LIBRARY
Michigan State
University

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRCIDateDuo.p65-p.15

 

 

MONEY, JUDGES, AND VOTES: THE EFFECTS OF CAMPAIGN SPENDING IN
STATE SUPREME COURT ELECTIONS

By

Christopher William Bonneau

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Political Science

2002

ABSTRACT
MONEY, JUDGES, AND VOTES: THE EFFECTS OF CAMPAIGN SPENDING IN
STATE SUPREME COURT ELECTIONS
By

Christopher William Bonneau

This dissertation is about campaign spending in state supreme court elections.
Specifically, I examine whether campaign spending by candidates for the state high court
bench promotes or inhibits electoral competition. Looking at all contested state supreme
court elections from 1990-2000, I find that, in some circumstances, campaign spending
promotes competition, while in other cases it retards it. The effects of campaign
spending on the vote total depend upon characteristics of the candidates, the state
electoral context, and institutional arrangements. The results both extend current findings
about the role of campaign spending in congressional and state legislative elections, and
speak to efforts currently underway in state legislatures to reform or eradicate judicial

elections.

Dedicated to my parents,

Edmund and J acklyn Bonneau

iii

ACKNOWLEDGMENTS

This dissertation could not have been written without the assistance of several
people. First, I would like to thank my dissertation committee of Melinda Gann Hall
(Chair), Paul Brace, Darren Davis, and Dave Rohde. By consistently challenging me and
being accessible to listen to my ideas, these individuals greatly improved the dissertation.
This project would not have been possible without the financial support of the National
Science Foundation (Grant: SES 0108906). I would also like to thank the Department of
Political Science at Michigan State University for providing an ideal environment in
which to attain a quality graduate education. I am grateful to the Departments of Political
Science at the University of Pittsburgh, Northern Illinois University, and the University
of North Texas for inviting me to present portions of this project. I received many useful
suggestions and ideas on those trips. Several fellow graduate students and faculty
members (both at Michigan State and elsewhere) were invaluable for their advice,
criticisms, and general friendship. I truly appreciate the contributions of: Charles
Finocchairo, Martha Humphries Ginn, Wonjae Hwang, Mark Hurwitz, Matt Kleiman,
Laura Langer, Mark Souva, and Rebecca Tothero. Finally, and most importantly, I could
not have asked for a better mentor than Melinda Gann Hall. Her advice, suggestions,
constructive criticisms, support, and patience were invaluable to the completion of this
dissertation. Her support throughout my four years at Michigan State was strong and

constant, and her professional career is one I can only hope to model.

iv

TABLE OF CONTENTS

List of Tables ........................................................................................ viii
Chapter 1: Introduction: Electing Judges ......................................................... 1
The Brief History of Electing Judges ...................................................... 3
State Supreme Court Elections and Democratic Theory .............................. 10
The Climate of State Supreme Court Elections ......................................... 15
Research Question .......................................................................... 17
Plan of the Dissertation .................................................................... 19
Chapter 2: Money and State Supreme Court Elections ....................................... 20
Money and the Composition of the Bench ............................................. 20
Money and the Actors in Judicial Elections ............................................ 22
The Role and Importance of Institutional Structures .................................. 27
The Role and Importance Of Candidates ................................................. 31
Conclusion .................................................................................. 35
Chapter 3: Research Design, Data, and Method ................................................ 37
The Importance of Comparative Research .............................................. 37
Research Design ............................................................................. 40
Data ........................................................................................... 44
Method ....................................................................................... 46
Conclusion ................................................................................... 51

Chapter 4: Describing State Supreme Court Elections ......................................... 52

State Supreme Court Elections Over Time .............................................. 52
Partisan versus Nonpartisan Elections .................................................... 62
State-by-State Comparisons ................................................................ 67
Conclusion ................................................................................... 74
Chapter 5: Explaining Campaign Spending ..................................................... 75
Specifying the Model ...................................................................... 75
Estimation Technique ..................................................................... 81
The Determinants of Campaign Spending .............................................. 84
Conclusion .................................................................................. 88
Chapter 6: The Effects of Money on Votes: Non-Institutional Model ...................... 89
Specifying the Models ..................................................................... 90
Specification ................................................................................ 98
Estimation Technique ..................................................................... 99
The Role of Money in State Supreme Court Elections ............................... 103
The Role of Independent Expenditures ................................................ 109
Conclusion ................................................................................. 1 12
Chapter 7: The Effects of Money on Votes: Institutional Model .......................... 114
Institutional Variables .................................................................... 114
Estimation Technique .................................................................... 119

The Role of Money and Institutions in State Supreme
Court Elections ............................................................................. 119

The Role of Money and Institutions in State Supreme

Court Elections: Selection System, Electoral Constituency
and Term of Office ........................................................................ 129

vi

The Role Of Independent Expenditures ................................................ 147

Conclusion ................................................................................. 151
Chapter 8: Conclusion ............................................................................ 152
Summary Of Results ...................................................................... 152
The Importance of Electoral Competition ............................................. 155
State Supreme Court Election Reform ................................................. 157
Future Research ........................................................................... 160
Appendix ............................................................................................ 163
References .......................................................................................... 175

vii

LIST OF TABLES

Table 1.1: Institutional Arrangements for States That Elect Their
Justices, 2000 .......................................................................................... 9

Table 4.1: Total Average Spending by Year, Contested General Elections
1990-2000 (number of elections in parentheses) ................................................ 53

Table 4.2: Total Average Spending by Year in 1990 dollars, Contested General
Elections 1990-2000 (number of elections in parentheses) .................................... 55

Table 4.3: Average Spending for Each Type of Candidate by Year, Contested
General Elections 1990-2000 ....................................................................... 57

Table 4.4: Average Percentage of the Vote by Year, Contested General Elections
1990-2000 ............................................................................................. 59

Table 4.5: Average Percentage of the Vote by Type of Election and Year,
Contested General Elections 1990-2000 .......................................................... 61

Table 4.6: Average Spending for Each Type of Candidate by Method
of Selection, Contested General Elections 1990-2000 .......................................... 63

Table 4.7: Average Percentage of the Vote by Type of Election and Method
of Selection, Contested General Elections 1990-2000 ......................................... 66

Table 4.8: Average Spending for Each Type of Candidate by State,
Contested General Elections 1990-2000 .......................................................... 69

Table 4.9: Average Percentage of the Vote by Type of Election and State,
Contested General Elections 1990-2000 (number of elections in parentheses) ............. 72

Table 5.1: Variable Descriptions for a Model of Campaign Spending in State
Supreme Court Elections ........................................................................... 80

Table 5.2: Variable Descriptions for the Selection Model of Campaign Spending
in State Supreme Court Elections ................................................................. 83

Table 5.3: Predicting Campaign Spending ...................................................... 85

Table 6.1: Variable Descriptions for a Model of Electoral Competition in State
Supreme Court Elections ........................................................................... 97

viii

Table 6.2: The Effects Of Money on Votes, Incumbent-Challenger
Contests ............................................................................................ 104

Table 6.3: The Effects of Money on Votes, Open Seats ...................................... 107

Table 6.4: The Effects of Money on Votes, Incumbent-Challenger
Contests, with Independent Expenditure Races Omitted ...................................... 110

Table 6.5: The Effects of Money on Votes, Open Seats, with Independent
Expenditure Races Omitted ...................................................................... 111

Table 7.1: Additional Variable Descriptions for a Model of Electoral
Competition in State Supreme Court Elections .............................................. 118

Table 7.2: The Effects of Money on Votes, Incumbent-Challenger
Contests .............................................................................................. 120

Table 7.3: Comparison of Non-Institutional and Institutional Models of
Incumbent-Challenger Contests .................................................................. 124

Table 7.4: The Effects of Money on Votes, Open Seats ..................................... 126

Table 7.5: Comparison of Non-Institutional and Institutional Models Of
Open Seat Contests ................................................................................ 128

Table 7.6: The Effects of Money on Votes, Incumbent-Challenger
Races with Selection System Interaction Terms .............................................. 130

Table 7.7: The Effects of Money on Votes, Incumbent-Challenger
Races with District Interaction Terms .......................................................... 131

Table 7.8: The Effects of Money on Votes, Incumbent-Challenger
Races with Multi-Member Interaction Terms .................................................... 132

Table 7.9: The Effects of Money on Votes, Incumbent-Challenger
Races with Term Interaction Terms ............................................................. 133

Table 7.10: The Effects of Term Length on Spending, Incumbent-
Challenger Contests ............................................................................... 137

Table 7.11: The Effects of Money on Votes, Open Seats with Selection
System Interaction Terms ........................................................................ 140

Table 7.12: The Effects of Money on Votes, Open Seats with District
Interaction Terms .................................................................................. 141

Table 7.13: The Effects of Money on Votes, Open Seats with Multi-
Member Interaction Terms ....................................................................... 142

Table 7.14: The Effects of Money on Votes, Open Seats with Term
Interaction Terms .................................................................................. 143

Table 7.15: The Effects of Term Length on Spending, Open Seat
Races ................................................................................................. 146

Table 7.16: The Effects of Money on Votes, Incumbent-Challenger
Contests, with Independent Expenditure Races Omitted ..................................... 148

Table 7.17: The Effects of Money on Votes, Open Seats, with

Independent Expenditure Races Omitted ....................................................... 149
Table A.1: OLS Results for Table 5.3 .......................................................... 164
Table A2: Hausman Test Results for Table 6.2 .............................................. 165
Table A3: Hausman Test Results for Table 6.3 .............................................. 166
Table A4: OLS Results for Table 6.2 .......................................................... 167
Table A5: OLS Results for Table 6.3 .......................................................... 168
Table A6: OLS Results for Table 6.4 ......................................................... 169
Table A.7: OLS Results for Table 6.5 ......................................................... 170
Table A8: Chow Test Results for Table 7.2 .................................................. 171
Table A9: Hausman Test Results for Table 7.2 ............................................. 172
Table A.10: Hausman Test Results for Table 7.4 ............................................. 173
Table A. 1 1: OLS Results for Table 7.2 ........................................................ 174

CHAPTER 1

INTRODUCTION: ELECTING JUDGES

The 2001 race for the Louisiana Supreme Court was one of the most contentious
(and expensive) in the history of the state (Swerczek 2001). Democrat Court of Appeals
Judge John Weimer faced off against fellow Democrat (and district court judge) Mary
Hotard Becnel. The winner of the election would sit on the court for the final year of
retired Justice Harry Lemmon’s term before running for reelection in 2002. Court
Observers saw this election as having the potential to affect the balance of the court
between the pro-trial lawyer forces (who supported Becnel) and the pro-business troops
(who supported Weimer). While the race was conducted with civility, and turnout was
Ii ght, both candidates campaigned vigorously for the seat (Swerczek 2001). Despite
spending only $272,000 (compared with Becnel’s $828,000), Weimer won the race with
51% of the vote. What factors contributed to Weimer’s victory? Are these factors
consistent across all elections for open seats? In this election for an open seat, the
candidate who spent the most money did not win the election. Does this mean that
spending is unrelated to electoral success in open seat elections?

The 2000 elections for the Alabama Supreme Court were some of the most
expensive (and bitter) in state supreme court election history (Goldberg, Holman, and
Sanchez 2002; Omdorff 2002). With five seats up for election, both sides of the tort
reform battle (pro-business interests and trial lawyers) saw this as a perfect opportunity to
solidify support for their cause on the court. Republicans won both open seats (Roy

Moore defeated Sharon Yates and Robert Harwood defeated Joel Laird), and the one

Republican incumbent (Champ Lyons) was reelected easily. Incumbent Ralph Cook, a
Democrat, was ousted by Republican Lyn Stuart, 3 Circuit Court judge. Stuart spent
$1,254,450 on her campaign, outspending Cook, who only spent $437,482. Incumbent
John Henry England (Democrat) did not fare much better. He lost his reelection bid to
Thomas Woodall, a Republican Circuit Court Judge. England spent $500,681 on his
reelection bid, an amount dwarfed by the $1,107,839 spent by Woodall. Did the amount
Of money spent by Stuart and Woodall have anything to do with the defeat of Justices
Cook and England? Did Stuart and Woodall spend enough money to overcome the
advantages of incumbency? Or was the election decided along ideological and partisan
lines, with the Republican candidates winning in a decidedly Republican state?

In the 2000 Ohio state supreme court elections, Justice Alice Robie Resnick, a
Democrat, was targeted for defeat by pro-business interests. She was challenged by
Court of Appeals Judge Terrence O’Donnell. Despite the Chamber of Commerce (and
other similar groups) spending over $5 million on the campaign, Resnick won with 56%
of the vote (Brown 2000). At the same time, Justice Deborah Cook, a Republican, won a
less contentious race over Municipal Court Judge Timothy Black. Taken together, do
these results mean that campaign spending does not matter, at least where incumbents are
concerned? Does the incumbency advantage trump spending in nonpartisan states? Are
there differences between the effects of spending in partisan and nonpartisan states?
How else can these results be explained?

This dissertation examines the relationship between electoral competition and the
amount of money spent by candidates in state supreme court elections. The election of

judges is one of the most interesting, and distinctive, facts Of American political life.

Voters in forty-three states elect state court judges at some level (Carp and Stidham
2001), and thirty-eight states have some form of election for their justices on the state
supreme court. Despite the fact that a majority of state court judges are elected, we know
little about the factors that determine the outcomes of these elections. This dissertation

represents one of the first comprehensive, systematic forays into this topic.

The Brief History of Electing Judges

The election of judges is an almost uniquely American phenomenon (Schotland
1985). The early American states selected judges much like their former British rulers:
they were appointed for life terms (although there were often provisions for
impeachment).l Some states had the judges appointed by the legislatures, while others
provided for gubernatorial appointment. In short, the selection of state court judges
paralleled the selection of federal judges (Sheldon and Maule 1997).

Originally, judges were selected in this way because the judiciary was considered
a weak institution. Consequently, courts were extremely deferential to legislatures, since
they were heavily dependent on them (Sheldon and Maule 1997). Interestingly, now
selection schemes like this are touted as promoting the independence of judges from both
the electorate and other political actors. Judges should be free to interpret the law and
make judicial decisions independent of as many constraints and political factors as

possible. Being appointed by the governor or legislature for a life-term frees judges from

 

' Of the twelve states that currently appoint justices to the high court bench, only one (Hawaii) is not one of
the original states of the union, or is a state that was once part of one of the original states (Maine was
initially part of Massachusetts, and Vermont was originally part of New York. Both Maine and Vermont
have their governor appoint their state supreme court justices.) Further, only two of the original thirteen
states have true elections for the state supreme court (Georgia and North Carolina).

paying attention to factors irrelevant to their jobs as judges.2 By not conditioning their
continued employment on things like elections, judges will be free to interpret the law
and make rulings based on legal criteria (such as the facts of the case, the law, and
precedent) and not on how they think the public will react.

It is important to note that proponents are appointed schemes are not really
arguing for complete “independence” from all political actors. As was discussed above,
judges who are appointed (and need to be reappointed) can hardly be said to be
independent from the appointing authority (legislature, governor, or both). Thus, the
“independence” argument is really an argument for independence from the electorate.
Certainly, with the rare exception of those states that provide for life-tenure, judges who
are dependent upon either the governor or the legislature for reappointment are not
completely “independent.” Presumably, an appointed judge who makes decisions that
are contrary to the policy preferences of the actor (or actors) who has (have) the power to
reappoint her could be subject to the same sanction as the elected judge who makes
decisions contrary to the preferences of the electorate: loss of her job. In fact, this
dependence on the legislature or governor (the lack of independence Of judges) was one
of the reasons given to move away from the appointment of judges to their election
(Sheldon and Maule 1997).

The appointment of judges was common until about 1830. This coincided with
the demise of the Federalists, and the rise of J acksonian democracy. “J acksonian

democracy meant that the average citizen could not only use the extended franchise to

 

2 Interestingly, many appointive schemes in the states do not have life-tenure for their judges. Indeed, only
Rhode Island has such a system, while three other states (Massachusetts, New Hampshire, and New Jersey)
have terms of office for this judges that expire when the judges reach age 70. In some other states (such as
Delaware, Hawaii, Massachusetts, and Vermont), the governor must limit her selection to one of the
candidates recommended by a judicial nominating commission (Hall and Bonneau 2000).

pick his leaders, but he could share responsibility for governing” (Sheldon and Maule
1997, 3). In terms of courts, this philosophy led to elected judges and shorter terms of
office. There were three primary reasons why judicial elections gained popularity. First,
there was the issue of judicial review. Since judges were invalidating laws enacted by the
legislature, it was argued that they should be chosen by the electorate, just like the
legislature.3 Second, the legal profession hailed the movement toward an elected
judiciary, seeing it as “an opportunity to provide the judiciary with its own separate
constituency, thus ensuring it some measure of independence from the legislature”
(Sheldon and Maule 1997, 4). As an independent branch of government, it was argued
that judges should not be agents of the legislature. Third, and finally, popular elections
allowed for the removal of incompetent and arrogant judges. While judges could be
impeached under appointive schemes, impeachment was thought of as too drastic a
punishment, and thus was used sparingly (Friedman 1985, 129—130).

The primary motivation for electing judges was that judges are political Officials,
and as political officials they should derive their political power from the people, and not
from a co-equal branch of government. Indeed, the process of electing judges is the only
process that can make them independent from the legislature and governor. Further,
competitive elections provide for electoral accountability: judges, like legislators, are
accountable to the electorate for their decisions and actions. Judges are thus less likely to
be out-of-step with the political climate of their state. After all, supporters of elections
would say, if a state provides for capital punishment, why should a judge who refuses to

uphold the death penalty be allowed to serve? This judge is refusing to apply the law as

 

3 It is interesting to note that even states that did not change the method by which they chose judges “added
restraint by shortening the terms for judges" (Sheldon and Maule 1997, 4), which also serves to reduce
judicial independence, albeit in a different way.

it is written, and is substituting her own personal policy preferences for the law. While it
would be extremely difficult to remove this judge in an appointed state, the voters can
remove this renegade judge in her next election.

Critics of judicial elections make two central arguments. First, they claim that
judges should not be held accountable to the electorate. Judging is not an intrinsically
political act like legislating, and judges should not be penalized by the electorate for their
rulings (e. g., Marks and Hoke 2001; Mikva and Sessions 2001; Pittsburgh Post-Gazette
2001a, 2001c; Wenzel 2001). Second, and relatedly, opponents of judicial elections
argue that even if it were conceded that electing judges is not, de facto, inappropriate, in
their current form judicial elections are ineffective mechanisms for holding judges
accountable. Specifically,

Reforrners argue that partisan elections, characterized by lackluster

campaigns devoid of issue content, are disconnected from substantive

evaluations of candidates or other meaningful considerations relevant to

the judiciary, which renders them ineffective as a means of accountability

(Hall 2001a, 316)

Regardless of whether judges should, or should not, be independent from the electorate,
the debate is irrelevant if elections fail to provide electoral competition, a necessary
component electoral accountability.

In response to criticisms like these, judicial elections have undergone two
significant evolutions since the 1830’s. First, while judicial elections were initially
partisan, there soon became allegations of corruption and control of the bench by political
machines. Around 1900, Progressives began pushing for judicial elections to be

nonpartisan. In these types of elections, candidates would run in a competitive election,

but without party labels. Supporters of nonpartisan elections argue that they provide for

electoral accountability, while not subjecting judges to the vagaries of partisan tides.
Thus, voters are able to remove a judge from office, but a judge will not be removed
simply by belonging to the wrong political party. By removing partisan considerations,
the hope was that more qualified jurists would be elected to the bench.4

However, removing partisan labels had the effect, perhaps unintended, of making
these races less competitive (see Chapters 4 and 7). By taking a crucial piece of
information away from the voters, states electing their judges on nonpartisan ballots
served to strengthen the incumbency advantage. Thus, not only are judges independent
from the legislature (since they are elected), but they are also less accountable to the
electorate (because the elections are less competitive).

A second reform, first proposed around 1913, is commonly referred to as “merit
selection.” While there are several variations of this selection scheme, in its most
common form a justice would be initially appointed to the bench by the governor (from a
list of names—usually three—submitted by a nominating commission) and this justice
would serve for a period of time (usually one or two years) before facing the voters in a
nonpartisan retention election. Voters would then be asked, “Should Judge X be
retained?” If the judge received more “yes” votes than “no,” then she would retain her
seat and serve for a fixed period of time (6 to 12 years, depending on the state) before

facing the voters again. If the justice was defeated, then the governor would fill the

 

4 Interestingly, the Progressives appear to be wrong on both counts. Glick and Emmert (1987) and
Bonneau (2001a) find no difference in the quality of judges between selection systems, and Hall (2001a)
finds that nonpartisan elections are driven by political factors, just as partisan elections. However,
Schaffner, Streb, and Wright (2001) find that, in local and state legislative elections, voters in nonpartisan
elections rely less on partisan cues and more on incumbency.

vacancy just as she did originally. Like nonpartisan elections, this was designed to
ensure a more qualified bench and remove political considerations from selecting judges.5

Merit selection is by far the system that gives the judges maximum independence
from both the legislature and the electorate. Once judges are appointed, they are not
dependent on the legislature (or governor) for reappointment. Further, since they run
unopposed for retention, they are also, according to all measurable indicators,
independent from the electorate. Indeed, these candidates very rarely lose. While
presuming to combine the “best” features of appointed (independence) and elected
(accountability) schemes, merit selection has the effect of making judges independent
from both other political actors (since they are not dependent on them for their
reappointment) and the electorate (since they very rarely lose their bids for retention).

Thus, the means by which judges are selected has moved from being independent
from the electorate but highly dependent on the legislature and governor (appointed), to
being accountable to the electorate and independent from the legislature and governor
(elections), to being independent from both the legislature and governor and the
electorate (merit).

In addition to the method by which they select their state supreme court justices,
states also vary on other institutional factors, such as term of office, electoral

constituency, and structure of the seat. In Table 1.1, the states that elect their high court

justices are displayed, along with their current institutional arrangements.

 

5 The evidence on the success of this plan is similar to the success of nonpartisan elections: there is no
evidence of a more qualified bench (Glick and Emmert 1987; Bonneau 2001a), and these races are not
immune from political considerations (Hall 2001a).

Table 1.1: Institutional Arrangements for States That Elect Their Justices, 2000

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

State Selection Term of Electoral Seat
System Office Constituency Structure

Alabama Partisan 6 Statewide Single-Member
Alaska Retention 10 Statewide Single-Member
Arizona Retention 6 Statewide Single-Member
Arkansas Partisan 8 Statewide Single-Member
California Retention l2 Statewide Single-Member
Colorado Retention 10 Statewide Single-Member
Florida Retention 6 Statewide Sligle-Member
Georgia Nonpartisan 6 Statewide Single-Member
Idaho Nonpartisan 6 Statewide Single-Member
Illinois Partisan/Retention* 10 District Single—Member
Indiana Retention 10 Statewide Single-Member
Iowa Retention 8 Statewide Single-Member
Kansas Retention 6 Statewide Single-Member
Kentucky Nonpartisan 8 District Single-Member
Louisiana Partisan 10 District Single-Member
Maryland Retention 10 District Single-Member
Michigan Nonpartisan" 8 Statewide Multimember
Minnesota Nonpartisan 6 Statewide Single-Member
Mississippi Nonpartisan 8 District Single-Member
Missouri Retention 12 Statewide Sigh-Member
Montana Nonpartisan/Retention*** 8 Statewide smile-Member
Nebraska Retention 6 District smile-Member
Nevada Nonpartisan 6 Statewide Single-Member
New Mexico Partisan/Retention*** 8 Statewide Single-Member
N. Carolina Partisan 8 Statewide Single-Member
N. Dakota Nonpartisan 10 Statewide Single-Member
Ohio Nonpartisan“ 6 Statewide Single-Member
Oklahoma Retention 6 Statewide Single-Member
Oregon Nonpartisan 6 Statewide Single-Member
Pennsylvania Partisan/Retention* 10 Statewide Multimember
S. Dakota Retention 8 Statewide Single-Member
Tennessee Retention 8 Statewide Single-Member
Texas Partisan 6 Statewide Single-Member
Utah Retention 10 Statewide Single-Member
Washington Nonpartisan 6 Statewide Single-Member
W. Virginia Partisan 12 Statewide Multimember
Wisconsin Nonpartisan 10 Statewide Single—Member
Wyoming Retention 8 Statewide Single-Member
* = originally elected in partisan elections; retention elections after

initial election
** 2 candidates nominated in partisan primaries/conventions
*** _

retention election only if unopposed

Regardless of form, elections to the state high court bench have been around since
1830, with no sign of disappearing. Given their importance, and the constant attempts to
reform (or eradicate) them, it is more than a little surprising that we know little about

these elections.

State Supreme Court Elections and Democratic Theory

Our lack of knowledge about judicial elections is troublesome, especially since
elections are one of the cornerstones of democracy. American democracy is founded (at
least in part) on the notion that officials are to be elected by their constituents (Pomper
1968). In states that have elections for state supreme court seats, justices are to be
selected by the voters of the state (or district) in which they reside, just as legislators.

One of the most important facts about elections is that they provide for
competition. That is, they provide the opportunity for voters to choose some candidates
over others. This is important because, “[T]he degree to which voters are not offered
choices on the ballot raises significant questions about the health of a democracy” (Squire
2000, 131). Competition can be conceptualized in two different ways.

First, a necessary condition of competition is simply the presence of alternatives.
That is, there must be more than one candidate from which to choose. Thus, under this
conceptualization, any election for a seat on the state high court bench is considered
“competitive” if more than one candidate seeks the seat. This definition has the
advantage of defining a race as competitive before the votes are tallied. Additionally, this
recognizes the fact that “[E]ven facing a weak challenger can cause a representative some

concern” (Squire 1989a, 283). However, the simplistic nature of this conceptualization is

10

also its biggest drawback; it confuses “competitive” with “contested.” Applying this
logic to congressional elections, the 2000 US. Senate race in Massachusetts between Ted
Kennedy and Jack Robinson was “competitive,” even though Kennedy won with 73% of
the vote. A competitive race must be contested; but a contested race does not have to be
competitive. The Kennedy race just mentioned was certainly contested, but could hardly
be said to be competitive.

An alternative conceptualization of a “competitive” race focuses on the
percentage of the vote obtained by both candidates. Under this definition, races can only
be said to be “competitive” after the election is complete. The lower the margin of
victory for the winner, the more competitive the race. Using this definition, the
Kennedy-Robinson race is viewed as less competitive than the 2000 US. Senate race in
Michigan between Debbie Stabenow and Spencer Abraham (in which Stabenow won
with 49% of the vote), although it is more competitive than if Kennedy were not
challenged at all. Instead of conceptualizing “competition” as dichotomous (present or
not), this definition views competition as a continuum, with races being more or less
competitive, depending on the level of support provided to each candidate by the voters.

In this dissertation, I focus on the second definition of competition. The reasons
for my choice are simple: I am interested in what factors promote (or inhibit)
competitive elections. While one can do this using a dichotomous measure (Bonneau and
Hall 2001), doing so requires me to consider the Kennedy and Stabenow races as
equivalent—they are both “competitive.” However, I am interested in what makes races
more or less competitive. That is, why did Kennedy win so easily, while Stabenow was

in a dogfight? More specifically, I focus on the effects of campaign spending on

11

competitiveness. Understanding the electoral process is centered (to some degree)
around this question: what factors explain competition, and can competition be promoted
(or inhibited) by changing certain conditions? This dissertation focuses on competition in
elections to the state high court bench.

It is important to study and understand electoral competition because its presence
(or absence) serves as a constraint (or lack thereof) on electoral actors. From the
congressional literature, we have learned that “[E]ven ‘safe’ representatives who face
challengers live with some fear that they may be defeated, and this wariness may
constrain their actions” (Squire 1989a, 282). For example, a congressional incumbent
who is facing competition in her bid for reelection may feel compelled to support
legislation she may otherwise oppose for fear of losing her seat. The more serious the
competition, the more constraint she will feel.

The same holds true for state supreme court incumbents: The presence of
competition may alter her behavior on the bench. In a series of studies, Brace and Hall
found that justices alter their behavior in the presence of elections (Brace and Hall 1990,
1995, 1997; Hall 1987, 1992, 1995). “[A]ppointive selection procedures are
associated with considerably less dissent [on state supreme courts]” (Brace and Hall
1990, 65; see also Hall 1992). The fact that they have to face the electorate and may be
subject to electoral competition alters the behavior of state supreme court justices on the
bench. More fundamentally than dissent rates, Hall (1995, 498) finds that, “The
willingness of supreme court justices actually to condemn individual defendants to death
may rest, at least to some extent, on general electoral conditions and the justices’

individual experiences with electoral politics.” Likewise, the frequency of facing the

12

voters affects behavior on the bench: “[T]he political result of shorter terms of office is a
decidedly more conservative bench, at least on the issue of the death penalty” (Brace and
Hall 1997, 1223; see also Brace and Hall 1995).

These examples point to the fundamental importance of studying electoral
competition: competition is essential for elected officials to behave in a way congruent
with the preferences of the electorate. The more competitive the election, the more
congruent a justice’s behavior will be with the electorate. Elections provide the
opportunity for the public to control the composition of the state supreme court bench
(and thus, indirectly, the outcome of cases); competitive elections are the realization of
this provision. This explains the findings by Hall (1995) that justices are more willing to
uphold capital punishment sentences based on their proximity to reelection.

It is important to note that what has been found to affect the behavior of actors
seeking election or reelection is solely due to the presence of competition. It has little to
do with what the electorate knows, or does not know, about the candidates. A
competitive election can be an effective mechanism by which to affect behavior even in
the absence of knowledge by the voters. “[E]ven in situations Of incomplete information,
there are powerful equilibrating forces pushing public policy in the direction of the
representation of public sentiment” (Ferejohn 1990, 5). All that is required is that the
voters are minimally concerned about the acts of the political actors (Key 1961; Pomper
1968; Ferejohn 1990). Indeed, “[T]he voter need not know what is wise, but only what is
personally satisfactory or obnoxious” (Pomper 1968, 39; see also Key 1961, 467). Thus,
despite the fact that voters are often ignorant both of the issues and the candidates (Key

1961), competition is vital to efficacious elections, if for no other reason than because it

13

is by the electoral process that voters can retain or remove Officials. Voters “can quickly
and bloodlessly dismiss an offensive official and thereby end his power, prestige, and
profit. No explanations need be given by the electorate, and no appeal can be taken from
its decisions (Pomper 1968, 254). For this to be true, there must be competition.

Clearly, electoral competition is vital for elections maintaining their legitimacy
and fulfilling their raison d’etre: allowing voters to choose their elected officials and
holding these officials accountable for their actions. While the necessity of competition
is clear, it is less clear how competition can be promoted. By the same token, while we
know much about the determinants of competition in congressional and state legislative
elections, these processes remain unknown in state supreme court elections. In this
dissertation, I focus on one key potential influence on competition in state supreme court
elections: campaign spending. To the extent that state supreme court election
competitiveness is affected by campaign spending, then the ability of the electorate to
make meaningful choices is compromised.

This dissertation has very practical implications for the debate over campaign
finance regulation currently raging in both state legislatures and the US. Congress. If
campaign spending by incumbents is directly related to their percentage of the vote in a
positive manner, then incumbent spending will serve to decrease competition, and add to
the already large advantages enjoyed by incumbent (see Jacobson 1997, 19-34 for a
summary of this literature). Similarly, if spending by challengers erodes the incumbent’s
percentage of the vote (and thus boosts the electoral support of the challenger), then
campaign spending would serve to promote competition. By limiting the amount of

money raised (and thus spent), states will serve to further strengthen the electoral position

14

of incumbents. The same holds true for candidates for open seats: campaign spending
may either promote or inhibit competition, and the states can structure these effects by
the types of campaign finance laws adopted. Finally, it may be the case the campaign
spending has no effect on competition, which means that promoters (and detractors) of
competition in judicial elections will have to look elsewhere for ways to either promote
or discourage competition. Also unknown is what effects, if any, institutional
arrangements have in structuring these relationships: Does it matter if the election is
partisan or nonpartisan? Occurs statewide or in a district? Occurs in a single-member
district or a multi-member district? At any rate, understanding the effects of spending on
electoral outcomes is necessary before measures intended to increase, or decrease, the

competitiveness of elections can be seriously considered.

The Climate of State Supreme Court Elections

To say that state supreme court elections, and the funding thereof, is a popular
political issue is an understatement. There has been an abundance of proposed
legislation, press coverage, and rhetoric about judicial elections in recent years. In every
state that elects judges to its high court bench, legislators, candidates, and the opinion
pages of local newspapers have called for either radical reform of the electoral process or
the eradication of judicial elections altogether.

In December of 2000, the Chief Justices of fifteen states, as well as others
interested in the politics of judicial elections, gathered in Chicago at a special summit to
discuss issues plaguing the conduct of judicial elections. This meeting, which culminated

in a Call to Action, was primarily designed to address the “growing concerns about the

15

million dollar war chests, attack advertising and even outright distortion of an opponent’s
record that seems to have become more widespread ...” (Glaberson 2000b, 1).6 This
summit comes on the heels of two 60 Minutes stories (1987, 1998), a F rontline
investigation (1999), and countless law review articles and other reports in the popular
media, all of which assume a strong, direct link between campaign expenditures and
election outcomes (American Bar Association 1998; Hansen 1998; Glaberson 2000a,
2000b).

Most recently, the Justice at Stake campaign released a report on the 2000 judicial
elections. This report, which examines contributions and expenditures in all states that
conducted state supreme court elections in 2000, concludes that

[T]he year 2000 signaled a dangerous turning point for America’s courts,

documenting the growing systematic, and unprecedented infusion of big

money and special interest pressure into the election of Supreme Court

justices across the country (Goldberg, Holman, and Sanchez 2002, 4).

The report links the increased campaign fundraising and expenditures to the
compromising of the fair and impartial administration of justice. Indeed, it states that,
“2002 could be a decisive year for the struggle to keep courts fair and impartial”
(Goldberg, Holman, and Sanchez 2002, 4). The report further implores both politicians
and voters to help stem back the tide of expensive and contentious campaigns for the
state high court bench, else “the threat to fair and impartial justice will grow, perhaps
rapidly” (Goldberg, Holman, and Sanchez 2002, 6).

It is not just judges and the media who are calling for the reform or elimination of

judicial elections. Governors in states such as Michigan (Bell 2001; Dickerson 2001a)

and Pennsylvania (Pittsburgh Post-Gazette 2001b) are proposing replacing elections for

 

6 The papers presented at this summit, as well as the Call to Action, were subsequently published in the
June 2001 Loyola of Los Angeles Law Review.

16

the state high court with appointments by the governor. The state legislatures in
Wisconsin and North Carolina are considering providing full public campaign financing
for some candidates (Glaberson 2001). Further, some states have even changed the
method by which they select supreme court justices, at least in part to insulate members
of those courts from the electorate and the rigors of campaigning for office. Since the
1960’s, sixteen states have changed their method of selecting supreme court justices (Hall
2001a). Moreover, some states have changed their selection system quite recently. In the
mid-1990’s, Mississippi changed from partisan elections to nonpartisan elections, and
Tennessee moved from partisan elections to retention elections. Additionally, Arkansas
moved from a partisan election scheme to nonpartisan elections, a change that takes
effect in 2002.

Clearly, the issue of judicial elections (and the funding thereof) is one that is
occupying increasing amounts of time on the public and legislative agenda. Underlying
many of these claims is that the amount of campaign spending serves to decrease
electoral competition. That is, opponents of judicial elections claim that elections are
being won by the candidate who spends the most money, and the more money that is
spent, the larger the electoral victory. It thus becomes even more important to
systematically examine and understand the dynamics of these elections, especially since

states are changing their methods of selection on the basis of untested, unverified claims.

Research Question
Former North Carolina Supreme Court Chief Justice Henry Frye recently said,

“‘Someone needs to do a careful analysis to find out exactly what things make a

17

difference in judicial elections’” (Berry 2001, 37). This dissertation is about electoral
competition in state supreme court elections. I am interested in what factors promote or
retard electoral competition for the state high court bench. Central to my analysis is the
role of campaign spending in these elections. Specifically, I ask: Does campaign
spending affect the percentage of the vote received by state supreme court candidates? In
order to answer this question, I examine all contested elections for the state high court
bench between 1990-2000.

This is an important question for the reasons detailed above: allegations abound
about the negative effects of campaign spending on the electoral process. However, none
of these claims have been tested in any systematic manner. It may in fact be the case that
campaign spending decreases competition, and the more money spent by the winner of
the election, the larger her percentage of the vote will be. Alternatively, it may be the
case that campaign spending promotes competition. The more money that is spent, the
higher voter awareness of the candidates and their respective positions will be, especially
for candidates who are relatively unknown. This would serve to weaken the incumbency
advantage (as well as other advantages such as name recognition), and lead to more
competitive races. Further, these effects may vary depending on the type of candidate
(incumbent or non-incumbent) as well as the institutional setting in which the election is
occurring. Addressing competition in state supreme court elections is essential to a

complete understanding of both the electoral and judicial process.

18

Plan of the Dissertation

In Chapter 2, I lay out a general argument about electoral competition and the
influence of campaign spending in state supreme court elections. Why should we expect
money to matter in these elections? What is it about campaign spending that should
make it so relevant to electoral outcomes?

My research design, data, and method are discussed in Chapter 3. In addition to
specifying my design, I also discuss the procurement of the data used in this project, as
well as my methodology for assessing the effects of money on votes.

Chapter 4 describes the context of state supreme court elections in the time period
studied here (1990-2000). Here, I look at patterns of competition and spending, and
break them down by year, selection system, type of election, and state.

Chapter 5 examines the determinants of campaign spending in state supreme court
elections. What factors promote campaign spending, and which factors retard it? Are
these factors the same factors that determine electoral competition?

In chapters 6 and 7, I empirically assess the influence of campaign spending on
election outcomes. In Chapter 6, I look at elections more generally, while in Chapter 7 I
explore the institutional differences between state supreme court elections. Does it
matter if the election is partisan or nonpartisan? District or statewide? Single- or multi-
member district? Drawing on the theory developed in this chapter as well as Chapter 2, I
specify multivariate models designed to uncover the influence of campaign spending on
state supreme court elections.

Finally, in Chapter 8, I summarize the major results of the dissertation, as well as

discuss some directions that future studies should pursue.

19

CHAPTER 2

MONEY AND STATE SUPREME COURT ELECTIONS

This dissertation focuses on competition and the role of money in state supreme
court elections. Specifically, I examine the relationship between candidates, contributors,
and voters. State supreme court elections, like elections for all other offices, revolve
around these three actors. Further, the relationships between these actors may vary
depending on institutional and political context. That is, the relationships between
candidates, contributors, and voters may be different in partisan elections compared to
nonpartisan elections, single-member seats compared to multi-member seats, statewide
elections compared to district elections, or highly electorally competitive states compared

to less electorally competitive states.

Money and the Composition of the Bench

Most fundamentally, elections dictate who will sit on the state supreme court
bench. Hall (2001a, 326) has shown that the outcomes of judicial elections are shaped by
“candidate- and issue-based forces.” Additionally, Bonneau and Hall (2000, 2001) have
shown that not only does the presence of a quality challenger decrease the incumbent’s
percentage of the vote, but also that the presence of a challenger (and quality challenger)
can be understood in predictable ways. However, none of these studies considers the role
of campaign expenditures. That is, we do not know if the amount of money spent by
candidates in state supreme court elections affects their electoral support, and thus serves

to promote or discourage electoral competition. Further, we do not know the extent to

20

which the influence of campaign spending varies under different institutional
arrangements.

It is important to understand the effects of campaign spending on state supreme
court electoral outcomes because campaign spending may influence electoral results in
unforeseen and unintended ways. On one side of the debate over the funding of elections
are those who argue that campaign spending is a form of free speech and thus that
campaign spending ought not, and cannot, be constitutionally regulated. Further, it is
believed that “[m]ore speech means more information, and more information produces an
enlightened and active citizenry” (Coleman and Manna 2000, 757; see also Palda 1994;
Smith 1996, 1999; Brubaker 1998). This is especially important in judicial elections
since these elections are characterized by low levels of information and saliency to the
electorate (Klots 1955; Johnson, Schaefer, and McKnight 1978; McKnight, Schaefer, and
Johnson 1978; Dubois 1979, 1984; Griffin and Horan 1979; Schotland 1985; Aspin and
Hall 1987, 1989; Champagne and Thielemann 1991; Jackson and Riddlesperger 1991;
Moog 1992; Arrington 1996; Reid 1996; Klein and Baum 2001). Viewed in this way,
campaign spending by candidates can have positive as well as negative effects and may
lead to more competitive elections, since the more information voters have about both
candidates, the closer the race should be (Jacobson 1980).

On the other side of the debate are those who believe that “campaign spending,
especially in unlimited amounts, is clearly the bane of democracy” (Coleman and Manna
2000, 758; see also Wertheimer and Manes 1994; Ferguson 1995). These people believe
that not only do candidates waste time to campaign and raise money (time that could be

better spent performing duties related to their office), but also that incumbents are

21

systematically advantaged and hence electoral competition is reduced.1 Moreover,
spending campaign money on “cynical, negative, and misleading campaign
advertisements” leads the public to become “distrusting, or, worse, apathetic and
uninvolved, and [thus] campaign spending fails to enlighten, engage, or educate the
public (Coleman and Manna 2000, 758). That is, not only does campaign spending serve
to protect those in office and decrease electoral competition, but it also contributes to the
general apathy of the electorate and distrust of the government.

While these two views clearly conflict, they do share one assumption: campaign
spending affects both the function of elections and electoral outcomes. Both camps argue
that campaign spending either promotes or inhibits electoral competition. This

dissertation tests this assumption in elections to state supreme courts.

Money and the Actors in Judicial Elections

Elections are composed of three primary actors: the candidates who are running
for office, the contributors to their campaigns, and the voters. Money affects each actor
in a different way.

Money is a necessary condition for candidates to run for Office (Heard 1960;
Alexander 1972; Jacobson 1980). This holds true regardless of the office being sought.2
In order to be competitive, candidates must raise and spend money (Jacobson 1997;
Cassie and Breaux 1998). In contested judicial elections between 1990-2000, the mean

spending for incumbents was $405,483, while it was $225,740 for challengers. Further,

 

' Given the high percentage of incumbents that are reelected (Hall 2001a), this is just as true in state
supreme court elections as it is in legislative elections.

2 While this statement is truer under some conditions (such as nonpartisan elections), it is no less true for
partisan elections. This finding has been confirmed extensively in congressional elections (all of which are
partisan), and there is no reason to suspect it would not hold in other partisan elections.

22

the average spending for winners of open seats was $379,908, compared to an average of
$283,265 for losers of open seat races. Interestingly, although the minimum amount
spent by a challenger who defeated an incumbent was only $500, the mean spending for
all challengers who unseated incumbents was $399,717.

Why might money be so necessary for success in state supreme court elections?
Simply put, money has been found to buy candidates access to voters, both in legislative
races (Jacobson 1978, 1980, 1990, 1997; Becker and Dunwoody 1982; Green and Krasno
1988, 1990; Squire and Wright 1990; Banaian and Luksetich 1991; Breaux and
Gierzynski 1991; Gierzynski and Breaux 1991, 1993, 1996; Squire 1992; Biersack,
Hermson, and Wilcox 1993) as well as state supreme court elections (McKnight,
Schaefer, and Johnson 1978; Dubois 1984, 1986a; Thielemann 1993; but see Arrington
1996). While being exposed to a candidate does not necessarily mean that a voter is
more likely to vote for that candidate, in general, voters are more likely to vote for
candidates with whom they are familiar (Jacobson 1980; Aspin and Hall 1987, 1989;
Alvarez 1997). Thus, in this regard, spending money benefits those candidates with
whom voters are unfamiliar (generally nonincumbents), and thus promotes competition.
With rare exceptions, state supreme court elections take place in a shroud of anonymity
and i gnorance.3 Not only are the candidates anonymous, but voters may also be ignorant
of the fact that an election is even occurring (Schotland 1985; Arrington 1996). This is
especially true in states like Idaho and Wisconsin, where elections for the state supreme

court occur in the spring, and not in the fall with elections for other statewide offices.

 

3 While this is less true in partisan elections (since the voters do have the political party affiliation of the
candidates), it is still true to the extent that the candidates themselves are unknown to most of the voters
(Champagne and Thielemann 1991).

23

Money therefore allows candidates to publicize their candidacies. This is more
important in elections for state supreme court because, unlike elections for governor or
congress, candidates for the high court bench do not receive much “free” publicity
(Thielemann 1993; Arrington 1996). A candidate for governor can receive free publicity
by announcing a new policy initiative, or by criticizing the incumbent. The same is true
for candidates seeking a seat in Congress (Jacobson 1980).4 Candidates for the state
supreme court are much more restricted in their ability to campaign by the canons of
judicial ethics, which prohibit candidates from commenting on cases (and topics) upon
which they may eventually decide.5 Hence, while a congressional candidate can receive
free news coverage by proposing mandatory life sentences for convicted drug traffickers,
a candidate for the state high court cannot (or at least could not, until Republican Party of
Minnesota v. White). Further, most state supreme court races have to compete with
elections for other state and federal offices. This, coupled with the relative obscurity of
state high court candidates in the first place, make publicizing one’s candidacy essential
to success—and it makes such publicity expensive.

The second actor in state supreme court elections is the contributor. Contributors
are the actors who decide to whom to give their money and how much they are going to

give. Given the importance of campaign spending to the potential success of a candidate,

 

4 Gierzynski and Breaux (1991, 1993) find that most candidates for the state legislature also cannot count
on receiving much free publicity.

5 Note that the recent United States Supreme Court decision in Republican Party of Minnesota v. White
(2002; 247 F 3d 854) invalidates most of these restrictions on candidate speech. Now that candidates
running for the state high court bench can campaign much like candidates for other elected offices, the
influence of campaign spending may increase (and thus we would also expect the amount of money spent
to increase). This is because campaign spending will allow the candidates to provide relevant information
to the voters about their views that was previously banned. If spending is effective even when candidates
cannot provide much meaningful information, it will be even more effective when meaningful information
can be provided. Because spending will be more effective, we would thus expect candidates to raise and
spend more money, because it is an easy way to increase their electoral support.

24

contributors play a vital, if not always visible, role in state supreme court elections.
There are several possible motivations for contributing money. Some people (or groups)
give money to candidates who share a similar ideology or party affiliation (Welch 1979;
Gopoian 1984; Su, Neustadtl, and Clawson 1995). In this case, a trial lawyer would
donate money to a candidate (likely a Democrat) because they both have a similar
philosophy regarding corporate liability and punitive damages. Others are issue-givers;
they give money to candidates who agree with them on a particular issue (or issues)
(Gopoian 1984; Munger 1989; Hall and Wayman 1990; Grier and Munger 1991; Romer
and Snyder 1994). A pro-life organization will give money to those candidates (likely
Republicans) who agree with them on restricting a woman’s right to choose. Some give
money because they personally know the candidate, or know someone who knows the
candidate (Thielemann 1993). And still others are motivated to contribute based on the
candidate’s merits; that is, they truly believe that the candidate is the most qualified for
thejob.

Contributors play a crucial role in the competitiveness of state supreme court
elections. Candidates must be able to spend (and thus need to raise) money in order to be
competitive. Evidence presented earlier suggests this is even more important for
nonincumbent candidates: the viability of their candidacy may depend, to a large extent,
on their ability to publicize their candidacy and increase their name recognition with the
voters. Contributors play a key role in the electoral process by enabling some candidates
(and handicapping others) to be competitive by providing them with one of they key tools

they need to be successful: money.

25

The voters are the third actor in state supreme court elections affected by money.
To the extent that judicial elections are low information, low-salience elections, the
electorate usually possesses little (if any) information on the candidates (Klots 1955;
Johnson, Schaefer, and McKnight 1978; McKnight, Schaefer, and Johnson 1978; Dubois
1979, 1984; Griffin and Horan 1979; Schotland 1985; Aspin and Hall 1987, 1989;
Champagne and Thielemann 1991; Jackson and Riddlesperger 1991; Moog 1992;
Arrington 1996; Reid 1996; Klein and Baum 2001). This is compounded by the fact that
candidates for judicial office are restricted in what they can say by the canons of judicial
ethics.6 What does this have to do with campaign spending? The more money raised and
spent by a candidate, the more she can get her name out there to the public. Jacobson
(1980, 31) found that

Although congressional voters have relatively little information about the

candidates, both the extent and content of information they do have a

decisive effect on how they vote. Normally, incumbents are much better

known than their challengers and so win with relative ease. Challengers

have a fighting chance only if they are able to use the campaign period to

neutralize the incumbent’s informational advantage—which is built up at

the public expense over his term of office.
Familiarity and name recognition, in addition to political party affiliation, are essential to
a successful candidacy (e.g., Abramowitz 1975, 1988; Jacobson 1978, 1980, 1981, 1997;
McKnight, Schaefer, and Johnson 1978; Dubois 1984, 1986a; Schotland 1985;
Champagne 1986; Thomas 1989; Banaian and Luksetich 1991; Breaux and Gierzynski
1991; Gierzynski and Breaux 1991, 1993; Squire 1992, 1995; Biersack, Hermson, and

Wilcox 1993). Increasing visibility costs money, and thus to provide voters with a

meaningful choice in judicial elections, candidates must raise and spend money.

 

6 Although the status of these canons are in doubt after Republican Party v. White (see footnote 5).

26

There is evidence to suggest that campaign spending does indeed increase the
knowledge level of the electorate, and thus contributes to competitive elections.
Examining elections to the US. House in 1994 and 1996, Coleman and Manna (2000,
783) conclude that:

[C]ampaign spending neither decreases political trust, efficacy, or interest

in and attention to campaigns. Spending does contribute to knowledge

and affect. Accurate perceptions of the incumbent’s record are generally

improved by incumbent spending and reduced by challenger spending, in

practice typically producing a net result of more accuracy and more

competitiveness
Further, Alvarez (1997, 204) finds evidence that “uncertainty [about the candidates]
generally diminishes across the course of a presidential campaign in response to issue and
substantive information.” Thus, contrary to those who bemoan the increased campaign
spending in judicial elections, there is some reason to believe that voters benefit from the
candidates spending money (if races for the state high court bench are similar to

presidential and House elections), and the higher the spending, the higher the benefits to

the electorate, and the more the competitive the election.

The Role and Importance of Institutional Structures

After ignoring institutions for many years, scholars of judicial politics have
recently started reexploring the importance of institutional structures to the politics of the
judiciary (Hall 1987, 1992, 1995, 2001a, 2001b; Hall and Brace 1989, 1992, 1999; Brace
and Hall 1990, 1995, 1997, 2000a, 2000b, 2001). This follows in the tradition of
legislative scholars (e.g., Mayhew 1974; Shepsle and Weingast 1987; Rohde 1991; Cox
and McCubbins 1993; Hall 1996; Lee and Oppenheimer 1999) as well as those who study

the bureaucracy (e.g., Moe 1987; Miller, Hammond, and Kile 1996). The conclusion of

27

these studies is clear: institutional arrangements matter. Institutional structures shape not
only the behavior of the actors within the institution, but also the context in which these
actors operate. There are good reasons to suspect that institutional structures will affect
both the amount of campaign spending as well as electoral competition and the
effectiveness of campaign spending.

The most basic difference in institutional arrangements among states is the
method by which the justices are elected. Other things being equal, candidates should
need to raise and spend more money in nonpartisan states compared to partisan states.
There are two reasons for this. First, having the party label next to the candidate’s name
provides the voter with a crucial piece of information to guide her decision (Klein and
Baum 2001; Schaffner, Streb, and Wright 2001). Second, and relatedly, there is the issue
of a partisan baseline vote. Since candidates in partisan states can receive votes from
voters who vote a straight-party ballot, while those in nonpartisan states cannot,
candidates in partisan states have an electoral advantage absent for candidates in
nonpartisan states (and thus partisan races should be more competitive). Money should
have more of an effect in nonpartisan races, because the lower the amount of information
provided to the voter, the more of an effect the campaign (and therefore money) should
have.

For example, consider a nonincumbent candidate running for the state high court
bench in a state with statewide, partisan, single-member district elections (such as
Alabama, Arkansas7, New Mexico, North Carolina, and Texas). If the findings from the

congressional literature hold, then this challenger can increase her percentage of the vote

 

7 Arkansas is switching to single-member statewide, nonpartisan elections in 2002. However, since
partisan elections were used there for the time period in this study, Arkansas is still classified as a partisan
election state.

28

(make the election more competitive) simply by spending more money. The more money
the challenger spends, the higher her level of electoral support will be.

However, what if the same nonincumbent candidate were running for election in a
state with statewide, nonpartisan, single-member district elections (like Georgia, Idaho,
Minnesota, Montana, Nevada, North Dakota, Ohio, Oregon, Washington, or Wisconsin)?
While not much research has been conducted on nonpartisan elections, existing evidence
indicates that campaign spending by a candidate serves to increase her percentage of the
vote, for both incumbents-and nonincumbents (Gierzynski, Kleppner, and Lewis 1998).
Thus, unlike for partisan elections (where incumbent spending does not matter),
incumbent candidates, like nonincumbent candidates, are able to increase their percentage
of the vote by spending more money. This suggests that the incumbency advantage
(Jacobson 1978, 1981, 1985, 1989, 1997; Dubois 1984, 1986a; Abramowitz 1988; Banks
and Kiewiet 1989; Cassie and Breaux 1998; Erikson and Palfrey 1998; but see Squire and
Wright 1990) may be larger in nonpartisan states than in partisan states.

The effects of money should also be contingent on whether the race is being run
statewide or in a district. Other things being equal, there should be more campaign
spending and this spending should be more effective (translate more easily into electoral
support) in district elections as opposed to statewide elections. While no differences have
been found in a similar circumstance between US. Senate elections, which are statewide,
and US. House contests, which are run in districts (Jacobson 1978, 1980, 1985;
Abramowitz 1988), this could be due to the differences in the offices, and not to the
electoral constituency. For state supreme court elections, the office is a constant, while

the electoral constituency varies. While the total amounts of spending are likely to be

29

lower in district-based elections (Gierzynski 1998), the fact that the district is more
compact (both geographically and demographically) should lead to spending being more
effective (Dubois 1986a; Hogan and Hamm 1998; Reid 1999). This should also lead to
more competition, at least under certain conditions. Hall (2001a) found that district-
based elections promote competition in nonpartisan elections, but not in partisan
elections. Higher levels of spending, then, may lead to more competition in district-based
elections as opposed to statewide elections.

The importance of spending also depends on whether the election is for one seat
(single-member) or more than one seat (multimember). In multimember races, all
candidates run against all other candidates for two or more seats; in single-member races,
only one seat is at stake. Competition in multi-member races has been found to be higher
than in comparable single-member districts (Cox and Morgenstem 1995). Further, these
races tend to be more expensive than single-member district elections (Gierzynski 1998).
This is because there are more candidates running in these elections. A candidate will
have greater difficulty attaining name recognition if she is running against three other
candidates as opposed to only one other candidate, other things being equal. Thus, not
only should multi-member races be more expensive, but higher levels of spending should
also better promote electoral competition in multimember districts than in single-member
districts.

Finally, both electoral competition and the effectiveness of campaign spending
should vary depending on the term of office at stake. Other things being equal, longer
terms should be more competitive than shorter terms (although Hall (2001a) found no

such effects), and campaign spending should be more effective in influencing a

30

candidate’s percentage of the vote when longer terms of office await the victor. Further,
races involving a longer term of office should also be more expensive. This is because
when state supreme courts have longer terms of office, there are fewer opportunities for
candidates to attain a seat on the bench.

Clearly, the amount of campaign spending, electoral competition, and the
effectiveness of campaign spending are all likely to be conditional on institutional
arrangements. Thus, failing to take into account institutional factors leads to an
incomplete understanding of the political process. This is one of the most compelling
reasons to study state supreme courts. While we know much about the dynamics of
elections and of campaign spending from both congressional and state legislative
scholars, we do not know the extent to which those findings are artifacts of the particular
institutional arrangements common to both Congress and state legislatures (Gierzynski
1998). Examining elections to state supreme courts allows for the testing of existing
explanations of the electoral process and campaign spending under different institutional

arrangements.

The Role and Importance of Candidates

Campaign spending should affect how well a candidate does in state supreme
court elections, because, quite simply, campaign expenditures allow for the provision of
information to voters in a low-information setting. It has already been established that
state supreme court elections are generally low-information, low-salience elections (Klots
1955; Johnson, Schaefer, and McKnight 1978; McKnight, Schaefer, and Johnson 1978;

Dubois 1979, 1984; Griffin and Horan 1979; Schotland 1985; Aspin and Hall 1987,

31

1989; Champagne and Thielemann 1991; Jackson and Riddlesperger 1991; Moog 1992;
Arrington 1996; Reid 1996; Klein and Baum 2001). In order to make their candidacies
known to the electorate, candidates must spend money. The better known the candidate
is, the less money she has to spend. Additionally, the better known a candidate is, the
higher her percentage of the vote should be, other things being equal. Most basically,
“Money is necessary because campaigns do have an impact on election results, and
campaigns cannot be run without it” (Jacobson 1980, 33).

Just because candidates are spending money does not necessarily mean that they
are spending the money on advertising and other activities publicizing their candidacies.8
In order for money to be effective, it must be spent on those campaign activities that
make the electorate aware of the candidate. A recent study of state supreme court
elections in eight states9 from 1989-2000 found that only 16.6% of the over $55 million
dollars spent was not on activities associated with advertising a candidacy (Goldberg,
Holman, and Sanchez 2002, 10). Further, in 2000, fully 14% of the money raised by
candidates for the state high court bench was used to purchase television airtime
(Goldberg, Holman, and Sanchez 2002, 14). It seems safe to conclude that candidates
spend most of their campaign funds promoting their election.

In addition to the differential effects of spending based on institutional
arrangements, characteristics of the candidates also should affect the efficacy of
campaign expenditures. There is no reason to expect that campaign spending should

benefit incumbents, incumbents who are facing voters for the first time, and non-

 

8 Other activities candidates spend money on include gifts, contributions to other campaigns, administrative
costs, and so forth (Goldberg, Holman, and Sanchez 2002).

9 The states included in this figure are Alabama, Illinois, Louisiana, Michigan, Montana, North Dakota,
Pennsylvania, and Wisconsin.

32

incumbents equally. A predominant finding of the congressional elections literature is
that incumbent spending is unrelated to their percentage of the vote, or related in a
negative direction, such that the more incumbents spend, the lower their percentage of the
vote (Jacobson 1980, 1990; Ansolabehere and Gerber 1994; but see Green and Krasno
1988, 1990; Thomas 1989; Banaian and Luksetich 1991; Levitt 1994). There are three
possible reasons for this.

First, small shifts in the vote may actually mean the difference between victory
and defeat for the incumbent (Green and Krasno 1988). This means that incumbents
raise and spend more money the more threatened they are, but since challengers spend
more effectively (since they are relatively unknown), it is more difficult for incumbent
expenditures to be translated into votes compared to challenger expenditures.

Second, incumbents occasionally overestimate the strength of their opponent,
because of uncertainty. Sometimes incumbents will wage full scale campaigns even
though their opponent is quite weak. A couple of state supreme court races illustrate this
nicely. In 1990 in Kentucky, Charles Leibson outspent Henry Triplett $135,041 to
$1,500. Leibson won with 72.3% of the vote. In 2000 in Nevada, Nancy Becker spent
$180,729 to Gary Backus’ $930. Becker received 60.8% of the vote. It could be argued
that Leibson and Becker overestimated the strength of their opponents.

Third, and finally, methodological issues such as endogeneity could be obscuring
the true value of incumbent spending, an issue I tackle in Chapter 3.

While a large majority of congressional incumbents have already won election,
this is not true for incumbent state supreme court justices. Some incumbent state

supreme court justices are facing the voters for the first time. This occurs because some

33

justices resign their office before the end of their term, and a replacement is appointed to
fulfill the remainder of the term of office. Hence, there are candidates who run as
incumbents despite the fact that they have never faced the electorate. Indeed, from 1990-
2000, almost 23% of the justices up for reelection initially attained their seats by
appointment. Since these incumbents have never faced the voters before, they may not
have all of the advantages of incumbents who have previously won elections. If one of
the advantages of incumbency is that the incumbent has previously won an election, and
thus been endorsed and is known by the electorate (Jacobson 1997), then incumbents who
have not yet been elected should experience more electoral competition than their
previously elected counterparts as well as spend more money. Further, campaign
spending will be more effective for appointed incumbents than it will for previously
elected incumbents, since appointed incumbents resemble non-incumbents in the sense
that they have not yet been elected.

In contrast to incumbent spending, non-incumbent spending has been found to be
highly significant and negatively related to how well the incumbent does. That is, the
more nonincumbents spend, the better they do (Jacobson 1978, 1980, 1985, 1990;
Abramowitz 1988, 1989; Green and Krasno 1988; Sorauf 1988, 1992; Squire 1989b;
Banaian and Luksetich 1991; Gierzynski and Breaux 1991, 1996; Ansolabehere and
Gerber 1994; Thielemann and Wilhite 1995 ; but see Erikson and Palfrey 2000). This is
generally because more non-incumbents begin the race in relative obscurity, and as they
spend more money, they become better known. This leads to higher levels of campaign
spending as well as a higher percentage of the vote. That is,

The more information voters have about a candidate, other things being
equal, the more likely they are to vote for him. Not that familiarity

34

invariably breeds approbation; awareness does not always carry a positive
valence. But it is generally much better for a candidate to be known than
to be unknown by voters (Jacobson 1980, 36-37)
Since judicial elections occur in a more anonymous setting than congressional elections,
non-incumbent candidates for state supreme courts should benefit even more from

increased spending. This holds true for both challengers to incumbents and for

candidates for open seats.

Conclusion

Judicial elections are phenomena that remain largely unexplored. Indeed, until
recently, much of what is known came from four states (but see Dubois 1980; Bonneau
2001b; Bonneau and Hall 2000, 2001; Hall 2001a, 2001b): California (Schotland 1985;
Dubois 1986a, 1986b), Illinois (Nicholson and Nicholson 1994; Aspin 1998), North
Carolina (Moog 1992; Arrington 1996; Reid 1996, 1999), and Texas (Schotland 1985;
Jackson and Riddlesperger 1991; Champagne 1992; Thielemann 1993; Champagne and
Cheek 1996; Hansen 1998; Check and Champagne 2000). While this work has been
extremely informative, the findings are limited to those states. Theories of elections and
campaign financing have not been systematically tested across all states that elect judges
to their high court bench.

This chapter highlighted both the institutional differences that exist among the
states as well as candidate specific factors that should affect both the amounts and
efficacy of campaign spending. It is important to understand the dynamics of state
supreme court elections because the stakes are so high. Calls for election reform and

eradication are becoming commonplace, calls that are based on untested claims made by

35

opponents of an elected judiciary. Put plainly, we have no idea what forces (specifically
campaign spending) promote or inhibit electoral competition in state supreme court
elections. Further, we have no idea what explains the amount of campaign spending that
occurs in state supreme court elections. The following chapters will explore the effects
of institutional arrangements and candidate-specific forces on electoral competition and

campaign spending in state supreme court elections.

36

CHAPTER 3

RESEARCH DESIGN, DATA, AND METHOD

This dissertation examines all elections to state high courts over a ten-year period
and addresses a fundamental question about electoral competition: Does money affect
the electoral performance of state supreme court candidates, and do these effects vary
across alternative institutional arrangements? Unlike previous studies, my central focus
is on the role of campaign spending. Specifically, I examine the effects of campaign

spending on electoral competition.

The Importance of Comparative Research

As discussed quite extensively in Chapter 2, elections for the state high court
bench vary on several institutional dimensions. State supreme court races vary in terms
of such characteristics as type of ballot, form of electoral constituency, and term of office
to name a few. This variation offers a unique opportunity to examine the effects of
institutional variation on campaign spending.

While institutional differences can complicate matters both theoretically and
empirically, it is essential to test theories in a variety of institutional settings. As
Przeworski and Teune (1970, 86) state: “All events are unique. But this does not imply
that their explanation cannot be based on general theories. Unless uniqueness is seen in a
highly literal sense in which every property of an event is in a class by itself, even unique
events do not defy theoretical explanation.” This dissertation tests existing knowledge of

electoral competition and campaign spending in new environments. If there is no

37

variation in the amount of competition and the effects of campaign spending across
institutional settings in fully specified models, then one can conclude that these
institutional differences have no effect on either electoral competition or campaign
spending. Specifically, we can then conclude that the effects of campaign spending on
competition cannot be influenced simply by changing institutional arrangements. If,
however, significant differences are found between different institutional structures,
ceteris parabis, then the effects of campaign spending on competition can be altered by
changing institutional arrangements. The implication of this is that our current
understandings of campaign finance would need to be rethought and take into account the
role of institutional arrangements in structuring campaign spending.

In addition to the difference in institutional structures among elections to the state
high court bench, there are also institutional differences between congressional elections
and state supreme court elections. These differences may limit the ability of
congressional research to speak to state supreme court elections. First, state supreme
court elections are generally held statewide, while congressional elections are conducted
in districts. Based on the discussion in Chapter 2, this may affect both electoral
competition and the influence of money. Second, while the public is generally ignorant
of congressional candidates, this ignorance is magnified in state supreme court elections
(Klots 1955; Johnson, Schaefer, and McKnight 1978; McKnight, Schaefer, and Johnson
1978; Dubois 1979, 1984; Griffin and Horan 1979; Schotland 1985; Aspin and Hall
1987, 1989; Champagne and Thielemann 1991; Jackson and Riddlesperger 1991; Moog
1992; Arrington 1996; Reid 1996; Klein and Baum 2001). Third, and finally, while the

term of office for congressional offices is two years, it is minimally six years for state

38

supreme courts. Thus, seats on the state high court bench may be valued more highly by
those who seek them than seats in Congress are valued by those candidates.

When conducting comparative research, it important that there are enough
similarities to make the comparisons meaningful. All of the variation that exists among
the states would be meaningless if they made the units of analysis (state supreme court
elections) so different from each other that the comparisons would be nonsensical. That
is, when conducting comparative analysis, it is important to make sure that the events are
comparable. Fortunately, while there are differences between the states that make their
examination desirable for the purposes of testing theories of campaign spending, there are
also similarities that make this research design appropriate. Here, I examine contested
elections to state courts of last resort. Thus, each candidate analyzed in this paper seeks a
seat on the state high court bench. Successful candidates will attain a seat on the highest
legal authority in their state, and authoritatively decide cases on a variety of legal issues.
Most of these elections are competitive. That is, there are at least two candidates for each
seat, and each candidate has a non-trivial chance of winning. Additionally, candidates in
all of these races have been limited in terms of what they can, and cannot, say (i.e., they
are compelled by the canons of judicial ethics to essentially run “issueless” campaigns)
(Schotland 2001).1 Thus, these elections are similar enough across states that

comparisons between them are not meaningless.

 

‘ Again, this is likely to change after the recent United States Supreme Court decision in Republican Party
v. White (see Chapter 2, footnote 5).

39

Research Design

Pooled Cross-Sectional Time Series Design. My research design has observations
both across space and over time. Cross-sectional time-series data have repeated
observations on fixed units—in this case, the fixed units are the states (Beck and Katz
1995). This pooled cross-sectional time series design has several advantages. As
Stimson (1985, 916) wrote:

Pooling data gathered across both units and time points can be an

extraordinarily robust research design, allowing for the study of causal

dynamics across multiple cases, where the potential cause may even

appear at different times in different cases. Many of the possible threats to

valid inference are specific to either cross-sectional or time-serial design,

and many of them can be jointly controlled by incorporating both space

and time into the analysis.

Cross-sectional studies involve observations over many units at a given period of
time. Examples of cross-sectional studies include most research in voting behavior as
well as comparative politics. Researchers who utilize cross-sectional designs are
interested in covariation, which is “presumed to be produced by unobserved causal
processes operating at some time before the data were gathered” (Stimson 1985, 917). In
contrast, time series analysts model the causal process in the longitudinal data. Time
series designs are common in economics and in studies that use economic data. In these
designs, time is theorized to be an important factor in the causal process of the
phenomenon of interest.

Instead of making the choice between cross-sectional analysis and time-series
analysis, I combine both methods of analyses and perform pooled cross-sectional time-

series analysis. This research design capitalizes on the strengths on both forms of

analysis. Specifically, “The main advantage to combining cross-sections and time series

40

in this manner is to capture variation across different units in space, as well as variation
that emerges over time” (Sayrs 1989, 7). While performing pooled cross-sectional time
series analysis raises its own methodological issues (which will be addressed as
necessary), this design is most appropriate to address the questions posed by this
dissertation.

Time Period. I look at all contested state supreme court elections from 1990—
2000. This time period was selected for several reasons. First, these are the most recent
elections. Hence, the results of this study cannot be criticized for being out-of-date, or
not taking into account the changing nature of judicial elections. Second, the nature of
judicial elections has changed over this period. These races have experienced a rise in
the incidence of contestation over the ten years of this study. In 1990, 60% of state high
court races contained more than one candidate, while over 80% of these races in 1996,
1998, and 2000 were challenged. Further, while the average margin of victory has
ranged from 58% (1990 and 2000) to 51% (1994), the total spending in these races (in
actual dollars) has increased from a low of $364,298 (1990) to a high of $859,518 (1998).
This rise in cost has coincided with increased activity by interest groups, who sometimes
spend millions of dollars in independent expenditures in support of (or opposition to)
candidates (Goldberg, Holman, and Sanchez 2002). The 1990’s also represent a high
point in the criticism of judicial elections, especially as far as the influence of money is
concerned (Schotland 2001). Thus, in general, judicial elections were more likely to be
less expensive, less contested, and less controversial events at the beginning of the

decade than at the end.

41

Sample. I examine all contested elections in states that elect justices on either a
partisan or nonpartisan ballot. The reason I limit my sample to contested races is that I
am interested in the effects of campaign spending on electoral competition. For
uncontested seats, the victor always receives 100% of the vote, regardless of how much
money she spends. Further, while candidates who run unopposed may still raise and
spend money, the fact that they are not being challenged means that they will raise and
spend less (and devote less effort to raising) money than if they were challenged. The
1992 race in Montana is a good example of this. Justice Bill Hunt ran uncontested and
spent $12,057. Fellow incumbents Jean Tumage and Karla Gray, running in contested
elections, spent $209,304 and $56,658 respectively.2 This suggests that the presence of a
challenger affects the amount of money spent by candidates. For these reasons,
uncontested races are omitted.

Justices who stand for retention are also excluded from this study. While one
could argue that justices up for retention are more similar to justices in contested partisan
and nonpartisan races than they are to justices running unopposed in these elections
(since justices in retention races can lose their bid to retain their seat by not gaining a
majority of “yes” votes, while candidates running unopposed have no chance of losing),
they are not studied here for a couple of reasons. First, of the sixteen states in which
justices stand for retention, eight states (Alaska, Arizona, Colorado, Iowa, Nebraska,
South Dakota, Utah, and Wyoming) either do not require campaign finance reports or
prohibit campaigning by justices seeking retention altogether. Additionally, in another

six states (California, Florida, Kansas, Maryland, Missouri, and Oklahoma), while

 

2 Tumage spent significantly more than Gray because he was challenged by a sitting justice for the chief
justice position.

42

candidates are not prohibited from campaigning and spending money, candidates in these
elections filed reports indicating no campaign activity. Thus, there is very little
information available on retention races.

Further, while retention races are similar to contested partisan and nonpartisan
elections in that an incumbent has the potential to lose the election, they are different
from contested elections in other important respects. First, the average vote for the
winning candidate in retention elections is significantly higher than the average vote for
the winning candidate in both partisan and nonpartisan elections. Aspin et al. (2000)
report that the mean affirmative vote for retention candidates from 1964-1994 was
74.9%. Further, for the period under examination here (1990-2000), the mean affirmative
vote was 69.8%. This is quite high compared to the percentage of the vote received by
the winning candidate in both contested partisan (56.7%) and nonpartisan races (55.6%)
during this time. Second, and relatedly, very few incumbents lose in retention races.
While 27 of 110 (24.6%) incumbents were defeated in contested partisan races and 7 of
105 (6.7%) incumbents were defeated in contested nonpartisan races during this time,
only 3 of 177 (1.7%) supreme court justices were defeated in their bids for retention from
1990-2000. Taken together, it is clear that retention elections are significantly different
from contested partisan and nonpartisan elections, and hence they are excluded from this
study.

There are twenty-two states3 that elect their justices on partisan and nonpartisan
ballots, and I examine all of them here, with two exceptions. North Dakota, while

electing justices on a nonpartisan ballot, does not require candidates to file campaign

 

3 Texas has two state supreme courts, one that deals exclusively with civil claims (Texas Supreme Court)
and one that only hears criminal cases (Texas Court of Criminal Appeals). Both courts are included here.

43

expenditure reports.4 New Mexico is omitted because of data availability issues. In New
Mexico, campaign finance records are destroyed five years after the election occurs. This
eliminates all records, except for the 1998 and 2000 elections. When I attempted to get
the records for the 1998 election, I was told they could not be located. After successive
attempts to obtain this data were unsuccessful, I decided to omit New Mexico from the
analysis. Fortunately, New Mexico only had one contested election in 1998 and none in

2000. Consequently, its omission should not cause any problems for this analysis.

Data

In order to specify models of state supreme court elections properly, I collected
data on both the characteristics of the elections and of the candidates. I extensively use
Hall’s (2000) dataset on state supreme court elections from 1980-1995.5 Using a variety
of sources, Hall compiled a database of state supreme court elections from 1980-1995,
which consists of all state supreme court races from this time period. This database also
contains a variety of state and electoral data, and is the only comprehensive dataset on
state supreme court elections.

I supplemented the Hall dataset in two ways. First, I assisted in the updating of
this data through the year 2000, so the dataset now contains all elections to the state high
court bench over a twenty-year period. Second, I collected data on campaign
contributions and expenditures for each contested election in the dataset from 1990-2000.
These data were obtained by contacting the appropriate office in each state (usually

located in the Secretary of State’s office) and purchasing copies of the actual campaign

 

4 Interestingly, North Dakota does require candidates to file reports detailing their list of contributors and
the amount of each contribution.
5 See Hall (20013) for a discussion of this dataset.

44

finance reports. It took over four months for the data to be collected, due to difficulties in
tracking it down as well as the labor-intensive nature of photocopying and mailing all of
the appropriate records. Fortunately, for elections in most states since 1996, the
campaign finance data are available on the intemet, and this facilitated the data collection
immensely.

The campaign finance reports were coded for total amounts spent by the
candidates. While each state has different reporting requirements and different filing
periods, they all require candidates to file a final report of all expenditures. Since I am
primarily interested in the total spending for each candidate in these elections, the fact
that the states have different filing requirements and periods dOes not impede my ability
to analyze the data in a systematic fashion.

Just as with federal campaign finance reports, candidates for state supreme court
must report the total amounts that they (and their campaign committees) have raised and
spent. Omitted from these reports are independent campaign expenditures by people and
interest groups. As in federal elections, these expenditures may be playing an important
role in judicial elections, especially since the amount of independent expenditures in
elections for the state high court is increasing with every successive election cycle
(Goldberg, Holman, and Sanchez 2002). Also like federal elections, we have little idea
how much is being spent, by whom, and for whom, because there are no disclosure
requirements for such expenditures. This is potentially a serious problem. While there is
no way in which independent expenditures can be included in the models (because of the
lack of disclosure), there is an indirect way its influence can attempt to be taken into

account. Media and other reports (i.e., Goldberg, Holman and Sanchez 2002) can be

45

searched systematically to identify those races with a significant amount of independent
expenditures. These races can then be excluded, and the models can be estimated with
these races omitted. If my results are robust despite the exclusion of races with a lot of
independent expenditures, then this is evidence, albeit indirect, that omitting independent

expenditures does not bias my results.

Method

Coefi‘icients. When attempting to assess the influence of campaign spending on
the vote, the dependent variable is the percentage of the vote received by the incumbent
(incumbent-challenger contests) or the winner (open seat races). The reason why
incumbent-challenger contests are analyzed separately from Open seat races is that, as
discussed above, there is good reason to expect that money affects these races differently
(Sorauf 1988, 1992). This, coupled with the other differences between incumbent-
challenger races and open seat contests (such as amounts of campaign spending and level
of electoral competition), suggest that separate analysis is appropriate.

While the dependent variable is continuous (vote total), models attempting to
assess the affects of campaign spending on votes are susceptible to a potential
simultaneity problem that can render the ordinary least squares (OLS) results both biased
and inconsistent. The argument is as follows: Candidates (challengers in incumbent-
challenger contests and any candidate for open seats) thought to have a legitimate shot at
winning are likely to attract more contributions than those thought to have little chance at
victory. Thus, the better the candidate’s chances, the more money she will receive (and

consequently be able to spend). The relationship between challenger spending and their

46

percentage of the vote is potentially reciprocal: money may help win votes, but the
expectation that a candidate will receive votes also helps to bring in money. To the
extent that this is true (that expected votes influence spending), then using OLS
regression will underestimate the effects of spending on votes, since the dependent
variable is the percentage of the vote received by the incumbent. The same holds true for
incumbents, but in the opposite direction: since the higher their expected vote, the less
money incumbents spend, if expected votes influence spending, then the OLS estimates
will overestimate the effects of incumbent spending on votes. Jacobson (1980, 1985) and
others (Welch 1981; Green and Krasno 1988; Gerber 1998) attempt to account for this
problem by estimating two-stage least squares (28LS) models. In order to do this, one
needs to identify measurable variables that will affect contributions (and thus
expenditures) without independently affecting the vote. This is no simple task. Further,
Jacobson (1985, 39) suspects that “no simultaneous equation model of this type can be
suitably identified.” Indeed, “After 20 years of research, the appropriate solution remains
elusive” (Jacobson 1997, 38). At the state legislative level, Giles and Pritchard (1985)
were unable to identify a satisfactory instrumental variable to test for a reciprocal effect,
and several congressional studies continue to use OLS despite its potential problems
since 2SLS is no less problematic (Squire 1989b; Jacobson 1996).

The use of 2SLS is problematic for theoretical reasons as well. In order for there
truly to be a reciprocal relationship, the vote at time t must affect the money raised at
time t— I . However, this is nonsensical. Rather, as Gierzynski and Breaux (1991, 209)
point out, “the expectation of how a candidate will perform affects the amount of money

that candidates can raise and subsequently spend.” Thus, the relationship is not

47

Expenditures“ 4:) Votes,
but
Expectations” => Expenditures“ => Votes,
Gierzynski and Breaux (1991) take into account the expectations of the closeness of the
race in order to correct for potential endogeneity problems. They argue that expectations
of the closeness of the race before the campaign begins are a way to take into account the
expected vote. Using state legislative races, they use the previous electoral margin of the
incumbent to take into account expectations of competitiveness. That measure, however,
is inappropriate for state supreme court elections for a couple of reasons. First, the
minimum term of office for state supreme court seats is six years, unlike the two-year
term of state legislators. While it is eminently sensible to argue that an election in 1992
is affected by the election in 1990, it is difficult to conceive of how an election in 1996 is
affected by an election in 1990, or even earlier. Second, unlike state legislative
incumbents, a large proportion of incumbents are facing the electorate for the first time,
and thus have no previous electoral margin. Using a measure of prior electoral
performance would mean excluding these incumbents, who make up a non-trivial
proportion of the bench. Third, and relatedly, using previous vote total also necessitates
the exclusion of all races for open seats.
In order to take into account expectations of the closeness of the race in state
supreme court elections, I use a measure of competitiveness of past state supreme court
elections. Specifically, I code for whether there was a close election (decided by 55% of

the vote or less) in the last election cycle for the state high court. While far from perfect,

48

this gives an indication Of how competitive state supreme court races have been, and
signals to contributors whether a challenger has a potential shot at winning.

Another potential measure of the expectations of the closeness of the race is if
there was an incumbent defeat in the most recent state supreme court election. The
models in Chapters 5 and 6 were also estimated using this measure (coded 1 if an
incumbent was defeated in the most recent state supreme court election), and there were
no changes in the substantive results. I prefer (and thus use) the prior close race measure
as opposed to the recent incumbent defeat measure because I am most interested in
electoral competition. In the most recent state high court race, an incumbent could have
won with 51% of the vote. While this incumbent was not defeated, the fact that she
narrowly won is an indicator of the competitiveness of the race, and may thus affect
future state supreme court elections

Endogeneity is a statistical problem as well as a theoretical problem. Specifically,
we can perform a Hausman test to check if a variable is truly endogenous (Gujarati 1995;
Greene 2000; Wooldridge 2002). If the results of this test indicate that there is no gain in
model efficiency by performing 2SLS as opposed to OLS, then there is no reason to use
2SLS since OLS performs just as well. Before settling on an estimation technique, the
models will be estimated by both ZSLS and OLS, and Hausman tests will be performed.

Standard Errors. Still another potential problem involves the OLS standard
errors. In order to satisfy the assumptions of OLS, the error processes must be assumed
to have the same variance and all of the error terms must be independent of each other.
In cross-section time-series models, this is a heroic assumption. While this assumption

does not affect the coefficients, it does affect the standard errors. This is important

49

because incorrect standard errors “will lead us to be either too confident or insufficiently
confident about whether our findings might merely be statistical artifacts” (Beck and
Katz 1995, 636). Because my dataset includes multiple observations from the same state
both over time and in a given year, in the strictest sense, observations within states might
not be truly independent. To ensure that I am obtaining correct standard errors, I employ
Huber/White/Sandwich robust variance estimators, set to recognize the panel structure of
the data. These estimators are robust to assumptions about within-group (state)
correlation.

Selection bias. It is also possible that the models estimated here are subject to a
selection bias since I am only analyzing contested races. By only looking at contested
races, if the process by which races are contested is ignored, the coefficients of the
models I estimate can be biased and the standard errors inefficient (Achen 1986; Breen
1996; Langer 2002). This is because there are factors that determine whether a race is
contested at all that are not being modeled. In order to fully understand the effects of
money on electoral competition, all races need to be analyzed, not just those that are
contested. That is, I need to specify the conditions under which an election is contested
before estimating the models on the determinants of campaign spending and the effects of
campaign spending on the vote.

Fortunately, James Heckman (1979) has developed a two-stage process that takes
into account “the nonrandom nature of the observed phenomenon in stage one and
removes the resulting biases in stage two” (Langer 2002, 87). In the context here, this
process takes into account the nonrandom nature of being challenged for election and

thus allows for the unbiased estimation of the effects of campaign spending in state

50

supreme court elections. Thus, I estimate the models using Heckman’s two-stage

correction for selection bias.

Conclusion

This dissertation looks at all states that both have contested judicial elections and
require candidates to file campaign finance reports. Looking at twenty states over a ten-
year time period allows me to assess the determinants of electoral competition as well as
the impact of campaign spending on electoral competition both across states and over
time. I employ this research design in order to best test existing knowledge of campaign
finance in state supreme court elections.

The following chapters empirically examine the determinants of campaign
spending as well as effects of campaign spending on electoral competition in state
supreme court elections. Specifically, I ask:

I. What are the determinants of campaign spending?

2. What factors promote (or discourage) electoral competition?

3. Do the effects of money vary depending on institutional arrangements?

Answering these questions will give us a good start into understanding the effects

of campaign spending on state supreme court elections.

51

CHAPTER 4

DESCRIBING STATE SUPREME COURT ELECTIONS

State supreme court elections possess extraordinary variation both across states
and over time, in terms of both electoral competition and campaign spending. As the
following tables demonstrate, electoral competition and campaign spending vary based
on year, candidate, and type of election. The extent to which this variation can be

understood systematically is the subject of subsequent chapters.

State Supreme Court Elections Over Time

Not surprisingly, state supreme court elections have, on average, become more

expensive over the period of this study, as Table 4.1 shows.

52

Table 4.1: Total Average Spending by Year, Contested General Elections 1990-2000
(number of elections in parentheses)

 

 

Year All Partisan Nonpartisan
Elections Elections Elections

1990 $364,348 (30) $404,937 (18) $303,464 (12)
1992 $576,268 (34) $663,063 (18) $512,373 (16)
1994 $748,398 (31) $1,108,665 (16) $364,113 (15)
1996 $579,336 (34) $722,388 (16) $452,177 (18)
1998 $860,990 (31) $1,147,952 (16) $554,898 (15)
2000 $801,108 (42) $846,465 (19) $763,639 (23)

 

53

Looking at all contested elections in this period, spending has increased, on
average, from 1990-2000. In general, each successive election has been more expensive
than the previous election. This holds true for both partisan and nonpartisan elections.

One can also see from Table 4.1 that partisan elections, in the aggregate, are
always more expensive than nonpartisan elections. Indeed, it was only in 2000 that
nonpartisan elections reached the spending levels achieved in partisan elections in 1992.
However, while it is too early to tell if this is a trend or just an anomaly, the gap between
spending in partisan elections and nonpartisan elections in 2000 was the smallest it has
ever been.

While Table 4.1 demonstrates that state supreme court elections are becoming
more expensive, these spending figures do not factor in inflation. It could be the case that
the rise in costs is due in large part to inflation. In Table 4.2, I adjust the spending figures

in Table 4.1, and report the spending amounts in 1990 dollars.

54

Table 4.2: Total Average Spending by Year in 1990 dollars, Contested General Elections
1990-2000 (number of elections in parentheses)

 

 

Year All Partisan Nonpartisan
Elections Elections Elections

1990 $364,348 (30) $404,937 (18) $303,464 (12)
1992 $525,809 (34) $571,631 (18) $474,259 (16)
1994 $645,161 (31) $958,840 (16) $310,570 (15)
1996 $472,792 (34) $589,570 (16) $368,989 (18)
1998 $672,148 (31) $896,272 (16) $433,081 (15)
2000 $599,008 (42) $632,022 (19) $571,736 (23)

 

55

As Table 4.2 demonstrates, while the campaign spending increases are not as
dramatic when measured in 1990 dollars, they are still becoming more expensive over the
decade of the 1990’s. Candidates for the state high court bench are still spending more
money in 2000 than they were in 1990 even when spending is measured in constant
dollars. A couple of examples illustrate this pattern nicely.

Justice John Comyn won his partisan election to the Texas Supreme Court over
Democrat Gene Kelly in 1990, spending $374,659. In his bid for reelection, he spent
$864,160 ($705,245 in 1990 dollars) to defeat Democrat Patrice Barron. Justice Deborah
Cook defeated J. Ross Haffey in a 1994 nonpartisan election for the state high court
bench in Ohio, spending $431,978 ($370,748 in 1990 dollars) in the process. She was
reelected in 2000, defeating Timothy Black, and spent $640,817 ($478,473 in 1990
dollars) in her bid to retain her seat.

Interestingly, Table 4.2 indicates that the most expensive races, on average, were
in 1994 and 1998 (especially in partisan states). One potential reason for this may be the
fact that these are mid—term elections. The absence of a presidential party candidate may
free up more money for contributors to donate to state high court candidates. It will be
interesting to see if this pattern continues in 2002.

Table 4.3 looks at campaign expenditures by type of candidate over time.

56

Table 4.3: Average Spending for Each Type of Candidate by Year, Contested General
Elections 1990-2000

 

 

Year Incumbent Challenger Winner of Loser of
Open Seat Open Seat

1990 $276,787 $79,073 $294,688 $91,372

1992 $288,903 $239,940 $321,455 $376,686
1994 $474,653 $182,013 $504,750 $382,348
1996 $352,823 $245,619 $263,126 $207,116
1998 $544,891 $253,233 $496,098 $496,911
2000 $476,009 $378,046 $384,716 $321,087

 

57

As can be seen from Table 4.3, spending by all types of candidates is higher in
2000 than it was in 1990. This holds even if we look at spending in 1990 dollars (not
shown). Given what we saw in Table 4.1, it is not surprising that candidates are spending
more money at the end of the decade as opposed to the beginning (since elections are
more expensive), but this leads to the question of the effects this spending has on
competition. Are candidates spending more money in their electoral bids because it
increases their percentage of the vote? Or is the spending by state supreme court
candidates fruitless, yielding no gain in electoral support? Or does the efficacy of
spending vary depending on the type of candidate? These questions will be answered
over the next two chapters.

This dissertation focuses on the effects of campaign spending on competition.
Thus, it is natural to ask: Have races for the state high court bench become more
competitive over the period of this study, or simply more expensive? The answer to this

question is provided in Table 4.4.

58

Table 4.4: Average Percentage of the Vote by Year, Contested General Elections 1990-
2000

 

 

Year All Partisan Nonpartisan
Elections Elections Elections
1990 56.64 55.73 58.00
1992 54.83 55.25 54.35
1994 53.11 52.66 53.58
1996 56.08 55.77 56.39
1998 57.87 53.98 62.02
2000 57.93 60.69 55.65

 

59

While state supreme court races were becoming more competitive from 1990 to
1994, this trend was reversed beginning in 1996, and now the pattern is toward less
competitiveness. Indeed, state supreme court elections, on average, were more
competitive in 1990 than in 2000. Looking at partisan and nonpartisan elections, no
patterns are evident. Partisan elections are, in general, more competitive than nonpartisan
elections, but the average percentage of the vote received by the winning candidate was
over 60% in the year 2000 for contested elections that occurred in partisan states.
Likewise, after having relatively competitive elections from 1990-1996, in 1998
contested nonpartisan elections were won by an average of over 62%. Looking at the
examples discussed earlier reinforces these aggregate findings.

Recall that in the two races discussed above, the incumbent candidate spent more
in her second election than in her first. Does this increased spending yield more or less
competition? In Texas, Supreme Court Justice John Comyn, despite spending more
money in 1996 than in 1990, saw his percentage of the vote drop from 55.67% to
51.99%. In 1994, Ohio Supreme Court Justice Deborah Cook won with 68.05% of the
vote, but she only received 51.88% of the vote in 2000. In her case, higher levels of
spending also led to a lower level of electoral support. While these are only two cases,
these examples add plausibility to the claim that campaign spending may actually
promote competition (or at least does not inhibit it).

In Table 4.5, the average percentage of the vote received by incumbents and

winners of open seats are displayed.

60

Table 4.5: Average Percentage of the Vote by Type of Election and Year, Contested
General Elections 1990-2000

 

 

Year Incumbent-Challenger Open Seats
1990 56.30 57.23
1992 55.35 53.73
1994 54.32 51.42
1996 54.22 60.02
1998 58.98 55.16
2000 57.99 57.81

 

 

' Average vote for incumbent-challenger contests is the average percentage of the vote for the incumbent;
for open seats, it is the average percentage of the vote for the winner.

61

There is no discernible pattern toward more competitive elections in terms of type
of election from 1990-2000, confirming what we saw in Table 4.4. While contested
elections in 2000 were more expensive than elections in 1990, contested elections in 1990
were more competitive, both for incumbent-challenger contests and races for open seats.
Additionally, while 1994 was the most competitive year overall (although 1996 was
slightly more competitive for partisan races), Table 4.3 shows that the 1994 election was
characterized by large spending discrepancies between incumbents and challengers and
the winner and losers of open seats. Thus, while state supreme court races are becoming

more expensive, they are not also necessarily becoming more competitive.

Partisan versus Nonpartisan Elections

In addition to seeing how state supreme court elections vary over time, it is also
interesting to see the variations by selection system. The tables above indicate that
partisan elections, on average, are more expensive than nonpartisan elections. In Table

4.6, I break down this spending by type of candidate.

62

Table 4.6: Average Spending for Each Type of Candidate by Method of Selection,
Contested General Elections 1990-2000

 

 

Selection Incumbent Challenger Winner of Loser of
Method Open Seat Open Seat
Partisan $524,356 $257,372 $452,075 $381,806
N onpartisan $306,702 $228,624 $279,053 $212,894

 

63

Incumbents in contested partisan elections spend the most money, according to
Table 4.6, while losers of open seats in contested nonpartisan elections spend the least.
Incumbents spend the most money, regardless of the type of election. Further, winners
outspend losers in both partisan and nonpartisan elections. Table 4.6 also shows that the
spending patterns of candidates are the same in contested partisan and nonpartisan
elections (incumbents spend more than challengers, winners spend more than losers) even
though there is more spending in partisan elections overall. A few examples make this
point clear.

Alabama Supreme Court Justice Mark Kennedy was challenged by Harold See in
1994. Kennedy outspent See $730,875 to $293,802. In the same election year, Terry
Butts and Robert Russell battled for an open seat on the Alabama high court bench.
Butts, who won the election, outspent Russell $968,097 to $156,395. In this partisan
election state, consistent with the aggregate information seen in Table 4.6, the incumbent
outspent the challenger and the winner of an open seat outspent the loser.

The state of Washington elects its Supreme Court Justices in nonpartisan
elections. In 1998, Justice Barbara Madsen spent $40,687 in her successful bid for
reelection, outspending her opponent, Jim Bates, who only spent $24,806. There was
also a contest for an open seat on the Court that year, which saw Faith Ireland defeat Jim
Foley, outspending him $87,580 to $8,209 in the process. Again, the incumbent spent
more than the challenger, and the winner of an open seat outspent the loser. Also note
that while this pattern was true for both partisan and nonpartisan elections, partisan
elections involved higher levels of spending than the nonpartisan elections, consistent

with Table 4.6.

64

In Table 4.7, we can see the average percentage of the vote for each type of

election and method of selection.

65

Table 4.7: Average Percentage of the Vote by Type of Election and Method of Selection,
Contested General Elections 1990—2000

 

 

Selection Method Incumbent-Challenger Open Seat
Partisan 55.06 56.96
Nonpartisan 57.39 54.27

 

66

The average percentage of the vote in incumbent-challenger contests is
significantly different from the average percentage of the vote in open seat contests for
both partisan and nonpartisan elections. Incumbent-challenger contests in partisan
elections are almost 2% more competitive than open seat races. Curiously, this is
reversed in nonpartisan states, with open seats being over 3% more competitive than
incumbent-challenger contests. Indeed, the least competitive elections are incumbent-
challenger races in nonpartisan states. Thus, not only is campaign spending conditional
on type of election, but electoral competition is as well.

Returning to the specific races discussed above, Justice Kennedy defeated See in
the incumbent-challenger race in Alabama in 1994 with only 50.30% of the vote. In
contrast, in the open seat race, Terry Butts received 53.57% of the vote, defeating Robert
Russell. Consistent with the evidence just discussed, the open seat race was less
competitive than the incumbent-challenger contest. In the nonpartisan races in
Washington in 1998, this was reversed. Faith Ireland won the open seat race over Jim
Foley with 68.13% of the vote, while Justice Barbara Madsen won her reelection bid over

Jim Bates with only 60.79% of the vote.

State-by-State Comparisons

The final way in which the campaign spending and competition data will be
described is by state. In addition to the variations over time, by type of candidate, and by
selection system, the political context of each state in this study affects both spending and
electoral competition. Even though two states may select justices in the same manner

(i.e., partisan elections), there may be more spending or more competition in one state

67

because of factors unique to the state. For example, consider Louisiana and Illinois, two
states that elect their justices by partisan elections in districts. As the following tables
will demonstrate, races in Louisiana are both more expensive and more competitive.
However, there are also more uncontested races in Louisiana. Thus, while the contested
races in Louisiana are more expensive and competitive than those in Illinois, it is also less
likely that a race in Louisiana will be contested. This indicates that electoral factors
unique to each state also affect campaign spending and competitiveness in state supreme
court elections.

In Table 4.8, the average spending for each type of candidate is broken down by

state.

68

Table 4.8: Average Spending for Each Type of Candidate by State, Contested General
Elections 1990-2000

 

 

State Incumbent Challenger Winner of Loser of
Open Seat Open Seat
Partisan
Alabama $602,762 $517,456 $1,010,295 $761,620
Arkansas $60,855 $26,214 $244,295 $250,105
Illinois -------------------- $258,276 $248,638
Louisiana $924,472 $650,901 $397,608 $362,205
N. Carolina $202,342 $107,124 $250,924 $56,849
Pennsylvania -------------------- $991,216 $1,121,768
Texas (Civil) $1,008,019 $318,614 $813,726 $483,023
Texas (Crim.) $102,064 $40,523 $67,651 $66,761
W. Virginia $711,193 $180,138 $400,038 $235,295
Nonpartisan
Georgia $128,591 $77,334 --------------------
Idaho $129,825 $88,423 $65,057 $99,134
Kentucky $155,457 $392,516 $122,407 $232,758
Michigan $619,722 $442,439 $578,928 $250,025
Minnesota $108,489 $4,381 $57,061 $8,954
Mississippi $267,910 $271,565 $191,461 $71,465
Montana $138,406 $111,384 $291,918 $178,240
Nevada $373,261 $253,787 $427,562 $134,194
Ohio $562,642 $329,791 $391,553 $451,198
Oregon $110,325 $114,194 $159,938 $107,378
Washington $80,329 $46,715 $129,489 $113,689
Wisconsin $367,316 $326,159 $405,349 $461,053

 

69

Incumbents spend more money than challengers, except in Kentucky, Mississippi
and Oregon. However, if only the second-place vote-getter, as opposed to all
challengers, is considered, then challengers in Mississippi spend $231,231 and
challengers in Kentucky spend $105,197, both lower than the amount spent by
incumbents. (Challengers continue to spend more than incumbents in Oregon.)

Table 4.8 also reveals that winners generally spend more than losers, although
there are notable exceptions (Arkansas, Idaho, Kentucky, Ohio, Pennsylvania, and
Wisconsin). If only the second-place vote-getter—the primary loser—is considered, the
results remain the same, except for in Wisconsin, where the primary loser spent, on
average, $148,960, lower than the amount spent by the incumbent.

In terms of highs and lows, incumbents on the Texas Supreme Court spent the
most money in their reelection bids, while incumbents in Arkansas spent the least.
Challengers for the state supreme court in Louisiana spent the most money in their
election attempts, while those in Minnesota spent the least. The Louisiana statistic is
interesting, considering candidates there run in districts and not statewide. Perhaps
challengers spend this much money because the incumbency advantage in stronger in
Louisiana than in other states. Looking at open seats, winners of elections in Alabama
spent the most and winners in Minnesota the least. Finally, in terms of the losing
candidates in open seat elections, those in Pennsylvania spent the most, and, once again,
those in Minnesota spent the least.

Clearly, there is much variation in campaign spending by state. A good example
of this is the incumbent spending variable. Both the highest amounts of spending (Texas)

and the lowest (Arkansas) come from partisan election states. The state in which the

70

election is occurring is an important source of variation, beyond that which can be
accounted for by institutional variables, such as selection system.
Are there also state-by-state variations in terms of electoral competition? The

answer can be found in Table 4.9.

71

Table 4.9: Average Percentage of the Vote by Type of Election and State, Contested
General Elections (number of elections in parentheses) 1990-2000

 

 

State Incumbent-Challeger Open Seats
Partisan

Alabama 57.14 (11) 56.39 (6)
Arkansas 65.18 (3) 55.41 (3)
Illinois ---------- (0) 54.89 (8)
Louisiana 47.18 (3) 51.85 (3)
North Carolina 49.40 (11) 58.72 (1)
Pennsylvania ---------- (0) 52.43 (4)
Texas (Civil) 62.14 (17) 56.29 (3)
Texas (Criminal) 52.73 (12) 59.12 (8)
West Virginia 50.28 (2) 66.91 (4)
Nonpartisan

Georgia 55.66 (3) ---------- (0)
Idaho 49.53 (2) 57.21 (1)
Kentucky 50.19 (5) 47.77 (4)
Michigan 56.34 (11) 47.21 (3)
Minnesota 63.77 (9) 62.32 (1)
Mississippi 50.76 (9) 62.08 (2)
Montana 57.11 (3) 54.20 (3)
Nevada 51.12 (3) 56.62 (3)
Ohio 60.30 (11) 57.77 (3)
Oregon 61.98 (2) 51.38 (1)
Washington 57.27 (10) 55.07 (6)
Wisconsin 63.66 (4) 56.94 (2)

 

72

As Table 4.9 shows, in some states, incumbent-challenger contests are more
competitive, while in other states, this label belongs to races for open seats. Further,
some partisan states have more competitive incumbent-challenger races (such as
Louisiana and North Carolina), while others have more competitive open seat contests
(such as Alabama and Arkansas). It is important to remember that only contested races
are included, so these data do not speak to whether an incumbent (or other candidate) is
likely to be challenged. However, these data do indicate that in competitive elections,
incumbents in some states are likely to have quite a battle on their hands.

Looking at the highs and lows, the most competitive state for incumbent-
challenger contests is Louisiana (which is somewhat surprising since these elections
occur in districts, and not statewide), while Arkansas is the least competitive, with
incumbents being reelected with 65% of the vote, on average. Open seats in Michigan
are the most competitive in that category, while those in West Virginia are not very
competitive at all—the winners receive over 2/3 of the vote. This is interesting because
both Michigan and West Virginia elect their justices in multi-member districts. Just as
with campaign spending, there are interesting variations between states that cannot be
accounted for by institutional factors.

In sum, the data demonstrate that there is extraordinary variation between states
both in campaign spending and electoral competitiveness. Incumbents generally
outspend challengers, on average, and winners usually outspend losers. However, there
are cases where challengers do outspend incumbents, and where losers outspend winners.
Further, in some states incumbent-challenger contests are most likely to be highly

competitive, while in others it is the races for open seats.

73

Conclusion

State supreme court elections vary tremendously over time, by type of election,
and by state. This variation occurs for both campaign spending and competition.
Further, in terms of campaign spending, patterns of spending vary depending on the type
of candidate.

This chapter described the campaign spending of candidates for the state high
court bench, as well as electoral competition. In addition to the variation that has
occurred over time, I found that both spending and competition varied by institutional
arrangements as well as by state. However, what this chapter did not address is what are
the determinants of campaign spending. Can the amount of money spent in a race be

understood systematically? This question is the focus of the next chapter.

74

CHAPTER 5

EXPLAINING CAMPAIGN SPENDING

We have just seen in Chapter 4 the wide variation in campaign spending by type
of election, by year, and by state. What explains this variation? Why do some races
involve more spending than others? In this chapter, I look at the predictors of campaign
spending, and then, in the next two chapters, focus my attention on the effects of

campaign spending on state supreme court election outcomes.

Specifying the Model

Dependent variable. The dependent variable in this analysis is the log of the total
amount of money spent in the race by all candidates (Log of Total Spending).

Independent variables. Several independent variables are hypothesized to affect
the total amount of money spent in elections. They can be grouped into three broad
categories: characteristics of the seat, institutional characteristics of the court, and the
state and electoral context.

Characteristics of the seat: Most fundamentally, whether or not the election is for
an open seat should affect the amount of spending (Open Seat). Sorauf (1988) has found
that, in state legislative races, open seat races are more expensive than incumbent—
challenger contests. In Chapter 4, we saw that in some states open seat races were more
expensive, while in others, incumbent-challenger races involved more spending. Other
things being equal, however, I hypothesize that open seat races will involve more

spending, since both candidates need to spend money in order to become known to the

75

electorate, while this is not the case in incumbent-challenger races (where only the
challenger has this difficulty). Further, it may be easier for candidates for open seats to
raise (and thus spend) money compared to the difficulties a challenger to an incumbent is
likely to encounter (e.g. Jacobson 1980, 1985, 1997; Green and Krasno 1988).

Even in incumbent-challenger contests, there should be a difference between
incumbents who have previously won election and those who are facing the voters for the
first time. These incumbents do not have all of the benefits of incumbency (Jacobson
1997). Incumbents without electoral experience will need to spend more money than
those candidates who have already been legitimized by the electorate. Thus, I expect that
there will be more spending in races where the incumbent is facing the voters for the first
time (Appointed First) than in other races.

Campaign spending should also be related to how close the race is expected to be.
The closer the race is expected to be, the more money should be spent, other things being
equal. Unfortunately, we do not have good measures of expectations of the closeness of
the race. However, the actual margin of victory has been found to be a good surrogate
for the expected margin of victory. That is, the closer the race is expected to be, the
closer it actually is (Jacobson 1980, 1985; Gierzynski and Breaux 1991).1 Thus, the
lower the margin of victory (Margin of Victory), the higher the amount of money that
should be spent.

Partisan control of the court, as for other political offices, is something that is
sought after by political parties. Large amounts of money have been spent trying to

obtain and retain control of the court (Glaberson 2000a; Dickerson 2001a; Omdorff 2002;

 

' Indeed, this is the reason given for estimating these kinds of models using ZSLS (see discussion in
Chapter 3).

76

Phillips 2002). If partisan control of the court is at stake in the elections (Control of
Court), then more money should be spent, other things being equal.

Finally, the amount of money spent on one supreme court race is at least partially
dependent on how many other races are on the ballot. Money is a finite resource, and the
more seats that are on the ballot (Number of Seats), the lower the amount of spending in
any one race should be.

Institutional characteristics: A basic institutional difference among states that
elect their judges is whether the election is partisan or nonpartisan. In Chapter 4, we saw
that nonpartisan races, in general, involved less spending than partisan elections. Thus, I
expect that partisan races (Partisan) will involve more spending than nonpartisan races.

In some states, judges are elected in geographical districts, as opposed to
statewide. By definition, these judges have a more geographically compact electoral
constituency. This should decrease the cost of a campaign, other things being equal
(Gierzynski 1998). Thus, I hypothesize that there will be less spending in elections that
are held in districts (District) rather than statewide.

Given the differences expected between partisan and nonpartisan states, we would
also expect these differences to hold in district elections as well as statewide elections.
Thus, consistent with previous studies that have found different dynamics between
partisan and nonpartisan district-based elections (Hall 2001a), I include an interactive
term taking into account the partisan nature of the district election (DistPart). As with
statewide elections, I expect that there will be more spending in partisan district elections

than in their nonpartisan counterparts.

77

Another institutional difference among the states is whether the election is for a
single-seat or whether candidates run in multi-member districts. Multi-member districts
have been found to involve more campaign spending than their single-member
counterparts. “In multimember districts candidates will likely be less well known and
will be in more competition for voters’ attention than candidates in single-member
districts. This means they need to spend more money” (Gierzynski 1998, 25). I expect
the same to hold true here: multi-member districts (MultiMember) should involve higher
levels of spending than single—member districts.

Finally, the term of office for state supreme court office varies among the states.
Terms of office range from 6 to 12 years. Seats on the bench in states that have longer
terms of office are more valuable than seats in states that have shorter terms, because
longer terms of office provide for more job security. Thus, since seats in states with
longer terms of office should be more sought after, I expect that there will be more
campaign spending when a longer term of office is at stake (Term).

State and Electoral Context: The final set of independent variables involves the
state and electoral context of the race. First, the size of the voting age population should
affect the levels of campaign spending. Simply put, other things being equal, the higher
the voting age population (Voting Age Population), the more money that should be spent,
since there are more people to reach (Squire 1989b; Gierzynski 1998; Partin 2002).

Second, the general competitiveness of state supreme court elections in the state
should affect the amount of money spent in a particular election. If there was recently a
close election for the state high court bench in a state, candidates may begin raising

money earlier and thus have more money to spend for their election. Thus, I expect that

78

if there was a close race (decided by 55% of the vote or less) in the most recent state
supreme court election (Prior Close Race), there will be more spending in the current
race, other things being equal.

Finally, the composition of the docket may affect campaign spending. Some
states deal primarily with criminal or governmental cases, while others have a large
proportion of their docket occupied by tort cases. The battle over tort reform has pitted
trial lawyers against corporations and other business interests. What both sides have in
common is their propensity to spend money to put candidates sympathetic to their point
of view on the bench (e.g., Glaberson 2000a; Omdorff 2002; Phillips 2002). This can be
seen clearly in Table 4.8 looking at the two Texas courts of last resort: The amounts
spent in races for the Texas Supreme Court (which only handles civil cases) dwarf
significantly the amount of money spent for seats in the Texas Court of Criminal Appeals
(which has jurisdiction only over criminal cases). One possible reason for this spending
disparity is the differential nature of the dockets (since salary, term of office, staff, and
other institutional issues are the same). Thus, I expect that in states with larger
proportions of tort cases on their docket, there will be more spending (Tort Docket) than
in states where a proportionately smaller percentage of their docket is occupied by tort
cases.

Finally, I include a dummy variable for the year of the election, minus one year,
to control for any temporal effects (1990, 1992, 1994, 1996, 1998).

For convenience, Table 5.1 lists all of the variables included in the analysis, along

with their exact measurement.

79

Table 5.1: Variable Descriptions for a Model of Campaign Spending in State Supreme

Court Elections

 

Variable

Variable Description

 

Log of Total Spending
Characteristics of the Seat
Open Seat

Appointed First

Margin of Victory

Control of Court

Number of Seats

Institutional Characteristics
Partisan

District
MultiMember

Term

State and Supreme Court Context

Voting Population
Prior Close Race

Tort Docket

Control Variables
1990, 1992, 1994, 1996,
1998

 

log of total campaign spending by all candidates
in the race

1 if the race was for an open seat

0 otherwise

1 if the incumbent was originally appointed to
the court

0 otherwise

margin of victory (%) for the winner

1 if partisan control of the court is at stake

0 otherwise

number of state supreme court races on the ballot

1 if the election was a partisan election

0 otherwise

1 if the election was held in a district

0 otherwise

1 if the election was held in a multimember district
0 otherwise

length (in years) of the term of office

voting age population of the state/district (10003)

1 if recent judicial election was decided by

55% of the vote or less

0 otherwise

proportion of the docket (1995) involving tort cases

1 if election occurred in the designated year
0 otherwise

80

Estimation Technique

Since the dependent variable is continuous, ordinary least squares regression is
appropriate. I use Huber/White/Sandwich robust standard errors, set to recognize the
panel structure of the data.

As mentioned in Chapter 3, there is a potential selection bias since I only observe
the dependent variable (total campaign expenditures) when the race is contested. Thus,
uncontested races are omitted. To the extent that what makes races contested is
nonrandom (a likely scenario, see Bonneau and Hall 2001), the estimates I obtain may be
biased. Thus, I need to take into account the nonrandom process of races being
contested. This can be done by finding some variables that affect the probability of
observation (contestation) but not the dependent variable of interest (total campaign
spending). In other words, I need to construct a model that predicts whether a state
supreme court race will be contested, and thus included in the dataset. The results from
this model can then be used to obtain unbiased estimates of the model of total spending
(Heckman 1979).

Fortunately, this has already been done. Bonneau and Hall (2001) exarrrine the
probability that an incumbent will be challenged in state supreme court races from 1988-
1995. The probability of a race being contested was found to be related to the following
factors: the incumbent being initially appointed, the ranney index, the number of lawyers
in the state, the salary of the justice, the term of office, and whether the race was partisan,
district, or partisan and district. Here, I use these variables found to predict contestation
in the selection equation to determine whether race was contested. Then, I estimate the

model of total campaign spending discussed above, with the selection bias corrected.

81

This procedure provides consistent, asymptotically efficient estimates for all the
parameters in the model. The variables included in the selection model and their exact

measurement are shown in Table 5.2

82

Table 5.2: Variable Descriptions for the Selection Model of Campaign Spending in State

Supreme Court Elections

 

Variable

Variable Description

 

Contested

Independent Variables
Appointed First

Ranney Index

Lawyers
Salary

Institutional Characteristics
Partisan

District
DistPart

Term

1 if the race was contested (and thus included in the
dataset)
0 otherwise

1 if the incumbent was originally appointed to
the court

0 otherwise

Ranney index of state partisan competition, as
calculated and reported by Holbrook and Van
Dunk (1993)

number of lawyers in each state in 1990
supreme court justice salary in 1995

1 if the election was a partisan election

0 otherwise

1 if the election was held in a district

0 otherwise

1 if the election was held in a partisan district
0 otherwise

length (in years) of the term of office

83

The Determinants of Campaign Spending

Table 5.3 displays the results for the model of campaign spending estimated using
Heckman’s two-step correction for selection bias. These results are essentially the same
as the results obtained by OLS regression without the correction for selection bias (as

seen in the Appendix, Table Al).

84

Table 5.3: Predicting Campaign Spending

 

 

 

Variable Coefficient Robust 2 P > |z|
Std. Error
Open Seat 0.118 0.125 0.944 0.345
Appointed -0.050 0.187 -O.267 0.790
First
Margin of -0.022 0.006 -3.653 0.000
Victory
Control of 0.166 0.145 1.146 0.252
Court
Number of —0.265 0.103 -2.579 0.010
Seats
Partisan -0.737 0.367 -2.006 0.045
District -0.240 0.507 -0.474 0.636
DistPart 0.062 0.643 0.097 0.923
MultiMember -0.267 0.350 -0.762 0.446
Term 0.283 0.075 3.748 0.000
Voting Age 0.000 0.000 0.913 0.361
Population
Prior Close 0.217 0.171 1.266 0.205
Race
Tort Docket 0.052 0.006 9.347 0.000
1990 -0.727 0.167 -4.360 0.000
1992 -O.386 0.308 -1.255 0.209
1994 -O.306 0.284 -l.075 0.283
1996 -0.442 0.121 -3.666 0.000
1998 0.036 0.155 0.232 0.816
Constant 11.602 0.998 11.620 0.000
Dependent variable: log of total spending in the election
Mean of dependent variable: 12.825
Number of observations 2 467
Censored observations = 265
Uncensored observations = 202
Log likelihood = 470.35?
x2 (8) 28.40
Prob > x2 = 0.000

85

Contrary to what was hypothesized, the total amount of campaign spending is not
related to whether the election was for an open seat or was an incumbent-challenger
contest. Thus, contrary to the state legislative findings, the type of election is not a
significant predictor of campaign spending in state supreme court elections. Further, the
type of incumbent also does not matter. Combined, these results suggest we need to look
elsewhere to understand the determinants of campaign spending.

As expected, the closeness of the race affects spending. The lower the margin of
victory for the winner, the more money that is spent. A decrease of 1% in the margin of
victory leads to an increase of 0.02% in total spending. Thus, an election that involved
total spending of $100,000 with a 5% margin of victory would involve $108,000 if the
margin of victory was 1%. Somewhat surprisingly, campaign spending is not related to
whether partisan control of the court is at stake. High levels of partisan competition and
spending for court seats are apparently independent of whether control of the court is up
for grabs. However, as expected, the number of seats on the ballot is a significant
predictor of spending. Other things being equal, the fewer the number of seats on the
ballot, the more money that will be spent on a race. One fewer state supreme court
election on the ballot leads to an increase in total spending of 0.27%. Using the example
above, a race with 2 other seats on the ballot that involved total spending of $100,000,
would involve spending of $127,000 if there was only one other seat on the ballot.

In terms of institutional arrangements, two variables are significant. First, less
money is spent on races in partisan statewide races than in the base category, nonpartisan
statewide races. This somewhat surprising given the results presented in Chapter 4

(partisan races were, on average, always more expensive than nonpartisan races), but that

86

could have been due to other factors (such as the competitiveness of partisan races).
Once other relevant variables have been controlled for, partisan statewide races, on
average, are less expensive than their nonpartisan counterparts. Campaign spending is
unrelated to whether the election occurred in a district (either partisan or nonpartisan) or
was held in a multi-member district. Finally, as hypothesized, the longer the term of
office, the more expensive the race. This is likely because the seat is more valuable, and
there are fewer opportunities to attain the high court bench. Increasing the term of office
by two years leads to an increase in spending of 0.56%. Other things being equal, if
$100,000 was spent on a race for a 6—year term, $156,000 would be spent if an 8-year
term of office was at stake, $243,360 for a 10-year term, and $379,642 for a 12-year
term.

In terms of the state electoral and supreme court context, both the size of the
voting age population and whether there was a close race in the most recent state supreme
court elections were insignificant. However, as expected, the percentage of the docket
composed of tort cases was significant. The higher the percentage of the tort cases
decided by the court, the more expensive the race was, other things being equal.
Specifically, a 1% increase in the proportion of tort cases on the docket leads to a 0.05%
increase in total spending. This suggests that candidates are more easily able to raise
(and thus spend) money in states where there is a large percentage of tort cases decided,
and that contributors are aware that it is important for them to have sympathetic members

of the state high court bench ruling on cases that typically involve large sums of money.

87

Conclusion

In this chapter, I sought to explain the determinants of campaign spending in state
supreme court elections. The evidence indicates that there is more campaign spending
when the race is expected to be close, there are fewer seats on the ballot, the race is
nonpartisan, the term of office is relatively long, and a higher percentage of the docket is
occupied by tort cases. The evidence presented indicates that the amounts of campaign
spending in state supreme court elections can be understood systematically, and is a
function of characteristics of the seat, institutional characteristics, and the state and
electoral context.

Now that I have analyzed the predictors of campaign spending, I turn my attention
to the effects of campaign spending on electoral competition. Does campaign spending
affect electoral competition? Do these effects vary based on institutional arrangements?

These questions are the subject of the next two chapters.

88

CHAPT ER 6

THE EFFECTS OF MONEY ON VOTES: N ON-INSTITUTIONAL MODEL

In Chapter 4, we saw that the patterns of competition and campaign spending vary
over time, by type of election, and by state. In Chapter 5, I analyzed the determinants of
campaign spending. However, we have yet to see what difference all of the money spent
in state supreme court elections makes. In this chapter, I specify multivariate models in
order to explain the variation just discussed. Specifically, looking at general elections for
the state high court bench, I address whether money affects electoral outcomes (Chapter
6), and whether the effects of money vary depending on institutional arrangements
(Chapter 7). Because of the different dynamics present in elections for open seats
(Sorauf 1988, 1992), I analyze these contests separate from incumbent-challenger
contests.

While the importance of institutional arrangements in structuring both judicial
behavior (Hall and Brace 1989, 1992; Brace and Hall 1990, 1995, 1997; Hall 1992, 1995)
and judicial elections has been demonstrated convincingly (Hall 2001a), I omit them in
this chapter to demonstrate how omitting institutional variables affects the conclusions
one reaches about the determinants of competition as well as the role of campaign
spending. Just as omitting institutional variables leads to incorrect inferences in models
of state supreme court decision-making, by comparing models estimated without
institutional variables with models that include these variables, we can see how incorrect

inferences about both competition and campaign spending would be made.

89

Specifying the Models

Dependent Variables. The dependent variable in the incumbent-challenger
models is the percentage of the vote received by the incumbent (Incumbent Vote). In the
open seat models, the dependent variable is the percentage of the vote received by the
winning candidate (Winner Vote).

Independent Variables. The key independent variables of interest in this analysis
are the level of spending by the candidates in the election. For incumbent-challenger
races, consistent with the congressional and state legislative research, the amount spent
by the incumbent (Log of Incumbent Spending) should not positively affect the electoral
support of the incumbent (Jacobson 1978, 1980, 1990; Ansolabehere and Gerber 1994;
but see Green and Krasno 1988, 1990; Thomas 1989; Levitt 1994). However, contrary to
the findings about incumbent spending, challenger spending (Log of Challenger
Spending) has been found to significantly affect the vote of the incumbent in a negative
direction (Jacobson 1978, 1980, 1990; Abramowitz 1988, 1989; Squire 1989b;
Gierzynski and Breaux 1991, 1996; Sorauf 1992; Ansolabehere and Gerber 1994;
Thielemann and Wilhite 1995; but see Erikson and Palfrey 2000). That is, the more
money spent by the challenger, the lower the incumbent’s percentage of the vote. I
expect the same to hold true here.

In terms of open seats, the spending of both candidates should positively impact
their electoral support (Log of Winner Spending and Log of Loser Spending). This is
because neither candidate has the advantages of incumbency. In order to become better
known, candidates need to spend money (Jacobson 1978, 1980, 1990, 1997; McKnight,

Schaefer, and Johnson 1978; Becker and Dunwoody 1982; Dubois 1984, 1986a; Green

90

and Krasno 1988, 1990; Squire and Wright 1990; Banaian and Luksetich 1991; Breaux
and Gierzynski 1991; Gierzynski and Breaux 1991, 1993, 1996; Squire 1992; Biersack,
Hermson, and Wilcox 1993; Thielemann 1993; but see Arrington 1996), and the more
they spend, the better they should perform with the electorate.1

While I am primarily interested in the effects of campaign spending on the
percentage of the vote received by the candidates, in order to properly specify these
models, other factors hypothesized to impact the incumbent’s (or winner’s) percentage of
the vote must also be included. These variables fall under two general categories:
candidate characteristics and the state and electoral context.

Candidate characteristics. In addition to the spending variables, there are three
other candidate characteristics that should affect the electoral support of the candidates.
First, the quality of the candidates should matter. Regardless of whether a candidate is
challenging an incumbent, or running for an open seat, candidates with prior judicial
experience should fare better than candidates without such experience (Bonneau and Hall
2000, 2001), just as candidates with prior elected experience perform better than
candidates without such experience in legislative races (Jacobson 1980; Green and
Krasno 1988; Van Dunk 1997). Other things being equal, I expect that a challenger with
prior judicial experience (Quality Challenger) will reduce the incumbent’s percentage of
the vote. Likewise, the winner of an open seat having prior judicial experience will

increase her percentage of the vote (Quality Winner), while the winner’s percentage of

 

' Consistent with the congressional literature, I elect to use measures of each candidate’s spending here
instead of the difference between candidate spending because the effects of a candidate’s campaign
spending depends on which candidate is spending the money. Thus, campaign spending by the challenger
may matter while that of the incumbent may not. Or challenger spending may be negatively related to the
incumbent’s percentage of the vote (as hypothesized), while incumbent spending is positively related to her
level of electoral support. Using a measure of the difference in spending would not allow me to assess the
effects of each candidate’s campaign spending on electoral competition.

91

the vote will be decreased if the losing candidate has prior judicial experience (Quality
Loser).

The political party of the incumbent (or winner) is also important, as it might also
influence her level of electoral support. In some states, being affiliated with a particular
party (such as the Democratic Party) may help a candidate’s chances of winning election,
while in others it will hinder the candidate’s hopes. Thus, I need to devise a measure that
takes into account a candidate’s party, conditional on the current partisan climate of the
state. There are a few different ways to do this. First, I could use the statewide vote
during the last presidential election (or current election if it is occurring in the same year).
Thus, for a Democratic candidate running in Arkansas in 1992, this variable would be
coded as 1 since the Democratic presidential candidate received a majority of the popular
vote in Arkansas. The same would hold true for those Democratic candidates running in
1994, since the last presidential election was in 1992. Second, I could do the same thing,
except use the gubernatorial results. Third, and finally, I could use the results from US.
Senate races.2

Here, I develop a composite measure that takes into account all of the measures
above along with the political party of the justice (Partisan Consonance). This measure
takes into account the degree of partisan consonance among the statewide elected

officials in the state. If the incumbent justice (or winning candidate), and the winning

presidential, gubernatorial, and senatorial candidates are all of the same party, then there

 

2 Alternatively, I could see if there was any evidence of “coattails” at either the presidential or gubernatorial
level. That is, I could examine if state supreme court candidates benefit if a presidential or gubernatorial
candidate of their political party wins election on the same ballot. When the measures of coattails were
included, however, they failed to achieve statistical significance. Given the high degree of collinearity
between the coattail variables and the measure of partisan consonance, both variables cannot be included in
the final model. Given the theoretical importance of the partisan consonance measure, the importance of
what it measures, and the failure of any coattail effects to achieve significance, 1 opt to include the partisan
consonance measure in my final specification.

92

is a high degree of partisan consonance, and competition should be decreased for
members of the majority party.3 Likewise, if the justice is of the opposite party of the
winning presidential, gubernatorial, and senatorial candidates, then this lack of partisan
consonance should lead to more competition.4

Hall (2001b) recently found that older justices are more likely to retire rather than
stand for reelection. While older justices are more experienced, the findings by Hall
(2001b) suggest that these justices may be electorally vulnerable, and thus choose to
retire rather than face the electorate. Older justices may be more likely to be perceived
(and characterized) as out-of-touch with the electorate, as well as be perceived as unable
to keep up with the workload on the court. Thus, other things being equal, I expect that
older justices (Age) will be more likely to experience competitive races.

For incumbent-challenger races, as has been discussed earlier, not all incumbents
are alike. Incumbents who have previously won election have faced the electorate before
and have had their candidacies approved. This is not the case for those who are facing
their first election. Thus, incumbents who have not yet faced the electorate (Appointed
First) may be more likely to receive less electoral support than their previously elected
colleagues, although Hall (2001a) found no evidence of such an effect.

Statefi and Electoral Context. Whether the incumbent or winner was challenged in
the primary could also affect her percentage of the vote. Other things being equal, if an

incumbent (or winner) had to run in a primary as well as a general election, her

 

3 If the election was held in the same year as the election for the state supreme court election, then I use
these results. However, if there is no election for the office in the same year as the state supreme court
election, then I use the results from the most recent election. Thus, elections in 1996 use the 1996
presidential results, as do elections in 1998.

An alternative way to code this variable is to use the percentage of partisan consonance, as opposed to an
ordinal scale. When the percentage measure is used, there are no changes to the substantive results.

93

percentage of the vote in the general election should be lower. This is because having to
face a primary challenge is an indication of candidate vulnerability, and vulnerable
candidates should receive fewer votes than non-vulnerable candidates. Hence, I include a
variable that indicates whether the incumbent or winner was challenged in the primary
(Compete).

As discussed extensively in Chapter 3, in order to account for the potential
endogeneity problems, I need some measure of state supreme court competitiveness. For
the reasons discussed earlier, I include a variable indicating whether there was a close
state supreme court election (decided by 55% of the vote or less) in the most recent
electoral cycle (Prior Close Race). As a measure of state supreme court competitiveness,
it is expected that a previous competitive state supreme court race will lead to a lower
percentage of the vote for the incumbent (or winner).

Whether partisan control of the court is at stake is also likely to affect
competitiveness (Control of Court). Specifically, if partisan control of the court is at
stake in the election, then the race should be more competitive, as both parties and
candidates will attempt to either gain control of the court or solidify their control of the
court.

Unlike US. House races, the electoral constituencies of state supreme court
candidates vary in size. Some candidates run statewide, while others run in districts.
Even more fundamentally, some states are bigger than others. That is, spending $1,000
in Alabama is not the same as spending $1,000 in Texas. Thus, in order to control for the

size of the electorate, I include a variable measuring the size of the voting age pOpulation

94

(Voting Population), consistent with studies of campaign finance in the US. Senate
(Squire 1989b, 1992) and state gubernatorial races (Partin 2002).

Like other elected officials, state supreme court justices may be held accountable
for issues perceived to be under their control. There has been a significant body of
literature demonstrating that the voters make retrospective decisions on incumbent
governors and legislators based on the state of the economy (Atkeson and Partin 1995;
Lowry, Alt, and Ferree 1995; Niemi, Stanley, and Vogel 1995; Svoboda 1995; Carsey
and Wright 1998). Further, the same has been found in both presidential and
congressional elections (e. g., Jacobson 1997). Applying this to state high court elections,
Hall (2001a) found that incumbent justices performed worse the higher the murder rate in
their state was, indicating that these incumbents were held responsible for the state of
public safety. Consistent with this, I expect that there will be more competition in
incumbent-challenger races the higher the murder rate in the year prior to the election
(Murder Rate).

Just as the percentage of the high court docket occupied with tort cases affected
the total amount of money spent, it also should affect electoral competition. Given the
amount of money at stake in these types of cases, both trial lawyers and corporations and
defense lawyers will want to ensure the election of judges sympathetic to their cause.
Specifically, the higher the percentage of tort cases on the docket (Tort Docket), the more
electoral competition there should be.

Finally, I include a dummy variable for the year of the election, minus one year,

to control for any temporal effects (I 990, 1992, I 994, 1996, I 998).

95

For convenience, Table 6.1 provides a complete list of all of the variables and

their exact measurement.

96

Table 6.1: Variable Descriptions for a Model of Electoral Competition in State Supreme

Court Elections

 

Variable

Variable Description

 

Vote

Candidate Characteristics
Log of Incumbent Spending

Log of Challenger Spending
Log of Winner Spending
Log of Loser Spending
Quality Challenger

Quality Winner

Quality Loser

Partisan Consonance

Age
Appointed First

State and Electoral Context
Compete

Prior Close Race

Control of Court

Voting Population
Murder Rate

Tort Docket

Control Variables

1990, 1992, 1994, 1996,
1998

percentage of the vote in the general election for the
incumbent (or winner for open seats)

log of total campaign spending for the
incumbent

log of total campaign spending for the
chaHenger

log of total campaign spending for the winner
log of total campaign spending for the loser
1 if incumbent was challenged by a
candidate with prior judicial experience

0 otherwise

1 if winner had prior judicial experience

0 otherwise

1 if loser had prior judicial experience

0 otherwise

degree of partisan consonance between the
incumbent (winner) and the presidential,
gubernatorial, and senatorial candidate

age of the incumbent

1 if the incumbent was originally appointed
to the Court

0 otherwise

1 if the incumbent or winner was challenged

in both the primary and general election

0 otherwise

1 if recent judicial election was decided by

55% of the vote or less

0 otherwise

1 if partisan control of the court is at stake

0 otherwise

voting age population of the state/district (10005)
State murders and nonnegli gent manslaughter per
100,000 population, lagged one year

proportion of the docket (1995) involving tort cases

1 if election occurred in the designated year
0 otherwise

97

Specification

There are two popular specifications for campaign spending models. First,
campaign spending can be modeled linearly, and the amount of spending can be simply
entered into the model. However, this assumes that the campaign spending is related to
votes in a linear fashion. That is, one additional dollar of spending has the same effect on
electoral support, regardless of how much was spent previously. Thus, if a candidate has
already spent $100,000, the linear specification assumes that one more dollar of spending
will have the same effect on her percentage of the vote than if she had only spent $100.
This does not make much sense, as spending should have more of an effect on the
percentage of the vote at lower levels of spending than at higher levels of spending.

A popular alternate specification is the semi-log model (Welch 1976; Jacobson
1996; Partin 2002). Semi-log models allow for diminishing marginal returns. Thus,
these models allow for the fact that spending has differential effects on the level of
electoral support depending on how much money was spent previously. There is one
major drawback to semi-log specifications. The natural log of zero is undefined. While
this is not a major problem in the congressional literature (since very few candidates
spend $0), in the state supreme court races examined here, 13 candidates (12 challengers
and 1 loser of an open seat) spent no money. In order to account for this, and to avoid
dropping these races, I assume that these candidates spent $1. While spending $1 is
different than spending $0, this change should not affect the results. Indeed, when the
races where a candidate spent $0 are omitted, the same results are obtained.

I also estimate the models using the linear specification. The substantive results

are identical: the same variables are significant under both specifications. Because the

98

semi-log model allows for diminishing marginal returns, and thus is theoretically more
appropriate, I use this specification, although it is important to note that the results are

invariant to specification.

Estimation Technique

Endogeneity problem. As discussed in Chapter 3, models of campaign spending
may be susceptible to endogeneity problems. To the extent that one (or more) of the
independent variables is endogenous, then the assumptions of ordinary least squares
regression (OLS) are violated and the OLS estimates are both biased and inconsistent. In
order to correct for this problem, an instrumental variable must be found for the
endogenous regressor that is both correlated with the endogenous variable and
uncorrelated with the error term.

Common practice in congressional elections is to use the incumbent’s prior
spending total as an instrument for current incumbent spending (Jacobson 1980, 1985;
Green and Krasno 1988), and then estimate the models using two-stage least squares
(ZSLS). Thus, incumbent spending in 1976 is used as an instrument for incumbent
spending in 1978. However, given that the terms of office for state supreme court
justices are a minimum of 6 years, and that almost one quarter of all incumbents are
facing voters for the first time, prior incumbent spending is an unsatisfactory instrument
for elections to the high court bench. Looking at races for the US. Senate (a much better
comparison to state supreme court races in terms of electoral constituency and term of
office), Squire (1989b, 1992) simply used OLS regression. This choice is defensible

because, in small samples, the ZSLS estimates are also biased (although they remain

99

consistent). Additionally, Partin (2002) used OLS to estimate the effects of spending on
the vote total of gubernatorial candidates, and Jacobson (1996) has recently done so for
congressional races. Further, as Jacobson has extensively documented (1980, 1985,
1997), the substantive findings about campaign spending hold regardless of whether the
models are estimated by OLS or ZSLS.

While there are theoretical reasons to suspect that my measure of a prior close
race sufficiently alleviates the potential endogeneity problem, endogeneity is a statistical
problem as well as a theoretical one. Even if there is no reason to think a variable is
endogenous, it may, in fact, be endogenous. In order to test for the endogeneity of
incumbent spending, I need to find an appropriate instrumental variable.

Unlike congressional elections, not all state supreme court seats are equal. State
supreme court seats vary on such things as salary, workload, staff, and term of office, to
name a few. Consequently, becoming a member of the court in some states is more
desirable than becoming a member of the court in others. The more valuable the seat, the
more money one is likely to spend in retaining that seat. Thus, I use a measure of state
supreme court professionalism developed by Brace and Hall (2001) as an instrument for
incumbent spending. This measure is an index compiled by confirmatory factor analysis
that includes that following factors: number of clerks, the difference between justice
salary and average employee earnings, the number of authorized supreme court justices
per 1000 residents, and the size of the docket. This index is correlated with incumbent
spending, but should not independently affect the vote total for the incumbent. Hence, it

is an appropriate instrument.

100

In order to test whether 28LS is more efficient than OLS (and thus should be
used), a Hausman test must be performed. The Hausman test compares the coefficients
from both models (2SLS and OLS) and determines whether the difference in the
coefficients is systematic. If the null hypothesis (that the difference between the two
models is not systematic) cannot be rejected, then there is no endogeneity problem. In
this case, incumbent spending can be said to be exogenous, and the model can be
estimated using OLS. Otherwise, 2SLS is to be preferred, using state supreme court
professionalism as an instrument for incumbent spending. Performing the Hausman test
on the models indicates that there is no gain in efficiency from using ZSLS (these results
can be seen in the Appendix, Table A2). Hence, these models will be estimated using
OLS regression.

Note that thus far the discussion has been centered only on instrumenting
incumbent spending. That is, challenger spending (as well as spending by both
candidates for open seats) has been assumed to be exogenous. There are some good
reasons to suggest that challenger spending is indeed exogenous. First, challengers
generally have difficulty raising money, and this difficulty is not likely to change much if
their expected percentage of the vote rises (Green and Krasno 1988). Second, recall that
in order for challenger expenditures to be endogenous, they must be correlated with the
error term (that part of the incumbent’s vote total not captured by the model). However,

To the extent that money and its trappings—a large staff, a new computer,

a professional campaign manager—inspire additional financial backing,

the causal process will be confined to the independent variable (CE)

[challenger expenditures] itself. Only fund-raising that is based on

information not contained in the independent variables will lead to biased
estimates for challenger spending. (Green and Krasno 1988, 904)

101

The story for open seats is a little more difficult. The dependent variable for these
races is the vote percentage for the winning candidate. If the expenditures by the winning
candidate are endogenous, then this means we will likely overestimate the affects of
spending of the vote total of the winner. Likewise, we are likely to underestimate the
effects of the spending of the loser. While both candidates for open seats begin the
campaign as non-incumbents, it is unknown how easy it is for these candidates to raise
money. Given the spending totals seen in Chapter 4, it is possible that as the expected
vote total rises, it is easier for the winning candidate to raise more money. To check for
this, I used the state supreme court professionalism index as an instrument for winner
spending, and performed a Hausman test, which indicates that OLS regression is
appropriate (see Appendix, Table A3).

In sum, the results of the Hausman tests confirm my theoretical argument that
ZSLS is not necessary. For both incumbent-challenger races and open seat contests, I
was unable to reject the null hypothesis of no systematic differences in the coefficients.
That is, 2SLS is not more efficient than OLS. Thus, because of both theoretical and
empirical reasons, I will estimate the models here using OLS regression.

Selection bias. In addition to the potential endogeneity problem, there is the
recurring issue of the potential selection bias given the nature of the sample. Since the
dependent variable (percentage of the vote) is only observed in contested races, if the
process by which races are contested is nonrandom (which is likely; see Bonneau and
Hall 2001), then omitting uncontested races could bias the coefficients and make the

standard errors inefficient (as discussed in detail in Chapters 3 and 5).

102

To correct for the possible selection bias, the models are estimated using
Heckman’s two-step correction procedure. Just as I did in Chapter 5, I use Bonneau and
Hall’s (2001) model predicting supreme court contestation to model the process by which
a supreme court races will be contested.5 The OLS results, without correcting for
selection bias, are presented in the Appendix (T ables A.4-A.7). As one can see by

comparing the tables, the same results generally obtain.

The Role of Money in State Supreme Court Elections
Table 6.2 shows the 015 results from the multivariate model of incumbent-

challenger contests.

 

5 The variables included in the selection model can be found in Table 5.2.

103

Table 6.2: The Effects of Money on Votes, Incumbent-Challenger Contests

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. 2 P > |z|
Error

Log of Incum. 0.159 0.943 0.169 0.866
Spending

Log of Chall. -l.357 0.464 -2.926 0.003
Spending

Quality Chall. -4.606 2.217 -2.078 0.038

Partisan 0.394 0.999 0.395 0.693
Consonance

Age 0007 0.002 -3.347 0.001

Appointed -2.933 1.388 -2.112 0.035
First

Compete -0.549 1.1 17 -0.492 0.623

Prior Close -3.730 1.603 -2.327 0.020
Race

Control of 0.446 1.346 0.331 0.740
Court

Voting 0.001 0.000 2.609 0.009
Population

Murder Rate 0.148 0.243 0.609 0.542

Tort Docket 0.111 0.075 1.485 0.138

1990 -3.527 1.486 -2.373 0.018

1992 -3.765 2.076 -1.814 0.070

1994 -2.727 2.601 -1.049 0.294

1996 -3.825 2.959 -1.292 0.196

1998 3.889 2.212 1.758 0.079

Constant 64.145 9.007 7.122 0.000

Dependent variable: percentage of vote for incumbent

Mean of dependent variable: 56.31

Number of observations = 392

Censored observations = 265

Uncensored observations = 127

Log likelihood = -624.354

x2 = 13.25

Prob > x2 = 0.039

104

Most relevant for the purposes here is the performance of the two spending
variables. While the amount of money spent by the incumbent does not affect her
percentage of the vote, the amount spent by the challenger does lead to a decrease in the
percentage of the vote for the incumbent. This is as expected and is consistent with the
existing congressional and state legislative literature. Challengers can increase their
electoral support simply by spending more money. Specifically, a 1% increase in
spending by challenger decreases the incumbent’s percentage of the vote by 0.014%. In
order to raise her level of electoral support 0.1%, a challenger would have to increase her
spending by 7%. If a challenger spent $100,000, she would need to spend $107,000 to
raise her percentage of the vote by 0.1%. This suggests that more spending, at least by
challengers, serves to promote competition. The presence of a challenger with prior
judicial experience also decreases the incumbent’s percentage of the vote by over 4.5%.
Thus, while campaign spending matters, so does prior judicial experience. The results
also show that the age of the incumbent leads to more competitive elections, being
initially appointed decreases the incumbent’s electoral support by almost 3%, and that a
prior close judicial race decreases the incumbent’s percentage of the vote by almost 4%.
Interestingly, neither partisan consonance nor partisan control of the court seems to affect
competition, indicating that state supreme court elections may be insulated to some
degree from the politics of the state.

The size of the voting age population is significant, indicating that the larger the
size of the voting age population, the less competitive the race. This is not surprising,
given that the larger the electoral constituency, the more money that a challenger would

have to spend to effectively reach the electorate. Given the difficulties challengers have

105

in raising money (Green and Krasno 1988), it is easy to see how they would fare more
poorly, other things being equal, in larger constituencies. Contrary to the results of Hall
(2001a), the murder rate does not appear to be related to electoral competition.
Incumbents do not appear to be held accountable for issues of public safety in these
elections. Finally, the percentage of the docket devoted to tort cases is also insignificant.
Contrary to expectations, elections for seats where the winner will spend a substantial
portion of her time deciding tort cases are not more competitive than seats where the
winner will spend significant amounts of time deciding other types of cases.

The story for open seats is somewhat different, as seen in Table 6.3.

106

Table 6.3: The Effects of Money on Votes, Open Seats

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. 2 P > |z|
Error

Log of Winner 1.033 1.032 1.002 0.317
Spending

Log of Loser -1.708 0.479 -3.564 0.000
Spending

Quality 2.290 1.753 1.306 0.192
Winner

Quality Loser -0.848 2.825 -0.300 0.764

Partisan -0.205 0.850 -0.241 0.810
Consonance

Compete -4.762 2.803 -1.699 0.089

Prior Close 2.444 3.768 0.649 0.517
Race

Control of 0.226 1.802 0.125 0.900
Court

Voting 0.000 0.000 0.850 0.395
Population

Tort Docket -0.047 0.103 -0.460 0.646

1990 2.817 4.201 0.671 0.502

1992 0.662 2.853 0.232 0.817

1994 0.663 1.818 0.365 0.715

1996 4.007 3.889 1.030 0.303

1998 -2.816 4.618 -0.610 0.542

Constant 59.284 18.470 3.210 0.001

 

Dependent variable: percentage of vote for winner
Mean of dependent variable: 55.87

Number of observations = 328
Censored observations = 265
Uncensored observations = 63

Log likelihood = —333.284
x2 = 3.71
Prob > x2 = 0.716

107

The amount of money spent by the winner fails to achieve significance.
However, the amount spent by the loser is statistically significant and in the hypothesized
direction. This suggests that the amount of money spent by the loser leads to a decrease
in the winning candidate’s percentage of the vote (and thus promotes competition). A
1% increase in campaign spending by the loser leads to a decrease of 0.017% in the
winner’s percentage of the vote. In order to raise her percentage of the vote by 0.1%, a
challenger would need to increase her spending by 5.8%. Thus, if she had spent
$100,000, then she would need to spend $105,800 to increase her level of support by
0.1%. The lack of significance for the spending by the winner is somewhat puzzling:
why should campaign spending not affect the electoral support of winners? The most
likely explanation is that both candidates in open seats begin the campaign on relatively
equal footing. They both spend money to gain recognition, and spending increases each
candidate’s visibility equally. Thus, each additional dollar spent by the winner leads to
an increase in her percentage of the vote, but each additional dollar spent by the loser
leads to a decrease in the winner’s percentage of the vote. If spending by the losing
candidate costs the winner more votes than spending by the winning candidate gains her
votes (which appears to be the case here), then campaign spending by the winner will
appear to have no effect on how well she does, while campaign spending by her opponent
will be significant.

Somewhat surprisingly, the prior judicial experience of the candidates does not
matter in open seat races. Neither the quality of the winner nor of the loser significantly
affects her percentage of the vote, nor does anything else, with the exception of

competition in the primary. Having a competitive primary costs the winning candidate

108

almost 5% of the vote, other things being equal, although this result is only marginally
significant and should be treated cautiously.

These results confirm the need to analyze incumbent-challenger races separately
from open seat races. In addition to the theoretical differences between these two types

of races, Tables 6.2 and 6.3 show how these races are empirically different as well.

The Role of Independent Expenditures

In Chapter 3, I discussed the problem of independent expenditures, and how
failure to include the amount of money spent by independent groups could affect the
results. In order to attempt to take this into account, I identified state supreme court
elections that were singled out by the media as involving large amounts of independent
expenditures (Goldberg, Holman, and Sanchez 2002). Thus, I estimate the models above
omitting the following races, based on the Goldberg, Holman, and Sanchez (2002) study:
2000 Alabama, 2000 Michigan, 2000 Mississippi, and 2000 Ohio. The results with these

races omitted are shown in Tables 6.4 and 6.5.

109

Table 6.4: The Effects of Money on Votes, Incumbent-Challenger Contests, with
Independent Expenditure Races Omitted

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. 2 P > |z|
Error

Log of Incum. 0.110 1.023 0.107 0.915
Spendirg

Log of Chall. -1.206 0.469 -2.569 0.010
Spending

Quality Chall. -5.332 2.075 -2.569 0.010

Partisan 0.117 1.088 0.108 0.914
Consonance

Age -0.006 0.002 -3. 138 0.002

Appointed -3.483 1.596 -2. 183 0.029
First

Compete -0.595 1.212 -0.491 0.623

Prior Close -4.876 2.112 -2.308 0.021
Race

Control of 0.909 1.841 0.494 0.621
Court

Voting 0.001 0.000 2.321 0.020
Population

Murder Rate 0.260 0.31 1 0.838 0.402

Tort Docket 0.085 0.089 0.962 0.336

1990 -5.554 2.720 -2.042 0.041

1992 -5.109 3.266 -1.564 0.118

1994 -4.411 3.351 -1.316 0.188

1996 -5.341 3.481 -1.534 0.125

1998 2.517 3.042 0.827 0.408

Constant 65.356 10.299 6.346 0.000

Dependent variable: percentage of vote for incumbent

Mean of dependent variable: 56.31

Number of observations = 380

Censored observations 2 265

Uncensored observations = 115

Log likelihood = -573.505

x2 = 27.90

Prob > x2 = 0.000

110

Table 6.5: The Effects of Money on Votes, Open Seats, with Independent
Expenditure Races Omitted

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. 2 P > |z|
Error

Log of Winner 0.936 1.131 0.827 0.408
Spending

Log of Loser -1.761 0.480 -3.672 0.000
Spending

Quality 2.122 1.826 1.162 0.245
Winner

Quality Loser -0.794 2.637 —0.301 0.763

Partisan -0.559 0.815 -0.686 0.492
Consonance

Compete -4.948 2.395 -2.067 0.039

Prior Close 2.828 4.833 0.585 0.558
Race

Control of 0.263 2.061 0.128 0.899
Court

Voting 0.000 0.001 0.219 0.827
Population

Tort Docket -0.061 0.093 -0.651 0.515

1990 2.883 4.998 0.577 0.564

1992 1.323 3.087 0.429 0.668

1994 1.405 2.469 0.569 0.569

1996 4.274 5.178 0.825 0.409

1998 -2.848 5.939 -0.480 0.632

Constant 64.961 18.552 3.502 0.000

Dependent variable: percentage of vote for winner

Mean of dependent variable: 55.87

Number of observations = 326

Censored observations = 265

Uncensored observations = 61

Log likelihood = -315.489

x2 3.55

Prob > x2 = 0.615

111

As can be seen from Tables 6.4 and 6.5, omitting those races that contained a lot
of independent expenditures does not change the substantive conclusions at all.
Campaign spending by the challenger still increases her percentage of the vote; campaign
spending by the loser of an open seat contest still increases her percentage of the vote.
Additionally, all of the other variables retain their significance (or insignificance). This
provides evidence, albeit indirect, that independent expenditures are not affecting the

results.

Conclusion

Consistent with the congressional elections literature, campaign spending by the
challenger serves to decrease the percentage of the vote of the incumbent, while
incumbents are not able to increase their electoral support by increasing their spending.
For open seats, the percentage of the vote of the winner is decreased by campaign
spending by the loser, but her own spending has no effect on her electoral support. These
results hold despite alternative specifications and taking into account the potential effects
of selection bias. This suggests that campaign spending by some types of candidates can
promote electoral competition.

Of course, noticeably missing from these models are institutional variables. The
models estimated and discussed here treat all state supreme court elections as alike,
regardless of their partisan nature, electoral constituency, and term of office. Doing so
yields results similar to those of congressional and state legislative elections. However,
given both the theoretical importance of institutional arrangements, as well as the

differential levels of spending and competition witnessed in Chapter 4, omitting these

112

variables leaves us with an incomplete understanding of state high court elections. I

address this deficiency in Chapter 7.

113

CHAPTER 7

THE EFFECTS OF MONEY ON VOTES: INSTITUTIONAL MODEL

The evidence thus far provides some limited support for the claim that campaign
spending can increase electoral competition. Spending by both challengers and the losers
of open seats leads to more competitive elections. This is consistent with the
congressional and state legislative literature. Further, this central result is invariant to
model specification.

Yet, to stop here would be to paint an incomplete picture. The analysis in Chapter
6 treated all state supreme court elections the same. However, one of the most
distinguishing characteristics of elections for state high court is that they vary on
fundamental institutional characteristics such as type of ballot on which they are elected,
electoral constituency, and term of office. In this chapter, I explore the role these
institutional differences have on both electoral competition and the effects of campaign

spending on how well candidates perform with the electorate.

Institutional Variables

The most fundamental institutional difference is the type of election. As has been
noted earlier, some states elect their state supreme court justices on partisan ballots, while
others do so on nonpartisan ballots. The only difference between these two types of
elections is that the political party affiliation of the candidate is listed on partisan ballots

and omitted on nonpartisan ballots. However, this difference can be significant. As we

114

have seen in Chapter 4, partisan elections are both more expensive and more competitive
than nonpartisan elections, in the aggregate.

There are two different methods to take into account the partisan nature of the
election. First, I could include a dummy variable in each model (incumbent-challenger
and open seats) to denote the type of election. If this variable is significant, then we can
conclude that the type of election has a significant effect on the vote total. Further, by
interacting campaign spending with type of election, we can ascertain if the influence of
campaign spending also varies by type of election.

A second possibility is to estimate partisan and nonpartisan elections separately.
In addition to possessing the same characteristics as including a dummy variable,
estimating these models separately allows one to assess the difference in the determinants
of electoral competition as well as the differences in campaign spending between these
types of races. Given the differences in competitiveness and campaign spending
described in Chapter 4, as well as the results by Hall (2001a), there are good reasons to
suspect that there may be differences in both the determinants of electoral competition
and the efficacy of campaign spending between partisan and nonpartisan elections.

However, before estimating the models separately, one can test to see if the
models are structurally different using the Chow test (Gujarati 1995). This procedure
tests the entire set of parameters between two data sets—in this case, partisan and
nonpartisan elections (Kennedy 1998). Both the partisan and nonpartisan models were
estimated separately, as was a model that combined both types of elections. As the
results from the Chow test indicate (Appendix, Table A8), there is no statistical

difference between the partisan and nonpartisan models. Consequently, I do not estimate

115

the models separately, and instead insert a dummy variable (Partisan) to take into
account the nature of the race.

Besides the type of election, there are three other institutional variables that
should affect both competition and campaign spending. As discussed earlier, not all
supreme court candidates run in statewide elections. In Illinois, Kentucky, Louisiana,
and Mississippi, candidates run in districts. Hall (2001a) found that candidates running
nonpartisan district elections received over 18% less of the vote (although this result did
not hold in partisan elections), suggesting that district elections may have different effects
on competition in different selection systems.1 This may also affect the influence of
campaign spending on electoral competition. Specifically, campaign spending should be
more effective in district-based elections compared to statewide elections (Dubois 1986a;
Hogan and Ham 1998; Reid 1999). Thus, I include a dummy variable to take into
account the electoral constituency of the election (District).

The presence of a multi-member race should also affect electoral competition.
Competition in multi-member districts has been found to be higher than in comparable
single-member districts (Cox and Morgenstem 1995). Thus, I include a dummy variable
to control for these situations (MultiMember). A word is in order about how vote totals
and campaign spending are adjusted to account for multi-member districts. I adjust the
percentage of the vote received using the method first used by Jewell (1982): I divide
each candidate’s percentage of the vote by the total number of votes in the district, and
then multiply this result by the total number of seats in the district. This procedure has

generally been recognized by state politics scholars (Tucker and Weber 1987; Weber,

 

I This was not found when an interaction term for district, partisan elections was included here. For this
reason, along with the nature of the interaction terms to be included in the following sections, I do not
include an interaction term for district, partisan elections in the analysis.

116

Tucker, and Brace 1991; Hall 2001a) as producing vote totals equivalent to the electoral
support for candidates in single-member districts (Hall 2001a).

For campaign spending totals, I perform a procedure suggested by Moncrief
(2001). I define the “challenger” as the last person to get elected. Here, since there are
always only two seats at stake in state supreme court multi-member elections, the person
with the second highest vote percentage is the key opponent because this person’s
percentage of the vote is the minimum amount necessary to win election. Thus, for the
highest vote-getter, the 2nd highest vote-getter is the “challenger.” For the second highest
vote-getter, the “challenger” is the person who finished third, since this is the candidate
the second highest vote-getter had to beat. This procedure best approximates the amount
of money spent by a “challenger.”2

Finally, the term of office may affect the incumbent’s (or winner’s) percentage of
the vote. Longer terms should be more attractive to candidates since there is increased
job security (Bonneau and Hall 2001). Thus, there should be more competition for seats
that have longer terms of office associated with them. To the extent that there is greater
competition, the election should be closer. I expect that the longer the term of office
(Term), the lower the incumbent’s (or winner’s) percentage of the vote will be.

In addition to these variables, the models estimated here contain the same
variables used to estimate the models in Chapter 6 (see Table 6.1 for these variables and
their measurement). For convenience, Table 7.1 provides a list of the institutional

variables and their exact measurement.

 

2 Since multi-member races are different from single-member races, and because my measures of
percentage of the vote received and campaign spending are approximations, I also estimated the models
omitting multi-member races. There was no change in the substantive conclusions.

117

Tal
Ste

Table 7.1: Additional Variable Descriptions for a Model of Electoral Competition in
State Supreme Court Elections

 

Variable Variable Description

 

Vote = percentage of the vote in the general election
for the incumbent (or winner for open seats)
Institutional Factors

Partisan = 1 if the election was held in a partisan state
0 otherwise

District = 1 if the election was held in a district
0 otherwise

MultiMember = 1 if the election is in a multimember district
0 otherwise

Term = length (in years) of the term of office

118

Esl

Estimation Technique

Endogeneity. The potential endogeneity problem that existed in Chapter 6 is also
present for the models estimated here. In order to check for endogeneity, I used the state
supreme court professionalism index discussed in Chapter 6 as an instrument for
incumbent and winner spending, and performed a Hausman test. The results of the
Hausman tests (presented in the Appendix, Tables A9 and A.10) indicate that OLS
regression is appropriate for both incumbent-challenger races and open seat contests, and
thus this technique is used to estimate the models here.

Selection bias. Also present is the potential selection bias since only contested
races are analyzed. As in Chapters 5 and 6, I model the process by which state supreme
court races are contested using the model developed by Bonneau and Hall (2001). While
I present and discuss the results from the Heckman procedure, the results from the OLS

regression are presented in the Appendix (Table All).

The Role of Money and Institutions in State Supreme Court Elections

Table 7.2 shows the OLS results from the multivariate model of incumbent-

challenger contests.

119

Table 7.2: The Effects of Money on Votes, Incumbent-Challenger Contests

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. z P > |z|
Error

Log of Incum. -0. 178 0.891 -0.200 0.841
Spending

Log of Chall. -1.515 0.494 -3.065 0.002
Spending

Quality Chall. -3.867 2.156 -l.794 0.073

Partisan -0.380 1.019 -0.373 0.709
Consonance

Age 0006 0.002 -3.014 0.003

Appointed -3.294 1.090 -3 .022 0.003
First

Compete 1.637 1.024 1.599 0.110

Prior Close -4.638 1.437 -3.227 0.001
Race

Control of 0.387 1.685 0.230 0.818
Court

Voting 0.000 0.000 2.068 0.039
Population

Murder Rate 0.342 0.308 1.11 1 0.267

Tort Docket 0.169 0.057 2.937 0.003

Partisan -4.425 2.486 -1.780 0.075

District 6575 3.058 -2. 150 0.032

MultiMember -0.274 2.734 -0.100 0.920

Term 2.879 0.785 3.668 0.000

1990 -4.428 1.813 -2.443 0.015

1992 -4.546 2.189 -2.077 0.038

1994 -3.035 2.869 -1.058 0.290

1996 -3.650 3.072 -1.188 0.235

1998 4.061 2.135 1.902 0.057

Constant 55.532 8.933 6.216 0.000

Dependent variable: percentage of vote for incumbent

Mean of dependent variable: 56.31

Number of observations = 392

Censored observations = 265

Uncensored observations = 127

Log likelihood = -618.488

x2 (6) = 19.28

Prob > x2 = 0.004

120

The results for incumbent-challenger contests with the institutional variables
included generally echo those obtained for all incumbent-challenger omitting such
variables (Table 6.2). Campaign spending by the challenger has a significant negative
effect on the percentage of the vote received by the incumbent; the more the challenger
spends, the better she does (and the worse the incumbent does). For every 1% increase in
challenger spending, the percentage of vote received by the incumbent will decrease by
over 0.015%. To increase her level of electoral support by 0.1%, the challenger must
spent an additional 6.7%. If she has already spent $100,000, then she will need to spend
$106,700 to increase her percentage of the vote by 0.1%, other things being equal.
Additionally, incumbent spending remains insignificant.

The challenger also performs almost 4% better if she has prior judicial
experience, and over 3% better if the incumbent was initially appointed. Further, the
older an incumbent gets, the worse she performs with the electorate. A prior competitive
race also leads to more competition suggesting that past state supreme court elections
affect current races. Specifically, if there was a competitive race in the most recent state
supreme court election, an incumbent will receive 4.6% less of the vote, other things
being equal. Electoral competition is unrelated to whether the incumbent faced primary
competition, and if partisan control of the court is at stake. Also of note is the fact that
partisan consonance is insignificant. This suggests that incumbent-challenger races for
the state high court bench are somewhat insulated from the general political climate of
the state. However, these races are not insulated from the type of work performed on the
court. The higher the proportion of the docket occupied by tort cases, the better the

incumbent does. This is contrary to what was hypothesized. A likely reason for this is

121

that the relevant interests (trial lawyers or corporations) devote significant resources to
protecting incumbents sympathetic to their cause. While they also may support
challengers, they first want to take care of the justices already on the bench. If this is the
case, then we would expect them to contribute heavily to incumbent’s reelection
campaigns, and thus strengthen the incumbency advantage. Thus, while the percentage
of the docket occupied by tort cases contributes to the level of campaign spending in the
race (Chapter 5), it does not contribute to electoral competition. Indeed, it even
discourages it.

Three of the four institutional variables attain significance. First, other things
being equal, there is more competition in partisan races. Incumbents will receive 4.4%
less of the vote in partisan races compared to nonpartisan races, other things being equal.
This is not surprising, given that we saw in Chapter 4 how these races were more
competitive than nonpartisan races, in general. District incumbent-challenger elections
are more competitive than their statewide counterparts. Other things being equal, an
incumbent will perform over 6.5% worse in a district-based election as opposed to
statewide election.3 Electoral competition is unrelated to whether the election occurred in
a multi-member district, but is related to the term of office. Somewhat surprisingly,
longer terms of office lead to less electoral competition (incumbents receive a higher
percentage of the vote, other things being equal). This may be because 90% of the
elections in the dataset are for 6- and 8-year terms of office. Nonetheless, this finding is

contrary to expectations.

 

3 Hall (2001a) found this for nonpartisan district elections but not for partisan district elections.

122

Comparing Tables 6.2 and 7.2 shows the need to include institutional variables in
studies of judicial elections. Table 7.3 compares the significance of the coefficients

displayed in Tables 6.2 and 7.2.

123

Table 7.3: Comparison of Non-Institutional and Institutional Models of Incumbent-

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Challenger Contests
Variable Table 6.2 Table 7.2
Log of Incumbent Spending NS NS
Log of Challenger Spending - -
Quality Challenger - -
Partisan Consonance NS NS
Age — -
Appointed First - -
Compete NS NS
Prior Close Race - -
Control of Court NS NS
Voting Population + +
Murder Rate NS NS
Tort Docket NS +
Partisan -
District —
MultiMember NS
Term 4-
1990 - NS
1992 - NS
1994 NS NS
1996 NS NS
1998 - NS
+ significant in a positive direction

significant in a negative direction
not significant

NS

124

Table 7.3 shows quite clearly the erroneous inferences that can be made by
omitting institutional variables. While all of the variables that were significant in Table
6.2 remain significant in Table 7.2, the percentage of the docket occupied by tort cases
becomes statistically significant in the institutional specification. Additionally, electoral
competition is affected by whether the race is partisan, in a district, and for a longer term
of office. Table 7 .3 is clear and convincing evidence of the importance of including
institutional variables.

The results for open seat races are shown in Table 7.4.4

 

4 I attempted to estimate the Heckman model for the open seat races. However, since there are only 63
races, and given the number of independent variables in the model, I was unable to do so in all cases,
especially for the interactive models that follow. Hence, in order to keep the results comparable, I estimate
the open seat models using OLS regression and do not correct for the potential selection bias. I did test for
selection bias, however, using an alternative procedure. First, I ran the selection model on the open seat
races. I then took the estimated probability from this analysis and included it as an independent variable in
the OLS regression. If there was selection bias, then this variable will be significant and will also be
controlling for the bias. If the variable is insignificant, then there was no bias (Brace 2002). The results
from the model indicate that there was no selection bias, and thus I report the results from the OLS
analysis.

125

Table 7.4: The Effects of Money on Votes, Open Seats

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust t P > It]
Std. Error

Log of Winner 0.130 1.567 0.083 0.935
Spending

Log of Loser -1.928 0.486 -3.969 0.001
Spending

Quality 0.679 2.184 0.311 0.760
Winner

Qualit)I Loser 1.606 3.450 0.466 0.648

Partisan -0.702 0.998 -0.704 0.492
Consonance

Compete -2.778 2.603 -1.067 0.302

Prior Close 5.223 4.705 1.110 0.283
Race

Control of -0.022 2.275 -0.009 0.993
Court

Voting -0.000 0.001 -0.182 0.858
Population

Tort Docket 0.004 0.107 0.034 0.974

Partisan -2.429 2.365 -1.027 0.320

District 0.579 3.074 0.188 0.853

MultiMember 12.584 6.237 2.018 0.061

Term 0041 0.698 -0.059 0.954

1990 2.270 4.894 0.464 0.649

1992 1.239 3.056 0.405 0.691

1994 0.049 3.167 0.015 0.988

1996 2.730 5.539 0.493 0.629

1998 -2.024 4.696 -0.431 0.672

Constant 75.703 17.924 4.223 0.001

 

Dependent variable: percentage of vote for incumbent
Mean of dependent variable: 55.87

N = 63

F(15, 16) = 30.66
Prob > F = 0.000
R2 = 0.3863

Root MSE = 8.997

126

Just as in Table 6.3, the loser spending variable remains significant, indicating
that campaign spending by the losers of open seats costs the winner electoral support.
For every 1% increase in spending by the losing candidate, the winner’s percentage of the
vote decreases by almost 0.02%. A challenger would need to spend an additional 5% to
raise her percentage of the vote 0.1%. Additionally, running in a multi-member district
increases the percentage of the vote received by the winner. Winners receive over 12.5%
more of the vote when they run in multi-member districts. The quality of the candidates,
degree of partisan consonance, state supreme court electoral context, presence of primary
competition, partisan control of the court, and size of the voting age population are all
insignificant. Further, with the exception of multi-member districts discussed above,
none of the institutional differences attain significance. Neither the type of election,
electoral constituency, or term of office is significantly related to competition for Open
seat state supreme court races.

In Table 7.5, the results from the non-institutional and the institutional models of

open seat races are compared.

127

Table 7.5: Comparison of Non-Institutional and Institutional Models of Open Seat
Contests

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Table 6.3 Table 7.4
Log of Winner Spending NS NS
Log of Loser Spending - -
Quality Winner NS NS
Quality Loser NS NS
Partisan Consonance NS NS
Comete - NS
Prior Close Race NS NS
Control of Court NS NS
Voting Population NS NS
Tort Docket NS NS
Partisan NS
District NS
MultiMember +
Term NS
1990 NS NS
1992 NS NS
1994 NS NS
1996 NS NS
1998 NS NS
+ significant in a positive direction

significant in a negative direction
not significant

NS

128

For the most part, the results are the same between the two models. Loser
spending is significant in both models. However, the presence of primary competition,
significant in the non-institutional model, loses its significance when institutional
variables are included. Additionally, the presence of a multi-member race also affects
electoral competition. Thus, as for incumbent-challenger contests, institutional variables
need to be included in models of state supreme court elections in order to fully

understand the dynamics of these elections.

The Role of Money and Institutions in State Supreme Court Elections: Selection
System, Electoral Constituency, and Term of Office

Now that we have seen how campaign spending affects electoral competition, the
next logical inquiry is to ascertain whether the efficacy of campaign spending varies
depending on the institutional arrangements of the states. That is, are the effects of
campaign spending different in partisan elections versus nonpartisan elections, district
versus statewide elections, and single-member versus multi-member elections? Tables
7.6, 7.7, 7.8, and 7.9 show the results of the interactive models for incumbent-challenger

races.

129

Table 7.6: The Effects of Money on Votes, Incumbent-Challenger Races with Selection
System Interaction Terms

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. 2 P > |z|
Error

Log of Incum. 0.104 1.386 0.075 0.940
Spending

Log of Chall. -0.920 0.368 -2.500 0.012
Spending

Log of Incum. -0.327 1.478 -0.221 0.825

Spen. x Part.

Log of Chall. -1.253 0.317 -3.952 0.000

Spen. x Part.

Quality Chall. -3.465 2.020 -1.715 0.086

Partisan 0.056 1.072 0.052 0.959
Consonance

Age 0004 0.001 -3.112 0.002

Appointed -3.288 1.056 -3.113 0.002
First

Compete 2.384 1.034 2.306 0.021

Prior Close -4.499 1.484 —3.031 0.002
Race

Control of 0.994 1.553 0.640 0.522
Court

Voting 0.000 0.000 1.187 0.235
Population

Murder Rate 0.206 0.312 0.660 0.509

Tort Docket 0.175 0.056 3.124 0.002

Partisan 13.591 16.686 0.815 0.415

District -6.360 3.172 -2.005 0.045

MultiMember -0.065 2.972 -0.022 0.983

Term 2.467 0.775 3.183 0.001

1990 -2.756 1.984 -1.389 0.165

1992 -3.300 1.938 -1.703 0.089

1994 -1.800 2.918 —0.617 0.537

1996 -2.262 3.036 -0.745 0.456

1998 4.338 2.157 2.011 0.044

Constant 48.360 15.622 3.096 0.002

Dependent variable: percentage of vote for incumbent

Mean of dependent variable: 56.31

Number of observations = 392

Censored observations = 265

Uncensored observations = 127

Log likelihood = ~612.341

x2 (6) = 15.15

Prob > x2 0.019

130

Table 7.7: The Effects of Money on Votes, Incumbent-Challenger Races with District
Interaction Terms

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. z P > |z|
Error

Log of Incum. —0.202 0.887 -0.227 0.820
Spending

Log of Chall. -1.488 0.508 -2.929 0.003
Spending

Log of Incum. 1.803 4.498 0.401 0.689

Spen. x Dist.

Log of Chall. -1.492 1.611 -0.926 0.355

Spen. x Dist.

Quality Chall. -3.753 2.072 -l.811 0.070

Partisan -0.434 1.042 -0.416 0.677
Consonance

Age 0006 0.002 -3.062 0.002

Appointed -3.177 0.987 —3.219 0.001
First

Compete 1.758 1.097 1.603 0.109

Prior Close —4.847 1.554 -3.118 0.002
Race

Control of 0.294 1.656 0.177 0.859
Court

Voting 0.000 0.000 2.044 0.041
Population

Murder Rate 0.338 0.297 1.139 0.255

Tort Docket 0.164 0.061 2.676 0.007

Partisan -4.450 2.455 -1.812 0.070

District -11.397 39.961 -0.285 0.775

MultiMember 0.076 2.600 0.029 0.977

Term 2.910 0.794 3.666 0.000

1990 -4.940 2.009 -2.459 0.014

1992 -4.754 2.163 -2.198 0.028

1994 -3.330 3.038 -1.096 0.273

1996 -3.878 3.233 -1.200 0.230

1998 3.830 2.266 1.690 0.091

Constant 56.120 9.103 6.165 0.000

Dependent variable: percentage of vote for incumbent

Mean of dependent variable: 56.31

Number of observations = 392

Censored observations = 265

Uncensored observations = 127

Log likelihood = -618.112

x2 (6) 20.44

Prob > x2 0.002

131

Table 7.8: The Effects of Money on Votes, Incumbent-Challenger Races with Multi-
Member Interaction Terms

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. 2 P > |z|
Error

Log of Incum. -0.558 0.933 -0.598 0.550
Spending

Log of Chall. -1.469 0.505 -2.909 0.004
Spending

Log of Incum. 12.081 6.056 1.995 0.046

Spen. x Mult.

Log of Chall. 0.962 5.475 0.176 0.861

Spen. x Mult.

Quality Chall. -4.052 2.217 -1.828 0.068

Partisan -0.293 1.073 -0.273 0.785
Consonance

Age 0006 0.002 -3.068 0.002

Appointed -3.310 1.085 ~3.051 0.002
First

Compete 1.468 1.006 1.460 0. 144

Prior Close -4.487 1.452 -3.090 0.002
Race

Control of 0.176 1.703 0.103 0.918
Court

Voting 0.001 0.000 2.210 0.027
Population

Murder Rate 0.451 0.301 1.495 0.135

Tort Docket 0.194 0.056 3.465 0.001

Partisan -5.481 2.518 -2.177 0.030

District 6385 3.255 -1.961 0.050

MultiMember -165.843 126.471 -1.31 1 0.190

Term 2.742 0.760 3.609 0.000

1990 -4.284 1.959 -2.187 0.029

1992 -4.690 2.294 -2.044 0.041

1994 —3.255 3.006 -l.083 0.279

1996 -3.365 3.126 -1.077 0.282

1998 4.690 2.149 2.183 0.029

Constant 59.344 9.650 6.149 0.000

Dependent variable: percentage of vote for incumbent

Mean of dependent variable: 56.31

Number of observations = 392

Censored observations 2 265

Uncensored observations = 127

Log likelihood = -617.305

x2 (6) = 26.28

Prob > x2 = 0.000

132

Table 7.9: The Effects of Money on Votes, Incumbent-Challenger Races with Term

Interaction Terms

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. z P > |z|
Error

Log of Incum. -6.214 3781 -1.644 0.100
Spending

Log of Chall. 7.061 3.720 1.898 0.058
Spending

Log of Incum. 1.020 0.566 1.801 0.072

Spen. x Term

Log of Chall. -l .421 0.572 -2.484 0.013

Spen. x Term

Quality Chall. -3.293 1.975 -1.667 0.095

Partisan -0.583 0.967 -0.603 0.546
Consonance

Age -0.006 0.002 -3.180 0.001

Appointed -2.529 1.033 -2.450 0.014
First

Compete 2.222 1.066 2.084 0.037

Prior Close -4.535 1.572 -2.884 0.004
Race

Control of -0.492 1.482 -0.332 0.740
Court

Voting 0.001 0.000 2.301 0.021
Population

Murder Rate 0.479 0.317 1.514 0.130

Tort Docket 0.180 0.060 3.021 0.003

Partisan -6.222 2.990 -2.081 0.037

District -7.209 4.027 -1.790 0.073

MultiMember -0.101 2.392 -0.042 0.966

Term 7.285 6.399 1.138 0.255

1990 -5.818 2.121 -2.743 0.006

1992 -5.806 2.348 -2.473 0.013

1994 -4.283 3.042 -1.408 0.159

1996 -4.369 3.162 -1.382 0.167

1998 4.281 2.239 1.912 0.056

Constant 27.884 41.957 0.665 0.506

Dependent variable: percentage of vote for incumbent

Mean of dependent variable: 56.31

Number of observations = 392

Censored observations = 265

Uncensored observations 2 127

Log likelihood = -615.045

)6- (6) 21.12

Prob > x2 = 0.002

133

Table 7.6 indicates that campaign spending by the challenger serves to promote
electoral competition regardless of whether the race is partisan or nonpartisan. However,
it has more of an effect in nonpartisan races. An additional 1% increase in spending by
the challenger in nonpartisan elections leads to a decrease of 0.013% in the incumbent’s
percentage of the vote; in partisan elections, the same increase in spending corresponds to
a 0.009% decrease in the incumbent’s electoral support. To increase her level of electoral
support 0.1%, a challenger in a nonpartisan election would need to increase her spending
7.7%, while she would need to raise her level of spending by 11.1% in partisan elections.
Table 7.7 indicates the importance of including a multiplicative term that takes into
account whether the election occurs in a district or statewide. The results indicate that the
results for challenger spending discussed in Table 7.2 only apply to statewide elections.
Specifically, a 1% increase in challenger spending corresponds to a 0.014% decrease in
the incumbent’s level of electoral support in statewide elections. If a challenger spent
$100,000, she would need to spend $107,000 to increase her percentage of the vote 0.1%
(a 7% increase in spending). Indeed, campaign spending has no effect on competition in
district-based elections, regardless of which candidate is spending the money. This may
be because there are other ways for a candidate to publicize her candidacy (such as going
door-to-door or meeting people at shopping centers) without spending money, since
districts are more geographically compact.

In Table 7.8 we see that challenger spending is only effective at promoting
competition in single-member districts. An additional 1% of spending by the challenger
will cost the incumbent 0.015% of her percentage of the vote. This means the challenger

would have to increase her spending by 6.7% to increase her level of electoral support by

134

6.7%. Curiously, incumbent spending serves to decrease competition in multi-member
districts, but not in single-member districts. Incumbents running in multi-member
districts can increase their percentage of the vote by 0.12% simply by spending an
additional 1%. To increase her electoral support by 0.1%, the incumbent only has to
spend an additional 0.8%. This seems to indicate that when more than one seat is at
stake, incumbents can increase their level of electoral support simply by spending more
money. One reason for this might be that since there are several candidates vying for a
limited number of seats, incumbents need to spend money to make themselves stand out
from the other candidates. Rerninding people which candidate is the incumbent serves to
distinguish the incumbent from every other candidate, and thus provides her with an
electoral advantage.

In Table 7.9, we see the results when campaign spending is interacted with the
term of office. Interpreting the effects of term of office on spending is not
straightforward. Interpreting interaction terms with interval-level data requires one to
determine the partial effect of the variables on the dependent variable (Friedrich 1982;
Wooldridge 2000). For example, suppose that the model is

Vote total = [30 + fllspending + figterm + [33spending *term + u
the partial effect of spending on vote total (holding all other variables constant) would be
[3; + B3term. By substituting interesting values of term into the equation, we can
ascertain the effects of spending, conditional on the term of Office, on vote total. Then, in
order to test for statistical significance, we need to substitute (term — X) *spending for the

interaction term, and re-run the regression, where X is the relevant value of term

135

(Wooldridge 2000). In Table 7.10, the effects of different terms of office on campaign

spending are displayed.

136

Table 7.10: The Effects of Term Length on Spending, Incumbent-Challenger Contests

 

 

 

Term Length Incumbent Spending Challenger Spending
6 -0.095 -1.465***
(0.940) (0.492)
8 1.945 -4.307***
(1.243) (0.492)
10 3.984* -7.149***
(2.185) (2.065)
12 6024* -9.990***
(3.251) (3.194)
*** = p < 0.01
** = p < 0.05
* = p < 0.10

137

The interpretation of these coefficients is as follows: “At a term length of X
years, campaign spending by the (incumbent/challenger) has a statistically
(significant/insignificant) impact on the percentage of the vote received by the
incumbent.” We can see that challenger spending promotes competition regardless of the
term of office at stake. Further, the longer the term of office, the greater the impact
challenger spending has on promoting electoral competition. For a six-year term, a 1%
increase in challenger spending leads to a 0.014% decrease in the incumbent’s percentage
of the vote, other things being equal. This same increase in spending leads to a 0.043%
decrease for a 8-year term of office, 0.071% for a 10-year term, and 0.099% for a 12-year
term of office. Thus, to increase her percentage of the vote by 0.1%, a challenger who has
spent $100,000 will need to spend $107,000 for a six-year term, $102,300 for an eight-
year term, $101,400 for a ten-year term, and only $101,010 for a twelve—year term. The
longer the term, the more effective the campaign spending of the challenger.

Incumbent spending seems to be significant only when longer terms of office are
at stake, and even then it is only marginally the case. It is also important to note that
when one looks at the relative ability of the incumbent to increase her electoral support, it
is clear that challenger spending is still more effective. For 8—year terms of office, a 1%
increase in incumbent spending increases her percentage of the vote by 0.04%, while the
same 1% of additional spending by the challenger decreases the incumbent’s electoral
support by 0.071%. Further, since incumbents generally spend more money than
challengers (see Chapter 4), an additional 1% of challenger spending is, in general, much
less actual spending than an additional 1% of incumbent spending. Taking the average

amount of spending for both incumbents and challengers from Chapter 4 (Table 4.8) in

138

the State of Washington illustrates this point nicely. An additional 1% of spending by an
incumbent in Washington is $803.29 ($80,329 x 0.01), and this would correspond to a
0.04% increase in her level of electoral support. However, for a challenger to decrease
the incumbent’s percentage of the vote by 0.071%, she would only have to spend $467.15
($46,715 x 0.01). Thus, while incumbent spending in marginally significant when longer
terms of office are stake, we should not make too much of this result.

Now, we turn to open seat races. Tables 7.11, 7.12, 7.13, and 7.14 display the

results from the interactive models.

139

Table 7.11: The Effects of Money on Votes, Open Seats with Selection System

Interaction Terms

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. t P > It]
Error

Log of Winner -3.133 2.344 -1.337 0.200
Spending

Log of Loser -1.919 1.236 -1.553 0.140
Spending

Log of Winner 4.292 2.645 1.623 0.124

Spen. x Part.

Log of Loser -0.237 1.312 -0.181 0.859

Spen. x Part.

Quality 0.615 2.402 0.256 0.801
Winner

Quality Loser 2.296 3.961 0.580 0.570

Partisan -0.489 1.326 -0.369 0.717
Consonance

Compete -3.044 3.433 -0.887 0.388

Prior Close 6.540 4.953 1.320 0.205
Race

Control of -0. 109 1.929 -0.057 0.956
Court

Voting -0.000 0.001 -0.070 0.945
Population

Tort Docket -0.042 0.107 -0.394 0.699

Partisan —53.051 36.251 -1.463 0.163

District 0.797 2.969 0.268 0.792

MultiMember 1 1.009 6.126 1.797 0.091

Term 0.287 0.671 0.427 0.675

1990 1.686 4.766 0.354 0.699

1992 0.301 2.680 0.112 0.912

1994 -0.481 3.377 -0.142 0.889

1996 2.645 5.749 0.460 0.652

1998 -3.191 4.958 -0.644 0.529

Constant 113.793 33.764 3.370 0.004

 

Dependent variable: percentage of vote for winner
Mean of dependent variable: 55.87

N = 63

F (15, 16) = 27.97
Prob > F = 0.0000
R2 = 0.4125

Root MSE = 9.015

140

Table 7.12: The Effects of Money on Votes, Open Seats with District Interaction Terms

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. t P > It]
Error

Log of Winner -1.832 1.370 -1.338 0.200
Spending

Log of Loser -2.049 0.441 4.646 0.000
Spending

Log of Winner 5.713 2.099 2.722 0.015

Spen. x Dist.

Log of Loser 3.881 1.364 2.845 0.012

Spen. x Dist.

Quality 3 .406 2.122 1.605 0.128
Winner

Quality Loser 1.336 3.300 0.405 0.691

Partisan -1.384 1.385 -0.999 0.333
Consonance

Compete -2.330 3.300 -0.706 0.490

Prior Close 5.228 4.013 1.303 0.211
Race

Control of 2.029 1.904 1.065 0.303
Court

Voting -0.000 0.000 -0.418 0.682
Population

Tort Docket 0.089 0.088 1.007 0.329

Partisan -1.836 1.855 -0.990 0.337

District -118.139 23.006 -5. 135 0.000

MultiMember 12.451 5.378 2.315 0.034

Term 0.536 0.540 0.993 0.336

1990 4.433 4.043 1.096 0.289

1992 3.927 5.532 0.710 0.488

1994 1.056 3.563 0.297 0.771

1996 0.179 4.727 0.038 0.970

1998 -2.211 4.558 -0.485 0.634

Constant 92.874 12.406 7.486 0.000

 

Dependent variable: percentage of vote for winner
Mean of dependent variable: 55.87

N=63
F (15, 16) = 99.79
Prob > F = 0.0000
R2 = 0.5149

Root MSE = 8.192

141

Table 7.13: The Effects of Money on Votes, Open Seats with Multi-Member Interaction

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Terms
Variable Coefficient Robust Std. t P > M
Error

Log of Winner 0.994 1.783 0.558 0.585
Spending

Log of Loser -1.891 0.530 -3.568 0.003
Spending

Log of Winner -15.760 16.446 -0.958 0.352

Spen. x Mult.

Log of Loser 6.269 14.434 0.434 0.670

Spen. x Mult.

Quality 1.436 2.074 0.692 0.499
Winner

Quality Loser 1.014 3.496 0.290 0.776

Partisan -1.027 1.134 —0.905 0.379
Consonance

Compete -2.357 2.992 -0.788 0.442

Prior Close 2.967 4.010 0.740 0.470
Race

Control of -1.058 2.406 -0.440 0.666
Court

Voting 0.000 0.000 0.487 0.633
Population

Tort Docket 0.005 0.095 0.050 0.960

Partisan -2.773 2.534 -1.094 0.290

District 2.339 3.398 0.688 0.501

MultiMember 143.083 85.546 1.673 0.114

Term 0349 0.832 -0.419 0.681

1990 2.658 4.927 0.539 0.597

1992 1.553 2.952 0.526 0.606

1994 0.517 3.000 0.172 0.865

1996 4.411 5.304 0.832 0.418

1998 0720 4.383 -0.164 0.872

Constant 66.229 21.500 3 .080 0.007

 

Dependent variable: percentage of vote for winner
Mean of dependent variable: 55.87

N = 63

F (15, 16) = 61.56
Prob > F = 0.0000
R2 = 0.4133

Root MSE = 9.001

142

Table 7.14: The Effects of Money on Votes, Open Seats with Term Interaction Terms

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. t P > M
Enor

Log of Winner -9.233 4.215 -2.191 0.044
Spending

Log of Loser -4.449 1.727 -2.577 0.020
Spending

Log of Winner 1.201 0.481 2.499 0.024

Spen. x Term

Log of Loser 0.410 0.268 1.531 0.145

Spen. x Term

Quality 1.753 1.928 0.909 0.377
Winner

Quality Loser 1.070 3.569 0.300 0.768

Partisan -0.727 1.089 -0.668 0.514
Consonance

Compete -2.891 3.027 -0.955 0.354

Prior Close 6.244 4.197 1.488 0.156
Race

Control of 1.657 2.070 0.801 0.435
Court

Voting -0.000 0.001 -0.843 0.412
Population

Tort Docket 0.080 0.069 1.151 0.266

Partisan -0.748 2.454 -0.305 0.764

District -0.055 2.635 -0.021 0.984

MultiMember 8.131 5.846 1.391 0.183

Term -20.076 5.627 -3.568 0.003

1990 3.326 4.640 0.717 0.484

1992 1.912 3.787 0.505 0.620

1994 0.887 3.356 0.264 0.795

1996 1.795 5.096 0.352 0.729

1998 -4.483 5.198 -0.862 0.401

Constant 220.766 42.966 5.138 0.000

 

Dependent variable: percentage of vote for winner

Mean of dependent variable: 55.87

N=63

F (15, 16): 38.38

Prob > F = 0.0000

R2 = 0.4472

Root MSE = 8.745

143

The results for open seats, like those for incumbent-challenger contests, show that
institutions can affect the influence of campaign spending on the electoral support of the
winner. In Table 7.11, we see that the influence of campaign spending on the outcome of
open seat elections is not contingent on selection system. Indeed, campaign spending
does not appear to be significant at all (although a likely reason for this is the collinearity
introduced by the interaction term; campaign spending by the losing candidate—Table
7.4—likely still matters).

Table 7.12 displays the results for district-based elections interactive model.
Competition is promoted by campaign spending by the loser in statewide elections. A
1% increase in campaign spending by the loser leads to a 0.02% decrease in the winner’s
percentage of the vote, other things being equal. The losing candidate can increase her
percentage of the vote by 0.1% if she spends 5% more. However, this effect is reversed
in district-based elections. Indeed, campaign spending by either the winner or the loser
will serve to decrease electoral competition in district-based elections. A 1% increase in
the spending by the winner increases her level of electoral support by 0.057%, while the
same increase in spending by the loser increases the winner’s percentage of the vote by
0.039%. This result is contrary with what was found for incumbent-challenger contests:
campaign spending by neither the incumbent nor the challenger significantly affected the
incumbent’s percentage of the vote, for both partisan and nonpartisan elections). This is
further evidence that open seat races are governed by different dynamics than incumbent-
challenger contests.

In Table 7.13, we see that campaign spending by the loser in single-member

districts significantly decreases the electoral support of the winner, and thus promotes

144

competition. Specifically, the winner’s percentage of vote can be reduced by 0.019% if
the loser increases her spending by 1%. Likewise, the winner’s percentage of the vote
can be reduced by 0.1% if the loser increases her spending by 5.3%. Campaign spending
has no affect on competition in multi—member elections.

Table 7.14 looks at the term of office. The effects term length has on campaign

spending are shown in Table 7.15.

145

Table 7.15: The Effects of Term Length on Spending, Open Seat Races

 

 

 

Term Length Winner Spending Loser Spending
6 -2.028 -1.998***
(1.564) (0.388)
8 0.374 -1.168*
(1.017) (0.596)
10 2.776*** -0.347
(1.213) (1.064)
12 5.177*** 0.473
(1.938) (1.577)
*** = p < 0.01
** = p < 0.05
* = p < 0.10

146

The results presented in Table 7.15 are interesting. While campaign spending by
the losing candidates promotes competition with shorter terms of office (six and eight
years), campaign spending by the winner serves to decrease competition when longer
terms of office are at stake (ten and twelve years). This could be a result of the electoral
dynamics in the state with longer terms of office (Illinois, Louisiana, Pennsylvania, and
Wisconsin), or it could be the case that spending retards competition when longer terms
of office are at stake (although there is no theoretical reason to think this is the case).

Overall, the general lesson from these models is clear: institutions affect how
campaign spending can influence electoral support. Some institutional arrangements
promote competition (single-member districts) while others retard it. And, under other
institutional arrangements (district-based elections, longer terms of office), the effects

depend upon the type of race.

The Role of Independent Expenditures

Tables 7.16 and 7.17 show the results of the models estimated in Tables 7.2 and

7.4, omitting the races that contained large amounts of independent expenditures.

147

Table 7.16: The Effects of Money on Votes, Incumbent-Challenger Contests, with
Independent Expenditure Races Omitted

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. 2 P > |z|
Error

Log of Incum. -0.191 0.999 —0.191 0.848
Spending

Log of Chall. -1.347 0.530 -2.543 0.011
Spending

Quality Chall. -4.901 2.045 -2.396 0.017

Partisan -0.437 1.157 -0.377 0.706
Consonance

Age -0.006 0.002 -2.817 0.005

Appointed -3.975 1.372 -2.898 0.004
First

Compete 1.434 1.080 1.327 0.184

Prior Close -5.551 1.950 -2.847 0.004
Race

Control of 0.618 2.135 0.289 0.772
Court

Voting 0.001 0.000 2.239 0.017
Population

Murder Rate 0.397 0.374 1.061 0.289

Tort Docket 0.144 0.070 2.066 0.039

Partisan -5.564 3.521 -1.580 0.114

District -4.374 4.039 -1.083 0.279

MultiMember -0.986 3.049 -0.323 0.746

Term 3.019 0.886 3.409 0.001

1990 -6.019 2.592 -2.322 0.020

1992 -5.427 3.286 -1.652 0.099

1994 -4.149 3.466 -1.197 0.231

1996 -4.928 3.434 —1.435 0.151

1998 2.887 2.560 1.128 0.259

Constant 55.689 10.593 5.257 0.000

Dependent variable: percentage of vote for incumbent

Mean of dependent variable: 54.94

Number of observations = 380

Censored observations = 265

Uncensored observations = 115

Log likelihood = -568.652

x2 (6) = 17.34

Prob > x2 = 0.008

148

Table 7.17: The Effects of Money on Votes, Open Seats, with Independent Expenditure
Races Omitted

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust Std. t P > M
Error

Log Of Winner 0.119 1.533 0.078 0.939
Spending

Log of Loser -1.928 0.490 -3.933 0.001
Spending

Quality 0.671 2.226 0.302 0.767
Winner

Quality Loser 1.557 3.400 0.458 0.653

Partisan -0.685 1.071 -0.640 0.531
Consonance

Compete -2.816 2.684 - 1.049 0.310

Prior Close 5.168 5.274 0.980 0.342
Race

Control of -0.057 2.424 -0.024 0.981
Court

Voting -0.000 0.001 -0.114 0.910
Population

Tort Docket 0.003 0.108 0.024 0.982

Partisan -2.531 3.173 -0.798 0.437

District 0.646 3.452 0.187 0.854

MultiMember 12.571 6.371 1.973 0.066

Term 0021 0.732 -0.028 0.978

1990 2.366 5.264 0.450 0.659

1992 1.286 3.025 0.425 0.676

1994 0.102 3.352 0.030 0.976

1996 2.810 6.078 0.462 0.650

1998 -1.912 5.885 -0.325 0.749

Constant 75.663 18.259 4.144 0.001

 

Dependent variable: percentage of vote for winner
Mean of dependent variable: 55.88

N = 61

F (15, 16) = 37.16
Prob > F = 0.0000
R2 = 0.3864

Root MSE = 9.212

149

Omitting those races that involved a lot of independent expenditures does not
affect the efficacy of campaign spending in incumbent-challenger races, as seen in Table
7.16. Campaign spending by the challenger continues to promote electoral competition,
while spending by the incumbent remains statistically insignificant. Finally, all of the
other variables significant in Table 7.2 are significant in Table 7.16, with two exceptions.
In the results for all races, competition was promoted by an election occurring in a district
and in a partisan state. These variables are not significant when races involving a large
amount of independent expenditures are omitted. This suggests that the results for the
district variable in Table 7 .2 are driven by elections in Mississippi (since it is the only
state with both district elections and a large amount of independent expenditures), while
Alabama is a prime determinant of the significance of the partisan variable achieving
significance, since it is the only state with both partisan elections and a large amount of
independent expenditures (although the significance of the partisan variable is marginal
in Table 7.2).

As Table 7.17 shows, excluding those races heavily infused with independent
expenditures does not significantly affect the results for open seat races. The spending by
the losing candidate continues to promote competition, while the spending by the
winning candidate does not affect her percentage of the vote. Further multi-member open
seat races contribute to less electoral competition. However, there is one difference:
when races with a large amount of independent expenditures are excluded, the fact that
there was a close race in the most recent state high court election loses its significance.
This suggests that previous close races may be an indication that there will be a large

amount of independent expenditures in subsequent races.

150

Overall, consistent with the results obtained in Chapter 6, excluding those
elections with a lot of independent expenditures does not significantly affect the

conclusions reached about the efficacy of campaign spending.

Conclusion

This chapter sought to ascertain the affects of institutional variables on
competition and on the efficacy of campaign spending. The institutional variation
present in state supreme court elections makes them ideal laboratories for studying the
effects of institutions on political processes.

In sum, one cannot escape the conclusion that institutions matter. Not only do
institutional arrangements structure competition (either by promotion or retardation), but
they also affect the efficacy of campaign spending. The results suggest that those who
lament the lack of competitive races for the state high court bench, or who are concerned
with the effects of campaign spending can alleviate their concerns by changing

institutional arrangements. More will be said about this in Chapter 8.

151

CHAPT ER 8

CONCLUSION

This dissertation has explored the role of campaign spending in promoting (or
retarding) electoral competition in state supreme court elections over a ten-year period.
The results of the empirical analysis suggest that many factors influence both campaign
spending and electoral competition in elections to the state high court bench:
characteristics of the candidates, the state and electoral context, campaign spending, and
institutional factors. In this chapter, I conclude by tying together the empirical results
with the theory developed in Chapters 1 and 2, as well as talk about the implications of

these results and directions for future research.

Summary of Results

This project was one of the first to examine state supreme court elections
systematically (but see Dubois 1980; Bonneau 2001b; Hall 2001a). In Chapter 4, we saw
that both competition and campaign spending varied by type of selection system, by type
of race, by state, and over time. In general, partisan elections were more competitive and
more expensive than nonpartisan elections. Incumbent-challenger contests in partisan
elections were more competitive than their open seat counterparts, while this was
reversed in nonpartisan states.

Chapter 5 examined the determinants of total campaign spending for a state
supreme court seat. Higher levels of campaign spending were found to be related to the

margin of victory for the winning candidate, the number of state supreme court seats on

152

the ballot, whether the election was held in a partisan or nonpartisan state, the length of
the term of office, and the percentage of the docket occupied by tort cases. Specifically,
the lower the margin of victory, the fewer the number of seats on the ballot, the higher
the percentage of the docket occupied by tort cases, and elections held in nonpartisan
states and for longer terms of office all lead to higher amounts of campaign spending by
the candidates, other things being equal.

In Chapter 6, I sought to ascertain the effects campaign spending had on the
percentage of the vote received by the incumbent (or, in the case of open seats, the
percentage of the vote received by the winner). However, I did not include any of the
institutional variables in these models. Hence, these models treated all state supreme
court elections the same. The principal reason for doing this was so that there was a
baseline model against which the institutional models could be compared. That is, I
wanted to estimate the models without institutional variables to better show the effects of
institutions when they were included in subsequent models.

The central finding of Chapter 6 was that, consistent with the congressional and
state legislative literature, challenger spending served to decrease the electoral support of
the incumbent, and spending by the losing candidate decreased the winner’s percentage
of the vote. That is, competition was promoted by campaign spending by certain
candidates. Additionally, competition was promoted in incumbent-challenger races by
the presence of a quality challenger, the age of the incumbent, whether the incumbent
was initially appointed to the bench, and if there was a prior close state supreme court
election. The size of the voting age population served to decrease competition, indicating

that the incumbency advantage is magnified in larger states. For open seat elections, in

153

addition to campaign spending by the loser, if the winner was challenged in a primary,
she was more likely to receive a lower percentage of the vote, other things being equal.
These results hold even if races that involved a lot of independent expenditures are
omitted, as well as under different model specifications and estimation techniques.

The extent to which this central result held in different institutional circumstances
was the subject of Chapter 7. The differences between the non-institutional models of
incumbent-challenger contests were displayed in Table 7.3. In addition to the same
variables attaining significance that did in the non-institutional models, in the institutional
model the percentage of the docket occupied by tort cases decreases electoral
competition, while partisan races and district—based elections promote electoral
competition. In terms of open seats, in addition to loser spending retaining its
significance, multi—member districts tend to discourage competition. Further, there was
no difference in campaign spending between partisan and nonpartisan elections.

Looking at the effects of institutional variables on the efficacy of campaign
spending yielded mixed results. Challenger spending promoted competition in both
partisan and nonpartisan races, as well as in statewide races, single-member districts, and
for longer terms of office. For open seats, campaign spending by the loser led to more
competition in statewide races, as well as elections that occurred in single—member
districts. In district-based elections, campaign spending by both the winner and the loser
served to decrease competition. Further, challenger spending served to promote
competition when shorter terms of office awaited the victor, while incumbent spending
served to decrease competition when longer terms of office were at stake, likely for the

reasons discussed in Chapter 7.

154

In sum, the results suggest that campaign spending will have the greatest
influence in promoting competition in statewide, single-member district elections,
regardless of the type of election (partisan or nonpartisan) or whether the election is an
incumbent-challenger race or a race for an open seat. Also, for incumbent-challenger
races, the longer the term of office at stake, the more effective campaign spending will be

in promoting competition.

The Importance of Electoral Competition

In Chapter 1, I discussed the role of competition in electoral and democratic
theory. Electoral competition is important because, by its very nature, it serves as a
constraint on the actions of elected officials. Regardless of the level of knowledge or the
participation of the electorate, electoral competition brings elected officials in line with
the preferences of their constituents (e. g., Squire 1989a). The same holds true for state
supreme court justices: electoral competition affects not only judicial behavior in terms
of dissent rates, but also affects the outcome of cases (Brace and Hall 1990, 1995, 1997;
Hall 1987, 1992, 1995). Thus, electoral competition plays a key role in understanding
political life.

The link between electoral competition and judicial behavior cannot be
understated. Consider three justices sitting on a state supreme court. These justices are
similar in their ideology (liberal) in a conservative state. Justice A just narrowly won her
reelection with 51% of the vote; Justice B is up for reelection in two years; and Justice C
does not face the voters for another four years. Electoral competition will affect the

judicial behavior of these judges, each in a different way. Suppose there is a case before

155

the court in which the justices are inclined to support one position, while the electorate
supports the opposite position. Justice A, given that she has just narrowly won election,
may be more inclined to support a position different than her own preferred position, and
consistent with the electorate that just returned her to office. Justice B, coming up for
election in two years and having just seen the competitive race that Justice A was in, is
unlikely to take a position on a controversial issue that is at odds with the preferences of
the electorate if she hopes to retain her seat. Finally, Justice C is least likely to be
affected by electoral politics. Even though she has to face the electorate to retain her
seat, this election is not for another four years. She does not face any immediate sanction
for her position.

The above example illustrates how electoral politics can affect different justices
sitting on the same high court. Electoral politics can also affect state supreme court
outcomes across states. Justices in states with little electoral competition and longer
terms of office can act in ways that are not consonant with the electorate with more
frequency than justices in states with higher levels of competition and shorter terms of
office. Thus, we would expect state supreme court decisions in states with shorter terms
and higher levels of competition to be more congruent with the preferences of the
electorate than state supreme court decisions in states with longer terms of office and
lower levels of electoral competition. Clearly, the presence (or absence) of electoral
competition plays a vital role in the behavior of state supreme court justices. This makes
understanding the factors that influence electoral competition imperative to

understanding the politics of state supreme courts.

156

State Supreme Court Election Reform

Of course, all of this relates back to the independence versus accountability
debate discussed in Chapter 1. The results suggest that under some circumstances
campaign spending promotes state supreme court justices being held accountable by the
electorate. That is, campaign spending is not necessarily a negative aspect of state
supreme court elections. While it is true that raising and spending money raises the
appearance of impropriety in some cases, the acts of raising and spending money also
promote electoral competition, at least when this is done by candidates relatively
unknown to the electorate. Given the recent United States Supreme Court decision in
Republican Party v. White, campaign spending is likely to become even more important
in promoting electoral competition as restrictions on what candidates can, and cannot say,
are lifted.

It is important to note that this dissertation does not speak to the debate over
whether judges should, or should not, be elected. Indeed, this project is agnostic about
the normative debate. Whether or not one method of judicial selection is “better” than
another is a question I do not address. I treat state supreme court elections as a given, and
seek to develop models that explain electoral competition in them. And, given that there
are state supreme court elections, the results indicate that campaign spending can lead to
judges being held more accountable by the electorate. Given the primacy of electoral
accountability to the rationale behind judicial elections, this is a significant finding.

The findings of this study point to several implications for the conduct, and

potential reform, of state supreme court elections. First, it should be noted that the

157

insignificance of incumbent spending does not mean that incumbent spending is
irrelevant or inconsequential. Surely, incumbents would have a difficult time getting
reelected if they spent no money. The finding that incumbent spending does not affect
their vote total is best interpreted in light of the advantages with which they begin the
campaign. Challenger spending is seen as a sign of strength; the more money a
challenger spends, the more viable her candidacy. Contrary to challengers, incumbent
spending is not a sign of strength; rather, since they begin the campaign with large
advantages in terms of experience, visibility, and so forth, high levels of incumbent
spending are a sign of weakness (Jacobson 1980, 1985, 1997). The more threatened an
incumbent feels (and thus the more likely it is that her race will be close), the more
money she spends. Thus, the more money an incumbent spends, the lower her vote total
is likely to be. This is not due to the amount of money spent by the incumbent, however,
but is a result of the competitive nature of the race.

The election of judges arose out of the need for courts to be independent from
legislatures. Somewhat ironically given the current climate, the legal profession hailed
the election of judges (Sheldon and Maule 1997). Now, there is growing concern about
the election of judges, with many arguing that judicial elections must either be heavily
regulated or eradicated (Schotland 1985 , 2001; Loyola of Los Angeles Law Review 2001;
Goldberg, Holman, and Sanchez 2002). As was discussed earlier, there is much debate in
state legislatures over the current climate of state supreme court elections. The results
presented here suggest that any attempt to limit the amount of money spent in these
elections will lead to less, and not more, competitive elections. Incumbent candidates are

not able to increase their level of electoral support by spending more money, but

158

challengers are; winning candidates are not able to increase their vote total, but losers are.
Those who decry the amounts of campaign spending in state supreme court elections
(Hansen 1998; Glaberson 2000a, 2000b; Loyola of Los Angeles Law Review 2001;
Goldberg, Holman, and Sanchez 2002) have overlooked the fact that campaign spending
promotes electoral competition. Campaigns matter, and campaign spending is necessary
for candidates to make themselves known to the electorate (Jacobson 1980; Coleman and
Manna 2000). This is especially true since campaign spending benefits those candidates
with whom the electorate is unfamiliar (usually nonincumbents) (Jacobson 1980; Alvarez
1997). Attempts to curb the amount of campaign spending are likely to only reinforce the
incumbency advantage, and suppress electoral competition.

The results also suggest that certain institutional reforms can promote
competition, while others will have no effect or discourage it. Arkansas is moving to
nonpartisan statewide single-member districts in 2002 (from partisan statewide single-
member districts). The results here suggest that this will lead to less competition (since
Chapter 4 indicates that nonpartisan elections are generally less competitive than partisan
elections). However, there will be more spending (Chapter 5) in these nonpartisan races
than there was in partisan races. Further, there will be no change on the influence of
campaign spending on electoral outcomes (Chapter 7). Campaign spending by the
challenger (incumbent-challenger races) and the loser (open seat contests) will continue
to adversely affect the electoral performance of the incumbent (or winner). If the
motivation behind the reform was to decrease electoral competition (and thus increase
electoral independence), then it will likely be successful; however, if the purpose for the

change in selection system was to lower the cost of these races or to decrease the

159

influence of campaign spending on the electoral support of the candidates, then the
success of the reform is doubtful.

The results also provide some comfort to those who think that the incumbency
advantage is insurmountable. For a candidate seeking to unseat an incumbent, she can
increase her vote total simply by spending more money. Additionally, there are other
factors within her control that can contribute to her success. For example, if she has prior
judicial experience, she will gain more votes, as she will if there was a recent close race
for the state high court bench, if the incumbent is older, or if the incumbent was initially
appointed to her seat on the high court bench. This suggests that challengers will perform
better if they are somewhat selective in their decisions to run (Banks and Kiewiet 1989;
Jacobson 1989; Bonneau and Hall 2000, 2001). In addition to candidate characteristics,
the state and electoral context matters, as do the institutional structures of the election. If
all else fails, however, the results here suggest that a candidate who is able to spend

freely is able to increase her vote total by a significant amount.

Future Research

While this dissertation represents the most comprehensive examination of
electoral competition and the role of campaign spending in state supreme court elections
to date, it certainly is not the end of such inquiry. There is much more to be done before
we can fully understand the dynamics of elections to the state high court bench.

First, this dissertation has not said anything about campaign contributions. In
order for candidates to spend money, they must have money. In Chapter 2, I talked a

little bit about the importance contributors play in the electoral process. Contributors

160

give money for a variety of reasons; however, very few of them give money to candidates
who are not viable. It is an open question as to what makes a potential candidate viable.
Understanding what makes candidates viable is important because knowing what factors
contributors look for in a candidate would help us understand how candidates attract
money. Further, given a candidate with certain characteristics, we would be able to
predict how much money she would be able to raise (and thus spend) as well as how well
she would do against an incumbent. Not all candidates are equal, and understanding what
makes candidates more attractive to contributors would help us assess the potential
electability of these candidates.

Second, there is the issue of how money affects judicial decision-making. We
have seen in this dissertation how campaign spending can affect the electoral support of
the candidate. In addition to concern over the influence of money on the electoral
process, there are those who are worried about the influence of money on the judicial
decision-making process. After all, if a candidate won her election due to campaign
spending (at least in part), then she may reward her contributors once she ascends to the
bench. To date, there is little evidence that justices give preferential treatment to their
contributors (Bell 2002), although anecdotal evidence of such favoritism abounds
(Campbell 1998; Robison 1998; Choate 2000a, 2000b; Farmer 2001; Marks and Hoke
2002; Popkey 2001; Scarritt 2001; Wersal 2001). At the very least, an appearance of
such favoritism exists, something that has not gone unnoticed by judges, lawyers, and
even the general public 01V ohl 2000; Glaberson 2000a; Dickerson 2001b; Marks 2001;
Marks and Hoke 2001; Wickline 2001; Phillips 2002). Ascertaining if, and to what

extent, campaign contributors are rewarded for their contributions once their candidate

161

ascends to the bench is a fundamental question that goes to the heart of the integrity of
the bench.

Unlike congressional and state legislative elections, state supreme court elections
have been relatively ignored in the political science literature despite not only their
importance, but also their analytical leverage. This dissertation has shown that both
electoral competition and campaign spending, as well as the effects of campaign
spending on electoral competition, vary according to institutional arrangements, along
with characteristics of the candidate and state. While there is more to be done, this
dissertation has shown the relevance of institutional structures on such fundamental
electoral variables as competition and campaign spending, as well as the importance of

understanding electoral competition and its consequences for the state high court bench.

162

APPENDDC

163

Table A.1: OLS results for Table 5.3

 

 

Variable Coefficient Robust z P > |z|
Std. Error

Open Seat 0.119 0.172 0.693 0.497

Appointed -0.108 0.177 -0.611 0.548
First

Margin of -0.026 0.003 -8.599 0.000
Victory

Control of 0.243 0.171 1.419 0.172
Court

Number of -0.225 0.079 -2.860 0.010
Seats

Partisan -0.030 0.281 -0.107 0.916

District 0047 0.278 -0.168 0.868

DistPart -0.062 0.366 -0.169 0.867

MultiMember -0.071 0.359 -0.198 0.845

Term 0.201 0.073 2.768 0.012

Voting Age 0.000 0.000 0.969 0.344
Population

Prior Close 0.215 0.124 1.726 0.101
Race

Tort Docket 0.055 0.005 11.875 0.000

1990 -0.746 0.219 -3.407 0.003

1992 -0.454 0.256 -1.772 0.092

1994 -0.206 0.223 -0.926 0.366

1996 -0.465 0.143 -3.242 0.004

1998 -0.032 0.135 -0.236 0.816

Constant 10.989 0.698 15.741 0.000

 

Dependent variable: log of total spending in the election
Mean of dependent variable: 12.825

N=202

F(18, 19) = 180.35
Prob > F = 0.000

R2 = 0.6078

Root MSE = 0.7921

164

 

 

 

Table A2: Hausman Test Results for Table 6.2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable 2SLS OLS ZSLS — OLS S.E.

Log of Incum. 3.093 0.425 2.668 6.732
Spending

Log of Chall. -1.578 -1.323 -0.254 0.599
Spending

Quality Chall. -4.961 -4.287 -0.674 1.737

Partisan 0.351 0.754 -0.403 1.127
Consonance

Age 0005 -0.007 0.002 0.004

Appointed -3.186 -3442 0.255 .‘
First

Compete -0.269 -0.065 -0.204

Prior Close -3.786 -4.086 0.301
Race

Control of 0.272 0.790 -0.518 1.581
Court

Voting 0.000 0.000 -0.000 0.000
Population

Murder Rate —0.201 -0.049 -0.151 0.345

Tort Docket -0.030 0.096 -0.125 0.311

1990 -0.779 -3.276 2.497 6.079

1992 -0.890 -3.691 2.801 6.171

1994 -0.718 -2.511 1.793 3.165

1996 -2.161 -3.370 1.209 1.777

1998 4.252 3.540 0.712 2.130

Constant 40.623 66.297 -25.673 64.751

 

Ho = difference in the coefficients is not systematic

x2 (14) = (B. - B.)’ (V. — V0403. - B.)
where
B, = the coefficient vector from ZSLS
[‘10 = the coefficient vector from OLS
Vt = the covariance matrix of the coefficients from ZSLS
Vo = the covariance matrix of the coefficients from OLS

x2 (16) 0.12
Prob > 12 = 1.000

 

' The missing value here is not that surprising. “This covariance matrix is guaranteed to be positive
definite only asymptotically and assurances are not made about the diagonal elements. Negative values
along the diagonal are possible (Stata 1999, 11).

165

Table A3: Hausman Test Results for Table 6.3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable 281.8 OLS 2SLS — OLS S.E.

Log of Winner 11.067 0.919 10.148 18.013
Spending

Log of Loser -2.772 -1.763 -1.009 1.706
Spending

Quality 3.212 2.113 1.099 3.443
Winner

Quality Loser -2.778 -0.823 -1.955 4.716

Partisan -3.743 -0.566 -3.177 6.202
Consonance

Compete 1.870 -4.977 6.848 11.871

Prior Close 2.828 2.786 0.042 1.583
Race

Control of 0.433 0.265 0.167 1.424
Court

Voting 0.000 0.000 -0.000 0.000
Population

Tort Docket -0.498 -0.062 -0.436 0.798

1990 4.554 2.958 1.596 4.384

1992 2.932 1.403 1.529 4.993

1994 -4.675 1.496 -6.170 10.651

1996 4.175 4.338 -0.164

1998 -2.051 -2.777 0.726 .

Constant -37.746 65.316 -103.063 184.462

 

H0 = difference in the coefficients is not systematic

x2 (12) = (B. - B.)’ (V. — vo)“(B. - B.)
where
B, = the coefficient vector from 2SLS
[3,, = the coefficient vector from OLS
V, = the covariance matrix of the coefficients from 2SLS
Vo = the covariance matrix of the coefficients from OLS

x2 (15) 0.73
Prob > x2 = 1.000

166

Table A.4: OLS Results for Table 6.2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust t P > It]
Std. Error

Log of Incum. 0.425 1.019 0.417 0.682
Spending

Log of Chall. -1.323 0.531 -2.493 0.023
Spending

Quality Chall. -4.287 2.419 -1.772 0.094

Partisan 0.754 0.958 0.787 0.442
Consonance

Age -0.007 0.002 -3.104 0.006

Appointed -3.442 1.404 -2.451 0.025
First

Compete -0.065 1.242 -0.052 0.959

Prior Close -4.086 1.642 -2.489 0.023
Race

Control of 0.790 1.455 0.543 0.594
Court

Voting 0.000 0.000 1.898 0.075
Population

Murder Rate -0.049 0.187 -0.264 0.795

Tort Docket 0.096 0.083 1.155 0.264

1990 -3.276 1.537 —2. 131 0.048

1992 -3.691 2.354 -1.568 0.135

1994 -2.51 1 2.922 -0.860 0.402

1996 -3.370 3.236 -1.041 0.312

1998 3.540 2.443 1.449 0.166

Constant 66.297 8.966 7.394 0.000

 

Dependent variable: percentage of vote for incumbent
Mean of dependent variable: 56.31

N = 127

F (16, 17) = 423.68
Prob > F = 0.000
R2 = 0.4942

Root MSE = 8.281

167

Table A5: OLS Results for Table 6.3

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust t P > M
Std. Error

Log of Winner 0.919 1.127 0.815 0.427
Spending

Log of Loser -1.763 0.556 -3.172 0.006
Spending

Quality 2.113 2.123 0.996 0.334
Winner

Quality Loser -0.823 3.277 -0.251 0.805

Partisan -0.566 1.053 -0.537 0.598
Consonance

Compete -4.977 2.866 -1.737 0.102

Prior Close 2.786 4.397 0.634 0.535
Race

Control of 0.265 2.240 0.118 0.907
Court

Voting 0.000 0.000 0.368 0.717
Population

Tort Docket -0.062 0.116 -0.539 0.597

1990 2.958 4.799 0.616 0.546

1992 1.403 3.342 0.420 0.680

1994 1.496 2.610 0.573 0.575

1996 4.338 5.287 0.820 0.424

1998 -2.777 5.506 -0.504 0.621

Constant 65.316 16.616 3.931 0.001

 

Dependent variable: percentage of vote for incumbent
Mean of dependent variable: 55.87

N=63

F (15, 16) = 131.78
Prob > F = 0.000
R2=0.3110

Root MSE=9.118

168

Table A6: OLS Results for Table 6.4

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust t P > M
Std. Error

Log of Incum. 0.251 1.102 0.228 0.822
Spending

Log of Chall. -1.156 0.544 -2.125 0.049
Spending

Quality Chall. -5.058 2.347 -2.155 0.046

Partisan 0.422 1.087 0.388 0.703
Consonance

Age 0007 0.002 -2.934 0.009

Appointed -3.835 1.651 -2.323 0.033
First

Compete -0.145 1.352 -0.107 0.916

Prior Close -5.145 2.278 -2.259 0.037
Race

Control of 1.384 2.046 0.676 0.508
Court '

Voting 0.000 0.000 1.391 0.182
Population

Murder Rate 0.087 0.285 0.304 0.765

Tort Docket 0.079 0.097 0.815 0.426

1990 -5.639 2.899 -1.945 0.068

1992 -5 .484 3.503 -1.565 0.136

1994 -4.597 3.661 -1.256 0.226

1996 -5.304 3.761 -1.410 0.177

1998 1.887 3.255 0.580 0.570

Constant 69.074 10.432 6.621 0.000

 

Dependent variable: percentage of vote for incumbent
Mean of dependent variable: 56.31

N = 115

F (16, 17) = 1914.66
Prob > F = 0.000

R2 = 0.4874

Root MSE = 8.465

169

Table A.7: OLS Results for Table 6.5

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust t P > M
Std. Error

Log of Winner 0.927 1.103 0.840 0.413
Spending

Log of Loser -1.762 0.563 -3.130 0.006
Spending

Quality 2.118 2.160 0.980 0.341
Winner

Quality Loser -0.794 3.048 -0.260 0.798

Partisan -0.571 1.030 -0.554 0.587
Consonance

Compete -4.949 2.757 -1.795 0.092

Prior Close 2.844 5.087 0.559 0.584
Race

Control of 0.268 2.278 0.118 0.908
Court

Voting 0.000 0.000 0.317 0.756
Population

Tort Docket -0.061 0.108 -0.564 0.580

1990 2.871 5.426 0.529 0.604

1992 1.336 3.839 0.348 0.732

1994 1.426 3.268 0.436 0.668

1996 4.274 6.004 0.712 0.487

1998 -2.855 6.611 -0.432 0.672

Constant 65.200 16.472 3.958 0.001

 

Dependent variable: percentage of vote for incumbent
Mean of dependent variable: 55.87

N = 61

F (15, 16) = 131.780
Prob > F = 0.000

R2 = 0.3111

Root MSE = 9.317

170

Table A8: Chow Test Results for Table 7.2

 

 

 

 

 

Model Estimated Sum of Squares df
Partisan 6844.093 20
Nonpartisan 2760.339 20
Full 8018.447 20
Chow Test:

ESS(F) - (ESS(P) — ESS(N)) / l_(
ESS(P) + ESS(N) / n(P) + n(N) - 2(k)

[8018.447 — (6844.093 + 2760.339)l / 20
(6844.093 + 2760.339) / [(58 + 69) — 2(20)]

F (k, n(P)+n(N)-2(k) = F (20,87)
F (20, 87) = -0.718

Thus, cannot reject the hypothesis that the regressions are the same
(Gujarati 1995)

171

Table A9: Hausman Test Results for Table 7.2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable 2SLS OLS ZSLS — OLS SE.
Log of Incum. 1.138 -0.188 1.326 3.592
Spending
Log of Chall. -1.634 -1.522 -0.112 0.247
Spending
Quality Chall. —4. 103 -3.858 -0.245 0.665
Partisan -0.513 -0.384 -0.129
Consonance
Age 0005 -0.006 0.001 0.001
Appointed -2.950 -3.123 0.173 0.275
First
Compete 1.278 1.402 -0.124 0.514
Prior Close 4.570 -4.631 0.061 0.099
Race
Control of 0.190 0.421 -0.231 1.043
Court
Voting 0.000 0.001 -0.000
Population
Murder Rate 0.259 0.366 -0.107 0.345
Tort Docket 0.099 0.165 -0.066 0.181
Partisan -3.136 -3.616 0.480 0.979
District -6.357 -6.619 0.262 0.746
MultiMember 0.625 -0.113 0.738 1.798
Term 2.430 2.624 -0.194 0.468
1990 -3.329 -4.510 1.181 3.078
1992 -3.266 -4.633 1.366 3.991
1994 -2.336 -3.166 0.831 3.034
1996 -3.171 -3.775 0.604 1.767
1998 4.277 4.001 0.277 1.270
Constant 44.474 55.936 -11.462 31.797

 

H0 = difference in the coefficients is not systematic
x2 (10) = (B. - B.)' (V. — V.)"(B. - B.)
where
,= the coefficient vector from 2SLS
130 = the coefficient vector from OLS
V, = the covariance matrix of the coefficients from 281.8
Vo = the covariance matrix of the coefficients from OLS

x2 (19) = 0.022

 

2 In this case, the model does not meet the assumptions of the Hausman test. This can be interpreted as
“strong evidence that we cannot reject the null hypothesis. Such a result is not an unusual outcome for the
Hausman test, particularly when the sample is relatively small ...” (Stata 1999, 12).

172

Table A. 10: Hausman Test Results for Table 7.4

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable 2SLS OLS 2SLS — OLS SE.

Log of Winner 5.463 0.130 5.333 9.940
Spending

Log of Loser -2.253 -1.928 -0.325 0.998
Spending

Quality Winner 1.210 0.679 0.531 1.696

Quality Loser -0. 160 1.606 -1.767 5.044

Partisan -1.546 -0.702 -0.843 1.564
Consonance

Compete -0.499 -2.778 2.279 3.156

Prior Close 3.588 5.223 -1.635 2.922
Race

Control of -0.642 -0.022 -0.621 1.115
Court

Voting 0.000 —0.000 0.000 0.000
Population

Tort Docket -0.232 0.004 -0.235 0.455

Partisan -2.894 -2.429 0.465 .

District 5.781 0.579 5.201 9.219

MultiMember 11.198 12.584 -1.387 5.522

Term 0879 -0.041 -0.837 1.587

1990 2.341 2.270 0.071 .

1992 0.239 1.239 -1.000 2.164

1994 -3.431 0.049 -3.480 6.970

1996 4.609 2.730 1.879

1998 -0.920 -2.024 1.104 .

Constant 25.210 75 .703 -50.493 88.625

 

Ho = difference in the coefficients is not systematic

x2 (18) = (B. - B.)’ (V. — V.)“(B. - B.)
where
B, = the coefficient vector from 2SLS
0,, = the coefficient vector from OLS
V, = the covariance matrix of the coefficients from 281.8
Vo = the covariance matrix of the coefficients from OLS

x2 (18) = -0423

x

3 . . . .

In this case, the model does not meet the assumptions of the Hausman test. Thrs can be interpreted as
‘Strong evidence that we cannot reject the null hypothesis. Such a result is not an unusual outcome for the
I‘lausman test, particularly when the sample is relatively small (Stata 1999, 12).

173

Table A. 1 1: OLS Results for Table 7.2

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Coefficient Robust t P > M
Std. Error

Log of Incum. -0.188 0.965 -0.194 0.848
Spending

Log of Chall. -1.522 0.537 -2.832 0.011
Spending

Quality Chall. -3.858 2.367 -1.630 0.121

Partisan -0.384 1.128 -0.340 0.738
Consonance

Age 0006 0.002 -2.736 0.014

Appointed -3.123 1.244 -2.511 0.022
First

Compete 1.402 1.236 1.135 0.272

Prior Close -4.631 1.583 —2.925 0.009
Race

Control of 0.421 1.852 0.227 0.823
Court

Voting 0.001 0.000 1.887 0.076
Population

Murder Rate 0.366 0.325 1.124 0.277

Tort Docket 0.165 0.065 2.544 0.021

Partisan -3.616 2.256 -1.603 0.127

District -6.619 3.067 -2.159 0.045

MultiMember -0.113 3.011 -0.037 0.971

Term 2.624 0.662 3.961 0.001

1990 -4.510 1.851 -2.437 0.026

1992 —4.633 2.294 -2.019 0.060

1994 -3.166 2.989 -1.060 0.304

1996 -3.775 3.180 -1.187 0.252

1998 4.001 2.271 1.761 0.096

Constant 55.936 9.360 5.976 0.000

 

Dependent variable: percentage of vote for incumbent
Mean of dependent variable: 56.31

N = 127

F (16, 17) = 71.77
Prob > F = 0.000
R2 = 0.5516

Root MSE = 7.944

174

REFERENCES

Abramowitz, Alan I. 1975. “Name Familiarity, Reputation, and the Incumbency Effect
in a Congressional Election.” Western Political Quarterly 28 (December): 668-682.

Abramowitz, Alan I. 1988. “Explaining Senate Election Outcomes.” American Political
Science Review 82 (June): 385-403.

Abramowitz, Alan I. 1989. “Campaign Spending in United States Senate Elections.”
Legislative Studies Quarterly 14 (November): 487-507.

Achen, Christopher H. 1986. The Statistical Analysis of Quasi-Experiments. Berkeley:
University of California Press.

Alexander, Herbert E. 1972. Money in Politics. Washington, DC: Public Affairs
Press.

Alvarez, R. Michael. 1997. Information and Elections. Ann Arbor: University of
Michigan Press.

American Bar Association. 1998. Report and Recommendations of the Task Force on
Lawyers' Political Contributions: Part 11. Chicago, IL: American Bar Assocation.

Ansolabehere, Stephen and Alan Gerber. 1994. “The Mismeasure of Campaign
Spending: Evidence from the 1990 United States House Elections.” Journal of Politics
56 (November): 1106-1118.

Arrington, Theodore S. 1996. “When Money Doesn’t Matter: Campaign Spending for
Minor Statewide Judicial and Executive Offices in North Carolina.” Justice System
Journal 18 (3): 257-266.

Aspin, Larry T. 1998. “Campaigns in Judicial Retention Elections: Do They Make a
Difference.” Justice System Journal 20 (1): 1-15.

Aspin, Larry T. and William K. Hall. 1987. “The Friends and Neighbors Effect in
Judicial Retention Elections.” Western Political Quarterly 40 (December): 703-715.

Aspin, Larry T. and William K. Hall. 1989. “Friends and Neighbors Voting in Judicial
Retention Elections: A Research Note Comparing Trial and Appellate Court Elections.”
Western Political Quarterly 42 (December): 587-595.

Aspin, Larry T., William K. Hall, Jean Bax, and Celeste Montoya. 2000. “Thirty Years

of Judicial Retention Elections: An Update.” The Social Science Journal 37 (January):
1-17.

175

Atkeson, Lonna Rae and Randall W. Partin. 1995. “Economic and Referendum Voting:
A Comparison of Gubernatorial and Senate Elections.” American Political Science
Review 89 (March): 99-107.

Banaian, King and William A. Luksetich. 1991. “Campaign Spending in Congressional
Elections.” Economic Inquiry 29 (January): 92-100.

Banks, Jeffrey S. and D. Roderick Kiewiet. 1989. “Explaining Patterns of Candidate
Competition in Congressional Elections.” American Journal of Political Science 33
(November): 997-1015.

Beck, Nathaniel and Jonathan N. Katz. 1995. “What To Do (and Not To Do) with Time-
Series Cross-Section Data.” American Political Science Review 89 (September): 634-
647.

Becker, Lee B. and Sharon Dunwoody. 1982. “Media Use, Public Affairs Knowledge
and Voting in Local Elections.” Journalism Quarterly 59 (2): 212-218.

Bell, Dawson. 2001. “En gler to Ask for Appointed High Court.” The Detroit Free
Press, January 27.

Bell, Dawson. 2002. “Good News About Judicial Fairness Gets Overlooked.” The
Detroit Free Press, March 25.

Berry, Ertel. 2001. “Party Labels Outweighed War Chests in 2000 High Court Races.”
North Carolina Lawyers Weekly 14 (April 2): 37.

Biersack, Robert, Paul S. Hermson, and Clyde Wilcox. 1993. “Seeds for Success: Early
Money in Congressional Elections.” Legislative Studies Quarterly 18 (November): 535-
551.

Bonneau, Chris W. 2001a. “The Composition of State Supreme Courts, 2000.”
Judicature 85 (July-August): 26-31.

Bonneau, Chris W. 2001b. “Money and Votes in State Supreme Court Elections.”
Paper Presented at the Annual Meeting of the American Political Science Association,
August 30-September 2, 2001, San Francisco, California.

Bonneau, Chris W. and Melinda Gann Hall. 2000. “Challengers in State Supreme Court
Elections.” Paper Presented at the Annual Meeting of the Midwest Political Science
Association, April 27-30, 2000, Chicago, Illinois.

Bonneau, Chris W. and Melinda Gann Hall. 2001. “Incumbents, Challengers, and the

Politics of Judicial Elections.” Paper Presented at the Annual Meeting of the Midwest
Political Science Association, April 19-22, 2001, Chicago, Illinois.

176

Brace, Paul. 2002. Personal Communication. September 10.

Brace, Paul and Melinda Gann Hall. 1990. “Neo-Institutionalism and Dissent in State
Supreme Courts.” Journal of Politics 52 (February): 54-70.

Brace, Paul and Melinda Gann Hall. 1995. “Studying Courts Comparatively: The View
from the American States.” Political Research Quarterly 48 GVIarch): 5-29.

Brace, Paul and Melinda Gann Hall. 1997. ”The Interplay of Preferences, Case Facts,
Context, and Structure in the Politics of Judicial Choice.” Journal of Politics 59
(November): 1206-1231.

Brace, Paul and Melinda Gann Hall. 2000a. “Comparing Courts Using the American
States.” Judicature 83 (March-April): 250-266.

Brace, Paul and Melinda Gann Hall. 2000b. “Courts as Agents of Continuity and
Change: Winning an Losing in State Supreme Courts.” Paper Presented at the Annual
Meeting of the American Political Science Association, August 31-September 3,
Washington, DC.

Brace, Paul and Melinda Gann Hall. 2001. “‘Haves’ Versus ‘Have Nots’ in State
Supreme Courts: Allocating Docket Space and Wins in Power Asymmetric Cases.” law
and Society Review 35 (2): 393-417.

Breaux, David A. and Anthony Gierzynski. 1991. “‘It’s Money That Matters’:
Campaign Expenditures and State Legislative Primaries.” Legislative Studies Quarterly
16 (August): 429-443.

Breen. Richard. 1996. Regression Models: Censored, Sample Selected, or Truncated
Data. London: Sage.

Brown, TC. 2000. “Resnick Overcomes Ads, Wins Supreme Court Race.” The
Cleveland Plain Dealer, November 8.

Brubaker, Stanley C. 1998. “The Limits of Campaign Spending Limits.” The Public
Interest 133 (Fall): 33-54.

Campbell, Linda P. 1998. “Fund-Raising by Justices Gets Attention.” Fort Worth Star-
Telegram, September 27.

Carp, Robert A. and Ronald D. Stidham. 2001. Judicial Process in America, 5‘h edition.
Washington, DC: CQ Press.

Carsey, Thomas M., and Gerald R. Wright. 1998. “State and National Forces in

Gubernatorial and Senate Elections.” American Journal of Political Science 42 (July):
994-1002.

177

Cassie, William E. and David A. Breaux. 1998. “Expenditures and Election Results.” In
Campaign Finance in State Legislative Elections, edited by Joel A. Thompson and Gary
P. Moncrief. Washington, DC: CQ Press.

Champagne, Anthony. 1986. “The Selection and Retention of Judges in Texas.”
Southwestern Law Journal 40: 53-117.

Champagne, Anthony. 1992. “Campaign Contributions in Texas Supreme Court Races.”
Crime, Law and Social Change 17 (March): 91-106.

Champagne, Anthony and Kyle Cheek. 1996. “PACs and Judicial Politics in Texas.”
Judicature 80 (J uly-August): 26-27.

Champagne, Anthony and Greg Thielemann. 1991. “Awareness of Trial Court Judges.”
Judicature 74 (February-March): 271-276.

Cheek, Kyle and Anthony Champagne. 2000. “Money in Texas Supreme Court
Elections.” Judicature 84 (J uly-August): 20-25.

Choate, Alan. 2000a. “State GOP Executive Director Denounces Findings of Study,
Author.” Mobile Register, October 31.

Choate, Alan. 2000b. “Ads Link Supreme Court Justices to Firestone Controversy.”
Mobile Register, November 2.

Coleman, John J. and Paul F. Manna. 2000. “Congressional Campaign Spending and the
Quality of Democracy.” Journal of Politics 62 (August): 757-789.

Cox, Gary W. and Mathew D. McCubbins. 1993. Legislative Leviathan: Party
Government in the House. Berkeley: University of California Press.

Cox, Gary W. and Scott Morgenstem. 1995. “The Incumbency Advantage in
Multimember Districts: Evidence from the US. States.” Legislative Studies Quarterly
20 (August): 329-349.

Dickerson, Brian. 2001a. “Belatedly, Engler Seeks Court Reform.” The Detroit Free
Press, January 29.

Dickerson, Brian. 2001b. “Reason Rules Civil Reform on Judiciary.” The Detroit Free
Press, February 12.

Dubois, Philip L. 1979. “The Significance of Voting Cues in State Supreme Court
Elections.” Law and Society Review 13 (Spring): 757-779.

178

Dubois, Philip L. 1980. From Ballot to Bench: Judicial Elections and the Quest for
Accountability. Austin: University of Texas Press.

Dubois, Philip L. 1984. “Voting Cues in Nonpartisan Trial Court Elections: A
Multivariate Assessment.” Law and Society Review 18 (3): 395-436.

Dubois, Philip L. 1986a. “Penny For Your Thoughts? Campaign Spending in California
Trial Court Elections, 1976-1982.” Western Political Quarterly 39 (June): 265-284.

Dubois, Philip L. 1986b. “Financing Trial Court Elections: Who Contributes to
Califomia’s Judicial Campaigns.” Judicature 70 (June-July): 8-16.

Erikson, Robert S. and Thomas R. Palfrey. 1998. “Campaign Spending and
Incumbency: An Alternative Simultaneous Equations Approach.” Journal of Politics 60
(May): 355-373.

Erikson, Robert S. and Thomas R. Palfrey. 2000. “Equilibria in Campaign Spending
Games: Theory and Data.” American Political Science Review 94 (September): 595-
609.

Farmer, John. 2001. “Justice For Sale in Pennsylvania.” Newark Star-Ledger, August
17.

Ferejohn, John A. 1990. “Information and the Electoral Process.” In Information and
Democratic Processes, edited by John A. Ferejohn and James H. Kuklinski. Urbana, IL:
University of Illinois Press.

Ferguson, Thomas. 1995. Golden Rule: The Investment Theory of Party Competition
and the Logic of Money-Driven Political Systems. Chicago: University of Chicago
Press.

Friedman, Lawrence M. 1985. History of American Law. New York: Simon and
Schuster.

Friedrich, Robert J. 1982. “In Defense of Multiplicative Terms in Multiple Regression
Equations.” American Journal of Political Science 26 (November): 797-833.

Gerber, Alan. 1998. “Estimating the Effect of Campaign Spending on Senate Election
Outcomes Using Instrumental Variables.” American Political Science Review 92 (June):
401-411.

Gierzynski, Anthony. 1998. “A Framework for the Study of Campaign Finance.” In

Campaign Finance in State Legislative Elections, edited by Joel A. Thompson and Gary
F. Moncrief. Washington, DC: CQ Press.

179

Gierzynski, Anthony and David A. Breaux. 1991. “Money and Votes in State
Legislative Elections.” Legislative Studies Quarterly 16 (May): 203-217.

Gierzynski, Anthony and David A. Breaux. 1993. “Money and the Party Vote in State
House Elections.” Legislative Studies Quarterly 18 (November): 515-533.

Gierzynski, Anthony and David A. Breaux. 1996. “Legislative Elections and the
Importance of Money.” Legislative Studies Quarterly 21 (August): 337-358.

Gierzynski, Anthony, Paul Kleppner, and James Lewis. 1998. “Money or the Machine:
Money and Votes in Chicago Aldermanic Elections.” American Politics Quarterly 26
(April): 160-173.

Giles, Michael W. and Anita Pritchard. 1985. “Campaign Expenditures and Legislative
Elections in Florida.” Legislative Studies Quarterly 10 (February): 71-88.

Glaberson, William. 2000a. “Fierce Campaigns Signal a New Era for State Courts.”
New York Times, June 5, 2000.

Glaberson, William. 2000b. “State Chief Justices Plan to Meet on Judicial Candidates’
Abuses.” New York Times, September 8, 2000.

Glaberson, William. 2001. “States Take Steps to Rein In Excesses of Judicial
Politicking.” New York Times. June 15.

Glick, Henry R. and Craig F. Emmert. 1987. “Selection Systems and Judicial
Characteristics: The Recruitment of State Supreme Court Judges.” Judicature 70
(December-January): 228-235.

Goldberg, Deborah, Craig Holman, and Samantha Sanchez. 2002. The New Politics of
Judicial Elections. New York: Justice at Stake Campaign.

Gopoian, J. David. 1984. “What Makes PACs Tick? An Analysis of the Allocation
Patterns of Economic Interest Groups.” American Journal of Political Science 28 (May):
259-281.

Green, Donald Philip and Jonathan S. Krasno. 1988. “Salvation for the Spendthrift
Incumbent: Reestimating the Effects of Campaign Spending in House Elections.”
American Journal of Political Science 32 (November): 884-907.

Green, Donald Philip and Jonathan S. Krasno. 1990. “Rebuttal to J acobson’s ‘New
Evidence for Old Arguments.”’ American Journal of Political Science 32 (November):
884-907.

Greene, William H. 2000. Econometric Analysis, 4‘11 edition. Upper Saddle River, NJ:
Prentice Hall.

180

Grier, Kevin B. and Michael C. Munger. 1986. “The Impact of Legislator Attributes on
Interest-Group Campaign Contributions.” Journal of Labor Research 7 (Fall): 349-361.

Griffin, Kenyon N. and Michael J. Horan. 1979. “Merit Retention Elections: What
Influences the Voters?” Judicature 63 (August): 78-88.

Gujarati, Damodar N. 1995. Basic Econometrics, 3rd edition. New York: McGraw Hill.

Hall, Melinda Gann. 1987. “Constituent Influence in State Supreme Courts: Conceptual
Notes and a Case Study.” Journal of Politics 49 (November): 1117-1124.

Hall, Melinda Gann. 1992. “Electoral Politics and Strategic Voting in State Supreme
Courts.” Journal of Politics 54 (May): 427-446.

Hall, Melinda Gann. 1995. “Justices as Representatives: Elections and Judicial Politics
in the American States.” American Politics Quarterly 23 (October): 485-503.

Hall, Melinda Gann. 2000. State Supreme Court Elections Data, 1990-1995. East
Lansing, MI: Michigan State University.

Hall, Melinda Gann. 2001a. “State Supreme Courts in American Democracy: Probing
the Myths of Judicial Reform.” American Political Science Review 95 (June): 315-330.

Hall, Melinda Gann. 2001b. “Voluntary Retirements from State Supreme Courts:
Assessing Democratic Pressures to Relinquish the Bench.” Journal of Politics 63
(November): 1 112-1140.

Hall, Melinda Gann and Chris W. Bonneau. 2000. “The Wisconsin Judiciary.” In
Wisconsin Government and Politics, 7‘h edition, edited by Ronald E. Weber. New York:
McGraw-Hill.

Hall, Melinda Gann and Paul Brace. 1989. “Order in the Courts: A Neo-Institutional
Approach to Judicial Consensus.” Western Political Quarterly 42 (September): 391-407.

Hall, Melinda Gann and Paul Brace. 1992. “Toward an Integrated Model of Judicial
Voting Behavior.” American Politics Quarterly 20 (April): 147-168.

Hall, Melinda Gann and Paul Brace. 1999. “State Supreme Courts and Their
Environments: Avenues to General Theories of Judicial Choice.” In Supreme Court
Decision-Making: New Institutionalist Approaches, edited by Cornell W. Clayton and
Howard Gillman. Chicago: University of Chicago Press.

Hall, Richard L. 1996. Participation in Congress. New Haven: Yale University Press.

181

Hall, Richard L. and Frank W. Wayman. 1990. “Buying Time: Moneyed Interests and
the Mobilization of Bias in Congressional Committees.” American Political Science
Review 84 (September): 797-820.

Hansen, Mark. 1998. “A Run for the Bench.” ABA Journal 84 (October): 68-72.

Heard, Alexander. 1960. The Costs of Democracy. Chapel Hill, NC: University of
North Carolina Press.

Heckman, James J. 1979. “Sample Selection Bias as a Specification Error.”
Econometrica 47 (January): 153-162.

Hogan, Robert E. and Keith E. Hamm. 1998. “Variations in District-Level Campaign
Spending in State Legislatures.” In Campaign Finance in State Legislative Elections,
edited by Joel A. Thompson and Gary P. Moncrief. Washington, DC: CQ Press.

Holbrook, Thomas M. and Emily Van Dunk. 1993. “Electoral Competition in the
American States.” American Political Science Review 87 (December): 955-962.

Jackson, Donald W. and James W. Riddlesperger, Jr. 1991. “Money and Politics in
Judicial Elections: The 1988 Election of the Chief Justice of the Texas Supreme Court.”
Judicature 74 (December-January): 184-189.

Jacobson, Gary C. 1978. “The Effects of Campaign Spending in Congressional
Elections.” American Political Science Review 72 (June): 469-491.

Jacobson, Gary C. 1980. Money in Congressional Elections. New Haven: Yale
University Press.

Jacobson, Gary C. 1981. “Incumbents’ Advantages in the 1978 US. Congressional
Elections.” Legislative Studies Quarterly 6 (May): 183-200.

Jacobson, Gary C. 1985. “Money and Votes Reconsidered: Congressional Elections,
1972-1982.” Public Choice 47 ( 1): 7-62.

Jacobson, Gary C. 1989. “Strategic Politicians and the Dynamics of US. House
Elections, 1946-1986.” American Political Science Review 83 (September): 773-793.

Jacobson, Gary C. 1990. “The Effects of Campaign Spending in House Elections: New
Evidence for Old Arguments.” American Journal of Political Science 34 (May): 334-
362.

Jacobson, Gary C. 1996. ”The 1994 House Elections in Perspective.” Political Science
Quarterly 111 (Spring): 203-223.

182

Jacobson, Gary C. 1997. The Politics of Congressional Elections, 4th edition. New
York: Longman.

Jewell, Malcolm E. 1982. Representation in State Legislatures. Lexington, KY:
University Press of Kentucky.

Johnson, Charles A., Roger C. Schaefer, R. Neal McKnight. 1978. “The Salience of
Judicial Candidates and Elections.” Social Science Quarterly 59 (June): 371-378.

Kennedy, Peter. 1998. A Guide to Econometrics, 3rd edition. Cambridge, MA: MIT
Press.

Key, V.O., Jr. 1961 Public Opinion and American Democracy. New York: Alfred A.
Knopf.

Klein, David and Lawrence Baum. 2001. “Ballot Information and Voting Decisions in
Judicial Elections.” Political Research Quarterly 54 (December): 709-728.

Klots, Allen T. 1955. “The Selection of Judges and the Short Ballot.” Journal of the
American Judicature Society 38: 134-140.

Langer, Laura. 2002. Judicial Review in State Supreme Courts: A Comparative Study.
Albany: SUNY Press.

Lee, Frances E. and Bruce I. Oppenheimer. 1999. Sizing Up the Senate: The Unequal
Consequences of Equal Representation. Chicago: University of Chicago Press.

Levitt, Steven D. 1994. “Using Repeat Challengers to Estimate the Effect of Campaign
Spending on Election Outcomes in the US. House.” Journal of Political Economy 102
(August): 777-798.

Lowry, Robert C., James E. Alt, and Karen E. Ferree. 1998. “Fiscal Policy Outcomes
and Electoral Accountability in American States.” American Political Science Review 92
(December): 759-774.

Loyola of Los Angeles Law Review. 2001. “Symposium: National Summit on
Improving Judicial Selection.” Loyola of Los Angeles Law Review 34 (June): 1353-1512.

Marks, Lynn. 2001. “Perspectives: Odor in the Courts.” Pittsburgh Post-Gazette,
November 5.

Marks, Lynn A. and Julia H. Hoke. 2001. “Electing Judges: Had Enough, Yet?”
Philadelphia Daily News, November 9.

Mayhew, David R. 1974. Congress: The Electoral Connection. New Haven: Yale
University Press.

183

McKnight, R. Neal, Roger Schaefer, and Charles A. Johnson. 1978. “Choosing Judges:
DO the Voters Know What They’re Doing?” Judicature 62 (August): 94-99.

Miller, Gary J ., Thomas H. Hammond, and Charles Kile. 1996. “Bicameralism and the
Core: An Experimental Test.” Legislative Studies Quarterly 21 (February): 83-103.

Mikva, Abner and William 8. Sessions. 2001. “Guarantee the Independence of All
Judges.” Los Angeles Times, March 31.

Moe, Terry M. 1987. “An Assessment of the Positive Theory of ‘Congressional
Dominance.’” Legislative Studies Quarterly 12 (November): 475-520.

Moncrief, Gary F. 2001. Personal Communication. November 13.

Moog, Robert. 1992. “Campaign Financing for North Carolina’s Appellate Courts.”
Judicature 76 (August-September): 68-76.

Munger, Michael C. 1989. “A Simple Test of the Thesis that Committee Jurisdictions
Shape Corporate PAC Contributions.” Public Choice 62 (August): 181-186.

Nicholson, Marlene Arnold and Norman Nicholson. 1994. “Funding Judicial Campaigns
in Illinois.” Judicature 77 (May-June): 294-299.

Niemi, Richard 6., Harold W. Stanley, and Ronald J. Vogel. 1995. “State Economies
and State Taxes: Do Voters Hold Governors Accountable?” American Journal of
Political Science 39 (November): 936-957.

Omdorff, Mary. 2002. “State Justices Races Most Costly in US.” The Birmingham
News, April 1.

Palda, Filip. 1994. How Much Is Your Vote Worth? The Unfaimess of Campaign
Spending Limits. San Francisco: Institute for Contemporary Studies.

Partin, Randall W. 2002. “Assessing the Impact of Campaign Spending in Governors’
Races.” Political Research Quarterly 55 (March): 213-233.

Phillips, Thomas R. 2002. “When Money Talks, the Judiciary Must Balk.” Washington
Post, April 14.

Pittsburgh Post-Gazette. 2001a. “Editorial: Throw Out the Baby.” Pittsburgh Post-
Gazette, January 17.

Pittsburgh Post-Gazette. 2001b. “Editorial: Ridge the Reformer.” Pittsburgh Post-
Gazette, February 14.

184

Pittsburgh Post-Gazette. 20010. “Editorial: Judging an Opportunity.” Pittsburgh Post-
Gazette, April 13.

Pomper, Gerald M. 1968. Elections in America: Control and Influence in Democratic
Politics. New York: Dodd, Mead & Company.

Popkey, Dan. 2001. “Contributors Could Benefit from Eismann’s Role.” The Idaho
Statesman, March 8.

Przeworski, Adam and Henry Teune. 1970. The Logic of Comparative Social Inquiry.
Malabar, FL: Krieger.

Reid, Traciel V. 1996. “PAC Participation in North Carolina Supreme Court Elections.”
Judicature 80 (July—August): 21-29.

Reid, Tracie] V. 1999. “The Politicization of Retention Elections: Lessons from the
Defeat of Justices Lanphier and White.” Judicature 83 (September-October): 68-77.

Robison, Clay. 1998. “‘Payola Justice’ in Texas Revisited.” Houston Chronicle,
November 2.

Rohde, David W. 1991. Parties and Leaders in the Postreform House. Chicago:
University of Chicago Press.

Romer, Thomas and James M. Snyder, Jr. 1994. “An Empirical Investigation of the
Dynamics of PAC Contributions.” American Journal of Political Science 38 (August):
745-769.

Sayrs, Louis W. 1989. Pooled Time Series Analysis. Newbury Park, CA: Sage.

Scarritt, Tom. 2001. “Who Holds the Gavel in Alabama?” The Birmingham News,
April 1.

Schaffner, Brian F., Matthew Streb, and Gerald Wright. 2001. “Teams Without
Uniforms: The Nonpartisan Ballot in State and Local Elections.” Political Research
Quraterly 54 (March): 7-30.

Schotland, Roy A. 1985. “Elective Judges’ Campaign Financing: Are State Judges’
Robes the Emperor’s Clothes of American Democracy.” Journal of Law and Politics 2
(Spring): 57-167.

Schotland, Roy A. 2001. “Financing Judicial Elections, 2000: Change and Challenge.”

The Law Review of Michigan State University—Detroit College of Law 2001 (Fall): 849-
899.

185

 

   

Sheldon, Charles H. and Linda S. Maule. 1997. Choosing Justice: The Recruitment of
State and Federal Judges. Pullman, WA: Washington State University Press.

Shepsle, Kenneth A. and Barry R. Weingast. 1987. “The Institutional Foundations of
Committee Power.” American Political Science Review 81 (March): 85-104.

Smith, Bradley A. 1996. “Faulty Assumptions and Undemocratic Consequences of
Campaign Finance Reform.” Yale Law Journal 105 (January): 1049-1091.

9

Smith, Bradley A. 1999. “Some Problems with Taxpayer-Funded Political Campaigns.’
University of Pennsylvania Law Review 148 (December): 591-628.

Sorauf, Frank J. 1988. Money in American Elections. Glenview, IL: Soctt, Foresman.

Sorauf, Frank J. 1992. Inside Campaign Finance: Myths and Realities. New Haven:
Yale University Press.

Squire, Peverill. 1989a. “Competition and Uncontested Seats in US. House Elections.”
Legislative Studies Quarterly 14 (May): 281-295.

Squire, Peverill. 1989b. “Challengers in US. Senate Elections.” Legislative Studies
Quarterly 14 (November): 531-547.

Squire, Peverill. 1992. “Challenger Quality and Voting Behavior in US. Senate
Elections.” Legislative Studies Quarterly 17 (May): 247-263.

Squire, Peverill. 1995. “Candidates, Money, and Voters—Assessing the State of
Congressional Elections Research.” Political Research Quarterly 48 (December): 891-
917.

Squire, Peverill. 2000. “Uncontested Seats in State Legislative Elections.” Legislative
Studies Quarterly 25 (February): 131-146.

Squire, Peverill and John R. Wright. 1990. “Fundraising by Nonincumbent Candidates
for the US. House of Representatives.” Legislative Studies Quarterly 15 (February): 89-
98.

STATA. 1999. Stata Reference Manual Release 6, Volume 2. College Station, TX:
Stata Press.

Stimson, James A. 1985. “Regression in Space and Time: A Statistical Essay.”
American Journal of Political Science 29 (November): 914-947.

Su, Tie-ting, Alan Neustadtl, and Don Clawson. 1995. “Business and the Conservative

Shift: Corporate PAC Contributions, 1976-1986.” Social Science Quarterly 76 (March):
20-40.

186

Svoboda, Craig J. 1995. “Retrospective Voting in Gubernatorial Elections: 1982 and
1986.” Political Research Quarterly 48 (March): 135-150.

Swerczek, Mary. 2001. “Weimer to Sit on Supreme Court: Voting Light, Close for 6th
District Seat.” The New Orleans Times-Picayune, November 18.

Thielemann, Gregory S. 1993. “Local Advantage in Campaign Financing: Friends,
Neighbors, and Their Money in Texas Supreme Court Elections.” Journal of Politics 55
(May): 472-478.

Thielemann, John and Allen Wilhite. 1995. “Congressional Turnover: Negating the
Incumbency Advantage.” Social Science Quarterly 76 (September): 594-606.

Thomas, Scott J. 1989. “Do Incumbent Campaign Expenditures Matter?” Journal of
Politics 51 (November): 965-976.

Tucker, Harvey J. and Ronald E. Weber. 1987. “State Legislative Election Outcomes:

Contextual Effects and Legislative Performance Effects.” Legislative Studies Quarterly
12 (November): 537-553.

Van Dunk, Emily. 1997. “Challenger Quality in State Legislative Elections.” Political
Research Quarterly 50 (September): 793-807.

Weber, Ronald E., Harvey J. Tucker, and Paul Brace. 1991. “Vanishing Marginals in
State Legislative Elections.” Legislative Studies Quarterly 16 (February): 29-48.

Welch, W. P. 1976. “The Effectiveness of Expenditures in State Legislative Races.”
American Politics Quarterly 4 (July): 333-356.

Welch, William P. 1979. “Patterns of Contributions: Economic Interest and Ideological
Groups.” In Political Finance, edited by Herbert Alexander. Beverly Hills: Sage.

Welch, W. P. 1981. “Money and Votes: A Simultaneous Equation Model.” Public
Choice 36: 209-234.

Wenzel, Fritz. 2001. “Judicial Experts at UT Law Forum Divided on Value of Elected
Bench.” The Toledo Blade, March 31.

Wersal, Greg. 2001. “Counterpoint: See Why Minnesota’s Judiciary Could Use
Campaign Reform.” Minneapolis-St. Paul Star-Tribune, March 10.

Wertheimer, Fred and Susan W. Manes. 1994. “Campaign Finance Reform: A Key to

Restoring the Health of Our Democracy.” Columbia University Law Review 94 (May):
1126-1159.

187

Wickline, Michael R. 2001. “Arkansas Faces Shift in Judicial Elections.” Arkansas
Democrat-Gazette, May 20.

Wohl, Alexander. 2000. “Justice for Rent: The Scandal of Judicial Campaign
Financing.” American Prospect 11 (May 22): 34-37.

Wooldridge, Jeffrey M. 2000. Introductory Econometrics: A Modern Approach.
Cincinnati, OH: South-Westem College.

Wooldridge, Jeffrey M. 2002. Econometric Analysis of Cross Section and Panel Data.
Cambridge: MIT Press.

188