voters in contrast to less salient
issues being on the ballot. Recent morality policy literature is drawn upon in this part of
the analysis to develop a typology of direct democracy issues. Findings suggest that
more votes are cast on morality politics issues than non-morality issues.
RATIONALE FOR THE RESEARCH
Introduction and Background
This dissertation examines the direct democracy movement and its impact on the
most important act of political participation, the act of voting. Specifically, this research
looks at whether states that offer initiatives and referendums to their citizenry have a
higher level of voter turnout than those states that do not allow for direct democracy
options. Direct democracy in the form of initiatives and referendums entails trusting
citizens to vote for and against legislation, without state legislatures or local commissions
and councils as intermediaries. Historically, this has led to citizens voting directly on a
wide range of issues--from women's suffrage and government reform earlier this century
to environmental protection, affirmative action, health reform, gambling and taxation
issues in the last decade.
Direct democracy (sometimes called participatory democracy) is often seen as
complementing representative democracy, by allowing citizens more opportunities to
participate in the political system and influence public policy. A debate within traditional
democratic theory drives this research, because at the heart of direct democracy theory is
the notion that ordinary citizens can and should hold the power to make laws, while
representative democracy suggests "rule by the few on behalf of the many" (Lawrence,
1999: 12). In his recent argument for "strong democracy," for instance, Barber (1984:
284) has argued that direct democracy "can increase popular participation in and
responsibility for government, provide a permanent instrument of civic education,
and give popular talk the reality and discipline of power that it needs to be effective."
Those states that have strong initiative and referendum movements are literally
giving their citizenry more of an opportunity to participate in the political system. This
research examines whether citizens are taking advantage of the participation
opportunities by turning out at the polls in greater numbers than occurs in states without
direct democracy. Political participation in the form of voter turnout in the twenty-four
states having popular referendums and initiatives is compared to voter turnout in the
twenty-six states that do not offer these elements of direct democracy to their citizens
(see Figure 1 in the Appendix, from Smith (1998), for a map of the states which do and
do not provide direct democracy in the form of initiatives and referendums).
The remainder of this chapter discusses the rationale for the research, introduces
the relevant literature, summarizes the research questions to be analyzed and discusses
the significance of the project.
Statement of the Problem: The Gap in the Research
To study the impact of initiatives and referendums on turnout, two separate tracks
in the political science literature need to be combined. One deals with initiative and
referendum concerns, while the second examines issues of voter turnout, electoral
' The focus of this project is on voter turnout, but some references will be made to the overall issue of
political participation as well. Some of the historical literature, for instance, does not speak to voter turnout
per se, but often makes mention of participation. Voter turnout is treated here as being one aspect of the
larger political participation issue. Efforts are made to distinguish the concepts when appropriate.
structures and voter attitudes. This project bridges these two literatures, neither of
which frequently cite or reference the other. 2 This is done to gauge the impacts of direct
democracy on turnout--for it is argued here that they are linked.
The scholarly literature on the effects of direct democracy and voter turnout is
underspecified. The initial empirical work on the connections between direct democracy
and voter turnout was a multistate, bivariate analysis of turnout between the years of
1960 and 1978 (Everson, 1981). That research concluded "there is no compelling
evidence that initiatives have generally slowed the rate of turnout decline from 1970 to
1978, although the rate of decline for initiative states was less in 1978." (Everson, 1981:
In his frequently cited 1984 book on direct democracy, David Magleby echoes the
sentiments found in the previous research. He does not provide any new empirical
evidence associating direct democracy with voter turnout, but he does review the relevant
literature and suggests there is "very little evidence" the initiative process results in
"higher rates of turnout" (Magleby, 1984: 77).
More important for the purposes of this project, David Magleby, the leading
scholar in the field on this topic, acknowledges that "very little is known about the
aggregate rates of participation on propositions." This is an important point, for it
suggests that a multivariate analysis of this issue is sorely needed. Magleby calls this
2 To rule out the possibility of duplicating existing research (see Webster, 1998: 112) for this project a
thorough search of a number of scholarly databases was undertaken during the fall of 1999. These include:
The Academic Index (1988-Present), America: History and Life, EBSCOHost, FirstSearch PAIS,
FirstSearch ECONLIT, FirstSearch DISSERTATIONS, Info Trac, JSTOR Journals, LEXIS-N E XIS,
Project Muse Journals, Web of Science and the library holdings of the 10 state universities in Florida.
Searches were done using all (or part) of such key phrases as: direct democracy, referenda(ums), initiatives,
and voter turnout.
issue one of several dealing with direct democracy that seems "never to have been
resolved" (Magleby, 1984: 77).
The second area of literature from which this project draws is the plentiful
research on voter turnout. While many things are known about exactly who votes
(Wolfinger and Rosenstone, 1980; Rosenstone and Hansen, 1993; Verba, Schlozman and
Brady, 1995; Miller and Shanks, 1996), no theory has become dominant in explaining
why people are voting less and less (Leighley, 1995; Teixeira, 1992; Conway, 2000).
One solution often floated for increasing turnout is to increase efforts at more direct
democracy, but its impact on actual voter turnout remains in question. Numerous models
of voting are reviewed in Chapter 3, with particular emphasis given to one proposed by
The central problem dealt with in this project, then, is the synthesis of literatures
described above and an assessment of the impact of direct democracy ballot questions on
voter turnout. The impact of direct democracy is measured controlling for many of the
other relevant variables discussed in the voter turnout literature. Several important
research questions drive this project and can now be described in more detail.
Summary of Research Questions
The central research concern here is untangling the impact of direct
democracy on voter turnout, controlling for a number of other variables. It is expected
that states with direct democracy will have lower voter turnout rates than those states
without initiatives and referendums on the ballot. While initially counterintuitive, this
expectation is derived from a central thesis called the hyperdemocracyy" thesis. It is
modeled on the "election frequency hypothesis" developed by Boyd and
discussed in more detail in Chapter 3 and 4. Briefly, the hyperdemocracy thesis assumes
that the sheer amount of time and effort needed for voters to muddle through the electoral
process has now produced a kind of hyper-active democracy for the American voter.
One recent survey, for instance, has found that the average citizen "must elect seventy-
two different officials to represent them at various levels of government... from President
and Congress to school boards, mine inspectors, and county clerks..." (Project Vote
Smart, 1998). Similarly, Cronin (1989: 208) states: "Clearly...the presence of too many
issues on a ballot can overwhelm the voter." Direct democracy ballot questions, then,
add to the hyper-active atmosphere and may help to reduce voter interest and turnout.
One thing is certain, the number of ballot initiatives offered to voters is on the rise in the
last two decades (see Figure 2 in the Appendix, from Saunders (1999), for a breakdown
The hyperdemocracy thesis also suggests that voter turnout may be driven even
lower in states where a large number of initiatives and referendums are on the ballot
and/or in campaigns with large amounts of money being spent for advertising and
advocacy. An important distinction concerning direct democracy ballot questions is
made in this project, however. While aggregate measures of direct democracy should be
negatively associated with voter turnout (supporting the hyperdemocracy argument) it is
recognized that a more nuanced approach to the study of direct democracy is needed.
3 In describing the survey, the Project Vote Smart Resource Center Director, Heather Holdridge, suggested:
"It is a daunting task even for the most involved among us. Only one in four Americans can name both
their U.S. Senators and I doubt that we could find anyone able to name all of their elected officials." The
survey was conducted by Washington State University, Oregon State University and Northeastern
University, with a sample size of 904 households.
Even a cursory glance at any number of ballot propositions over the years
(Matsusaka, 1992; Bowler and Dovovan, 1998) suggests that voters have much greater
interest in certain kinds of ballot questions than in others. Matsusaka (1992) and others
(FACIR, 1986: 45; Bowler, Donovan and Happ, 1992), for instance, note that issues
developed by and referred by legislatures to voters have lower levels of turnout than
those issues placed on state ballots by the voter initiative process. This phenomenon,
commonly referred to as voter drop-off, is often seen when turnout on ballot propositions
is compared directly to voter turnout for candidate elections on the same ballot (Bibby
and Holbrook, 1996).
This important distinction has led to the development of a companion thesis to
hyperdemocracyy;" the "hypodemocracy" thesis. Briefly, hypodemocracy refers to
specific issues and ballot questions that increase voter turnout. Recently, this included
environmental issues in Florida (see Table 1),4 affirmative action questions in California,
gambling issues in Alabama, and an abortion ballot measure in Maine. The final chapter
of this project develops a typology of direct democracy issues in an attempt to identify
the types of issues that are most responsible for higher levels of voter turnout. Chapters 3
and 4 revisit the hyperdemocracy and hypodemocracy theses.
4 Table 1 compares voter turnout levels in Florida in 1996. It represents a rough first cut at systematically
understanding turnout variations at the state level. As the table suggests there was a drop-off of
approximately 14% between the number of voters who participated on the ballot questions concerning the
environment compared to those who voted on the other three ballot issues. This preliminary look at votes
in Florida represents the kind of analysis the project describes in more detail in Chapter 7. For more
information on drop-off see Cronin (1989: 67).
TABLE 1: NUMBER OF VOTES CAST IN FLORIDA 1996
COMPARING VOTES CAST ON BALLOT ISSUES
TO VOTES CAST ON ELECTIONS TO OFFICES*
Overall Turnout 5,444,245
Presidential Race 5,303,794
Fee on Everglades Sugar Production (Initiative) 5,102,822 (Note change in top 3 from
Responsibility for Paying Cost of Water Pollution (1) 4,991,461 bottom 3)
Everglades Trust Fund (I) 4,934,105
Tax Limitation (I) 4,870,400
Judicial Nominating Commissions (Legislative) 4,590,120
Constitution Revision Commission (L) 4,454,186
Retention of Justice Anstead 4,127,428
Retention of Justice Shaw 4,097,169
Retention of Justice Wells 4,018,517
* The only statewide races in 1996 were those involving the Presidential/Vice Presidential candidates and
the races concerning Supreme Court Justice retention.
This research builds several multivariate models of the impact of direct
democracy on turnout by combining measures of direct democracy in the states with
important socioeconomic, structural and attitudinal variables found in other subsets of the
voter turnout literature (Boyd, 1989). This is the first such study known to test the impact
of direct democracy in this fashion. Also, unique to this research, both individual-level
and state-level data are analyzed.
In addition, this research goes a step further in evaluating the impact of direct
democracy on turnout, as the project looks more closely at certain elections in a number
of states to study the impacts of particular kinds of initiatives and referendums on
turnout. Specifically, differences in voter turnout levels are gauged for elections in which
highly salient (or controversial) ballot questions are presented to voters in contrast to less
salient issues. Turnout for ballot questions dealing with affirmative action, right-to-
work laws (Everson, 1981) and gay rights are different from turnout in elections
concerning taxation and health reform, for instance. For some ballot questions, the
amount of money being spent, interest group participation and media coverage are also
This project speaks to larger questions associated with traditional theories of
democracy. While quantitative in nature, it does not lose sight of the larger qualitative
implications of direct democracy and voter turnout. Indeed, an interest in the qualitative
aspects of direct democracy stimulated the exploration of these topics for a research
project. These larger questions are addressed in Chapter 2, on the history of the direct
democracy movements and in the conclusions discussed in Chapters 6 and 7.
The definitions below are offered in the interest of clarity, as the literature on
direct democracy occasionally uses different terms to define similar subjects. Those
derived from specific literature are indicated.
The literature on direct democracy makes clear there are actually a number of
different definitions at play on the topic. Normally, an initiative refers to the specific
process that allows residents of a state who are registered to vote "to propose through a
petition process laws and constitutional amendments on a statewide ballot for approval or
rejection by popular vote" (Bonan, Haven and Horer, 1997: 2). Initiatives are considered
a "political safety valve" that allow people to act directly on issues which the legislature
has avoided. There is some evidence that just by allowing initiatives, legislatures may
stay in greater touch with public needs (Harrigan, 1998: 138).5
Indirect initiatives allow citizens to petition to require the legislature to consider
proposals and, if rejected, "puts the matter on the ballot." Direct initiatives bypass the
legislature and place the matter "directly before voters for approval or rejection" (Bonan,
Haven and Homer, 1997: 2). Nearly sixty percent of those measures put on state ballots
using the direct initiative process between 1898 and 1998 have been rejected (Waters,
2000). Both kinds of initiatives are allowed in twenty-four states. In many of the same
states, as well as some others, citizens can reject laws or amendments proposed by the
legislature, such as changes to the state constitution or bond measures to raise revenue.
This is commonly referred to as the constitutional or legislative referendum process and
is allowed in forty-nine states. Popular referendums, where residents have the power to
place an action of the legislature directly on a ballot, are allowed in twenty-two of the
twenty-four states with initiatives. In some states, petitioners can use the direct initiative
as a popular referendum, particularly if it is less restrictive to do so in the state (Magleby,
1994: 221-222; Waters, 1999: 125).
These subtle differences in the language of direct democracy are important here,
for this project deals exclusively with the twenty-four states having initiative and popular
referendums those comprising the direct democracy states. Turnout in these twenty-
5 An August, 1999, Field Institute survey of 1,010 California residents supported this position, as well.
Mark DiCamillo, director of the Field Poll, summarized the findings in this way: "When the Legislature
doesn't step up to the plate, propositions are seen as a good thing. There are still very few people who
consider them a bad thing" (Gledhill, 1999). Alan Rosenthal takes a strong position against the use of
direct democracy, however, calling it "sloppy democracy" that "is the wrong way to make policy"
(Rosenthal, 1997: 85). He prefers that more power be vested in legislative institutions than in citizen
four states is compared to turnout in the twenty-six states that do not offer these
elements of direct democracy to their citizens.6
This research examines whether the existence of initiatives and referendum on
state ballots leads to higher rates of voter turnout. As well, it examines whether or not
certain kinds of initiatives and referendum lead to higher turnout rates. The research,
however, is limited in scope. For instance, it does not examine city or county-level direct
democracy. Only state-level ballot questions are tackled here. Comparisons between
direct democracy in the United States and other nations are excluded from this study.7
As well, certain data collection issues have limited this research. As Chapter 5 of
the project lays out in more detail, limited data are available on this topic. This research
does not examine individual-level voter attitudes toward specific ballot measures, for
instance, even though the research on voter attitudes toward direct democracy in general
Finally, the time frame for this study has been limited. While direct democracy
issues have been on state-level ballots for a century now, this research deals only with
6 Seven states (California, Oregon, Washington, Colorado, North Dakota, Oklahoma and Arizona) have
accounted for roughly eighty percent of all ballot initiatives in the past century (Harrigan, 1998: 138).
7 For an example of comparative studies see Butler and Ranney (1994).
8 One exception to this is the discussion of work done by Arthur Lupia (1991, 1992 and 1994). His Caltech
doctoral dissertation included a post-election voter survey concerning ballot issues in California. This
allowed for some analysis of voter attitudes to a particular ballot question, but Lupia's findings are from a
rational choice / formal modeling perspective which do not lend themselves well to the kind of research
being done here.
state-level measures appearing since 1976. This, however, still allows for several
hundred ballot questions to be included in the research.
Chapter 2 focuses attention on the direct democracy movement in the states by
highlighting the history of the movement and how direct democracy has been studied in
the political science literature. Chapter 3 discusses the "second half" of the equation, so
to speak, looking at the relevant political participation and voter turnout literature. The
heart of the project is presented in Chapter 4. That chapter describes the synthesis, or
blending, of the direct democracy and voter turnout literatures. The rationale for building
a bridge between these literatures is discussed, along with a review of the overarching
questions guiding the research. Chapter 5 looks at the data, methodology and procedures
used: an individual-level and a state-level analysis of voter turnout. Chapter 6 analyzes
the findings in the data and includes a section dealing exclusively with the implications of
these findings. Chapter 7, the final chapter, develops a typology for direct democracy
ballot questions. The chapter concludes with a more nuanced look at certain direct
democracy ballot questions to reveal that some issues do act to spur voter turnout under
THE DIRECT DEMOCRACY MOVEMENT IN THE STATES
History of the Movement: The Progressive Era
The direct democracy movement is a product of the Progressive era. While a
number of political scientists have covered the history of direct democracy in the states
by discussing the early rise of the Populist Party and the Progressives (Smith, 1998), a
number of primary and secondary sources are revisited here to better gauge the specific
relationship between this period, the direct democracy movement and political
For one thing, the actors and issues of the Progressive era were not as
homogeneous as much of the current political science literature suggests. For instance, as
a number of historians have noted, direct democracy was one of several reforms used in
1 Occasionally discussions of direct democracy begin with an evaluation of changes made during the
colonial period (Broder, 2000; Cronin, 1989), or to trace its roots even earlier to Athenian democracy
(Manin, 1997). While it is recognized here, as Manin (1997) notes, that representative governments have
never given an institutional role to citizens, as the direct democracy states do, the discussion of direct
democracy and turnout is studied here in its twentieth-century context only. See Manin for a rich
discussion tracing the roots of direct democracy from the drawing of lots in Ancient Greece through the
Philadelphia Constitutional Convention of 1787. For an assessment of why certain states adopted direct
democracy and what reasons account for current usage see Price (1975) and Banducci (1998).
2 Too many recent political science researchers have failed to study a number of important aspects of the
Progressive era that are vital for understanding the impact of direct democracy on turnout. Bowler and
Donovan (1998) cover the history of the movement in seven pages mainly retelling the history quoted by
Magleby (1984 and 1994). Smith (1998) covers the history in two pages, relying heavily on Schmidt
(1989), Magleby (1984) and Cronin (1989). While these researchers have done an excellent job in
combing through the direct democracy literature, they have often tailored their citations and discussions to
more general explanations and research questions. As this research project looks specifically at the link
between direct democracy and voter turnout, a reexamination of primary and secondary sources is
the era between roughly 1890 and 1917 to transform government at the national, state,
and local levels. At the national level, the ascendance of Theodore Roosevelt to the
presidency in September of 1901 marked the beginning "of an epoch in American
political life" where new responses to industrialism, prohibition, the demand for women's
suffrage, and governments' relationship to the economy were developed (Kennedy, 1971:
vii). Even though the national government's role in Progressive reforms was dramatic,
many political actors, including Roosevelt, had honed their skills at the state level prior to
1901.3 The states were really the loci of several key tenets of the Progressive approach.
William Allen White, a Kansas newspaper editor, expressed this attitude toward
the states best when he declared the remedy for the ills of democracy was more
democracy (Kennedy, 1971: 19). 4 He stated that work done in the states is "most
fundamental" as "it affects more people than the work either in the cities or in the federal
government" (White, 1910: 25). This attitude, that state democratic institutions should be
more accessible and more responsive to the public, was embodied in several states'
adopting the initiative and referendum. Direct democracy was seen as a key to unlock the
special interest grip on a number of partisan state legislatures (Link and McCormick,
1983: 34; Derthick and Dinan, 1999: 97) and a way to secure more legislation that upheld
particular moral beliefs (Hofstadter, 1963: 4). Again, more democracy was viewed as the
3 Hofstadter (1963: 7) makes an interesting point on this subject by listing the ages of several Progressive
era political leaders in 1900: Robert La Follette was 45, Woodrow Wilson and Louis Brandeis were 44,
Roosevelt 42, and W.J. Bryan, 40.
4 Interestingly, a similar argument is offered in Barber's "strong democracy" thesis (1994: 282) three-
quarters of a century later, when he states "...it is more rather than less experience of government that will
insulate voters against manipulation and prejudice." In contrast, King (2000: 141) provocatively suggests
that "when Americans become dissatisfied with government, they call for more democracy. The more they
call for more democracy, the more of it they get. The more they get, the more dissatisfied they become."
answer to a number of societal ills--and important to the thesis presented here--
more democracy implies more political participation. More importantly, the state was
viewed as the initial place to bring more power to the people.
The initiative and referendum permitted voters themselves to propose and enact
legislation even against the will of the legislature and were "adopted in twenty-two states
...between 1898 and 1918" (Link and McCormick, 1983: 58). The first state to amend its
state constitution to adopt direct democracy provisions was South Dakota in 1898 (Smith,
1998). When Oklahoma joined the Union in 1907 it was the first state to do so with
direct democracy provisions already in its original constitution (Schmidt, 1989). The
first state to vote on a proposition was Oregon in 1904.
A consequence of direct democracy's early growth in the states has also been
considered a factor in why the states are seen being "in the vanguard of policy
development" and as "laboratories of policy experimentation" (Derthick and Dinan,
The Progressive Party Platform of 1912 also offers insight into the general
importance of direct democracy during this period. Besides calling for "equal suffrage"
and "direct election of United States Senators," the platform called for citizen rights to be
"secured by the initiative, referendum and recall..." (Progressive, 1912: 129). The
commitment to direct democracy has been called the "centerpiece" (Milkis, 1999: 7) of
the entire platform. The platform also suggested the Republican and Democratic parties
5 See Schmidt (1989) for a thorough overview of the state-by-state growth in the direct democracy
movement. Of Oregon he suggests that their 1904 votes were the product of seven years of organizing by
one person, in particular, controversial Populist leader William Simon U'Ren. U'Ren's motives are
discussed more fully in Broder (2000). The citizens of Oregon approved two ballot issues on June 6th of
had become tools of corrupt interests that were part of "an invisible government
owing no allegiance and acknowledging no responsibility to the people" (Progressive,
Many Progressives spoke directly to the link between initiatives, referendums and
political participation. For many, direct democracy was seen as the perfect vehicle in a
democratic society to give citizens the ultimate power; the right to control their own
destiny. A 1911 Fabian Society pamphlet expressed it this way: "...the Referendum
came into being gradually and naturally, not as an ascension of popular power, but as a
mere retention by the sovereign people of certain important powers in their own hands"
(summarized in Butler and Ranney, 1978: 26).
Butler and Ranney (1978), drawing partly on the work of Hofstadter (1955),
provide a strong summary of the Progressive position relative to voter turnout. They
suggest the Progressives felt "that truly democratic government consists of all the John
and Jane Q. Publics observing, discussing, pondering, deciding, and, finally, voting"
(Butler and Ranney, 1978: 28). Butler and Ranney suggest the Progressives felt that too
many political decisions "take place a long distance from most voters, both
geographically and psychologically," and that when the people can place "unmediated
votes," the government, in essence, moves closer to home (Butler and Ranney, 1978: 30).
1904: "The first gave voters the right to choose candidates for state office in primary elections (instead of
in party conventions); the second gave counties the right to ban liquor sales" (Schmidt, 1989: 8).
6 Expressing an antiparty mood was not new to the Progressive era. There is strong evidence that the
Gilded Age of political participation / party period concept of the late 19th century, where it is argued the
political parties held broad power, may be somewhat overstated. Recent debate among historians suggests
the existence of strong antiparty fervor (ironically, even expressed by new parties) through much of the
mid- and late 19th century. This is found to be the case at the state level, but perhaps more pronounced at
the local, township level. For more on this debate about party strength and antiparty sentiment see
Altschuler and Blumin, (1997), Watson (1997), and Formisano (1999).
As well, the Progressives appeared to have appreciated the fact that every aspect of
direct democracy, from signature gathering to voting results was a matter of public record
and available to all. In the Progressive's viewpoint a citizen's human potential is
maximized in an open, direct system. Voting will lead to "other forms" of political
participation in this logic and "...when popular votes become the true coin of political
power, people will make it a point to cast them at every opportunity. People will
participate in their government because they believe in it, and they will believe in it
because they participate in it and control it" (Butler and Ranney, 1978: 32-33). 7
Cronin (1989: 11) summarizes the Progressive position well in stating: The
"referendum, initiative and recall are nonviolent means of political participation" and that
"direct democracy (according to its supporters) increases voter interest and election-day
The positive links between direct democracy and voter turnout, however, have
been questioned by a number of historians. Kleppner (1987: 231), for instance, notes that
if "participation profiles....in primary elections" of the 1910s and 1920s "remotely
resembled those at later primaries, turnout was even lower and more sharply class-
skewed than at general elections." Kleppner contends that, at least in the case of the
direct primary, direct democracy was used as a tool by groups out of power that desired
to get back into power. McConnell (1971: 127) has suggested: "in practice as well as
7 Butler and Ranney (1978) admit that some writers feel the act of voting, in and of itself, does not improve
human potential, for it can be an easy task that requires little real commitment. Some have argued that
democracy can be (and should be) more than that. Butler and Ranney cite Pateman (1970), in particular,
for this take on the Progressive argument.
theory Progressivism faltered...the zeal for direct democracy via initiative and
referendum brought reforms, but the results were trivial beside the promise."
Recent evaluations of the Progressive era reforms aimed at improving voter
turnout have been increasingly negative. It has been suggested that direct democracy
reforms actually held "back the ability of marginal groups to gain entry into the political
system as fully participating voters" (McDonagh, 1999: 156) and "institutionally
empowered a theory of representation that virtually guaranteed an ethically barren public
discourse" (Ethington, 1999: 193). Mileur (1999: 278-280) summarizes these concerns
by saying: "Rather than a true citizen democracy, the Progressive reforms have produced
a much more complex system of politics that places greater and often discouraging
demands on voters." He goes on to state that Progressive reforms radically changed the
incentives for voting through civil service and corrupt-practices reforms that reduced
patronage, made it more difficult for third parties to form and "made a politics of mass
organization more difficult." Mileur claims that this has resulted in declining levels of
voter turnout and "the changed pattern of turnout in which poorer, less educated, working
class, minority, and immigrant populations are disproportionately missing from the
voting booth." All of these concerns are discussed in this dissertation and speak to the
heart of why it is important to successfully measure the true impact of direct democracy
on voter turnout.
In addition, much has been written lately on whether initiative and referendum
measures are the work of upper class interests (Smith, 1998; Broder 2000) or ordinary
8 Cronin (1989: 210) states later, however: "Few promoters of direct democracy ever claimed that everyone
would vote and that the public would vote with equal levels of enthusiasm on all measures put before them.
They merely wanted a 'safety valve' process...."
citizen interests. Who brought forward the earliest initiatives and referendums? Some
historical writers speak to this issue as they discuss the early Progressive movement. On
the one hand, some authors have suggested Progressive era reforms came from "above"
and from "the WASP middle class" of the period (Mann, 1975: 13). This view of
Progressive era reforms suggests that status and class concerns of the middle- and upper
class were dominant in shaping political change during the period. Link and McCormick
argue the initiative and referendum, for instance, were used "most effectively by well-
organized interest groups, such as labor unions, prohibitionists, and woman suffragists"
(1983: 58). 9 Those who suggest today that faux populism (Smith, 1998) and moneyed
interests (Broder, 2000) drive some direct democracy issues often discuss the topic using
the same kind of terminology.
On the other hand, some historians have argued that the Progressive era reforms
came from "below" and were born from the needs of "workingmen" and the "immigrant"
classes (Mann, 1975: 20).10 Cronin (1989: 50) has suggested early direct democracy
measures were promoted "by groups regarded as cranks--socialists and single -issue
groups, most notably the single-taxers" and that incumbent legislators tended to dismiss
these groups as too radical. He states that the first proponents of direct democracy were
the "leftist factions the socialists and the People's party." In rural states this movement
9 A recent argument that underscores why women may have supported the initiative and referendum
movement has been suggested by Clemens (1997). Her thesis suggests that since the party period of the
1800s was male dominated, women became more interest-group oriented. Progressive era support for
direct democracy, in the form of women's groups trying to gain political influence, is tied to antiparty
sentiment and frustration with male-dominated legislatures.
10 For more on this debate, see Mann (1975). For the position that Progressivism came from "above,"
Mann excerpts work by George E. Mowry. Kennedy (1971) also reaches this conclusion in his analysis of
Mowry and the writings of Richard Hofstadter. For the position that Progressivism came from "below,"
Mann includes work by Michael Paul Rogin and John L. Shover.
might have been headed by the Farmers' Alliance, for example. It is the language
associated with working people's movements and the needs of the downtrodden often
used by many initiative and referendum supporters today, as outlined in the "Anecdotal
Support" section later in this chapter.
A middle ground, however, is found in the work of David P. Thelen. In his study
of the origins of Progressivism in Wisconsin, Thelen argues that "direct democracy" in
that state grew out of the economic desperation following "the depression of 1893-1897"
(Thelen, 1972: 309). His thesis suggests that Progressive reforms were not necessarily
built around mere status or class concerns but were much more issue-based. He suggests
this "created a new mass politics that united men as consumers and taxpayers in
opposition to the old politics... (308). His work has been summarized as suggesting
that the Progressive coalition was formed of "many different groups," not just the
interests of particular classes (Mann, 1975: 31).
This summary of one of the debates surrounding the history of the Progressive
movement is important in understanding the direct democracy movement and recent
interpretations of just which groups now control the process. While it is recognized that
powerful special interests in some states may push particular reforms using the direct
democracy approach, and this certainly appears to be the position taken in most political
science analyses of the movement, Thelen's work on the early aspects of Progressivism
suggest that the truth may not be so cut and dried. In the final chapter of this research,
the alternative explanation alluded to by Thelen, that recent direct democracy ballot
questions may be both class-based and issue-based, is explored. The history of
Progressivism suggests that this approach to the question of "who is using the initiative?"
will shed light on why some initiatives stimulate higher levels of voter turnout
It is also important to note that even though the use of direct democracy is on the
rise in the last few decades, those states that use the initiative and referendum have
actually been doing so this entire century. In fact, nineteen of the twenty-four states
adopted the process of direct legislation between 1890 and 1920. Four states adopted the
process between 1920 and 1978. Only one state, Mississippi, has adopted the process
since 1978 (Magleby, 1994: 218-219).
To summarize the Progressive era impacts on direct democracy, it is perhaps best
to suggest that the causes for establishment of direct democracy procedures is much less
homogeneous then presented by the majority of political science literature discussing the
period. The era encompasses a number of interwoven political interests (upper and
middle class interests, strong party and antiparty interests), many of whom agreed on at
least one principle: the initiative and referendum would be used in their state. As noted
earlier, it was hoped by many that this would lead to better government, higher levels of
political participation and subsequently, increased voter turnout. The addition of direct
democracy in certain states during the Progressive era was adding an element to the
political landscape of these states that would forever alter the political opportunities
afforded to their citizens. Whether due to simple antiparty sentiments, frustration with
state legislatures or a normative desire to have more individuals involved in the
democratic process, for this brief period of American history, direct democracy can be
seen as a fundamentally different and controversial avenue to create a political
environment unique to certain states. This is why the analysis of voter turnout in these
states can make it easier to judge whether the experiment of direct democracy has payed
off. With this summary of the Progressive era showing the determinants of direct
democracy to be varied and heterogeneous, it is anticipated that the outcomes of the
analysis will be equally diverse.
The legacy of the Progressive era also extends to the diversity in direct democracy
ballot questions faced by voters in the time period being discussed here: 1976 2000.
The term limits movement, in particular, has been a Progressive-style reform widely used
in direct democracy states over the past decade. In fact, Louisiana is the only state with
term limits and no initiative process. Conversely, every state with the initiative process
has enacted some limits, the speed with which that occurred often being dependent on
how low the signature requirements were to place a term limits measure on the ballot.
Term limits were also most severe in states where voters had the easiest time getting the
initiative on the ballot (Francis and Kenny, 2000).
Progressive-style reform can also be seen in the number of tax initiatives (Smith,
1999), public employee pay questions, environmental regulations, and minimum wage
questions (Donovan, Bowler, McCuan and Fernandez, 1998) recently placed on statewide
There have been numerous direct democracy questions, however, which do not fit
so cleanly in any Progressive-era schema. These would include recent ballot questions
concerning transportation (Gerber, 1999), the movement to require color-blind
procedures in California college admissions and government contracting (Derthick and
Dinan, 1999) and nuclear-freeze concerns (Citrin, 1996). A review of ballot
questions voters were asked to consider in the Fall of 1999 and Spring of 2000 included
such issues as a ban on cockfighting in Oklahoma, video gambling in Alabama, a 'partial
birth abortion' ban in Maine and same-sex marriages in California. In each of these
cases, one would be hard-pressed to argue that these issues grow directly out of
Progressive era concerns. Instead, it suggests that the direct democracy process has
grown well beyond its original incarnation earlier this century and now encompasses a
broad array of issues and items, some far beyond what might have originally been
imagined by the champions of direct democracy.
This suggests, as well, that since the number of ballot topics has increased this
century, the potential for greater voter turnout on these topics might also have increased.
Greater attention can now be given to those researchers who have already studied the
relationship between direct democracy and voter turnout.
Researching Direct Democracy: Review of the Literature and Empirical Models
As suggested in the introduction to this project, the scholarly literature on the
effects of direct democracy and voter turnout is underspecified. The initial empirical
work on the connections between direct democracy and voter turnout was a multistate,
bivariate analysis of turnout between the years of 1960 and 1978 (Everson, 1981). In that
study, the dependent variable was voter turnout and the author compares turnout in states
with "at least one" direct or indirect initiative on the ballot to turnout in states with no
initiative on the ballot. The author initially concludes that turnout is slightly higher in
"initiative" states, but he takes the analysis one step further, to control for "region," by
then removing the southern states. When doing so, the original effects of initiatives drop
out and the research concludes, "the overall average is barely (less than .01) in favor of
initiative states." Relative to direct democracy's association with voter decline, it is
noted that "there is no compelling evidence that initiatives have generally slowed the rate
of turnout decline from 1970 to 1978, although the rate of decline for initiative states was
less in 1978" (Everson, 1981: 423). Everson suggests that this is caused by the fact a
number of states had extremely controversial items on their ballots in 1978, such as
Missouri's bitterly contested right-to-work initiative (Everson, 1981: 422).
As summarized previously, in his frequently cited 1984 book on direct
democracy, David Magleby echoes the sentiments found in the Everson research. He
does not provide any new empirical evidence associating direct democracy with voter
turnout, but he does review the relevant literature and suggests there is "very little
evidence" the initiative process results in "higher rates of turnout" (Magleby, 1984: 77).
Magleby compares presidential and midterm election turnout in California,
Massachusetts and Washington with the national averages between 1960 and 1982. He
finds these three states have higher turnout during these years, but no other variables are
controlled for in his analysis (Magleby, 1984: 79). More important for the purposes of
the research presented in this project, Magleby acknowledges that "very little is known
about the aggregate rates of participation on propositions" (Magleby, 1984: 77). This is
an important point, for it suggests that a thorough multivariate analysis of this issue is
needed. Magleby calls this issue one of several dealing with direct democracy that seems
"never to have been resolved" (Magleby, 1984: 77). Recent publications by Magleby
(1994) and Magleby and Patterson (1998) on direct democracy and voter turnout
simply cite the 1981 and 1984 studies already mentioned.
Since Magleby's book, Schmidt (1989: 27-28) has suggested voting on initiatives
has a "spillover effect" on candidate races, "thereby raising overall voter participation."
He compares "voter turnout in elections with and without ballot initiatives" and finds
turnout to be higher in states with direct democracy for the election years 1976-1984, but
this is also a bivariate analysis with no control variables introduced.
A recent unpublished work on direct democracy suggests "the literature on
referendum voting devotes relatively little attention to turnout." In its review of seven
major books on referendum voting written in the last ten years the researchers conclude
previous writers have "barely scratch(ed) the surface on turnout issues" (Tedin, Matland
and Weiher, 1999: 9-10).11
Theoretical Support for Direct Democracy's Impact on Voter Turnout
A substantial amount of literature, however, suggesting a positive association
between direct democracy and voter turnout can be found in the writing of advocacy
" Three other recent works have touched on the nexus between direct democracy and voter turnout. A
dissertation by Shadforth (1997: ii) found that initiatives and referendum could have a "positive effect on
levels of political efficacy, interest, knowledge, and participation." Her work uses Ordinary Least Squares
Regression and Multivariate Analysis of Variance (MANOVA) and is limited in the number of independent
variables considered. Beyond income, age, education, strength of party identification, and a measure of
congressional / presidential elections, no other variables are controlled for in her models. She concludes by
stating that her two methodologies produce "opposite" findings on the impact of initiatives and referendum
on her dependent variables. An unpublished conference paper by Smith (1999: 12) found "popular
referenda can increase turnout during midterm elections," but his model does not control for a number of
measures, such as a variable for Southern states, important ballot structure concerns, certain attitudinal
measures, and a number of key socioeconomic details, which need to be taken into consideration for any
thorough analysis of voter turnout. An unpublished conference paper by Tolbert, Grummel and Smith
(2000) uses pooled time series data for the fifty states to study elections between 1970 and 1996. They find
that direct democracy is associated with higher voter turnout in presidential and midterm elections.
groups and others that would like to see more initiatives and referendums at the
state level. As well, mainstream political theories such as the 'rational choice' approach
speak to the issue.12 While the empirical nature of these works may be questionable, the
influence of them on interest groups, state legislatures, and the media remains substantial.
This section of the project explores the writing of advocacy groups and explains the
rational choice perspective.
The Initiative and Referendum Institute states on its website that voter turnout
was 3.75% higher (in 1996) in "the top five most populated non-initiative states"
(Waters, 1998). The website also includes anecdotal snippets, such as "the Daily
News/WABC-TV poll conducted regarding a possible vote on the funding of a new
stadium for the Yankees showed that '...the stadium issue will increase voter turnout on
Election Day.'" (Waters, 1998). The Institute considers itself a non-profit, bipartisan
organization that lobbies heavily for the protection of direct democracy rights.
The New York Public Interest Group (NYPIRG) produced a report intended to
lobby the New York State legislature, contending turnout between 1976 and 1996 was
"consistently higher" in initiative and referendum states "than non initiative and
referendum states." The group suggests "empowering voters through the power of
initiative and referendum is closely associated with increased voter participation"
(Bonan, Haven and Homer, 1997: 6). Methodologically, however, their work is seriously
12 While not empirical in nature, political science professor Joseph F. Zimmerman offers a thoughtful and
cogent analysis of the benefits of direct democracy. He suggests it is one of several "democratizing
flawed. When questioned as to whether or not they controlled for region or any other
variables in the study, the senior author stated they did not.
Finally, a recent article on possible affirmative action ballot questions in Florida
began with a headline typical of those often found in the press relative to ballot questions
and turnout. 13 The headline suggested: "Ballot Proposal May Lift Turnout," and the
article went on to conjecture that "minority voting" could be boosted in an upcoming
election if an affirmative action ballot question was placed before voters and that this
increase would likely "favor Democrats." The article continued by quoting an African-
American, Democratic Congresswoman from the state, whom also suggested a "tidal
wave" of minority turnout was probable. The Republican governor of Florida, Jeb Bush,
countered: "If it's not on the ballot that would be great" (Associated Press, 1999).
The rational choice perspective
A more mainstream and theoretical argument concerning direct democracy and
voter turnout is found in some voter behavior research, such as summarized in Everson
(1981). His summary draws upon models of the "rational-activist" conception of voters.
In this model, voters decide whether or not to vote and for which candidate (or issue)
based on which action might give them greater expected benefits. The act of voting
occurs when greater gains are perceived for voting, as opposed to not voting (Niemi and
actions" citizens may take to "...set the stage for enlarged, more satisfying citizen participation" (1997:
13 A Quad-City (Iowa) Times story of June, 1999, had a similar take as it summarized a recent election
there in terms of "Grass-Roots Groups Help Bring Defeat." The focus of the article was on the "strong"
labor union turnout for a ballot question concerning an Iowa Constitution budget proposal (Tibbetts, 1999).
Weisberg, 1984; Conway, 2000). Citizens benefit from direct democracy in the
rational-activist model, if the following postulates are considered:
"1) Many voters see little direct connection between votes for candidates and
2) Therefore, there is little incentive for them to vote;
3) However, if citizens were allowed to vote directly on issues, the incentives to
participate would be much clearer;
4) Therefore, initiatives ought to increase voter turnout." (Everson, 1981: 417)14
A similar argument is also suggested by Mueller, drawing on the game theory
component of rational choice theory (1979: 66). He contends that the system of
representative democracy used in the United States was implemented because "town
meeting assemblies bec(a)me impossible with large numbers of voters." Mueller argues
"direct citizen participation in the democratic process at all levels" reduces "the scope for
strategic behavior to its bare minimum, by expanding the number of players of the game,
and literally 'isolating' each of them from one another."
Another summary of the rational choice perspective is offered by Schmidt (1989:
27) when he suggests that laws passed by initiative campaigns produce tangible results
for citizens. Schmidt adds: "A hypothetical rational voter, looking at the nationwide
results of the 1986 campaigns for the U.S. House of Representatives, in which only 2
percent of 394 incumbents were defeated, would not be likely to get involved in future
14 For a summary of several "rational" models of political participation, see Whiteley (1995) and Aldrich
(1991). See Hinich and Munger (1994: 180) for a brief discussion concerning direct democracy,
representative democracy and group theory that uses a rational choice perspective. For criticisms and
comments of the "rational" models, albeit not directly related to citizen lawmaking, see Green and Shapiro
(1994) and Conway (2000: Ch. 6).
elections ... the same voter might well be attracted by the 46 percent success rate of
state-level initiative campaigns in 1985-1986."
A further discussion of the rational choice literature and voter turnout is found in
Chapter 3. This brief overview, however, shows the appeal of direct democracy for some
adherents of rational choice theory and buttresses their argument that direct democracy
can lead to higher levels of voter turnout, in theory and in practice.
In summary, however, there are reasons suggested in the literature for both
believing and perhaps moreso in not believing direct democracy increases voter
awareness and participation. Each of the arguments to date, however, appears to leave
out crucial variables that might mediate the impact of direct democracy on voter turnout
levels the socioeconomic status and attitudes of the voters and variables relating to
electoral structures. For a discussion of these essential concepts, however, one must
move away from the abundance of initiative and referendum literature because these are
not adequately considered by those researchers. Chapter 3 summarizes a number of
models of voter turnout and makes an argument as to why any study of direct democracy
and voter turnout must add components from the voter turnout literature.
POLITICAL PARTICIPATION AND VOTER TURNOUT:
REVIEW OF THE LITERATURE AND EMPIRICAL MODELS
In the previous chapter, discussion focused on the history of the direct democracy
movement and how the impact of direct democracy on voter turnout has been handled in
previous research. It was argued that much of the research is underspecified or is done
by various research groups who often support direct democracy and have not been
thorough in their analysis of initiatives, referendums and turnout. The main failing of
much of that research is that incomplete models are used in various analyses.
In this chapter, discussion centers on a number of relevant models of political
participation. The chapter opens by discussing the dominant models of voter turnout
found in the political science literature. This is followed by an explanation of one
particular model, that offered by Richard Boyd, which has strongly influenced the
research reported here. The chapter closes by reviewing voter attitudes toward direct
Models of Voting
An integral part of explaining voter turnout and voter decline is understanding the
interdependence of a number of essential variables that help explain who is, or is not,
participating. Researchers now know much about who votes, and how various
socioeconomic, structural and attitudinal issues vary in their explanatory power (for an
alternate argument see Matsusaka and Palda, 1999).' For instance, higher
education and greater affluence are now understood to be positively related to the act of
voting. In contrast, minority status and living in the South are negatively associated with
voting turnout. Structural issues, such as the number and frequency of candidate choices
on the ballot and residency or registration requirements of a state, also impact voting
turnout. Citizens' attitudes, particularly party identification and feelings of efficacy, can
have equally dramatic effects on turnout among the voting age population.
Several models of political participation exist, including those associated with
socioeconomic concerns, psychological conditions, mobilization efforts and the
aforementioned rational choice model. A brief description of each of these models is
warranted before the model used in this research is described. The basic assumptions of
the socioeconomic status approach and the mobilization approach will be discussed in
greatest detail, along with the strengths and weaknesses of each.
The Socioeconomic Status and Political Psychology Approaches
The earliest studies of political participation in the United States in the 1940s
highlighted the importance of certain sociological characteristics on the act of voting.
Early research in this area appeared to help explain some patterns of voting among the
electorate, but soon came under fire from researchers who wished to emphasize more
election-specific and contextual variables. In particular, a person's party identification,
1 In this 1999 article, Matsusaka and Palda suggest "that whether or not a person votes is to a large degree
random." This position supports a rational choice perspective on voting and is, as the authors suggest,
"problematic for psycho/sociological approaches" (431). They suggest turnout may be driven by
"idiosyncratic costs like the weather, the traffic, personal health, and so on" (442).
and strength of that identification, was seen as extremely influential even though it was
downplayed in the work of the 1940s (Niemi and Weisberg, 1984).
Spearheaded by the Michigan Survey Research Center, research in the 1950s
placed its major emphasis on three psychological aspects: "The person's attachment to a
party, the person's orientation to the issues, and the person's orientation toward the
candidates" (Niemi and Weisberg, 1984: 12). The Michigan Model, as it became known,
suggested that a multitude of causes 'funneled' together to 'cause' a particular person to
vote. The Michigan Model included several of the sociological measures introduced in
the work of the 1940s, but a voter's party identification "became the core of the model"
(Niemi and Weisberg, 1984: 13). Later versions of the Michigan Model clarified the
notion that party identification is an important long-term factor and that political issues
and candidate behavior were short-term factors affecting vote choice and stimulating
The legacy of the Michigan Model remains a subject of particular debate within
the political science literature. This lively debate persists concerning the long-term
viability of the model and whether or not party identification still acts as a strong
determinant of when people choose to vote. Many scholars, in challenging the Michigan
Model, have contended that there is a 'decline of parties' and 'partisan voting' in the
American electorate (Wattenberg, 1996; Burnham, 1989). Larry Bartels (2000) has most
recently challenged this position. He finds that the impact of partisan loyalties on voting
behavior may be as strong in recent elections as it was in the original Michigan Model of
the 1950s. This recent debate reiterates that any strong model of voter turnout should
include some measures for party identification. In studies of direct
democracy and turnout, party identification should be considered.
Based on data collected in the 1960's, Verba and Nie (1972) continued the
tradition of multivariate analysis by developing an intriguing model of American
participation whose basic assumption was that individuals with resources such as time,
income, age, skills and education participated in higher rates than those without such
resources. In addition, these early socioeconomic status (SES) models suggested those
individuals with certain civic orientations and attitudes participated in greater political
activity (Leighley, 1995: 183).
Numerous other scholars have since expanded this model, with particular
emphasis being given to the links between levels of education and political participation.
Indeed, of the elements in the socioeconomic models, education appears to be the key
resource triggering greater participation. Nie, Junn and Stehlik-Barry (1996: 3) provide a
substantial list of studies from three decades suggesting educational attainment is "the
chief explanatory variable" in research on political behavior and attitudes; this is also
echoed in Jackson's (1995) discussion of the positive influence of education on
registration likelihood and turnout.
Besides education, to be discussed in more detail later in this chapter, some
researchers have chosen to focus on other aspects of the socioeconomic approach to test
specific hypotheses for variables such as age, 2 social group, race, and gender.
2 Bibby (2000: 314) has suggested : "The two personal characteristics that are most closely related to voter
turnout are age and education."
Miller (1992), and Miller and Shanks (1996) offer an in-depth
look at the issue of age and participation. They suggest that as the pre-New Deal voters
began to pass away, they were replaced with less habitual voters in the growing post-New
Deal years. This "generational replacement model" (Miller and Shanks, 1996: 52) offers
a reasonable explanation of turnout decline and suggests that increased education slightly
offsets the change in voting habits for the post-New Deal cohorts. Weaknesses of their
thesis include concerns about the specificity of the data used, their limited look at intra-
generational change and that they do not take into account the political context of
particular elections; such as state to state differences in participation laws and political
In contrast to these researchers, others have suggested voter decline is a product
of several simultaneous trends in addition to generational changes. As part of a broader
study of participation, Teixeira reaffirms the relationship of socioeconomic conditions to
participation by suggesting "socioeconomic upgrading (particularly educational)" has
pushed turnout higher, but that a decline in what he terms "social connectedness, as
manifested in a younger, less married, and less church-going electorate" has pushed
turnout lower. Lower turnout is also explained by "declining psychological involvement
in politics" in his research (1992: 49). Teixeira admits, however, that his particular
theory may not be capturing "other important factors affecting turnout" (1992: 49).3
Wolfinger and Rosenstone also include important socioeconomic variables in
their research as they struggle with the issue of voter turnout. They suggest, for example,
SFor a further summary of Teixiera's findings and the implications of his research see Conway, 2000: 174-
175 and Conway, 1993.
"the transcendent importance of education" on turnout and their data support the
"positive relationship between age and turnout" (1980: 102). They downplay, however,
the importance of social environments, free time and income on voting. They conclude
that a "rich person is no more likely to vote than someone with an average income"
(1980: 103). In addition, they suggest that a number of socioeconomic groups (African
Americans, Hispanics) are underrepresented at the polls disproportionate to their numbers
in the larger American society.
The finding that a number of social groups have lower levels of political
participation is an important subset of research within the socioeconomic status (SES)
literature. Tate reports that SES variables such as age, education and political attitudes
were generally associated with African American voter turnout in the 1984 and 1988
presidential elections, but that "political context and group-based political resources"
(1991: 1159) were also important. Lien argues that the basic SES model needed to
expand to more of an "ethnic group consciousness model" (1994: 238) to explain better
the political participation habits of the Asian and Mexican Americans she studied.
These studies suggest that basic SES models are helpful in gauging 'who'
participates in the culture, but that more complete models of political participation need
to be developed and tested to measure American minority groups. These studies also
suggest the importance of including group-based resource variables to complement the
standard SES approaches. Group-based and social context variables are significant in
studies by Leighley (1990) on elements of the 1976 American National Election Study,
Bennett's analysis of "noncollege young whites" (1991) and studies of African American,
Latinos and Anglo-Whites by Verba, Schlozman, Brady and Nie (1993).
Gender is also noted in the SES theoretical approach as having a major
impact on rates of political participation (Conway, 2000: 36-39). As part of a larger
study of voters in the 1984 elections, Leighley and Nagler (1992: 718) suggest race "and
sex have become much stronger predictors of turnout than they were in 1972, though
education is still the primary determinant of voting." Their research also highlights the
fact that women voters have increased their tendency to vote "relative to men" (1992:
724). Schlozman, Bums and Verba (1994), Schlozman, Bums, Verba and Donahue
(1995) and Verba, Bums and Schlozman (1997) suggest fewer differences between male
and female voters in a trio of pieces. These researchers, building on data they collected
in 1989 for their Civic Voluntarism Model, found that when given similar "resources"
(1994: 963), women's overall level of political activity was similar to men. They also
suggest men and women address similar issues, except on educational and abortion
concerns. Finally, they suggest, "gender differences in political orientation seem to be
specific to politics rather than the manifestation of general personal attributes."
Weaknesses of these three pieces could include their overreliance on a single data set (it
is a single snapshot of individuals) and that their models do not take into account some of
the important political activity surrounding the races they examine.
As noted earlier in the discussion of the Michigan Model, political psychology, in
the form of certain "beliefs, attitudes, and values" (Conway, 2000: 48) also plays a major
role in understanding patterns of political participation in addition to the SES theoretical
approach. This thread running through some of the literature is sometimes captured as a
measure reflecting civic duty, efficacy, or feelings about government attentiveness and
trust in government. While some researchers have suggested that political attitudes in the
United States may increase activities like voting in comparison to other nations
(Powell, 1986) 4 others suggest that, overall, American attitudes toward their own
conceptions of citizenship and participation are extremely diverse. Theiss-Morse, for
instance, found a complex mix of "four distinct perspectives on the participatory duties of
citizens" in her study of Minnesota residents (1989: 370). 5 Her findings suggest some
citizens view their role within the political system as important and influential, while
others view themselves as less influential and they are therefore indifferent.
Other analyses of political psychology suggest that attitudes toward government
and participation may be influenced by economic pressures (Bennett, 1998), family ties,
partisan identification and by the media (Conway, 2000: 52-64; Bibby, 2000: 322-325;
Hadley, 1978: 67-103). 6 In summary, political psychology remains a significant
theoretical approach and diverse measures of political psychology should be included in
models that assess the impact of direct democracy on voter turnout.
The original models of the 1960s and 1970s, as shown above, have been
readjusted and expanded over time. While weaknesses of the SES models range from
concerns about aggregate data used to describe individual level behavior to their
overemphasis in analyzing the behavior of Anglo-Americans, the strength and flexibility
of these models may be best understood looking at a particular work namely, by briefly
4 For a response see Jackman, 1987.
5 Minnesota is one of the twenty-six states that does not offer the initiative process to citizens.
6 It is acknowledged here that there remains some debate as to whether or not political attitudes precede
political behavior. As Leighley notes, the standard SES model "assumes that positive civic orientations are
causally prior to acts of participation," even though some literature suggests this may not be the case (1995:
returning to an example from the literature dealing with how measures of education
are treated within the SES approach.
Nie, Junn and Stehlik-Barry suggest that previous SES literature may misspecify
the impact of education. They question numerous studies and provide a more nuanced
argument concerning education and participation. In short, they contend that increased
education may be boosting political tolerance in the culture, but that increased education
(in and of itself) has not increased "overall participation" (1996: 141). They suggest the
reason for this counterintuitive finding is that increased education is "relative," in that
more citizens now have greater levels of education, but they are still participating in a
political culture with limited ("competitive") opportunities (1996: 187). The researchers
suggest that a measure of "educational environment," not simply years of education, is a
more reasonable measure of the impacts of educational achievement on participation.
This Nie, Junn and Stehlik-Barry work is helpful in understanding that the basic
SES theoretical approach continues to undergo significant change. If the findings in their
study are valid, it may mean for instance that numerous models, such as those discussed
earlier by Teixeira and Wolfinger and Rosenstone, would need to be revisited and may be
considered incomplete. This also gives additional justification for making the case that
the standard models of voter turnout would do well to include measures of direct
democracy. As Nie, Junn and Stehlik-Barry make the case that "educational
environment" impacts turnout, it could be argued that this is analogous to how aspects of
the political environment, like having direct democracy questions on a ballot, can impact
turnout. These elements of the political environment can now be discussed in more
The Mobilization Approach: The Role of Parties, Groups and Institutional
Even though the SES and political psychology approaches have become the
dominant theoretical approaches for the study of American political participation and
hence, receive considerable attention here, a growing body of research "emphasizes the
importance of mobilization as a major factor influencing participation" (Leighley, 1995:
181). The mobilization approach often complements the other approaches and makes the
basic assumption that participation is a "response to contextual cues and political
opportunities structured by the individual's environment" (Leighley, 1995: 188). It has
been more loosely defined as when elements of the political system "pull people into
politics" (Rosenstone and Hansen, 1993: 6). For purposes of this project, mobilization is
discussed broadly, as the efforts of political parties and groups along with the effects of
more formal legal and institutional structures that facilitate or impede participation.
Political parties are often noted in the literature as being important mobilizing
forces on the American electorate. (Bibby, 2000; Wielhouwer, 2000). Caldeira and
Patterson studied state legislative elections in Iowa and California to find that campaigns
and partisan competitiveness had significant effects on which voters participated. More
specifically, they contend that participation in these kinds of elections is not only the
consequence of the "demographic makeup" of a district, but also of "political activation"
(1982: 378). In their judgment, the activity of political parties and the intensity of a
campaign directly influence this activation. They conclude, in contrast to what is often
discussed relative to the SES approach, that looking at a voter's education level is simply
not sufficient to understanding why they might vote.
Similar sentiments on partisan mobilization are evident in the work of Hill
and Leighley (1993), who explain the role of party competition and ideology in
gubernatorial elections and in that of Wielhouwer and Lockerbie (1994), who suggest
that citizens contacted by political parties prior to the election have an increased
probability of voting. Hill and Leighley (1996), find that the Democratic Party has had
reasonable success mobilizing lower-class voters under certain conditions, such as in
presidential election years.
As the SES approach looks at particular social groups and their level of
participation, the mobilization approach looks at certain groups within the culture and
how they organize politically. Hill, Leighley, and Hinton-Andersson (1995) find an
important link between lower-class turnout and the generosity of welfare benefits
provided by particular states. In their study mobilization has important policy
implications (for a published correction of one of their findings, see Ringquist, Hill,
Leighley and Hinton-Andersson, 1997). Squire, Wolfinger and Glass (1987) suggest that
one group in particular--people who have recently moved--could be better mobilized to
participate within the political system if residency and registration requirements were
When mobilization of voters is made more difficult by registration laws and other
institutional issues, lower participation rates can result. Rhine (1992), Boyd (1989),
Highton (1997), Hill and Leighley (1999), Martinez and Hill (1999) and others have
recently studied this aspect of participation, all finding rates of participation and voter
mobilization influenced by the legal structure.
It should be noted, however, that there is some disagreement in the
literature on this subject, as at least one researcher has suggested "that restrictive
registration laws do not dissuade individuals with lower levels of education from voting
any more than individuals with higher levels of education" (Nagler, 1991). This research
is in direct contrast with a number of studies within the SES and mobilization theory
approaches. It directly challenges the earlier works noted by Wolfinger and Rosenstone
(1980) and some from the mobilization approach who argue that legal and institutional
impediments can have a negative impact on participation. In the research reported here,
variables measuring legal and institutional impediments are included in order to speak
directly to this debate.
Perhaps the seminal work on mobilization theory, which encapsulates many of the
topics discussed thus far, appeared in 1993 with Rosenstone and Hansen's Mobilization,
Participation, and Democracy in America. This work makes a number of claims about
the mobilization approach. First, it suggests political leaders "mobilize participation in
pursuit of their own advantage: to win elections, pass bills, amend rulings, and influence
policies." In times of change, leaders "exploit" any "new opportunities" and
accommodate any "new constraints" (1993: 232). In other words, the Rosenstone and
Hansen argument suggests that important political actors often initiate and maintain
various mobilization efforts. There is some difference, then, between this idea of
political participation and that offered by the SES and political psychology approaches
that often see individuals as instigating their own modes of participation based on their
resources, beliefs and attitudes. This is not to argue the two approaches are incompatible,
but to suggest they view the motivation of individual participation somewhat differently.
In addition, Rosenstone and Hansen speak to the "puzzle" of voter decline by
suggesting "at least half of the decline in electoral participation since the 1960's" can be
explained with the changing pattern of mobilization "by parties, campaigns, and social
movements" (1993: 213). Using National Election Study data, they also suggest that
patterns of "class inequality" (1993: 234) persist in the electorate and mobilizerss target
people" from the "upper echelons" of the society (1993: 241).
Their mobilization arguments are strong in that they offer an alternative approach
to viewing participation and they use historical contexts like Watergate (Rosenstone and
Hansen, 1993: 112) to buttress their basic arguments. One weakness is that they may be
exaggerating the power of mobilization over time, since their measures do not control for
the previous voting habits or past participation of individuals.
What is the relationship between direct democracy balloting and the mobilization
approach? How does the argument of Rosenstone and Hansen hold up when direct
democracy questions (which are often not associated with particular parties) and
candidate races are on the same ballot? This dissertation studies the candidate-less direct
democracy movement, but elements of the mobilization model are still considered to see
if they significantly impact voter turnout in direct democracy elections. As the
hypotheses in Chapter 4 describe, it is anticipated that mobilization features are important
in spurring increased voter turnout in initiative and referendum elections.
As a brief analysis of Nie, Jun and Stehlik-Barry was provided earlier in a
concluding discussion of the SES theoretical approach, this section concludes by
summarizing one work that attempts to blend important arguments from the SES
approach with essential elements of the mobilization approach. Verba, Schlozman, and
Brady's Voice and Equality (1995) expounds a new model for understanding
political participation in the United States, which the authors refer to as the Civic
Voluntarism Model. To get a better understanding of who is participating, they use a
two-stage survey method to gather information on over 15,000 Americans (1995: 33).
This initial group of 15, 000 respondents is narrowed to roughly 2,500 citizens who make
up the core of their original Citizen Participation Study. They suggest the Civic
Voluntarism Model is "complementary" to mobilization theories because it analyzes
citizen motivation, a citizen's capacity to participate and the various networks that recruit
and mobilize citizens. Suggesting that the home, parental education, and family church
attendance are all intertwined in formulating a citizen's political attitudes, they highlight
the complexity of political participation in American culture (1995: 458-459). They
suggest that looking at any single predictor of political participation may actually be
distorting the "interwoven" effects of family, school and an individuals' work
While other writers have spoken of the connection between political participation
and church attendance, (Miller and Shanks, 1996: 75) political participation and work
environment (Wolfinger and Rosenstone, 1980: 99), and political participation and
schooling (Bennett, 1991); the argument of Verba, Schlozman and Brady ties these
themes together. Their model, then, is most convincing as a description of why certain
individuals participate, even if it falls short in predicting participation, as they argue
(Verba, Schlozman and Brady, 1995: 443). Weaknesses in their work include the fact
that their model does not consider issues of political context and that the external validity
of their model in studying minority cultures is questionable. Past participation habits of
their survey respondents would also have been helpful to see if some of the benefits
of mobilization they discuss lose momentum when studied over time. In summary, Voice
and Equality represents a reasonable synthesis of several of the key assumptions in the
socioeconomic status, political psychology and mobilization approaches. It is an
important work for understanding exactly what measures should be included, when
possible, in any study of voter turnout.
Additional Review of the Rational Choice Perspective
One critic of Voice and Equality has been John Aldrich (1997: 421-423),
representing the rational choice approach to political participation. He has suggested that
the book lacks "a domain specific measure of political preference," in reference to "why"
citizens actually participate. Aldrich feels that a specific answer to why someone may
participate politically should be directly asked of respondents and that other questions,
which emphasize resources and solicitation are incomplete. Aldrich concludes that the
Civic Voluntarism Model is an "incomplete--but extremely valuable--part of many
theories." He would like to see their model deal more explicitly with political
participation being a "choice" that some individuals opt for, while others do not.
These statements by Aldrich provide a useful transition to a second review of the
rational choice approach to political participation (see Chapter 2 for the initial
discussion). In criticizing Verba, Schlozman and Brady for instance, Aldrich raises many
of the essential assumptions inherent in the rational choice philosophy. These researchers
look to see participation treated as a "choice" individuals decide upon and they would
like to see more participation models include this concept in their calculations.
Aldrich summarizes the rational choice position on voter turnout by
stating that "rational choice accounts of turnout are possible" (Aldrich, 1993: 246). He
goes on to suggest that voting is a low-cost, low-benefit activity and hence, has created a
problem for rational choice advocates who usually would suggest high turnout would be
a result of something with such little cost (Fiorina, 1990). To remedy this dilemma,
however, Aldrich suggests that turnout may not actually be a good candidate for rational
choice theorizing because it is not very complex. This is in contrast to his earlier
statement about turnout being a problem that is explainable with rational choice
theorizing. This kind of conundrum shows the problems in general that the rational
choice approach has in explaining participation. While writers such as Uhlaner (1989)
and Jackman (1993) have attempted to explain participation, in groups and at the
individual level, with a rational choice perspective, their findings are questionable.
Uhlaner's work, where rational choice is placed in a social, mobilizing context, for
instance, is taken to task by Green and Shapiro (1994: 53) who suggest her findings
cannot be replicated. Further weaknesses of the rational choice approach to the study of
participation, can be found in Green and Shapiro (1994).
After looking at these four theoretical approaches to political participation and
modeling voter turnout, which of them appears most significant and relevant to an
analysis of direct democracy? The socioeconomic status models appear to have aided our
understanding of participation the most, thus far. This is stated, however, with several
caveats. As noted in some of the literature in all the perspectives, participation is more
than voting. The study of voting certainly matters, but certain models of voting may
not be appropriate for studying other kinds of participation, such as writing letters or
getting involved with protest movements.
It might be best to think of these four approaches actually overlapping in a kind
of Venn diagram with "gray" areas of shared space highlighting the particular kinds of
participation they help describe. In this same regard, there may be no single best
approach to modeling political participation and subsequently, voter turnout. The best
approach appears to depend on which kind of participation is being described/modeled.
For instance, when modeling voting behavior (either past behavior or in hopes of making
some predictions about 'who' will vote in the next election) the SES/resources model
would be necessary, albeit not sufficient. Weaknesses of the SES and political
psychology approaches, however, continue to be that they deal with a number of highly
correlated variables. 7
On the other hand, if other forms of participation are to be modeled, such as who
is contacted, joins groups, or forms protests, then elements of the mobilization approach
are well suited for the task. As pointed out earlier, however, if one wanted to study the
participation habits for the growing number of minority groups in the United States, the
current SES and mobilization models would need to be enhanced. They would need to
include contextual/political culture measures and in the effort undertaken here to study
direct democracy, elements of a state's political culture are included for that reason.
7 When the number of variables in these models is reduced they can be easily criticized, however (see the
Abramson Aldrich / Hill Cassel debate described in Conway, 2000: 182).
If the empirical aspect of research on voter turnout is de-emphasized, but
pedagogy and teaching about the broader concept of political participation is valued, then
the rational choice model can be appropriate. It provides a number of 'what if scenarios
--prisoner's dilemma, concept of the median voter, game theory and others--which can
get one thinking about the numerous kinds of participation using creative and enticing
logic. Where it may have flaws empirically (as noted earlier by Green and Shapiro), it
still has relevance to understanding certain modes of participation.
As discussed in Chapter 2, the rational choice approach must be
considered in evaluating models of participation and voting. Several researchers
(Everson, 1981; Mueller, 1979; Hinich and Munger, 1994; Schmidt, 1989), as noted
earlier, have drawn conclusions about rational voters and their potential support for direct
democracy ballot questions. As the summary of the rational choice model provided by
Aldrich suggests, however, rational choice scholars often have a difficult time
empirically testing their ideas. For this reason, the research undertaken here avoids more
formal, rational choice modeling and opts for a more empirical study of direct democracy
and turnout. While the contributions of rational choice theory are noted and catalogued,
the model developed here draws more from the work of the SES, political psychology
and mobilization researchers, particularly the work of Richard Boyd that is discussed in
more detail below.
The main argument here is that each of the models described thus far would
benefit from adding measures of whether or not any direct democracy issues are on the
ballot in a given election.
The Boyd Hypotheses
This research emphasizes one model that already combines the important effects
of socioeconomic, structural and attitudinal variables (Boyd, 1981, 1986, 1989).
Beginning in 1981, with modifications coming in 1986, Richard Boyd developed several
working models showing the impact of various issues on turnout. To begin, Boyd
suggested that expanding election calendars have "increased the number of times electors
are called to the polls," contributing to lower turnout. Citing Anthony Downs (1957),
Boyd summarizes his earliest argument by stating that the cost of participating in any
single election may be declining, but "the cost of voting in a large proportion of them is
high and probably increasing" (Boyd, 1981: 142). Boyd elaborates these ideas more fully
in 1986 with the introduction of a more formal hypothesis: "The more frequently
elections are held, the less likely it is that an individual will vote in any given election."
(Boyd, 1986: 90) This election frequency hypothesis is coupled with one that considers
the impact of a salient statewide contest on voter turnout in a presidential election year, as
well. Both of these ideas are tested using CPS Election Study data from the 1980
election cycle and found to be significant explanations of voter decline. Boyd further
clarifies his ideas in 1989 by expanding his research to three national elections (1976,
1980 and 1984). That research also supports his earlier election frequency hypothesis
and goes on to conclude that the timing of primaries and the postwar shift of
gubernatorial races to the congressional election year have caused a small amount of
Additional studies of electoral structures and voter attitudes show their
impacts on participation. These include: Conway's study of midterm congressional
elections (1981); Darcy and Schneider's study of "confusing ballots" (1989); Hill's
Democracy in the Fifty States (1994); Fenster's study of same day registration (1994);
Erikson's study of the National Voting Rights Act (1995); Oliver's study of eligibility
restrictions (1996); and Highton's look at "easy registration" (1997). In each of these
works, an effort is made to show how influences such as political context and voter
attitudes impact turnout.
Many elements of the Boyd model are used in this project. It is introduced here to
provide the background for its further discussion in Chapter 4. In that chapter, particular
elements of Boyd's theory and research results are explored in more detail. His election
frequency hypothesis, for instance, has led directly to the hyperdemocracy thesis being
tested in this project. His research clarifies important elements of the SES and
mobilization approaches to studying voter turnout while leaving room to add elements
such as direct democracy.
Before explaining the specific research questions driving this project, however, it
is important to briefly review what is already known about voter attitudes toward direct
democracy. It is important to include measures of these attitudes in any model of voter
Additional Review of the Role of Voter Attitudes
Since it is generally agreed that strong support of one's party and positive feelings
about one's influence on the political system is positively associated with higher voter
turnout, what is known of voter attitudes toward direct democracy? It would appear
reasonable to suggest that if direct democracy did not result in higher turnout,
public attitudes toward direct democracy must be lukewarm, at best. When measured,
support for direct democracy remains mixed in the public's mind. The NYPIRG report,
mentioned in Chapter 2, cites a Rochester Democrat and Chronicle poll showing 66.3%
of New Yorkers believe adding initiatives and referendums to voter options would be "an
excellent or good idea" (Bonan, Haven and Homer, 1997: 7). In 1999, 63.2% of
Floridians in one poll agreed with the statement: "the country would be better off if more
citizens became directly involved in political decision making" (Florida Voter Poll,
Jack Citrin suggests voters can be "dazed and confused by a barrage of
commercials" relative to initiatives although the public often supports their use. Citrin
says voting on these issues is a "more difficult task than voting for a candidate," and that
turnout rates are generally lower in ballot-proposition elections than in candidate
elections (1996: 280). Wolfinger and Greenstein (1968) previously touched on this
confusion over direct democracy issues. They have suggested "asking voters to pass
judgment on substantive policy questions strains their information and interest..."
(Quoted in Harrigan, 1998: 139). Recent work by Bowler and Donovan proposes just the
opposite. They find voters "can and do think about and decide upon propositions in ways
that make sense and in ways that take advantage of readily available information" (1998:
165). This project attempts to untangle the impact of voter attitudes concerning direct
democracy on voter turnout, particularly as a typology of direct democracy issues is
developed. As this brief summary of the research implies, this is a fruitful area for
further research given some of the conflicting conclusions presented.
As this review of alternate theories and prior research suggests, in order to
properly study the impact of direct democracy on voter turnout, measures of
socioeconomic status, electoral structures, and voter attitudes should be included. This
chapter has summarized the major participation models and drawn our attention to the
important work of Richard Boyd. Exactly how direct democracy measures should be
included in models of voting is the focus of Chapter 4 of the project, where a synthesis of
the work discussed in Chapters 2 and 3 is undertaken.
Bringing the Two Research Areas Together
This work suggests that those researchers interested in building more complete
models of voter turnout should consider adding a variable measuring the presence or
absence of direct democracy measures on the ballot. A companion suggestion is that
researchers interested in studying the impact of direct democracy on voter turnout should
look to the voter turnout literature for clear ideas on how to proceed with their analysis.
To date, practitioners of these two areas of study seem to be looking past each other as
they attempt to explain elements of the voter turnout puzzle. The heart of this project lies
in reconciling these two areas of research and in building a model that provides evidence
as to whether direct democracy spurs turnout.
Democratic theory provides the underlying theoretical base for the project. As
noted in Chapter 1, differences between representative and direct democracy are often
discussed in the political science literature (Lawrence, 1999; Barber, 1984; Cronin, 1989;
Manin, 1997; King, 2000). One of the lynchpins of direct democracy theory is that
citizens need, desire and appreciate opportunities to participate directly in enacting laws,
constitutional amendments and/or tax measures. The current system in most states where
representative democracy is the norm does not push citizens to participate in the same
way as direct democracy. Voting on initiatives and referendums often requires voters to
transcend issues of party and personality to think about complex policy issues. Are they
taking advantage of the opportunities afforded by direct democracy? The best
way to answer this question is by building a bridge between the direct democracy and
voter turnout literatures.
The following conclusions can be offered suggesting what is now known about
the link between direct democracy and voter turnout:
Most studies of direct democracy have been underspecified. They often
simply compare overall turnout rates in states with direct democracy to
those states without the option, but rarely control for alternative
explanations of turnout. There exists no strong, published, multivariate
analysis of direct democracy and turnout. Indeed, the last published,
bivariate analysis of direct democracy and voter turnout appeared in the
This lack of empirical information about the impact of direct democracy
on voter turnout has been compounded by anecdotal evidence often
produced by subjective interest groups and the mainstream media.
While strong models of the electorate's turnout behavior exist in the
literature, none to date have added a measure for direct democracy in their
calculation. Presidential election year measures, congressional election
year measures, senatorial election year measures, registration measures
and the like are often added to voter models as contextual variables, but
the idea of testing whether direct democracy ballot issues also impact
turnout has been basically ignored.
*This suggests that a new set of models, which captures the complexity
of most current voter models, but adds a measure of direct democracy to
the calculation, is sorely needed.
The following section of this chapter explains in more detail the specific research
questions under consideration.
The Overarching Question: The Impact of Direct Democracy on Voter Turnout
The conclusions offered by Boyd and the other structural researchers noted in
Chapter 3 are thought provoking. Could Boyd's election frequency hypothesis be
expanded to include an election magnitude element? An election's magnitude, then,
would refer to the notion that voters may be literally inundated at the polls with ballot
questions and this may be contributing to the reduced levels of voter turnout witnessed in
the last few decades. Are citizens becoming more discontent with the political system, as
the system requires more from participants?
As introduced in Chapter 1, the hyperdemocracy thesis suggests higher levels of
turnout will not be found in direct democracy states. It states that the sheer amount of
time and effort needed for voters to work through decision-making about electoral issues
has now produced a kind of hyper-active democracy for American voters. One recent
survey, for instance, has found that the average citizen "must elect seventy-two different
officials to represent them at various levels of government...from President and Congress
to school boards, mine inspectors, and county clerks..." (Project Vote Smart, 1998).
Similarly, Cronin (1989: 208) states: "Clearly...the presence of too many issues on a
ballot can overwhelm the voter."' Direct democracy ballot questions, then, add to
the hyper-active atmosphere, and may help to reduce voter interest and turnout.
Multivariate modeling can help determine if states with direct democracy indeed have
lower turnout rates in comparison to states without this option.
A companion research question to the hyperdemocracy argument,
however, suggests that in certain elections particular ballot questions can incite higher
rates of turnout. As introduced in Chapter 1, the hypodemocracy thesis refers to specific
issues and ballot questions that increase voter turnout. Recently, this included
environmental issues in Florida, affirmative action questions in California, gambling
issues in Alabama, and an abortion ballot measure in Maine. Specifically, differences in
voter turnout levels are gauged for elections in which highly salient (or controversial)
ballot questions are presented to voters in contrast to elections that present less salient
issues. Turnout for ballot questions dealing with affirmative action, right-to-work laws
(Everson, 1981) and gay rights are different from turnout in elections concerning taxation
and health reform, for instance.
Previous research in this area shows mixed results. In his groundbreaking work
on direct democracy and turnout, Everson (1981: 424) suggested certain kinds of
initiatives may indeed lead to higher rates of turnout when there is "broad appeal" for the
issue and an absence of a "highly visible candidate election" on the ballot and when
"previous turnout has dropped off and an initiative can contribute to a return to the
Cronin also discusses the fact that some voters may skip referendums and initiatives all together and just
opt to vote in candidate races. This is commonly referred to as drop-off, falloff or roll-off. Cronin states
"technically, drop-off (which he also calls falloff) is the percentage of voters who come to the polls but fail
to vote on candidates or proposals found lower on the ballot" (1989: 67). Cronin suggests that a five to
fifteen percent drop-off/falloff rate is common in state elections. For a further discussion of roll-off see
Darcy and Schneider (1989).
norm." The problem is that Everson only offers these ideas as conjecture and does
not test them empirically. Matsusaka (1992), an economist, offers somewhat stronger
support for the argument that certain kinds of initiatives increase turnout in his rational
choice approach to ballot questions in California. His thesis suggests that "good
government" issues did not spur the same levels of turnout as "distributive" ballot
questions, for which he suggests voters have a greater interest. Here, then, is some
evidence that a typology of voter issues is vital to understanding which particular direct
democracy issues spur the greatest levels of turnout. In contrast, however, in an
unpublished piece, Saunders (1999) suggests that categorizing ballot measures as "easy"
or "hard" (borrowed from Carmines and Stimson) has little impact on whether campaign
spending affected initiative success. While not dealing with voter turnout exclusively,
that research is the only other work found which attempts to systematically categorize
In summary, while hyperdemocracy suggests that overall levels of turnout in
direct democracy states will be lower than turnout in non-direct democracy states,
hypodemocracy takes the concept to another level by recognizing that turnout could be
spurred by ballot measures on some types of issues within the twenty-four direct
democracy states. Chapter 6 summarizes the findings of this project relative to the
hyperdemocracy thesis. Chapter 7 develops a typology of direct democracy issues to
identify the types of issues that are most responsible for higher levels of voter turnout.
Additional Research Questions
A number of other questions are also considered in this project, however, as sub-
theses of the hyperdemocracy/hypodemocracy arguments. These can be grouped
according to the models of voting discussed in Chapter 3: socioeconomic impacts,
psychological/attitudinal impacts, mobilization impacts and structural/election-specific
Predicted Socioeconomic Impacts
Of the socioeconomic data, it is anticipated in the individual-level analysis that an
increase in age will positively affect turnout (at least until a leveling-off stage kicks into
effect, which is controlled). If individuals are married, have higher levels of education,
longer length of residence in the community, and live outside of the South, it is
anticipated they will have an increased probability of turnout. In contrast, minority
citizens and those without employment will show decreased levels of turnout.
For the state-level data set a measure concerning minority diversity within a state
is applied. It is anticipated that states with higher minority populations, as measured by
Hero (1998), will show decreased levels of turnout. A more complete discussion of this
measure follows in Chapter 5.
Predicted Psychological / Attitudinal Impacts
The main psychological and attitudinal measures included in this research concern
those gauging external efficacy, party identification and the strength of that party
identification. These are the key measures used by Boyd (1989) and it is anticipated
that each of these indicators has an important impact on voter turnout. Specifically,
individuals with high measures of external efficacy and individuals with strong partisan
leanings will be positively associated with higher voter turnout. Based on a previous
analysis of party identification, Republican voters should be more likely to vote than non-
Republicans (Conway, 2000: 168). These impacts should hold in states with direct
democracy and those without direct democracy.
In the state-level data set a secondary measure, a state's ideology, is used as a
substitute for the party and efficacy measures. It is anticipated that more conservative
states will have higher rates of turnout than their more liberal counterparts. A more
complete discussion of the ideology measure can be found in Chapter 5.
The state-level data set also includes a measure for state political culture (Elazar,
1994 and 1966). It is anticipated that states with a moralistic culture will have higher
rates of turnout than states leaning toward the individualistic and traditionalist cultures.
As Elazar (1966: 91) has stated, in the moralistic states, "it is the duty of every citizen to
participate in the political affairs of the commonwealth." A number of previous
researchers have included political culture in their analysis, as well (Kniss and Lindaman,
Predicted Structural / Election-Specific Impacts
Following Boyd's lead, it is predicted that requiring more days between
registration and voting will be negatively associated with turnout in the states under
consideration. This should hold for both the individual-level and state-level analysis.
Boyd's conclusions concerning the impact of senate and gubernatorial races are also
considered in the research. While Boyd found that gubernatorial elections increase
presidential year turnout, the hyperdemocracy thesis suggests just the opposite. The
presence of senate and gubernatorial (as well as other executive branch) elections should
be negatively associated with turnout in both the individual-level and state-level analysis.
It is recognized that this impact is counter-intuitive. Though senate and gubernatorial
races are highly visible, with good publicity, they should reduce turnout according to the
hyperdemocracy theory because they also add to the sheer number of items voters must
face. In the state-level data set measures are also included to gauge the impact of state
house and state senate races on the ballot. To be consistent with the above formulation, it
is anticipated that states with a higher percentage of contested state house and senate
races on the ballot for a given year will have lower rates of turnout than states without
such races. It is anticipated that voters from the South will come out to vote in fewer
numbers than their national counterparts. The South has consistently lagged the rest of
the nation in voter turnout rates and this is, in part, due to longstanding structural barriers
that have discouraged voting (Scher, 1997). In the individual-level and state-level
analysis a measure of South and non-South states is included. Non-southern states should
be associated with higher levels of turnout in both sets of analysis.
Two other structural/election-specific measures are considered in the state-level
data set: interest group strength in the state and party competition. Interest group
strength is considered because it is known to be closely associated with legislative power
and public policy outcomes in a number of states (Thomas and Hrebenar, 1996). Since
these same interest groups at the state level often take positions on direct democracy
questions (Broder, 2000; Gerber, 1999), their impact is gauged here, as well. Party
competition in the state is considered since high levels of competition should lead to
greater interest in state races and, subsequently, higher levels of voter turnout. The
source of each of these measures is discussed further in Chapter 5, and it is anticipated
that states with low levels of interest group domination, but high levels of party
competition will show increased rates of voter turnout.
Predicted Mobilization Impacts
As discussed in Chapter 3, the relationship between voter mobilization and voter
turnout is difficult to predict when it comes to the subject of direct democracy.
Normally, mobilization measures are considered where there are political actors and
parties involved in the election (Rosenstone and Hansen, 1993). In the case of direct
democracy these elements do not formally exist, but it is recognized that direct
democracy elections often occur simultaneously with candidate elections, and it would be
difficult to measure direct democracy's impact without studying candidate races.
Informally, political actors may publicly speak for or against a ballot question and
political parties may take a position on the subject, but this is a far different scenario than
voters usually face. Most often, voters have some opinion concerning political actors
and/or the party label under which they are running. These cues are not formally
provided with ballot questions. To measure mobilization, then, this project takes a
cautious approach. Following the work of Rhine (1992), Boyd (1989), Highton (1997),
Hill and Leighley (1999) and Martinez and Hill (1999), for the individual-level data set
being analyzed, registration laws are the sole measure gauging mobilization. It is
anticipated that states with more strict registration requirements will have lower
turnout. For the state-level analysis registration dates are also considered.
The state-level data set contains one direct measure of voter mobilization: the
level of union membership in the state (Tedin, Matland and Weiher, 1999). It is
anticipated that high levels of union membership will be positively associated with higher
levels of voter turnout. Since mobilization by parties is often considered a major impact
on voting, the measure of party competition used in the state-level data set (as discussed
above) also could be an indicator of mobilization within the state i.e. the parties in
competitive states understand the need to bring out their base and mobilize a large
number of voters so in addition to party competition as a structural variable, it is
expected to lead to higher turnout as a mobilization measure, as well.
While this chapter summarized the main research questions for this project, the
next chapter focuses on the specifics of measuring and testing these questions.
Consideration is given to the particular data being used, variable measurement and the
procedures used in the data analysis.
DATA, METHODS AND PROCEDURES
The discussion in Chapter 4 centered on the bridging of two different research
domains, namely, those associated with direct democracy and the voter turnout literature.
In this chapter, discussion turns to analyzing the data and methodology employed to test
the research questions also presented in Chapter 4. The following topics are covered: 1)
the two complementary approaches to be used; 2) the data sources used; 3) variable
measurement; and 4) the advantages and disadvantages of these approaches.
The first approach used incorporates survey data from the American National
Election Study, similar to the procedure employed by Boyd (1989), to look at direct
democracy and turnout in the states from 1976 to 1996. The second approach uses the
state as the unit of analysis and also builds a multivariate model of direct democracy and
turnout. Similar to recent research on voter registration and voter turnout (Highton,
1997; Martinez and Hill, 1999), this approach gauges the differences in turnout between
direct democracy states and those without such procedures.
The Individual-Level Approach: Data, Discussion of the Variables, Advantages
This research approach examines two subsets of the American National Election
Studies (ANES Sapiro, et al, 1998). The first data set analyzes a number of variables
affecting voter turnout in six presidential election years, 1976 to 1996. It has a sample
size of 8,556. The second data set looks at the same variables for the five congressional
election years between 1978 and 1994. It has a sample size of 7,850. The ANES studies
are rich in socioeconomic and attitudinal data for this large number of respondents, and
variables are added to the ANES data set from a number of sources to fully specify the
impact of electoral structure elements. For example, The Book of the States provides
information on the states' closing day for registration and gubernatorial races not found
in the ANES data.
The initiative data is drawn from a Congressional Research Report on direct
democracy, with the types of ballot issues categorized using the classification in Bonan,
Haven and Homer (1996). Whether or not a particular state had an initiative or popular
referendum on the ballot in any given year is added to the data set. As Table 2 indicates,
a slight increase in the number of states with direct democracy questions on their ballots
can be noted from 1976 (17 states) to 1996 (20 states). The average is 17.5 states.
1 The findings of the Congressional Research Report were cross-referenced with the findings of The
Initiative and Referendum Institute. The Institute now has a comprehensive website
(http://www.iandrinstitute.org) that categorizes and lists all ballot issues in all twenty-four direct
democracy states from 1898 to the present. The website also contains tables showing signature
requirements for each state, the total number of initiatives and referendums for each state, whether states
have indirect/direct initiatives, popular/legislative referendums and similar comparative concerns.
TABLE 2: NUMBER OF STATES WITH AT LEAST ONE INITIATIVE OR
REFERENDUM ON THE BALLOT 1976-1996
Year States with Initiative or Referendum Questions
Ave. : 17.5
Source: The Congressional Research Service Report 96-932 GOV, as reported in Bonan, Haven and
Homer, 1997, Chart 5.
The dependent variable used in this approach is 'voter turnout,' measured by
responses to the ANES question: "In the election, about half of the people voted and
about half of them didn't. Did you vote?" ANES codes these responses as "1" for "No,
did not vote" and "2" for "Yes, voted."
This approach uses logistic regression to gauge the impact of direct democracy on
voter turnout. As noted earlier, previous works, particularly by Everson and Magleby,
used no multivariate regression techniques to measure the combined impact of variables.
Logistic regression allows for the effects of a large number of variables to be interpreted
when the dependent variable is dichotomous.
The independent variables used in the individual-level analysis are as follows:
AGE: Age is measured in two ways for this data set. The first measure is simply the age
of the respondent as coded by the ANES. Since age often shows a non-linear relationship
(the very young and the very old are less likely to vote) the age of the respondent is
UNEMPLOYED: The ANES asks respondents: "are (you) working now, temporarily
laid off, or are you unemployed, retired, permanently disabled, or what? Are you doing
any work for pay at the present time?" For this project, respondents who answered, "not
employed" are coded with a one and all other respondents are coded with a zero. The
"not employed" category includes homemakers, retired individuals, and those who are
laid off, on strike, or permanently disabled.
MARRIED: Married respondents are coded with a one and all non-married, divorced,
separated, widowed and non-married partners are coded with a zero.
LENGTH OF RESIDENCE: The ANES asks respondents how long they have lived in
their present location. Respondents living in their residence from less than one year to
thirty-nine years were coded with the actual number of years (i.e. a person in their current
residence for one year was coded with a one, etcetera). Respondents who have lived in
their current residence from forty to eighty-nine years were coded with a forty-one and
residents in the same location for "all of life" are coded with a ninety.
EDUCATION: The ANES coding system for education was slightly altered here. A set
of dummy variables was developed to test how different levels of education impact
turnout. These dummy variables break the ANES respondents into the following groups:
a) a grade school education or less; b) if they completed high school; c) if they
completed some college work; and d) if they completed college or an advanced degree.
In Table 3 and 4, discussed in more detail in Chapter 6, the 'completed high school'
category is left out of the analysis and the other groups are included. Following the
discussion of education from the previous chapter, it is anticipated that respondents
with 'a grade school education or less' will have lower turnout than their 'some college'
or 'advanced degree' counterparts.
NONWHITE: This is one of two ANES measures used to capture race. For this variable,
they code the respondents' race as "White, Black or Other." White is coded here as zero
and the others are coded with a one.
HISPANIC: Respondents of Mexican-American, Chicano, Puerto Rican and "other"
Hispanic descent are coded with a one in the data; non-Hispanics are coded with a zero.
EFFICACY: Between 1976 and 1996 ANES gauged external efficacy by asking
respondents if they agreed, disagreed or neither agreed or disagreed with the following
similar statements: 1) "I don't think public officials care much what people like me
think" and 2) "Public officials don't care much what people like me think." Respondents
that agreed with these statements were coded here with a one, respondents who neither
agreed nor disagreed were coded with a two and respondents who disagreed were coded
with a three.
STRENGTH OF PARTISAN IDENTIFICATION: Respondents with independent or
apolitical leanings are coded one, those who lean independent were coded two, weak
partisans were coded three and strong partisans were coded four.
PARTY IDENTIFICATION: Respondents who consider themselves independents or
apolitical are coded zero, along with Democratic respondents. Republicans are coded
with a one.
INITIATIVES AND REFERENDUMS: If a state had an initiative or referendum on
their state ballot for a given election year, they were coded with a one. If no initiative
was on the ballot that year, the state was coded zero. For example, in the analysis of
the 1976 election, North Dakota had an initiative on the ballot and would have been one
of the seventeen states coded with a one for that election (see the discussion below for the
coding of ballot measures in the state-level data set).
SENATE RACE: If a state had a U.S. Senate race on their state ballot for a given year,
they were coded with a one. If no senate race was on the ballot that year, the state was
coded zero. In the rare instance (Kansas, 1996, for example) where both U.S. Senate
seats were on a state ballot, the state was still coded with a one.2
GUBERNATORIAL RACE: If a state had a gubernatorial race on their ballot for a given
year, they were coded with a one. If no gubernatorial race was on the ballot that year, the
state was coded zero.
CLOSING DATE: This variable measures the number of days away from the general
election day when a state closes its registration books. The actual number of days was
used for coding purposes. For example, the State of Florida, which changed its
registration requirement during this period, was coded with a forty-five for the 1976
election, a twenty-nine for the 1996 election and so forth.
NONSOUTH: This measure identifies the eleven secession states with a one (Alabama,
Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina,
Tennessee, Texas, and Virginia) and the non-secession states with a two.
2 Information concerning U.S. Senate races was taken from Sapiro (1998) and cross-checked with the FEC
The advantages of this approach are that a number of variables (party
and civic attitudes gauged by ANES) are included in the models. The large number of
cases provided by the ANES data is also a benefit to the project. The disadvantage of this
approach is that the individual-level data collected by ANES are not the best way to
measure the impact of certain variables on voter turnout in each state. The ANES does
not survey the same number of people in each state, for instance, and the results from this
method of analysis may not completely explain direct democracy's effect on turnout. For
this reason, a complementary approach is also used.
The State as the Unit of Analysis: Data, Discussion of the Variables, Advantages
While a micro-level analysis of voter turnout using ANES data enables testing of
an individual-level model, another approach for studying the impact of direct democracy
on turnout is to use each state as the unit of analysis. Therefore, a new data set has been
developed; similar in many ways to the one described above, to test the main research
questions for a single election year 1996. If turnout should be spurred in any year, it
should be a presidential election year, so this offers the supporters of direct democracy a
'best case scenario' for studying the significance of their issue on overall turnout rates.
For state-level data from the 1996 elections a number of sources are used. A
measure of social diversity, specifically the size of the minority population in each state
is taken from Faces of Inequality, by Rodney Hero (1998). This 'minority diversity'
number provides an excellent resource for gauging the socio-cultural homogeneity in
each state. Democracy in the Fifty States (Hill, 1994) provides the party competition
measure. In this book, Hill uses a modified version of the Ranney index to look at
party control of governorships and legislatures.
To substitute for the party and civic attitude variables available in the ANES data,
a secondary measure, the ideology score for each state (Wright, Erikson, and Mclver,
1985), is used. For a measure of interest group strength in each state the classification
system developed by Thomas and Hrebenar (1996) is used. Their system classifies states
into separate categories ranging from those in which interest groups are considered
'dominant' in the public policy process to those states where interest groups are
'subordinate.' Barone, Lilley and DeFranco (1998) provide the information needed on
the number of contested state house and senate races in 1996.
Since there is some evidence to suggest that union membership (Tedin, Matland
and Weiher, 1999) is an indicator of mobilization efforts, membership numbers for each
state are obtained from the Statistical Abstract of the United States (1997). For the
variables in this data set that are similar to the individual-level data, 'closing day' and
'non-south,' the same sources are used.
The dependent variable of 'voter turnout' for each state in this approach comes
from the Federal Election Commission website (2000). The number used is the percent
of turnout of the voting age population in each state. For example, Alabama had a voting
age population of 3,220,000 in 1996, with 2,470,766 of those individuals registered to
vote (76.73%). Their voter turnout on Election Day was 1,534,349, which is 47.65% of
the voting age population.
Unlike the individual-level study that used logistic regression to
study the variables, ordinary least squares is used in this analysis. This is best, as the
dependent variable here is the states' reported level of voter turnout, which is a
The independent variables in this model are as follows:
INITIATIVES: Ballot questions are measured in two ways for the state-level data set. In
the first instance, similar to the individual-level data set, states are coded one if they had
at least one question on the state ballot in 1996. In addition, to gauge whether the sheer
amount of ballot questions before the voters could impact turnout, as the hyperdemocracy
thesis suggests, another variable was created. This second variable was coded for the
actual number of initiatives and referendums on the ballot for each state. For example,
Alaska had three issues on the ballot, Arizona had four, and they were coded accordingly.
The states with the largest number of ballot questions in 1996 were California (17) and
Oregon (16). Four of the twenty-four direct democracy states had no issues on the ballot
in 1996 (Illinois, Mississippi, Oklahoma and Utah).
UNION MEMBERSHIP: The percent of workers in each state who are union members
was coded for this variable. The range was from a low of 3.7% in North Carolina to a
high of 26.8% of the workforce in New York. Of note, the national average in 1996 was
14.5%, down from 20.1% in 1983.
INTEREST GROUP STRENGTH: The categories used by Thomas and Hrebenar (1996)
are adopted here. States were coded with a zero if interest group strength was
subordinate or complementary to other aspects of the political system; coded one if the
strength was complementary; two if the strength was complimentary/dominant; and
four if the strength was dominant.
PARTY COMPETITION: Several researchers have developed measures of party
competition. Often the measure is based on the number of times a particular party carries
a state in the run for president, U.S. House and Senate seats and/or governor (Bibby,
2000). This project adopts the intuitive party competition score developed by Hill (1994)
for his book on Democracy in the Fifty States. Hill's measure draws on a modified
version of the Ranney index, which looks at competition at the gubernatorial and state
legislative level. Hill adds a component to take into account the results of the presidential
elections in each state and presents the results in an easy-to-understand whole number
index. States that are most competitive such as New Jersey and Ohio receive scores of 47
and 40, respectively; while less competitive states receive lower scores (example:
Mississippi (13) and Georgia (14)). The actual score given to each state by Hill is coded
in the data set.
U.S. SENATE RACE ON THE BALLOT: As noted in the discussion of the individual-
level data set, if a state had a U.S. Senate race on its state ballot in 1996 the variable is
coded a one. If no senate race was on the ballot, the variable was coded zero. Of note,
even though Kansas had two races on the ballot in 1996, it is still coded with a one.
Thirty-three states held U.S. Senate races in 1996.
EXECUTIVE OFFICES ON THE BALLOT: Fourteen states held executive branch
elections in 1996. Of those fourteen, eleven had at least the governor's race on the ballot
(Delaware, Indiana, Missouri, Montana, New Hampshire, North Carolina, North Dakota,
Utah, Vermont, Washington and West Virginia). Georgia's Secretary of State seat was
on the ballot, along with a number of executive posts in Oregon (3) and
Pennsylvania (3). For the data set, a state was coded according to the total number of
executive offices on the ballot, understanding that even though lower ticket races may not
have the same draw as a governor's race, the fact that an additional statewide office was
on the ballot should be considered. For example, of the fourteen states, Delaware was
coded with a two (races for governor and lieutenant governor); while West Virginia was
coded with a six (races for governor, attorney general, agriculture commissioner, auditor,
secretary of state, and treasurer).
CONTESTED STATE SENATE DISTRICTS ON THE BALLOT: The percentage of
state senate districts that were contested was coded in the data set. The percentage
ranged from zero in Alabama and Virginia, which held no state senate races in 1996 to a
number of states with seventy-five to one-hundred percent of seats being contested. To
clarify, this is not the total number of seats that could have been filled in 1996, but rather
the actual number of seats (out of that total) with an actual contest.
CONTESTED STATE HOUSE DISTRICTS ON THE BALLOT: This variable was
coded in the same way as the state senate races discussed above.
IDEOLOGY: Wright, Erikson and McIver's (1985) index of state ideology is frequently
used in the literature. Harrigan (1998: 103) takes their index and lists the states in an
easy-to-read table ranking them from most liberal (Vermont coded one here) to most
conservative (Idaho coded forty-eight here). Hawaii and Alaska are not included in the
index or ranking, but they are assigned the scores of fifteen and forty-five, respectively,
based on the number of times they voted for the Republican presidential candidate in the
previous eight elections (see Bibby, 2000: 73).
MINORITY DIVERSITY: Hero (1998) creates a measure of minority diversity for
each state by looking at the ratio of blacks, Latinos and Asians to the white population.
States range from Vermont (.004 minority diversity score) to California (.732 minority
diversity score). The score provided by Hero is used in the data set.
POLITICAL CULTURE: Dummy variables were created to gauge the impact of political
culture. The categorization used comes from Hero's (1998:12) ranking of Elazar's
(1984) original three labels: moralistic, traditionalistic and individualistic. For the
models here, the seventeen traditionalistic states are dropped from the analysis and the
impact of the moralistic and individualistic states is gauged.
CLOSING DATE: This variable measures the number of days away from the general
election day when a state closes its registration books. The actual number of days was
used for coding purposes. For example, Florida was coded with a twenty-nine and
Minnesota was coded with a zero.
NONSOUTH: This measure identifies the eleven secession states with a one (Alabama,
Arkansas, Florida, Georgia, Louisiana, Mississippi, North Carolina, South Carolina,
Tennessee, Texas, and Virginia) and the non-secession states with a two.
The advantage of having a state-level approach is that studies at the state level
may yield different answers to the main theses from those captured in an analysis of the
ANES data. Also, there is considerable support in the literature for testing research
questions in this manner. Conversely, the disadvantages of this approach are that party
identification and attitudinal variables cannot be captured. In addition, the large
numbers for study afforded in the ANES approach are drastically reduced when the 'n' of
the study comes down to just the fifty states.
These two complementary approaches provide the best hope for detecting subtle
differences in the actual impact of direct democracy on voter turnout. This chapter has
explained how these two data sets are organized and the next chapter discusses, in detail,
the results of the analysis on each data set.
3 Of the seventeen traditionalistic states, eleven are the southern states that seceded from the union at the
start of the Civil War. As Chapter 6 explains, two state-level models are run to judge the impact of the
'political culture' and 'nonsouth' measures separately. Both sets of results are reported.
ANALYSIS AND IMPLICATION OF FINDINGS
The aim of this research is to shed light on a significant question facing American
democracy. Do direct democracy states have higher rates of voter turnout than those
states that do not offer direct democracy to their citizens? Chapter 2 explained that
previous research into direct democracy and voter turnout has been underspecified, but
noted that there are sound theoretical reasons for believing that direct democracy could
spur turnout. Chapter 3 explained that contemporary voter turnout models would benefit
by adding a measure of direct democracy to the mix. Chapter 4, however, suggested that
bridging the direct democracy and voter turnout literature with a new, more specified
model gauging voter turnout would lead to the following results: States with direct
democracy would have lower levels of voter turnout than their non-direct democracy
counterparts. This counter-intuitive prediction was based on the concept of
hyperdemocracy that suggests direct democracy states have produced a kind of
hyperactive political environment that causes members of the electorate to shy away from
voting. Hyperdemocracy suggests that more ballot questions and ballot choices leads to
lower levels of voter turnout.
This chapter discusses the results of several multivariate models of voter turnout,
all of which measure the impact of direct democracy on turnout. To summarize the
results, they do not support the hyperdemocracy thesis and, in fact, suggest that direct
democracy may indeed have a positive impact on voter turnout. The exact findings and
the implications of these findings are discussed in the remainder of the chapter.
The Individual-Level Approach: Analysis of the Findings
The individual-level analysis provides the opportunity to study the impact of
direct democracy on voter turnout using a unique data set combining elements of the
American National Election Study (1976-1996) and various measures collected for each
state. This analysis follows the example set by Boyd (1989) in building a model of voter
turnout, but adds the element of direct democracy. Table 3 shows the results of the
analysis during presidential election years (1976, 1980, 1984, 1988, 1992 and 1996).
Each of the socioeconomic measures in the model behaves as expected. Increases
in age and residential stability are positively and significantly associated with increased
voter turnout. Being married also has a positive and significant impact on voting. As
expected, being unemployed is negatively associated with turnout and in a significant
manner. For nonwhite and Hispanic respondents, the relationship is not significant.
The education measure produces some interesting results. Breaking the education
measure into a series of dummy variables highlights the notion that not all levels of
education have equal impact on voter turnout. In fact, those respondents with less than a
high school education have significantly lower levels of turnout than those with more
than a high school degree. This finding suggests that future models of voter turnout
would benefit from breaking down education measures to best study their impact and that
standard models which include a simple measure for education may be underspecified as
Nie, Junn and Stehlik-Barry (1996) have suggested.
The political party and civic attitude measures in the model behave as anticipated.
High levels of efficacy and strong party identification are positively and significantly
associated with increased voter turnout. The relationship between Republican
respondents and turnout is not significant.
Of the structural variables, states outside of the old confederacy have significantly
higher rates of turnout than their southern counterparts and states that close their
registration books long before Election Day have significantly lower rates of turnout
compared to those with more liberal registration requirements. Each of these findings
jibes well with the majority of the literature.
Surprises come, however, when the other structural variables are considered.
Contrary to the findings of Boyd (1989), having a U.S. Senate race on the ballot during a
presidential election year appears to have an insignificant impact on voter turnout. In
none of the six election years included in the model does a Senate race have a significant
impact on turnout. While Boyd (1989) found that gubernatorial races had a positive and
significant impact on turnout, the model presented here is not so conclusive. In only one
year under consideration, 1988, did the presence or absence of a gubernatorial race have a
significant impact on voter turnout. Contrary to Boyd, however, the impact in 1988 was
negative. These findings suggest that further research may be needed to gauge the full
impact of Senate and gubernatorial races on turnout.
The impact of direct democracy on voter turnout is not what the hyperdemocracy
thesis suggests. The thesis contends that direct democracy questions should be
negatively associated with voter turnout, but the findings in Table 3 show that direct
democracy is only significantly associated with voter turnout in 1980. More importantly,
and in contradiction to the hyperdemocracy thesis, in 1980 states with direct democracy
had higher levels of turnout than their non-direct democracy counterparts. The
implication of this finding is explored in detail after the findings from the state-level
data set are discussed below.
Overall then, Table 3 highlights the notion that a well-specified model of
individual-level voter turnout contains many expected findings, but that a few surprises
exist. The majority of the socioeconomic, party, and attitudinal variables behave as
expected, but several of the structural variables offer contradictory findings from what is
found is some of the literature. Direct democracy's impact on voter turnout, at least in
this model, is only significant in 1980 and any other connections between direct
democracy and voter turnout cannot be supported from this particular analysis of the data.
An analysis of the congressional election years provides a complementary understanding
to the results found in Table 3.
TABLE 3: THE IMPACT OF INITIATIVES AND REFERENDUM ON VOTER TURNOUT
IN PRESIDENTIAL ELECTIONS: 1976-1996 (Logistic Regression)
Independent Variable B S.E.
Age .096*** .009
Age Squared -.001*** .000
Unemployed -.303*** .092
Married .504*** .060
Length Residence .004*** .001
Grade School or Less -.922*** .079
Some College .828*** .072
Advanced Degree 1.535*** .110
Nonwhite -.161 .083
Hispanic -.082 .109
Party ID and Civic Attitudes
Efficacy .251*** .031
Strength of ID .451*** .029
Republican ID .113 .062
Initiatives and Referendums (I&R)
I&R 80 .278* .114
I&R 84 .126 .128
I&R 88 -.161 .110
I&R 92 -.020 .454
I&R 96 .040 .443
Other Structural Variables
Senate 80 -.070 .127
Senate 84 -.100 .096
Senate 88 -.035 .091
Senate 92 .146 .093
Senate 96 -.089 .105
Gov 80 .023 .118
Gov 84 .197 .169
Gov 88 -.647*** .185
Gov 92 -.213 .174
Gov 96 -.032 .189
Nonsouth .249*** .070
Closing Day -.020*** .004
(Constant) -3.875*** .277
-2 Log Likelihood 7911.375
Model Chi-Square 1933.144***
% Correctly Classified 78.2%
***p< .001; **p< .01; *p< .05
Source: ANES data with some variables added by author (see text). Note: 73.8% of the total sample
reported voting. The model predicts 37.2% of the non-voters and 92.8% of the voters. The total %
correctly classified is reported in the table as 78.2%.
Table 4 shows the results of the individual-level analysis during a number of
congressional election years (1978, 1982, 1986, 1990 and 1994). The majority of
findings in this table mirror those found for the presidential election years.
The socioeconomic variables in the off-year election analysis have similar
impacts and are in similar directions to those found in the presidential election year
analysis. Being older, having a stable residency and being married are all positively
associated with voter turnout. Each of these variables has a statistically significant
impact, as well. As with the previous data set, for nonwhite and Hispanic respondents,
the relationship is not significant. Education again proves to be highly significant and
respondents with at least some college experience vote in higher numbers than those with
only limited schooling.
The political party and civic attitude variables in this model also mimic those
found in Table 3. High levels of efficacy and strong party identification correspond
significantly to higher levels of voter turnout. The relationship between Republican
respondents and voter turnout levels cannot be determined.
Table 4 shows that non-Southern states have higher levels of turnout than their
Southern counterparts and that states which close their registration books far in advance
of election day have significantly lower levels of turnout than states with more liberal
registration laws. While some previous research, as mentioned in Boyd (1989), suggests
that U.S. Senate races can have a positive impact on voter turnout in off-year elections,
the findings here are quite mixed. In 1982 and 1986 a U.S. Senate race on the ballot did
not have a significant impact on turnout rates. The picture gets murkier when 1990 and
1994 are considered. In each of these years Senate races were significantly associated
with higher levels of voter turnout, but in 1990 the impact was negative and in 1994
the impact was positive. This finding suggests that no particular 'rule of thumb' can be
used for judging the impact of Senate races during congressional election years.
The impact of gubernatorial elections on voter turnout is also difficult to gauge in
congressional election years. In 1982 and 1986 the relationship between gubernatorial
races and voter turnout is insignificant. In 1990 and 1994, however, states with
gubernatorial races on the ballot had significantly lower levels of turnout than those states
without such races.
The impact of direct democracy on voter turnout is not statistically significant in
any off-year election (as it was in the 1980 election). Taken together then, in the vast
majority of election years analyzed using this individual-level approach, direct
democracy appears to have an insignificant effect on voter turnout, controlling for a
number of other impacts. By moving next to the analysis of the state-level data set these
initial findings can be explored in greater detail.
' It has been suggested that this result could be a function of how competitive the contests were. In other
words, was there an incumbent in the race? While incumbency was not taken into account here, future
research could explore this issue further.
TABLE 4: THE IMPACT OF INITIATIVES AND REFERENDUM ON VOTER TURNOUT
IN CONGRESSIONAL ELECTIONS: 1978-1994 (Logistic Regression)
Independent Variable B S.E.
Age .126*** .009
Age Squared -.001*** .000
Unemployed -.444*** .104
Married .305*** .059
Length Residence .009*** .001
Grade School or Less -.886*** .080
Some College .631*** .074
Advanced Degree 1.058*** .082
Nonwhite -.048 .085
Hispanic -.212 .130
Party ID and Civic Attitudes
Efficacy .099*** .030
Strength of ID .420*** .029
Republican ID .057 .060
Initiatives and Referendums (I&R)
I&R 82 .151 .135
I&R 86 .073 .094
I&R 90 -.104 .137
I&R 94 .022 .090
Other Structural Variables
Senate 82 .101 .089
Senate 86 -.147 .104
Senate 90 -.415*** .089
Senate 94 2.888*** .217
Gov 82 .090 .088
Gov 86 -.014 .176
Gov 90 -.542*** .152
Gov 94 -4.128*** .225
Nonsouth .171* .071
Closing Day -.019*** .003
(Constant) -4.968*** .284
N = 7,850
-2 Log Likelihood 8098.327
Model Chi-Square 2721.580***
% Correctly Classified 73.3%
***p<.001; **p<.01; *p<.05
Source: ANES data with some variables added by author (see text). Note: 54.5% of the total sample
reported voting. The model predicts 67.2% of the non-voters and 78.4% of the voters. The total %
correctly classified is reported in the table as 73.3%.
The State-Level Approach: Analysis of the Findings
The state-level analysis provides the opportunity to study the impact of direct
democracy on voter turnout using a different combination of variables than those used in
the individual-level analysis. This data set captures the nuances of one election year,
1996, and gauges the impact of direct democracy while controlling for a number of
important contextual variables.
Tables 5A and 5B show the results of a regression model where the dependent
variable is the percentage of voter turnout by state in 1996. In this model, initiative states
were simply coded one if they had one or more initiatives on the ballot and zero if they
Four variables stand out as being significantly associated with voter turnout in
Table 5A. 1) States with high levels of party competition have higher levels of voter
turnout. 2) As expected, states with large minority populations tend to have lower levels
of voter turnout. 3) Surprisingly, states with a large percentage of state house seats on
the ballot in 1996 had lower levels of turnout than those with a lower percentage of races
on the ballot. As with the U.S. Senate race impact discussed previously, more research
appears necessary to figure out why more races on the ballot leads to lower levels of
turnout, as much of the literature suggests just the opposite should occur.2 4) Most
important, however, is the finding that those states with at least one initiative on the
ballot in 1996 had higher levels of voter turnout than states without a ballot question.
2 One possible explanation could be the competitiveness of the contests, similar to what was described in
the footnote above. In addition, whether or not there was an incumbent senator in the race could be a
This appears to give additional support to some of the individual-level findings
discussed previously and directly contradicts the hyperdemocracy thesis.
The other variables on this model are statistically insignificant, even though most
are in the expected direction. States with a high percentage of state senate seats on the
ballot are positively associated with higher levels of turnout than states with a lower
percentage of races, although not at significant levels. The same finding also holds for
states with a U.S. Senate race on the ballot. As the mobilization literature suggests, high
levels of union membership are positively associated with high levels of voter turnout,
but the relationship is not significant. States that are rated more conservative
ideologically are positively associated with higher levels of turnout than more liberal
states and this also holds, as expected, for those states categorized as having a moralistic
political culture. In neither of these cases, however, is the relationship statistically
significant. Similar to the individual-level data set, states with registration date deadlines
far from Election Day are negatively associated with voter turnout, as are states with at
least one executive office on the ballot in 1996 and individualistic in their political
culture. None of these relationships are statistically significant, however. States with
more dominant interest group influence are negatively associated with voter turnout, but
the relationship is not significant. Overall, this model suggests that the majority of
attitudinal and structural variables influencing voter turnout in the states in 1996 was not
as important as whether or not a state had an initiative on the ballot, a large minority
population and strong party competition.
Table 5B is identical to Table 5A except the variable for 'nonsouth' states is
included here instead of the variables measuring political culture. As mentioned earlier,
of the seventeen traditionalistic states, eleven are the southern states that seceded
from the union at the start of the Civil War. For this reason, it was considered important
to run the model two different ways and report both sets of findings.
THE IMPACT OF INITIATIVES AND REFERENDUM ON VOTER TURNOUT:
HAVING AT LEAST ONE INITIATIVE ON THE STATE BALLOT IN 1996
(Ordinary Least Squares)
Independent Variable Adj R-Square = .71 Adj R-Square = .70
B S.E. T B S.E. T
Presence of Initiative .034* .014 2.491 .042*** .012 3.431
Party Competition .001* .001 2.615 .001* .001 2.282
Minority Diversity -.150** .045 -3.303 -.146** .046 -3.171
State House Races -.069* .034 -2.057 -.050 .033 -1.497
State Senate Races .036 .030 1.208 .030 .030 .990
U.S. Senate Race .017 .014 1.217 .025 .014 1.844
Executive Races -.002 .003 -.720 -.001 .003 -.445
Interest Group -.014 .008 -1.737 -.015 .008 -1.913
% Union .003 .001 1.801 .002 .001 1.398
Ideology .000 .001 .562 -.000 .001 -.285
Closing Day -.001 .001 -1.635 -.002** .001 -3.043
Moralistic .013 .024 .555-
Individualistic -.021 .022 -.961 -
Nonsouth .009 .021 .450
(Constant) .522*** .048 10.774 .550*** .041 13.424
Source: Various measures compiled by author see text for explanation.
For this second model, there is only one significant change to report. While party
competition, minority diversity and initiatives still have an important impact on turnout;
the variable measuring the percentage of state house seats on the ballot is no longer
significant. Instead, and consistent with each of the individual-level models discussed
previously, the closing day for voter registration becomes significant. The direction of
the other variables in Table 5B is similar to those found and in Table 5A with the
exception of political ideology, which has a negative impact when the 'nonsouth'
variable is used. As anticipated, the non-southern states have higher rates of turnout than
their southern counterparts, even though the relationship is not statistically significant.
Overall, the models presented in Tables 5A and 5B provide strong evidence that
ballot initiatives can have a strong and positive impact on voter turnout when controlling
for a number of important state-level variables. Tables 6A and 6B show the impact of
direct democracy on voter turnout when the sheer number of ballot questions is
measured, instead of simply indicating whether or not at least one initiative was on the
ballot in 1996, as Table 5 reports.
When the total number of ballot questions for each state is considered there is
some change in the results from what is seen in Table 5. As Table 6A reports, when a
state has more than one initiative on the ballot, there is a positive association with higher
levels of voter turnout, but the relationship is not statistically significant. The
hyperdemocracy thesis suggests that more ballot questions should have a negative impact
on turnout. For the other variables in the model, only two are significantly associated
with voter turnout and the results mirror those found in Table 5A and 5B. High
levels of party competition in a state are positively associated with higher levels of
turnout and a large minority population is negatively associated with turnout. The
direction of the other variables in Table 6A mirrors those found in Table 5A.
Similar to what was discussed concerning Table 5, the model measuring the sheer
number of initiatives on the ballot was run a second time where the political culture
variable was dropped and the 'nonsouth' variable was included. The results of this model
can be found in Table 6B.
In this model the variable measuring the sheer number of initiatives on state
ballots in 1996 is statistically significant. Indeed, states with more ballot initiatives had
higher rates of voter turnout than those with a lower number of initiatives or no initiatives
at all. This suggests that because of direct democracy states like California and Oregon,
with a large number of questions on their ballots in 1996, had higher levels of voter
turnout than if they would not have had these questions on their ballots. This, again, is a
direct contradiction to the hyperdemocracy thesis, which suggests a greater number of
ballot questions should be negatively associated with turnout.
As found in Table 5B, party competition, minority diversity and the closing dates
for registration are significantly associated with levels of voter turnout. As well, the
insignificant variables in Table 6B mirror those found in Table 5B, except for the
ideology measure whose sign changes from negative to positive in Table 6B.
Before concluding the discussion of the individual and state-level data, there is
one important inconsistency which should be noted. In the individual-level data analysis
(Table 3) the impact of initiatives and referendums on voter turnout in 1996 is
statistically insignificant. The state-level analysis, however, presented in Tables 5
and 6, suggested that the presence of the initiative on state ballots in 1996 had a
significant impact on turnout. Why is the individual-level data suggesting one result
while the state-level data suggests another result for the same year? This question may
be answered by considering that this data analysis still leaves a number of issues
unexplained. In other words, this analysis may not be explaining the entire story
concerning direct democracy and voter turnout when studied in these complementary
The answer to the inconsistency in the findings may be that simply having direct
democracy issues on the ballot may not be enough of an encouragement to get the
electorate to the polls, yet, overall states with direct democracy still appear to draw more
voters than states without the process. The issue here concerns interpreting these findings
with the knowledge that there are two separate units of analysis being considered. The
state-level finding suggests that direct democracy has a positive impact on turnout in the
states that use the process, but the individual-level analysis suggests we may not know
'why' particular individuals in those states are drawn to the polls. One potential solution
for determining the importance of this 1996 finding is to replicate the state-level analysis
in future research to see if they match up with the findings in various years considered in
the individual-level tables.
THE IMPACT OF INITIATIVES AND REFERENDUM ON VOTER TURNOUT:
THE SHEER NUMBER OF INITIATIVES ON THE STATE BALLOT IN 1996 (OLS)
Independent Variable Adj R-Squared = .66 Adj R-Squared = .64
B S.E. T B S.E. T
Number of Initiatives .001 .002 .683 .003* .002 2.024
Party Competition .002** .001 2.744 .001* .001 2.445
Minority Diversity -.169** .051 -3.317 -.182*** .051 -3.546
State House Races -.059 .036 -1.651 -.040 .037 -1.102
State Senate Races .032 .033 .997 .027 .033 .831
U.S. Senate Race .012 .015 .793 .020 .015 1.406
Executive Races -.003 .003 -1.100 -.003 .003 -1.051
Interest Group -.011 .009 -1.272 -.014 .009 -1.734
% Union .002 .002 1.558 .001 .001 1.071
Ideology .000 .001 1.267 .000 .001 .426
Closing Day -.000 .001 -.974 -.001* .001 -2.465
Moralistic .030 .027 1.108 -
Individualistic -.016 .023 -.101 -
Nonsouth .005 .023 .260
(Constant) .500*** .055 9.058 .554*** .045 12.388
Source: Various measures compiled by author see text for explanation.
Implications and Discussion
The tables in this chapter and the analysis of these two complementary data sets
provide little support for the hyperdemocracy thesis laid out in Chapter 4 of this project.
In fact, this analysis suggests that direct democracy may lead to higher levels of voter
turnout, whether measured at the individual-level or at the state-level. These findings
appear to support the position of direct democracy advocates who have suggested for
decades that initiatives and referendums provide important opportunities for individuals
to participate within the political system and would stimulate higher rates of voter
turnout. The evidence here supports their contention.
The support found here for their position is even more important for the fact that
this project attempted a multivariate approach to studying the impact of direct
democracy, which has been called for since Magleby (1984) wrote about the subject
nearly two decades ago. While the goal of this research was, frankly, to suggest that
there are negative impacts associated with direct democracy and turnout, the findings
suggest that direct democracy makes an important contribution to understanding the
variation in voter turnout levels. This research should encourage others to immediately
include measures of direct democracy in their turnout models, as suggested in Chapter 4.
Many others, however, may be troubled by these same findings for a number of
reasons. As the review of the literature in Chapter 2 and 3 suggested, many mainstream
and respected political scientists appear dissatisfied with direct democracy. Rosenthal
(1997: 85) has called it "sloppy democracy," and he is joined by the likes of Smith
(1998), Kleppner (1987), Mileur (1999) and Broder (2000). For each of these writers,
direct democracy has questionable beginnings and even more questionable
There is an important irony to consider here, however. While critics of direct
democracy often contend that it corrupts the political system, weakens political parties
and undermines the work of state legislatures and the benefits of representative
democracy, few would argue they want to be in a democracy with such low levels of
voter turnout. Hence, if direct democracy actually helps to spur political participation, at
least in the form of increased voter turnout as suggested here, there may be a potential
benefit to direct democracy that should please even the harshest critics. In other words, if
initiatives and referendum do spur turnout, than those individuals and researchers
opposed to the process have to weigh whether or not the downside of direct democracy
outweighs the potential gains shown in voter turnout. Critics of direct democracy now
appear to be in somewhat of a theoretical bind. If greater levels of voter turnout, even
controlling for a number of other factors, are the result of having the direct democracy
option available to citizens, than it would be difficult to argue that Americans should not
be given the opportunities that direct democracy may afford.
Some of the numbers concerning the contemporary usage of direct democracy
further exacerbate the ironies. Consider the following: According to the Statistical
Abstract of the United States (1997: 28), in 1970 forty-four percent (44%) of the U.S.
population (89,103,000/203,302,000) lived in the twenty-three states (Mississippi
becomes the twenty-fourth state in 1992) with direct democracy. By 1997, forty-eight
percent (48%) of the U.S. population (126,888,000/265,284,000) lived in the twenty-four
states with initiatives and referendum. Hence, roughly half of the U.S. population now
has the option of voting in direct democracy elections. More importantly,
many of the high growth states going into the next decade are in the West and Southwest
and are the very same states with heavy usage of the initiative and referendum. In seems
fair to conclude then that the use of direct democracy will only grow in the coming
decades and become even more controversial.
These findings do not suggest, however, that all aspects of the direct democracy-
voter turnout nexus are explained in the findings. It could be argued, for instance, that
the casual link this dissertation examines direct democracy leads to higher voter turnout
- could be explained in alternative ways. In fact, it has been suggested that the "arrow"
could go in the other direction, so to speak. What if voter turnout and other aspects of a
states' political culture has led them to adopt direct democracy and this, in turn, helps to
explain their higher levels of turnout found in the data? Perhaps the causal sequence
described here is premature or describes simply one aspect of the direct democracy-voter
turnout debate. The findings here do not dispute the fact that these alternative
explanations are possible. These findings are not mutually exclusive from other
arguments which could suggest high levels of political participation may be the driving
force behind the adoption and use of direct democracy. These findings, however, do
attempt to help explain the impact of direct democracy on voter turnout and future
research could certainly look to the question of "which came first: voter
turnout/participation or direct democracy?" What is it about the history and culture of the
direct democracy states which has led them to adopt the process? These questions
deserve further attention and can lead to a better understanding of direct democracy and
In short, a review of the history presented in Chapter 2 suggested that the
use of direct democracy was one of several political tools adopted by certain states early
this century to help them deal with what they perceived to be a hostile, unfriendly
political system. Critics viewed state legislatures as corrupt and inattentive to certain
moral and philosophical positions that they believed were ill represented within their
states. The use of direct democracy swept quickly through a number of Western and
Southwestern states, but only one of the original thirteen colonies (Massachusetts in
1918) saw fit to adopt the process. Direct democracy became an important issue of
'states rights,' which could keep representative government at bay. As the contemporary
numbers suggest, however, more Americans than ever are now poised to vote in direct
democracy elections in the coming years.
The final chapter of this project takes the issue one step further by discussing a
typology of direct democracy questions. As stated in Chapter 1, not all direct democracy
questions impact voter turnout in the same way. The final chapter here discusses the
large variety of ballot questions that appear before voters and suggests that a typology of
ballot questions can be created using the morality policy literature as a basis for
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID ERI5C16EH_FXU7KQ INGEST_TIME 2013-10-24T22:39:22Z PACKAGE AA00017692_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC