THE IMPACTS OF PARTY COHESION ON DEMOCRATIC ACCOUNTABILITY
By
Shih-hao Huang

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Political Science - Doctor of Philosophy
2014

ABSTRACT
THE IMPACTS OF PARTY COHESION ON DEMOCRATIC ACCOUNTABILITY
By
Shih-hao Huang
This dissertation features party cohesion as a factor influencing how voters hold ruling
elites accountable. It consists of three essays that dissect the impacts of party cohesion on
democratic accountability in the United States. In a nutshell, this dissertation finds that when
party cohesion increases, the focus of politics is shifted toward partisan rivalries, and hence
the ruling elites may not be held accountable to the public as they pursue partisan goals.
The first chapter is an essay titled Income Inequality and Electoral Costs of Party
Loyalty in the United States, 1976-2008. The United States features a political system where
party loyalty has been traditionally seen as electorally costly to individual legislators.
However, if being loyal to the party is costly, why did legislators in the U.S. increase their
party loyalty to a high level during the past decades? I argue that the electoral cost of party
loyalty could be reduced, and I suggest that such a reduction of electoral cost was resulted
from the increases in income inequality. The evidence from the U.S. House elections between
1976 and 2008 indicates that party loyalty becomes less costly when income inequality
increases. This finding may not only complement the existing theories of party government,
but also shatter the conventional wisdom on the role of the party in personal-vote systems.
The second essay is titled Clarity of Responsibility and Clarity of Party Line: The
Impacts of Party Cohesion on Economic Voting in the United States, 1980-2008. Powell and
Whitten (1993) maintain that governing party cohesion is a stable feature of political systems
that enhances clarity of responsibility and economic voting. However, party cohesion is not
stable, and there are ebbs and flows in party cohesion. Further, research finds that at the
micro level, voters respond to partisan cues and become more partisan when the elites from

each party stand with their in-party members and that economic conditions matter less when
there are more partisan voters. With data from the American National Election Studies
(ANES) between 1980 and 2008, this essay demonstrates that party cohesion is a
double-edged sword. As party cohesion increases from a low to mid level, it may strengthen
economic voting by enhancing clarity of responsibility. When it is at a high level and party
line becomes clear, it may hinder economic voting.
Finally, the essay, Resisting Being Held Accountable: The Impact of Party Cohesion on
Concentration of Votes in the United States, 1992-2012, is the third chapter. Party cohesion
has been seen as a constraint on particularistic exchanges between politicians and voters.
However, using the concentration of votes as an indicator of particularistic exchanges
between parties and voters, I find that party cohesion has a positive impact on the
concentration of votes in presidential elections between 1992 and 2012. This finding suggests
that a smaller range of the electorate is served by the incumbent party when the party is more
cohesive. Further, I demonstrate that the concentration of votes received by an incumbent
party positively affects the vote share of the party. Hence, a higher level of party cohesion
does not improve accountability. Instead, it contributes to the ability of a party to resist being
held accountable to the interest of the public.
The three essays show how accountability may vary with the changes in party cohesion.
The evidence presented suggests that party cohesion may not improve accountability as what
political scientists have wished for. Rather, high levels of party cohesion may hinder voters
from holding ruling elites accountable for the public interest.

This dissertation is dedicated to my grandparents, father and mother.

iv

ACKNOWLEDGEMENTS

I deeply appreciate that Michigan State University offered me the opportunity to pursue a
doctoral degree years ago. Pursuing a doctoral degree at MSU has been a wonderful and
unforgettable journey in my life. Here, let me take the opportunity to express my appreciation to
people who have helped me successfully earn this degree.
My appreciation first goes to my advisor, Eric Chang. Prof. Chang has never given me up.
He has always believed that one day I can be a great researcher. Even when I have lost my faith
in myself, he kept encouraging me to study the topic that I am interested in. He was also the
mentor who taught me how to survive in this program. He gave me direction when I was lost.
Without him, I could have never had a chance to earn the degree.
Secondly, I appreciate my committee members, Prof. Corwin Smidt, Prof. Jeff Conroy-Krutz,
Prof. Ani Sarkissian. Prof. Smidt was my instructor for statistical methods. I also worked with
him as a TA and a RA. Observing his teaching style and his profession in methodology, I have
dreamed to become a researcher like him. Prof. Conroy-Krutz, who is a specialist in comparative
politics and African politics, taught me how to read and write papers. He has always worked hard
not only for his own research and classes, but also in reading students’ paper. He gave me a lot of
ideas and advice to improve my research. I took Prof. Sarkissian’s class in my first semester in
MSU. She was supportive and taught me how to dissect a paper. The summaries I wrote for her
class have been the main sources when I need the information about identity politics. She helped
me to have a good start in this program so that I have the confidence in getting through the
following challenges.
I thank all the other faculty members who gave me helps. My special thanks will go to Prof.
Jacoby, Prof. Schneider and Prof. Colaresi. They have pushed me to think more about statistical

v

methods. They gave me many instructions so that I may get through the comprehensive exam
and learn from the process. I also have to thank Prof. Williams and Prof. Lee, who supported me,
discussed research with me and was always nice to me.
I am also grateful for the support from my family. My parents gave me the strongest support
and back-up. I always know that they are always there whenever I need them. My sister, my
brother in law and their two kids also provided physical and mental support to me.
In addition to my family, I appreciate all of my friends in the U.S. Wenchin Wu, Hsin-hsin
Pan, Seoyoun Choi, Helen Lee and Fang-yu Chen were my closest friend in this program. I will
always remember the moments we shared problems with research as well as with our life.
Moreover, I will always be grateful for the moments I spent with Sabina, Rachel, Jonah, Jason,
Petra, Jack, Bob, Rob, Chungho, Sungmin, Masaaki, Masayoshi. My friends from church, “Dry
Dad,” “Dry Mom,” Yuju, Sean, Peilin, Bobo, Hana, Hsiaomin, Miaobin, Charles Liu, Hiya,
Johnathan, Joanna, Justin, Diana, Hsinghsiang, Tzan-cheng, Peichi, Nono, Yang-yang, Jerry,
Katie, Yian, Liz, Sarah, Fengchiao, Nora, have been standing by me in the past few years. I
appreciate their friendship and patience which gave me strength to get through all the stages. My
appreciation also goes to Taofang Huang and Chinghsing Wang, who have never been hesitated
expressing their concerns about me and encouragements to me. I also want to thank my
neighbors, Gieu and Tzufen. I enjoyed the time with them in the last two years.
My friends in Taiwan have always stood by me. Prof. Shing-yuan Sheng and Prof. Jih-wen
Lin have always cared about my progress. They have discussed with me about research as well
as the life in the U.S. Furthermore, I want to give special thanks to my friends from college,
Pei-fen Yao, Chien-ju Yu, Chia-ling Wang, Kate Lu, Curtis Wang, Chun-ming Yang, Yu-peng
Lan, Janet Chan, Viona Tsai; my cohorts from high school, Rick Lin and Eric Lu; and my cohorts

vi

from NCCU’s master program, “NorthernStar,” Jengtsun, Ray-Ou Yang, Lavita Lan. Without
them, I would have never been so joyful in overcoming all the difficulties during these years.
Last but not least, I thank myself for not giving up.

vii

TABLE OF CONTENTS
LIST OF TABLES ........................................................................................................................... ⅹ
LIST OF FIGURES .......................................................................................................................... xi
INTRODUCTION ............................................................................................................................. 1
Accountability .................................................................................................................................... 2
Party Cohesion and Accountability in Contemporary American Politics: An overview ................... 9
Chapters ........................................................................................................................................... 14
CHAPTER 1 .................................................................................................................................... 17
INCOME INEQUALITY AND THE ELECTORAL COST OF PARTY LOYALTY IN THE
UNITED STATES 1976-2008 ......................................................................................................... 17
Costs and Benefits of Party Loyalty ............................................................................................ 21
Income Inequality and Electoral Cost of Party Loyalty............................................................... 24
Enlarging Issue Cleavages ............................................................................................. 25
Sorting People into Partisan Groups .............................................................................. 27
Granting Party Activists More Leverage ....................................................................... 28
Regression Analysis ..................................................................................................................... 30
Findings.......................................................................................................................... 33
Robustness ................................................................................................................................... 40
Party Polarization and Aggregate Party Loyalty ............................................................ 40
Partisanship of Constituencies ....................................................................................... 41
Monotonically Rising Income Inequality ...................................................................... 45
Conclusion ................................................................................................................................... 49
CHAPTER 2 .................................................................................................................................... 51
CLARITY OF RESPONSIBILITY AND CLARITY OF PARTY LINE: THE IMPACTS OF
PARTY COHESION ON ECONOMIC VOTING IN THE UNITED STATES, 1980-2008........... 51
The Conventional Wisdom: Clarity of Responsibility ................................................................. 54
The Missing Story: Clarity of Party Line .................................................................................... 56
Regression Analysis ..................................................................................................................... 60
Findings ......................................................................................................................... 66
Discussion .................................................................................................................................... 77
Conclusion ................................................................................................................................... 81
CHAPTER 3 .................................................................................................................................... 83
RESISTING BEING HLED ACCOUNTABLE: THE IMPACTS OF PARTY COHESION
ON CONCENTRATION OF VOTES IN THE UNITED STATES................................................. 83
Party Cohesion and Concentration of Votes .............................................................................. 87
Concentration of Votes and Electoral Success ........................................................................... 96
Discussion ................................................................................................................................ 102
Conclusion ............................................................................................................................... 105

viii

CONCLUSION .............................................................................................................................. 107
APPENDIX .................................................................................................................................... 114
REFERENCES .............................................................................................................................. 129

ix

LIST OF TABLES

Table I.1: Ideal Accountability versus Party-Oriented Accountability .............................................. 5
Table I.2: Accountability Discussed in Chapters ............................................................................... 8
Table 1.1: Party Loyalty and Its Correlation with Vote Share ........................................................ 22
Table 1.2: Regression Analysis for the U.S. House Elections ......................................................... 35
Table 1.3: Models with More Control Variables ............................................................................. 44
Table 1.4: State-Level Income Inequality and the Electoral Cost of Party Loyalty ........................ 47
Table 2.1: Party Cohesion in the U.S. House of Representatives .................................................... 56
Table 2.2: The Associations between Variables .............................................................................. 62
Table 2.3: Regression Analysis........................................................................................................ 69
Table 3.1: Party Cohesion and Concentration of Votes ................................................................... 94
Table 3.2: Concentration of Votes and Vote Shares ...................................................................... 100
Table A.1: Summary Statistics for Chapter One ........................................................................... 115
Table A.2: Summary Statistics for Chapter Two ........................................................................... 116
Table A.3: State-Level Concentration of Votes, 1992-2012 ......................................................... 119
Table A.4: Control Variables in Chapter Three ............................................................................. 122
Table A.5: Web Links in Chapter Three ........................................................................................ 126
Table A.6: Summary Statistics for Chapter Two ........................................................................... 128

x

LIST OF FIGURES

Figure 1.1: Income Inequality and Issue Cleavages ....................................................................... 27
Figure 1.2: The Cost of Party Loyalty, by Income Inequality ........................................................ 38
Figure 2.1: Incumbent Party Cohesion and Economic Voting ........................................................ 72
Figure 2.2: Opposition Party Cohesion and Economic Voting ....................................................... 73
Figure 3.1: Party Cohesion and Concentration of Votes, 1992-2012.............................................. 90

xi

INTRODUCTION

In the discipline of political science, the benefits and costs of party cohesion are called into
question by two contradictory concerns. On one hand, numerous scholars believed that a high
level of party cohesion in Congress is key to improving accountability for collective goods.
When each party acts as a unitary actor, programmatic party platforms become the focus of
politics. Citizens may clearly see distinctive alternative platforms and hold parties responsible
for serving the best interest of the public (APSA 1950; Shugart and Carey 1992). On the other
hand, the rise of party cohesion after the 1970s ignited a concern about party wars between two
cohesive parties and stalemate politics where many national problems are left unsolved (Rae
2007; Sinclair 2006). By extension, this concern implies that when party cohesion increases,
parties would not be held accountable to the best interest of the public. Although these two
concerns are equally plausible, empirically it remains unclear whether, and how, party cohesion
affects the way in which voters hold parties or politicians accountable.
To shed light on this controversy, my dissertation, The Impacts of Party Cohesion on
Democratic Accountability, includes three studies that explore the influences of party cohesion
on voting behavior and the electoral fortunes of parties, as well as legislators. The first essay
investigates party cohesion at the level of individual legislators and asks how voters hold
legislators accountable for their loyalty to the party. The second essay examines how economic
accountability is conditional upon party cohesion. Finally, the third essay empirically tests
whether high levels of party cohesion make incumbent parties more accountable for the interests
of citizens nation-wide. In light of these three studies, I share the second concern above. That is,
when party cohesion increases, the focus of politics is shifted toward partisan rivalries, and hence
the ruling elites may not be held accountable to the public as they pursue partisan goals.
1

Accountability
Accountability is the extent to which citizens may hold the ruling elites accountable for their
performance in serving the interest of the ruled. Accountability can also be described in terms of
principal-agent relationship. Ruling elites as agents have duties to their principals, citizens; the
principals have the rights to punish and reward the agents for their actions. In this sense,
accountability is also the capacity of the principals to penalize or reward the agents after their
actions (Ferejohn 2007). As Manin, Przeworski and Stokes have pointed out, ―governments are
accountable if voters can discern whether governments are acting in their interest and sanction
them appropriately, so that those incumbents who act in the best interest of citizens win
reelection and those who do not lose them‖ (Manin, Przeworski and Stokes 1999, 40). In other
words, holding ruling elites accountable means that voters, through elections, reward the
incumbents who serve the best interest of citizens and punish those who do not.
Accountability is a virtue of democracy, a key element of well-functioning democratic
politics, as well as one of the important reasons why democracy is pursued. As William Riker has
pointed out, ―the function of voting is to control officials, and no more,‖ and this function is ―all
that is needed to protect liberty‖ (Riker 1982). In other words, democracy was pursued because
people believed that through competitive and fair elections, politicians can be controlled and held
responsible for what they do in office. With accountability, there is a chance for citizens to
replace unfavorable or incapable politicians with those who might better serve the interest of the
public. Without it, there would be no threats to ruling elites, and the voices of people would not
be taken seriously. Moreover, a democracy without accountability is not substantially different
from an authoritarian country where the ruling elites can resume their powers regardless of
governing performance. Even worse, without accountability, democratic politics may not be
sustained and consolidated in that people may try to replace the government through ways other
2

than elections.
However, accountability is far from guaranteed in any democracy. It is prevalent that
politicians and parties in democracies may win elections without serving public interest. For
example, in developing countries with short-term experience of democratic practice, ruling elites
may win elections by particularistic exchanges with voters, as well as by a patron-client network
(Shugart and Carey 1992; Kitchelt and Wilkinson 2007; Lynn 2008). Even in an aged,
established democratic country, politicians may manage to get reelected by winning personal
votes from pork-barreling activities and constituency services (APSA 1950; Arnold 1990; Carey
and Shugart 1995; Mayhew 1974). Therefore, lack of accountability is a common problem in
democracies.
Lack of accountability can not be discussed in absolutes. There is neither a democracy
without any accountability nor a democracy with perfect accountability. For example, as
mentioned previously, if parties or politicians in a political system can win elections with
particularistic services or clientelistic networks, such a political system can be considered as
lacking accountability. However, even in this case, accountability still exists in the sense that the
parties or politicians are still held accountable to particularistic interests. As this example shows,
lack of accountability is not about whether there is accountability or not, but about trade-offs
between different types of accountability.
Due to trade-offs between accountability of different kinds, finding a solution to lack of
accountability is difficult. Solutions to lack of accountability often offer trade-offs of their own;
they improve some aspects of accountability while leading to deficiencies in others. In the
existing literature, increasing party cohesion is one such solution. According to the literature,
lack of accountability may result from lack of party-oriented politics. In party-oriented politics,
parties act as unitary actors that can be held accountable collectively for its governing
3

performance. In addition, when parties act as unitary actors, the role of individual politicians is
not important. Programmatic party platforms are the focus of politics, and particularistic
exchanges that may reduce accountability are constrained due to the dominance of parties in
policymaking process (Cox 1987; Shugart and Carey 1992; Franzese, Jusko and Nooruddin 2007;
Lynn 2008). Moreover, voters are able to clearly see parties as distinctive alternatives over which
they can choose in elections (APSA 1950, Shugart and Carey 1992), and the clarity of
responsibility for governing performance can also be enhanced (Powell and Whitten 1993). From
this perspective, making legislators vote together with their in-party colleagues in the legislature,
or, put it simply, increasing party cohesion, is a solution to lack of accountability. However,
taking a closer look, what increases in party cohesion may improve is not the ideal accountability
defined previously. While high levels of party cohesion may indeed make parties the focus of
politics, they may lead to a severe partisan rivalry and stronger partisanship of the electorate,
resulting in party-oriented accountability. Party-oriented accountability means that parties and
their performance, instead of the public interest, are emphasized when voters hold ruling elites
accountable. Increases in party cohesion thus involve a tradeoff between party-oriented
accountability and the ideal accountability.
To further clarify the concepts of party-oriented accountability and the trade-offs between
party-oriented accountability and the ideal accountability, three questions are essential: (1) Who
is holding the ruling elites accountable? (2) Who are the ruling elites that are held accountable?
(3) What are the ruling elites held accountable for? In the case of party-oriented accountability,
partisan voters or particular sets of the electorate that strongly support a party are holding the
party accountable for partisan goals or the success of party platforms (See Table I.1). However,
the ideal accountability exists when citizens can hold ruling elites accountable for the best
interest of the public. In other words, to the first question, the ideal answer should be all citizens,
4

but not partisan voters or particular sets of voters; to the second question, the ideal answer can be
all kinds of ruling elites and is not limited to parties; to the third question, the public interest,
rather than particular partisan goals or platforms, is the ideal answer. Therefore, party-oriented
accountability is neither equivalent to nor improving the accountability that political scientists
have conventionally wished for.

Table I.1. Ideal Accountability versus Party-Oriented Accountability
Citizens

Elites

Performance

Ideal
Accountability

All Citizens

All kinds of Ruling Elites

Public Interest

Party-oriented
Accountability

Partisan Voters
or
Core Supporters of A Party

Parties

Partisan Goals
or
Party Platforms

In the literature, there are three variants of the trade-off between party-oriented
accountability and the ideal accountability (See Table I.2), and all of them are involved with
party cohesion. First, the extent to which voters hold their legislators accountable for party
loyalty involves a trade-off between the accountability for individual legislators‘ loyalty to their
party and the accountability for serving local interests. It is also a trade-off between what Carey
(2009) calls individual accountability and collective accountability. In Carey‘s definition,
―collective accountability operates primarily through parties and requires legislators bearing a
common party label to act in concert‖ (Carey 2009, 1). In other words, collective accountability
is party oriented. It emphasizes the role of parties and requires individual legislators acting
cohesively with their party. By contrast, individual accountability emphasizes whether individual
politicians are held accountable to their constituencies independently of parties. By extension,
individual accountability is undermined when legislators are held accountable, not for serving
5

local interests, but for their loyalty to the party in carrying out party platforms. Conversely, if
legislators are not held accountable for their loyalty to the party, collective accountability is not
likely to exist. Accordingly, in order to understand the trade-off between collective accountability
and individual accountability, it is necessary to investigate whether legislators are held
accountable for their party loyalty.
The second trade-off between party-oriented accountability and the ideal accountability can
be seen in the literature on economic accountability. Economic conditions are crucial in
determining the public welfare, and ideally, ruling elites should be held accountable for
economic conditions. In the tradition of economic voting, scholars have studied whether voters
reward ruling elites for good economic conditions and punish them for bad ones. In their findings,
economic voting is contingent upon various factors (Anderson 2007). Among others, one factor
that conditions economic voting is clarity of responsibility. The conventional view is that voters
are more likely to hold the incumbent party accountable for economic conditions when the
clarity of responsibility is enhanced. To have high clarity of responsibility requires that the
legislators of incumbent party act cohesively so that voters may attribute certain economic
performance to the party‘s unitary actions (Powell and Whitten 1993). This suggests that high
levels of party cohesion and party-oriented politics may improve economic accountability.
However, evidence shows that when the elites behave in a partisan manner and when partisan
cues are prevalent, voters may downplay economic conditions in the elections (Hetheington 2001;
Kayser and Wlezien 2011; Levendusky 2009; Rudolph 2006). Therefore, as high levels of party
cohesion lead to a severe partisan rivalry, voting decisions may be affected more by partisan cues
than by economic conditions. Party cohesion plays a significant role in the trade-off between
economic accountability and party-oriented accountability.
The third trade-off is about a core issue of accountability: Can citizens hold the incumbent
6

party accountable for serving a broad range of people? As mentioned, such accountability might
be hindered when ruling elites manage to obtain sufficient votes by particularistic exchanges
between them and certain subsets of the electorate. When such exchanges become the center of
politics, politicians are held accountable neither for the ―success in delivering collective goods,‖
nor for ―improving overall distributive outcomes along the lines favored by broad categories of
citizens‖ (Kitschelt and Wilkinson 2007, 2). Compared to individual politicians, parties have to
win elections in a nation-wide constituency, and their platforms have to target a broad range of
citizens. If politics become party-oriented, and if parties discipline their legislators well in
carrying out party platforms, particularistic exchanges may be constrained as individual
legislators are not able to carry out policies that serve particular sets of voters (Shugart and Carey
1992; Franzese, Jusko and Nooruddin 2007; Lynn 2008). Therefore, party cohesion is key to
constraining particularistic exchanges and improving accountability for the public interest.
However, a high level of party cohesion may enable a party to better target their certain sets of
supporters and win elections on the basis of votes from these supporters (McGillivray 2004).
Accordingly, an incumbent party might be held accountable to partisan or core voters of the party,
rather than the entire public. The tradeoff between party-oriented accountability and the ideal
accountability once again hinges on the levels of party cohesion.

7

Table I.2. Accountability Discussed in Chapters
Citizens

Chapter One

Local
Constituency

Elites

Individual
Legislators

Performance

Trade-offs

Party Loyalty

Individual
Accountability
v.s.
Collective
Accountability
Economic

Chapter Two

Chapter Three

Individual Voters

Entire Public

Incumbent Party

Incumbent Party

Economic
Condition

accountability
v.s.
Partisan Issues

Public interest
Collective Goods
v.s.
Particularistic interests

As of today, these trade-offs and the impacts of party cohesion on them have not been
completely addressed. One major deficiency of previous studies is that they failed to take into
account the variation of party cohesion across time in studying the impacts of party cohesion. In
other words, the existing literature has not empirically tested whether accountability varies with
the ebbs and flows of party cohesion. Utilizing data for party cohesion in the United States,
where party cohesion in Congress has surprisingly increased, my dissertation casts a new light on
the impacts of party cohesion on accountability. While it has been commonly believed that high
levels of party cohesion may improve accountability, I find little empirical support for such a
belief. In particular, the evidence suggests that some aspects of accountability that have been a
particular concern of political scientists, including economic accountability and accountability to
the entire public, are traded off for party-oriented accountability due to high levels of party
cohesion.

8

Party Cohesion and Accountability in Contemporary American Politics: An overview
A distinctive characteristic of contemporary American politics is that two parties act more like
two different unitary actors than before. Individual legislators in the Capitol vote more frequently
with their party, and, on average, more than 90 percent of legislators from each party vote with
their in-party colleagues in order to fight against the other side (Sinclair 2006). Within four
decades, American politics has become much different from what it was.
Before the late 1990s, political scientists could have never imagined such high levels of
party cohesion appearing in the United States. In most of literature on American Congress and
legislative behavior, American politics was characterized by candidate-centered elections in
which party labels and party votes were not decisive to electoral outcomes (Carey and Shugart
1995; Mayhew 1974); by individual legislators that focused on constituency service, acting as if
there was no party discipline and frequently voting across party lines in roll call votes (Mayhew
1974; Fiorina 1977; Sinclair 2006) and by within-party heterogeneity that often split each party
on important issues (Aldrich 1995; Cox and McCubbins 1993; Cox and McCubbins 2005;
Sinclair 2006).
Observing these phenomena, scholars were worried by the consequences of low party
cohesion. In 1950, American Political Science Association (a.k.a. APSA) addressed such a
concern in a report titled as ―Moving toward Responsible Party System.‖ This report viewed that
―when both parties are weakened or confused by internal divisions or ineffective organization, it
is the nation that suffers‖ (APSA 1950, 17). As the report stated, ―either major party, when in
power, is ill-equipped to organize its members in the legislative and the executive branches into a
government held together and guided by the party program,‖ and ―Party responsibility at the
polls thus tends to vanish.‖ The report even considered that this problem ―affects the very
heartbeat of American democracy‖ (APSA 1950, 5). The solution proposed by the report is a
9

party system that is ―democratic, responsible and effective—a system that is accountable to the
public, respects and expresses differences of opinion, and is able to cope with the great problems
of modern government.‖ Such a system requires ―first, that the parties are able to bring forth
programs to which they commit themselves and, second, that the parties possess sufficient
internal cohesion to carry out these programs‖ (APSA 1950, 17-8). To sum up, the APSA‘s report
stressed the importance of increasing party cohesion in order for the United States to have a
responsible party system.
A model of the system that the APSA‘s report urged to establish can be found in the United
Kingdom. Cox (1987) studies the history of the UK‘s party system in order to explain why
members of Parliament (MPs), unlike American members of Congress (MCs), vote loyally with
their party and why voters have a strong tendency to vote for parties instead of individual
legislators. He finds that the ―efficient secret‖—the fusion of the executive and legislative
powers in the Cabinet, as well as the extension of suffrage, are the driving forces. 1 In particular,
the Cabinet, with the institutional rights to threaten resignation and dissolving Parliament, may
pressure in-party MPs to vote cohesively in supporting the Cabinet‘s bills, lead to high levels of
party cohesion in the U.K. Furthermore, given high levels of party cohesion, the role of
individual legislators diminished. Such high levels of party cohesion in Parliament was causally
related to why the electorate recognized the dominance of party in determining legislative
outcomes and hence voted for candidates on the basis of choices between parties. All in all, due
to the secret of the Parliamentary system, the U.K. has developed party-based or party-oriented
politics.
For Americans, having a level of party cohesion that is comparable to the UK seemed like a
1

Bagehot (1865) proclaims that the fusion of executive and legislative powers is the secret that makes the British
government run efficiently.
10

dream that would never come true, given the lack of ―efficient secret‖ inherent in American
political institutions. Sharing the same concern with APSA and Cox, pundits have worked to
study explanations as to why party cohesion is low in the U.S., and how it can be increased
(Aldrich and Rohde 2000a; Aldrich and Rohde 2001; Cox and McCubbins 1993; Cox and
McCubbins 2005; Ladwig 2010; Rohde 1991). The focus of research on party cohesion has been
put on answering a ―how-so‖ question. However, as mentioned, party cohesion in the U.S. has
recently increased, and parties have become important actors in American politics. This recent
development in the U.S. indicates that the ―dream‖ has surprisingly come true. It is time for
political scientists to shift gears and answer a ―so-what‖ question.
To be clear, the recent development provides a great opportunity to empirically investigate
whether it is desirable for Americans to have high levels of party cohesion. In particular, both the
APSA‘s report and Cox‘s work indicate that party cohesion is closely related to how a
government and individual legislators are held accountable. Several questions can be derived
from these studies: Are American voters more able to hold the incumbent party accountable for
the public interest today than before? Is there party-oriented accountability in the making? Does
party-oriented

accountability hinder other aspects

of

accountability (e.g. economic

accountability)? Do American voters punish or reward their legislators after observing the
legislators voting cohesively with the party? The increases in party cohesion in a given political
system make it possible to answer these questions and test whether accountability varies with
levels of party cohesion
The above investigations require looking into the variation of party cohesion that can be
observed by voters. Accordingly, at the level of the party, I measure party cohesion as the
average percentage of legislators from a party voting together in each roll call vote. As for party
cohesion at the level of individual legislators, I measure party loyalty as the percentage of votes
11

in which a legislator has voted with the majority of the party out of all (contested) votes during a
term (a.k.a. Party Unity Scores). These measurements capture the idea of party cohesion that
directly affect policymaking and will be observed by voters. Although these measurements have
some problems with measuring the underlying ideological cohesiveness and might fail to
distinguish between different sources of observed cohesion (i.e. agenda setting, disciplinary
actions of a party, ideological cohesiveness and etc), they are valid measurements when the focus
is on the revealed behavior and its consequences.2
With the revealed party cohesion in the U.S. House of Representatives, my dissertation
consists of three essays that dissect the impacts of party cohesion on accountability. The three
essays can be connected by a single overarching story. Unlike the United Kingdom, the United
States does not have the ―efficient secret‖ that may lead voters to realize the importance of party
and drive them to vote for parties rather than individual legislators. However, the rise of income
inequality has driven American voters to care about the success of the party with which they
identify, despite the lack of institutions that may induce party voting. Due to this impact of
income inequality, voters are more likely to reward legislators who are loyal to their party and
punish the dissenting ones. It is then less costly for legislators to vote loyally with their party,
and legislators may be more willing to toe the party line.
The continuing rise of party cohesion ignited party wars in the Capitol and made the party
line between two parties clear. Due to the proliferation of partisan cues, voters are less likely to
hold the incumbent party accountable for the economy when there is a clear party line. In other
words, high levels of party cohesion may hinder voters from holding the incumbent party
accountable for the economy. What is worse, with a high level of party cohesion, an incumbent
2

Examples of the problems with party unity scores can be seen in Crespin, Rohde and Vander Wielen (2011),
Desposato (2005) and Krehbiel (2000).
12

party may be more capable of carrying out policies that target particular subsets of voters. In
return, these targeted voters may give the incumbent party disproportionately large support that
may cover the loss of votes from the failure to serve the public interest. As such, high levels of
party cohesion may enable ruling parties to resist being held accountable to the public interest.
In the next section, I briefly summarize the puzzle, argument, and methods in each of the
essays. These summaries will further show how these essays will contribute to the studies of
accountability.

13

Chapters
The first chapter is an essay titled Income Inequality and Electoral Costs of Party Loyalty in the
United States, 1976-2008. The United States features a political system where party loyalty has
been traditionally seen as electorally costly to individual legislators (Carey and Shugart 1995;
Carey 2009; Carson et. al. 2010). However, if being loyal to the party is costly, why did
legislators in the U.S. increase their party loyalty to a high level during the past decades? Instead
of believing that they were committing political suicide, I argue that the electoral cost of party
loyalty could be reduced so that a higher level of party loyalty would not significantly increase
the legislators‘ chances of losing elections. In addition, I suggest that such a reduction of
electoral cost was resulted from the increases in income inequality. As income inequality
increases, issue cleavages are enlarged, voters sort themselves into the two parties, and party
activists gain more leverage in the elections. Consequently, more voters care about the success of
the party and hence prefer a loyal legislator to a dissenting one. The evidence from the U.S.
House elections between 1976 and 2008 indicates that party loyalty becomes less costly when
income inequality increases. This finding may not only complement the existing theories of party
government, but also shatter the conventional wisdom on the role of the party in personal-vote
systems. Furthermore, this finding shows that income inequality affects the way voters hold
legislators accountable and suggests a new direction for future research on the politics of income
inequality. Last, this finding suggests that politically polarized behavior by the elites (e.g. House
Representatives standing firmly with their party in the House) is punished less than before,
implying that polarization also exists among the citizens. This may shed light on the debate over
whether polarization exists only among the elites.
The second essay is titled Clarity of Responsibility and Clarity of Party Line: The Impacts
of Party Cohesion on Economic Voting in the United States, 1980-2008. Powell and Whitten
14

(1993) maintain that governing party cohesion is a stable feature of political systems that
enhances clarity of responsibility and economic voting. However, party cohesion is not stable,
and there are ebbs and flows in party cohesion. Their study fails to take into account the variation
of party cohesion across time. In addition, they only consider governing party cohesion, while
they neglect the role of opposition party cohesion. Given that responsibility can not be clear as
many legislators from the opposition party work with the incumbent party, opposition party
cohesion should be taken into account in discussing clarity of responsibility. More importantly,
the interaction between incumbent and opposition party cohesion can not be explored without
including opposition party cohesion into analyses. The interaction between the two parties‘
cohesion may affect economic voting through a mechanism different from clarity of
responsibility. Research finds that at the micro level, voters respond to partisan cues and become
more partisan when the elites from each party stand with their in-party members (Hetheington
2001; Levendusky 2009; Rudolph 2006). In addition, it has been observed that economic
conditions matter less when there are more partisan voters (Kayser and Wlezien 2011).
Accordingly, I argue that a high clarity of party line, which is a result of concurrent high levels of
incumbent and opposition party cohesion, may lead to a weak relationship between voters‘
economic perceptions and voting intentions (a.k.a. weaker economic voting). With data from the
American National Election Studies (ANES) between 1980 and 2008, this essay demonstrates
that party cohesion is a double-edged sword. While any increases in either party‘s cohesion
might enhance clarity of responsibility, they simultaneously improve clarity of the party line. The
consequence is a curvilinear relationship between party cohesion and economic voting. As party
cohesion increases from a low to mid level, it may strengthen economic voting by enhancing
clarity of responsibility. When it is at a high level and party line becomes clear, it may hinder
economic voting. These findings not only contribute to our understanding of economic voting,
15

but also offer a new insight that high levels of party cohesion may lead to party-oriented
accountability and hinder economic accountability.
Finally, the essay, Resisting Being Held Accountable: The Impact of Party Cohesion on
Concentration of Votes in the United States, 1992-2012, is the third chapter. Ideally, politicians
should be held accountable by all the citizens for the best interest of the whole public. In reality,
one way for the politicians to resist public accountability is to earn sufficient votes from some
subsets of the electorate through providing particularistic services and goods. Party cohesion has
been seen as a constraint on such particularistic exchanges between politicians and voters (APSA
1950; Cox 1987; Shugart and Carey 1992). However, using the concentration of votes as an
indicator of particularistic exchanges between parties and voters, I find that party cohesion has a
positive impact on the concentration of votes in presidential elections between 1992 and 2012.
This finding suggests that a smaller range of the electorate is served by the incumbent party
when the party is more cohesive. I then continue to demonstrate that the concentration of votes
received by an incumbent party positively affects the vote share of the party. Hence, a higher
level of party cohesion does not improve accountability. Instead, it contributes to the ability of a
party to resist being held accountable to the interest of the public.
Through these three essays, my dissertation provides empirical evidence that helps to clarify
the relationship between party cohesion and democratic accountability. In particular, my
dissertation takes into account the variation of party cohesion across time in studying the
trade-offs between various types of accountability. It further demonstrates that high levels of
party cohesion do not affect the accountability in the way that the literature has suggested.
Instead, party-oriented accountability is improved by the rise of party cohesion, while the ideal
accountability—the extent to which citizens hold ruling elites accountable for the public interest
is diminishing in contemporary American democracy.
16

CHAPTER 1
INCOME INEQUALITY AND THE ELECTORAL COST OF PARTY LOYALTY IN THE
UNITED STATES 1976-2008

In the United States, which has a personal-vote electoral system, parties and party labels are
deemed as less important in elections than the cultivation of personalistic support (Carey and
Shugart 1995). Thus, legislators may not gain substantial electoral benefits from showing loyalty
to their party.3 In addition, voting loyally with one‘s party in Congress frequently conflicts with
serving the interest of local constituencies (Ansolabehere, Snyder, and Stewart 2001;
Canes-Wrone, Brady, and Cogan 2002; Carroll and Kim 2009; Carson 2005; Lebo, McGlynn and
Koger 2007). Therefore, party loyalty has been traditionally considered and empirically
demonstrated to be costly to legislators in the U.S. (Carson et al. 2010). Given that legislators
who vote loyally with their party in Congress may risk their personal votes and even their seats,
the U.S. political system has long been criticized for its lack of party cohesion.4
However, party loyalty in Congress has increased over recent years, which is at odds with
the conventional theory. If it is costly for legislators to vote loyally with their party, why would
many legislators have increased their party loyalty? While it may be the case that these

3

Carey and Shugart (1995) rank electoral systems by the extent to which a given system generates incentives to
cultivate personal votes. According to their ranking, the U.S. system is among those in which legislators have the
greatest incentive to cultivate their personal support for their own characteristics or constituency services, as
opposed to relying on party labels or party identification. Follow-up studies include those by Hicken and
Simmons (2008) and Keefer and Khemani (2009).

4

In this research, party loyalty means the extent to which an individual legislator votes loyally with his or her party
in Congress. It is measured as the percentage of roll call votes in which a legislator has voted with the majority of
the party out of all (contested) votes during a term. Additionally, I use the term party cohesion to refer to
aggregate loyalty in order to stress the difference between the aggregate loyalty of a party as a whole and the
loyalty of a given individual legislator. Aggregate loyalty means the overall level of the loyalty among the
members of a party (i.e. the mean or median of party loyalty in a party).
17

legislators were not single-mindedly seeking reelection, it is not likely that they were all
committing political suicide. An existing explanation in the literature is that the number of safe
districts has increased, and the representatives who feel ―safe‖ are more willing to risk losing
votes (Carson et al. 2010; Ladwig 2010, 500). While this argument sounds plausible, it does not
explain why more and more districts have become safe in the first place. Moreover, a legislator
from a safe district might not necessarily find it safe to toe the party line; neither party leaders
nor any individual legislator can precisely predict how many votes a legislator is losing in the
election for each roll call vote. In other words, no one is sure whether a legislator will still enjoy
safety in the next election. Hence, the existing literature is insufficient to explain why a legislator
would ever feel safe in voting loyally with the party if party loyalty is electorally costly.
This paper sheds new light on this puzzle by suggesting that increases in income inequality
may have changed the way the American voters hold legislators accountable for legislators‘ party
loyalty, leading to a reduction in the electoral cost of party loyalty. The literature on the politics
of income inequality has demonstrated that the widened gap between the rich and poor has
enlarged the issue cleavages among citizens (Gelman 2008), motivated voters to sort themselves
into the two parties (Gelman 2008; McCarty, Poole and Rothenthal 2006) and granted party
activists more leverage in elections. Accordingly, I argue that there would be more voters caring
about the success of the party with which they identify. These voters would appreciate, rather
than punish, loyal legislators. It follows that party loyalty would be less costly to legislators than
it was before. In other words, legislators are more loyal, not because they are more willing or
able to bear the electoral cost of party loyalty, but because the electoral cost of party loyalty has
lessened as income inequality has grown.
By investigating the association between income inequality and the electoral cost of party
loyalty in House elections from 1976 to 2008, I demonstrate that the electoral cost of party
18

loyalty decreases with income inequality.5 This finding echoes a quote from Jacobson and
Bovitz (1998, 21): ―Loyalty is evidently calculated; the general rule seems to be […] go with the
party only if the expected electoral penalty will not significantly increase your chances of losing
your seats.‖ In other words, legislators may vote loyally with their party in Congress when doing
so is not likely to cost them the next election. Compared to the previous literature, the findings in
this paper are more consistent with the observed rise of party loyalty in Congress.
Additional implications can be drawn from this paper. First, the findings in this paper imply
that a responsible party system is more likely to exist when there is a higher level of income
inequality. A responsible party system is one in which parties act cohesively to carry out party
platforms and can be held accountable collectively (APSA 1950; Ranney 1962); hence, the
government can be run in an efficient way (Cox 1987). While a responsible party system requires
a high level of party loyalty, the personal-vote system may constrain increases in party loyalty in
the U.S. However, the evidence here suggests that this constraint on party loyalty is less effective
when income inequality increases and the electoral cost of party loyalty is reduced. This finding
may not only complement the existing theories of party government, but also shatter the
conventional wisdom on the role of the party in personal-vote systems. Second, the findings
point out a political impact of income inequality that has not yet been fully explored. The
political impacts of income inequality in the United States and in the whole world have drawn
the attention of the political scientists and motivated a high volume of studies.6 As of today,
5

This period is the so-called post-reform era. During the 1970s, the U.S. experienced a series of congressional
reforms which increased the power of the leadership of the majority party and weakened committee chairmen
(Rohde 1991). These reforms contributed to a stronger party influence in Congress (Cox and McCubins 2005;
Cox and McCubbins 2007; Rohde 1991; Smith 2007). By focusing on this period, I am able to set aside the effect
of the congressional reforms on legislative behavior and look into the association between income inequality and
the electoral cost of party loyalty.

6

Income inequality has been examined as a factor affecting political participation (Gaventa 1980; Pateman 1971;
19

however, there is little work investigating the linkage between income inequality and the way in
which individual legislators are held accountable by voters. My paper indicates a new direction
for studies of income inequality. Third, the finding that politically polarized behavior by the
elites (e.g. House Representatives standing firmly with their party in the House) is punished less
than before implies that polarization also exists among the citizens. This may shed light on the
debate over whether polarization exists only among the elites.7

Schattscneider 1960; Solt 2008; Solt 2010), political polarization (Bartels 2004, 2008; Garand 2010; Gelman
2008; McCarty, Poole, and Rosenthal 2006), democratization (Acemoglu and Robinson 2006; Boix 2003) and
violence (Nepal, Bohara and Gawande 2011; Weede 1981).
7

There is an unsettled debate about political polarization in the U.S. While some scholars observe that American
citizens are polarized (Abramowitz and Saunders 1998, 2008), others argue that the polarization of citizens is a
myth (Fiorina, Abrams and Pope 2006). In addition, there are scholars taking the middle position that the elites
and the citizens are both polarized in one way or another (Gelman 2008; Levendusky 2009; McCarty, Poole and
Rosenthal 2006).
20

Costs and Benefits of Party Loyalty
It has been documented that legislators in the U.S. bear an electoral cost associated with their
party loyalty. The U.S. has a single member district (SMD) system with primary elections.
Legislators in such a personal-vote system are expected to have low party loyalty because party
votes are downplayed by legislators in elections and voting loyally with the party in Congress
may not benefit legislators (Carey and Shugart 1995). Empirically, scholars have also found that
the closer a legislator‘s ideological (policy) position is to the party, the fewer votes the legislator
gains in the next election (Ansolabehere, Snyder, and Stewart 2001; Canes-Wrone, Brady, and
Cogan 2002; Carson 2005). Arguing that ideology is different from party loyalty, Carson et al.
(2010) have directly demonstrated that party loyalty is electorally costly to legislators.
Despite the evidence of the electoral cost of party loyalty, some studies suggest that party
loyalty can also be electorally beneficial to legislators. Party cohesion (i.e. the aggregate loyalty
of a party‘s members) enhances the informativeness of a party label (Ashworth and Bueno de
Mesquita 2008), and this helps reduce voters‘ uncertainty about the candidates who carry that
party label (Snyder and Ting 2002). By extension, when a party is more cohesive, its party label
better provides voters with ―surety‖ about the individual candidates‘ actions in office (Grynaviski
2010), improving the credibility of candidates from that party (Adams, Merrill and Grofman
2005). A higher level of party loyalty in an individual candidate indicates a higher probability
that the candidate will help in carry out the party platforms. Conversely, a low level of loyalty
signals that the candidate will not be a reliable contributor to the success of the party platform.
Taken all together, party loyalty may bring both benefits and costs to a legislator. Table 1.1
shows that the correlation between legislators‘ party loyalty and vote shares changes over time.
From 1976 to 2008, the overall correlation between party loyalty and vote share is -.042 (p<.001),
suggesting that a higher level of party loyalty is associated with a smaller vote share. However,
21

there was a notable change during this period. From 1976 to 1984, the legislators with higher
loyalty indeed obtained a smaller vote share, as the literature suggests. Between 1986 and 1998,
the correlation coefficients were statistically insignificant. Starting from the election in 2000, the
correlations become positive and statistically significant. These changes indicate that party
loyalty may no longer be associated with a cost in terms of vote share. A roughly parallel
development is that the average and median party loyalty in the House gradually increased over
the period. If party loyalty were costly, any influences that drove party loyalty to rise would have
been offset by the electoral cost, and the increases in party loyalty would have been limited. The
rise in party loyalty thus adds to the evidence that the electoral cost of party loyalty may have
become smaller.

Table 1.1. Party Loyalty and Its Correlation with Vote Share
Term
Reelection Correlation
Median Party
Average Party
94
1976
-.243***
82.732(%)
75.238(%)
Year
Loyalty
Loyalty
95
1978
-.169**
80.154
73.935
96
1980
-.124*
84.086
77.177
97
1982
-.150**
83.221
77.526
98
1984
-.100*
85.862
80.315
99
1986
.003
86.680
82.577
100
1988
.068
88.568
84.114
101
1990
.006
86.802
82.562
102
1992
-.015
88.087
84.606
103
1994
.084
91.748
87.944
104
1996
.026
92.821
87.655
105
1998
.066
91.798
87.732
106
2000
.203***
91.729
87.831
107
2002
.151**
94.171
90.674
108
2004
.185***
94.658
92.244
109
2006
.115*
94.318
91.544
110
2008
.238***
96.022
93.199
Note: 1.σ: p<.1; *: p<.05; **:p<.01; ***:p<.001
2. The data for party loyalty is the party unity score from the Voteview.com
(2012), and the data for vote share is from the Office of the Clerk of
the U.S. House Representative (2012).

22

What could have led to these changes? According to the previous discussion, party loyalty
can be beneficial in that it signals how promising a legislator would help in carrying out party
platforms. Such an effect would take place and manifest only when voters care about whether
their favorable party platforms can be successfully carried out by the party. Thus, what led to the
proliferation of voters who care about the success of a party could be the same factor that led to
the reduction in the electoral cost of party loyalty. As follows, I argue and demonstrate that
income inequality is one of the driving forces.

23

Income Inequality and Electoral Cost of Party Loyalty
Income inequality per se seems to be irrelevant to party loyalty, yet it is empirically associated
with the electoral cost of party loyalty. Before the late 1970s, income inequality in the U.S.
remained at a relatively low level. The value of the Gini Index, a commonly adopted measure of
income inequality, fluctuated between 35 and 37. Between 1976 and 2008, it increased from 35
to 44.8 Because the Gini Index may not indicate the segment of society where changes in the
income distribution occur, I also examine ―the ratio of the upper bound value of the ninth decile
(i.e. the 10% of people with highest income) to that of the first decile (i.e. the 10% of people
with lowest income)‖ (OECD iLibrary 2013). 9 This alternative measurement shows the
difference in income between the rich and poor. During the years under investigation, the ratio
increased from 3.69 to 4.89, and the average ratio is 4.34. The pattern is similar to that of the
Gini Index. These facts and Table 1 jointly indicate that during the rise of income inequality, the
negative association between party loyalty and the vote shares of individual legislators gradually
changed to a positive one. In other words, the electoral cost of party loyalty is smaller when there
is a higher level of income inequality.
Could income inequality and the electoral cost of party loyalty be endogenous? Or, is the
empirical association between the two a spurious relationship which results from other factors?
In numerous studies, the rising inequality in the world, as well as in the United States, is
attributed to globalization and technological advancements (Alderson and Nielsen 2002;

8

The Gini Index ranges from 0 to 100, with 0 indicating perfect equality and 100 indicating maximal inequality.
Perfect inequality means that every unit gets the same amount of income, while maximal inequality means that
one unit gets all of the income. The data is drawn from the US Census Bureau (2012). In this paper, I present
family income inequality, but the trend for household income inequality is quite similar.

9

The data for the decile ratio is taken from the Organization for Economic Co-Operation and Development (OECD)
and the description of the decile ratio is from OECD iLibrary (2013).
24

Blanchflower and Slaughter 1999; Charles-Coll 2011; Krueger 2012; Neckerman and Torche
2007). Although partisan manipulation of macroeconomic policies could strengthen or mitigate
the rising trend of income inequality, it has not fundamentally stopped the rise of income
inequality (Bartels 2008).10 Furthermore, there is no existing work indicating any other factors
that could theoretically or empirically affect both national income inequality and the
district-level relationship between party loyalty and vote shares. It is reasonable to hypothesize
that income inequality serves as an exogenous condition that affects the electoral cost of party
loyalty.
Income inequality and the electoral cost of party loyalty are not associated simply by chance;
theoretical mechanisms that link the two have been suggested by literature on the politics of
income inequality. In short, increases in income inequality may enlarge issue cleavages, sort
people into partisan groups and grant party activists more leverage in the elections. As a result of
these mechanisms, more voters care about the success of a party and prefer a loyal legislator to a
dissenting one, leading to a smaller electoral cost of party loyalty. In what follows, I will discuss
each of these mechanisms in turn.

Enlarging Issue Cleavages
Since most policies are involved in the deployment and allocation of resources in society, the
rich and poor may possess different policy preferences. Page, Bartels and Seawright (2013)
provide evidence that the rich indeed view various issues differently than other people. They also
10

Bartels (2008) argues that there are partisan differences in facilitation of income growth for high and low income
people. In short, Democratic governments help the income growth for the poor, while Republican governments
facilitate the income growth for the rich. However, in my view, if the partisan differences are the major drivers of
income inequality, income inequality should be fixed at a certain level because the effect of one party‘s policies
could be offset by those of the other party.
25

find that policy preferences over economic regulation and social welfare are closely associated
with those over other issues. As such, the rich and poor may have different preferences not only
for distributive or redistributive policies, but also for policies that are seemingly irrelevant to the
distribution of wealth.11
While the poor have become poorer, the rich richer, as President Obama has pointed out,
―The middle class has shrunk‖ (Krueger 2012, 1). When the gap between the rich and poor is
enlarged, so are the differences in policy preferences between these two groups (Gelman 2008).
Figure 1.1 shows a graphical illustration of this process. When income inequality is relatively
low, there are two distinct types of people with different policy preferences (shown as ○ and □),
as well as some middle class people (*) with moderate policy preferences. When income
inequality is high, the distance between the two distinct types is enlarged. In addition, those with
moderate preferences gradually disappear. Given that the middle class has shrunk and the
intersection of policies preferred by the groups at the extremes has become smaller, within each
group, there are fewer moderate members who might support policies favorable to the contesting
group. Meanwhile, even though heterogeneity may still exist within each group with regard to
preferences on issues that are arguably less relevant to income (e.g. moral issues), the majority of
each group supports policies that are strictly distinct from those preferred by the contesting
group.12 All in all, when there is a high level of income inequality, the policies preferred by the

11

Page, Bartels and Seawright (2013) examine attitudes toward issues that are not directly related to the economy,
including scientific research, environmental protection, homeland security, defense spending, and education. They
find that even on those issues, the rich have different views than the general public.

12

If issues that are irrelevant to income had a large influences on the politics, partisan sorting according to income
levels would not have been observed in the previous literature (Gelman 2008; McCarty, Poole and Rothenthal
2006). Further, it would not have been observed in this paper that income inequality conditions the electoral cost
of party loyalty. Given the empirical evidence presented in the literature and in this study, it is reasonable to state
that most voters are influenced by issues that are relevant to income.
26

rich are much more different from those preferred by the poor. A result of this development is
that not only rival policies, but also moderate ones, become less acceptable to voters at each
extreme.

○

○

○ ○○＊＊＊＊＊□□□

□

□

Low Income

High Income
○

○ ○○○＊＊＊＊＊□□ □ □

□

A. Low Income Inequality
○○○○ ○ ○○○

○

＊

□

□

□ □

Low Income

□

High Income
○○

○ ○

○

＊

□

□ □□□ □ □□ □

B. Increasing Income Inequality
Figure 1.1. Income Inequality and Issue Cleavages

Sorting People into Partisan Groups
Gelman (2008) and McCarty, Poole and Rothenthal (2006) observed that issue cleavages were
more partisan and that voters were sorting themselves into different parties. As they reported, the
partisanship among voters incrementally changed along the line separating the rich and poor. The
phenomenon observed involved not only more voters identifying themselves with parties, and
fewer voters identifying themselves as independents, but also voters grouping themselves into
different parties according to their differences in income. In other words, given that rival and
moderate policies are less acceptable, voters are more motivated to sort themselves into parties in
order to pursue favorable policies and prevent unfavorable ones from being carried out. It
follows that the success of the other party, which may bring about unfavorable policies, is less
acceptable to the voters of either party. As a result, more voters care about which party wins
elections (Stonecash 2006).
27

The enlarged gap in policy preferences means that voters on each end of the policy
spectrum have moved away from the center, and they would find the party on the other side more
distant. Meanwhile, since a high level of party loyalty signals that a legislator often stands with
the party‘s positions, voters also find themselves closer to an in-party legislator who votes
loyally with the party. The party loyalty of an in-party legislator is thus favorable to the voters
who have a stronger sense of being members of the party.

Granting Party Activists More Leverage
When income inequality increases, there is a lower level of political participation (Gaventa 1980;
Pateman 1971; Schattscneider 1960; Solt 2008; Solt 2010). While regular voters withdraw from
the elections, party activists can still be mobilized by parties to participate in politics (Grynaviski
2010). As a consequence, party activists gain greater leverage on the electoral outcomes.
Empirically, two pieces of evidence support that party activists have gained greater leverage
in elections. First, it has been argued that party activists exert great influence in primary elections,
motivating legislators to take positions close to those of their party (Ansolabehere, Snyder, and
Stewart 2001; Aranson and Ordeshook 1972; Brady, Han, and Pope 2007; Coleman 1971).13
However, if voters in general elections had been punishing loyal legislators, generating a
substantial cost of party loyalty, the increases in party loyalty in Congress due to primary
elections would have been limited. Second, if party loyalty had remained costly in general
elections, the reelection rate of legislators would have been lower, given that many legislators
have increased their party loyalty. However, the continuing high rate of reelection indicates
13

It has also been argued that primary elections constrain the influence of the party leadership on the nomination of
candidates, leading to smaller incentives for legislators to be loyal to the party (Carey and Shugart 1995; Mayhew
1974; Shugart and Carey 1992). Since primary elections are not the focus of this research, I leave this controversy
for future studies.
28

otherwise.14 Therefore, the influence of party activists may have been extended to general
elections. Compared to other voters, party activists are more likely to care about the success of
the party and to prefer a loyal legislator over a dissenting one. When these party activists gain
more leverage in elections, party loyalty becomes less costly to legislators.
Through the above mechanisms, increases in income inequality may shift the focus of
voters toward the success of the party. Voters concerned with the success of the party, especially
party activists, appreciate the party loyalty of in-party legislators. In addition, dissenting
legislators who have not firmly supported the party‘s positions are seen as out-party legislators or
―traitors‖ who deserve to be punished in elections.
To go beyond the prima facie association between income inequality and the electoral cost
of party loyalty, I test by regress analysis of whether the relationship between party loyalty and
vote share is conditional upon the level of income inequality. If such conditionality exists, a
higher level of party loyalty will cost individual legislators fewer votes when income inequality
is higher, while it will cost more votes when income inequality is lower. Accordingly, the
regression analysis is designed to test the following hypothesis:
When national income inequality is at a higher level, the party loyalty of an individual
legislator has a smaller negative effect on the vote share received by that legislator. When
national income inequality is at a lower level, the party loyalty of an individual legislator has a
larger negative effect on the vote share received by that legislator.

14

The reelection rate for House representatives has never been lower than 85% since 1964. Between 1976 and 2008,
the reelection rate was, on average, 94.59 % (OpenSecrets 2014, www.opensecrets.org/bigpicture/
reelect.php).
29

Regression Analysis
The dependent variable is the vote share of each individual legislator. It is calculated as the
number of votes received by a legislator divided by the total votes cast in the legislator‘s district
in a given election. I multiply this number by 100 to make it a percentage. The data for the vote
shares is obtained from the U.S. House of Representatives (2012).15 Given that my theory
involves a cross-level interaction, I conduct both pooled OLS regression and random-effect
multilevel analysis.16
The key independent variables are Party Loyalty, Income Inequality and the interaction
between them. Party Loyalty is defined as the extent to which a legislator acts together with the
party. I adopt the most common measurement of party loyalty, which takes the percentage of
votes in which a legislator has voted with the majority of the party out of all (contested) votes
during a term. Although this measurement, which is also known as the Party Unity Score, has
some problems with measuring the underlying ideological cohesiveness between a given
legislator and the rest of his party, it is still a valid measurement when the focus is on the
revealed behavior and its consequences.17 The main concern in this research is how voters
punish and reward individual legislators after they observe the extent to which the legislators of a
party act together to carry out the party platform. Therefore, the observed behavior is much more
relevant to the purpose of this research than the underlying cohesiveness. The source of the data
is Voteview.com, maintained by Keith Poole, which provides a complete record for roll call votes
15

Office of the Clerk, U.S. House of Representatives. 2012. http://clerk.house.gov/member_info/ electionInfo/
index.aspx.

16

The random-effect model is a random-slope model because my theory involves the interaction between party
loyalty at the level of individual legislators and income inequality at the level of election years. However, the
results are not different in a meaningful way when random-intercept estimation is applied.

17

Examples of the problems with party unity scores can be seen in Crespin, Rohde and Vander Wielen (2011),
Desposato (2005) and Krehbiel (2000).
30

from 1879 to 2010. As shown in Table 1, from the 94th to the 110th House of Representatives, the
median party loyalty increased from 82% to 96%, and the average increased from 75% to 93%.
As mentioned, Income Inequality is measured by the Gini Index and the decile ratio of income
between the 90th percentile and the 10th percentile.18 I will present two separate sets of models
using the two different measurements. To test the hypothesis that the relationship between party
loyalty and vote share is conditional upon income inequality, I include the interaction between
Party Loyalty and Income Inequality. I expect that the negative effect of party loyalty will be
reduced by increases in income inequality, and therefore, the interaction should have a positive
coefficient.
In addition to the key independent variables of interest, I also control for some variables that
have been noted as relevant in the literature. First of all, as in most of the literature, I include
Ideological Extremism of each legislator, transformed from DW-NOMINATE scores. These
scores measure legislators‘ voting tendencies on the liberal-conservative spectrum. Carson et al.
(2010) argue that in voters‘ evaluations, party loyalty and ideology are two different dimensions,
and as noted before, ideological extremism has been seen as negatively related to the electoral
fortune of legislators. The data for DW-NOMINATE scores is drawn from Voteview.com, and I
take the absolute values of the scores so that a higher score indicates a more extreme ideology,
no matter whether the legislator is liberal or conservative.
In order to control for the personal characteristics of incumbent legislators, I include
Freshman, Party Leader and Previous Vote Share. These variables represent factors
conventionally regarded as affecting congressional elections. First, Freshman is coded as 1 if the
incumbent legislator is a freshman running for his or her first reelection, and 0 otherwise. The
18

As noted, I use the Gini Index for family income inequality, but using household income inequality does not
change the results of the regression analysis in a meaningful way.
31

variable indicates both seniority and electoral experience. I hypothesize that a freshman is
electorally vulnerable compared to other legislators. The data is drawn from Gary Jacobson‘s
dataset.19 Second, Party Leader is coded as 1 if the legislator is a party leader (i.e., Speaker,
majority leader, minority leader, whip, Democratic caucus chair or Republican conference chair),
and is coded as 0 otherwise. The data is taken from the website of the Clerk of the House of
Representatives. Since party leaders may have to spend more time and effort in the Capitol and
less in the districts, I hypothesize that they are more vulnerable than other legislators in the
elections. Third, Previous Vote Share is included to control for the effect of an individual
legislator‘s previous electoral results and the baseline variation of vote share among legislators.
Further, I control for the characteristics of the electoral competition faced by a given
legislator by including the variables Challenger’s Quality, Midterm and Presidential Approval.
Challenger’s Quality measures another crucial factor in U.S. Congressional elections (Carson
2005; Jacobson 2004). When facing an experienced challenger, incumbents obtain fewer votes
than they would otherwise. I include Jacobson‘s experienced candidate, which is a dummy
variable with 1 indicating that the challenger(s) previously held elected office and 0 otherwise.
At the national level, midterm elections (Midterm) and Presidential Approval have been thought
to have an effect on congressional elections. Compared to elections in the presidential election
years, midterm elections are thought to be more difficult for the legislators of the president‘s
party (hereafter, ―in-party legislators‖), but easier for those of the other party (hereafter,
―out-party legislators‖) (Erikson 1988). I code the variable Midterm as 0 when a given election is
in the year of a presidential election, 1 when it is a midterm election for an in-party legislator and
-1 when it is a midterm election for an out-party legislator. Presidential Approval is expected to
have a positive effect on the vote share of the legislators who are in the president‘s party.
19

I deeply appreciate Prof. Jacobson‘s generosity in offering his dataset and allowing me to use it in this research.
32

Conversely, it should have a negative impact on the vote share of the legislators from the other
party. I take the presidential approval rating that was measured the closest to election day for
each House election from Gallup (2013), and I subtract 50 from each number. I multiply the
resulting numbers by -1 for the out-party legislators.
Last, the economic conditions at the national level during the election years, in terms of
logged GDP per capita based on purchasing power parity (hereafter, Log GDP) and the
unemployment rate (Unemployment), are controlled for in the model as well. These two variables
are also coded with consideration of the difference between in-party and out-party legislators; I
multiply the values by -1 for those who are not in the president‘s party. I thus expect to find a
positive coefficient for Log GDP and a negative one for Unemployment. The former means that a
greater GDP is expected to help in-party legislators, while it costs out-party legislators votes. The
latter means that a higher unemployment rate is expected to cost in-party legislators votes, while
it benefits out-party legislators.20 The data for GDP is taken from the World Development
Indicators, and the data for unemployment rates is from the Bureau of Labor Statistics.21

Findings
In Table 1.2, Model Ⅰ, Ⅱ and Ⅲ are the models with income inequality at the national level
measured by the Gini Index, while Model Ⅳ and Model Ⅴ are the models with income
inequality measured by the decile ratio. Model Ⅰ is a pooled OLS regression without any
control variables. The negative coefficient for party loyalty and the positive coefficient for the
interaction between party loyalty and income inequality suggest that the electoral cost of party
20

I follow Carson et al. (2010) in coding Midterm, Presidential Approval, Log GDP and Unemployment so as to
account for the partisan differences in the effects.

21

The data for GDP is taken from NationMaster.com (2012) and The World Bank (2012), and the unemployment
rates are from Bureau of Labor Statistics (2012).
33

loyalty can be mitigated by increases in income inequality.
To take the analysis a step further, Model Ⅱ includes all the control variables. The
conditioning effect of income inequality remains. Namely, the effect of party loyalty differs
depending on the levels of income inequality. The negative effect of party loyalty (i.e., the
electoral cost of party loyalty) can be reduced by an increase in income inequality. According to
the coefficients, when the Gini Index is as low as 35, a one-unit increase in party loyalty costs a
legislator .151 percent of votes. As the Gini index increases by one unit, the cost of party loyalty
decreases by .016 percent. When the Gini Index increases to 44, a one-unit increase in party
loyalty only costs a legislator .007 percent of votes─the electoral cost becomes subtle. By
extension, other things being equal, in an election year in which the Gini Index is greater than
44.44, the effect of party loyalty will turn out to be positive. Model Ⅲ is a random-effect model
with the same variables that are included in Model Ⅱ. The coefficients are significant and
similar, which again supports my hypothesis.
Model Ⅳ is a pooled OLS regression with income inequality measured as the 90/10 decile
ratio. In this model, a similar relationship exists, except that the interaction between party loyalty
and income inequality is only significant at the .1 level. Model Ⅴ is a random-effect model using
the decile ratio. In this model, the interaction term is significant at .01 the level. Two
implications can be drawn from these models. First, even with a different measurement of
income inequality, the negative effect of party loyalty is reduced as income inequality increases.
Second, these coefficients suggest that when the people at the 90th percentile of income earn 4.80
times more than those at the 10th percentile, party loyalty will have a positive effect on legislators‘
vote share.

34

Table 1.2. Regression Analysis for the U.S. House Elections
Model Ⅰ

Model Ⅱ

Β (CRSE)
Loyalty
Inequality
Loyalty×Inequalit
y
Ideology
Freshman
Party Leader
Previous Vote
share
Challenger‘s
Quality

Β(CRSE)

Β (CRSE)

-1.54 (.23)***

-.71 (.29)*

-.76 (.18)***

-.48 (.22)*

-3.65 (.56)***

-1.69 (.68)*

-1.78 (.38)*** -11.20 (4.95)*

.04 (.01)***

.02 (.01)*

.02 (.01)***

.10 (.05) σ

Model Ⅴ
(RE)
β (SE)
-.54 (.14)***
-12.30
(2.94)***
.12 (.04)**

5.31 (2.29)* -5.70 (1.33)*** 5.65 (2.43)* 5.81(1.33)***
2.5 (.59)**
2.40 (.42)*** 2.43 (.58)** 2.41 (.42)***
-2.09 (1.13) σ -1.84 (1.12) -2.09 (1.14) σ -1.82 (1.12)
.47 (.03)***

.48 (.01)***

.47 (.03)***

.48 (.01)***

-7.93 (.51)*** -7.56 (.39)*** -7.98 (.51)*** -7.57 (.39)***
-3.56 (.86)**

-3.67 (.31)*** -3.55 (.86)**

-3.66 (.31)***

.14 (.04)**

.16 (.02)***

.140 (.04)**

.16 (.02)***

.10 (.19)

.10 (.09)

.12 (.19)

.11 (.09)

.01 (.12)

.01 (.06)

-.00 (.12)

.01 (.06)

106.98(27.33)* 110.03(14.85)*

Constant

R2

Model Ⅳ

(RE)
Β (SE)

Midterm
Presidential
Approval
Unemployment
Rate
Log GDP Per
Capita

N
N of Clusters

Model Ⅲ

87.35

*

**

(21.34)**

6548
17

6548
17

6548
17

6548
17

.02

.33

91.54
(12.22)***
6548
17

.33

sd (pcoh)

.02 (.01)

.02 (.01)

sd (cons)

1.72 (.83)

1.79 (.76)

sd (residual)

11.41 (.10)

11.42 (.10)

Chi Square(2)
Note: σ: p<.1; *: p<.05; **:p<.01; ***:p<.001

255.07***

245.76***

35

Figure 1.2 graphically illustrates the effect of party loyalty using the results in Models Ⅱ
and Ⅳ. The lines with a negative slope indicate that for the designated value of the Gini Index,
party loyalty is associated with an electoral cost; a legislator who shows a higher level of party
loyalty receives a smaller share in the election. A flatter line thus represents a smaller electoral
cost of party loyalty. In the left panel, party loyalty is less costly when the Gini Index has a
greater value. Although the maximum value of the Gini Index in this dataset is 44, this model
predicts that party loyalty could be beneficial to legislators when income inequality is 45 or
higher.22 The right panel displays a similar pattern from using the decile ratio to measure income
inequality. As the decile ratio increases from 3.6 to 4.8, the line flattens and gains a slight
positive slope. The two panels both indicate that the cost of party loyalty decreases as income
inequality rises.
Like the key independent variables, most of the control variables affect vote share in the
expected way. Across all of the models, the coefficients for Income Inequality are significant and
negative. Considering the interaction between income inequality and party loyalty, for the
legislators with a higher level of party loyalty, the negative effect of income inequality is smaller.
This finding is consistent with my theory that a loyal legislator is more favorable to voters when
income inequality is higher. The coefficients for legislators‘ ideology are positive and significant
across all the models. According to Carson et al. (2010), ideology may not affect vote shares
when party loyalty is included. However, ideology has a positive effect when both it and party
loyalty are included in the models. Among the other variables, Freshman, Challenger’s Quality,
Previous Vote Share, Midtem and Presidential Approval show significant effects. Freshman has

22

For simplicity, the difference in the Gini Index between the lines in the figure is held consistent at 2; this is why I
graphed a line for the value 45 instead of the maximum value in my dataset. However, this line is not meaningless,
given that 45 is the value of the Gini Index for 2011, and the value for 2012 is slightly higher than 45.
36

positive and statistically significant coefficients, meaning that a freshman in Congress running
for his or her first reelection has some advantage compared to senior legislators. Challenger’s
Quality has a negative and statistically significant effect, confirming the findings in the literature
that a legislator facing an experienced challenger receives a smaller share of votes (Jacobson
2004; Carson 2005; Carson et al. 2010). Further, as expected, the vote share received by a
legislator in the previous election has a positive effect on the vote share received in the present
election. As for Midterm, the coefficients show that in-party legislators receive smaller shares of
votes in midterm elections than in presidential election years, while out-party legislators receive
larger shares of votes in the midterm elections. The positive signs of the coefficients for
Presidential Approval indicate that a higher presidential approval rate benefits in-party
legislators but makes out-party legislators suffer.

37

63

63

Ratio (90/10)=3.6

62

62

Gini=35

Gini=37

61

Predicted Vote-Share

61

Ratio (90/10)=3.9

59

Ratio (90/10)=4.5

59

Gini=41

Ratio (90/10)=4.2

60

60

Gini=39

58

58

Gini=43

Ratio (90/10)=4.8

57

57

Gini=45

70

75

80

85

90

95

100

70

75

80

85

90

95

100

Party Loyalty

Party Loyalty

Figure 1.2. The Cost of Party Loyalty, by Income Inequality
Note:

Ideology and Previous Vote Share are set at their means; Party Leader and Freshman are
set to be 0; Midterm and Quality Challenger are set to be 1; Log GDP and
Unemployment rate are set at their means for In-party legislators; Presidential Approval
is set to be about 1 point above 50 for In-party legislators.

Three of the control variables are not significant at the .05 level. Party Leader is only
significant at the .1 level in the pooled models, but the negative signs indicate that being a party
leader may indeed shift the focus of a legislator away from serving constituents, which in turn
costs him or her votes. The coefficients for the two economic variables are not statistically
significant, suggesting that national economic conditions may not affect the vote shares of
individual legislators in House elections.
To summarize the findings, the relationship between party loyalty and vote shares is found
to be conditional on income inequality. The negative effect of party loyalty on vote share─the
electoral cost of party loyalty─is smaller when income inequality is higher. This finding is
logically consistent with the increasing trend of party loyalty. In other words, as legislators
38

increased their party loyalty over the past few decades, the electoral cost incurred by party
loyalty was also decreasing due to the rising income inequality.

39

Robustness
To ensure the robustness of the finding that the electoral cost of party loyalty is conditional on
income inequality, I perform additional tests with alternative model specifications below. Overall,
the effect of income inequality on the electoral cost of party loyalty is shown to be robust in
these models.

Party Polarization and Aggregate Party Loyalty
Party polarization at the elite level took place at the same time as the rise in income inequality
and the decrease in the electoral cost of party loyalty. Hence, the effect of income inequality on
the electoral cost of party loyalty might result from party polarization among elites. To account
for this possibility, I include the variables Aggregate Party Loyalty and the Polarization Index,
which measure party polarization on two different dimensions.23
First, as previously noted, Bartels (2008) asserted that there are partisan differences in the
facilitation of income growth for the rich and poor because the two parties implement different
macroeconomic policies. In my view, such partisan differences, if they exist, would manifest
when the legislators from the party of the president vote loyally with the party enough that
policies that affect income growth can be carried out. In addition, aggregate party loyalty might
also affect legislators‘ vote share by enhancing the informativeness and credibility of the party
label. To account for the possibility that aggregate party loyalty might affect both income
inequality and vote share, I include the average party loyalty of a given legislator‘s party for each
election year. 24 Second, ideological polarization among elites may further affect income
23

Aggregate party loyalty may show the extent to which the legislators from a party vote together to support the
party‘s position, while the polarization index shows how far one party‘s position is from the other party‘s position.

24

As an alternative, I also measured aggregate party loyalty as the average party loyalty of the president‘s party, and
multiplied it by -1 for the legislators from the opposition party in order to account for its different effects on
40

inequality and how voters view the elites in Congress. I include the Polarization Index that has
been applied by McCarty, Poole and Rosenthal (2006) to measure ideological polarization in
Congress. Specifically, the index measures the distance between the two parties‘ median
DW-NOMINATE scores in the House; the data is taken from the Voteview.Com.
In Table 1.3, the interaction term remains to be statistically significant after including
Aggregate Party Loyalty and Polarization (Model Ⅰ). This result suggests that the effect of
income inequality is robust to the inclusion of aggregate party loyalty and ideological
polarization. Moreover, the relationship between income inequality and the electoral cost of party
loyalty does not result from the elites‘ behavior in Congress.25

Partisanship of Constituencies
Another concern to be addressed is whether the electoral cost of party loyalty is conditional on
district partisanship instead of income inequality. Carson et al (2010) have found that district
partisanship, measured as the two-party vote share of a given legislator‘s party in the most recent
presidential election, is a condition that may reduce the cost of party loyalty. However, district
partisanship and the extent to which a loyal legislator is appreciated would be driven by the same
underlying desire for the success of the party. For example, the voters who desire the success of
the Democratic Party would support the Democratic presidential candidate, showing stronger
partisanship. These voters would also prefer a loyal legislator to a dissenting one, leading to a
smaller electoral cost of party loyalty for the incumbent legislator. Therefore, district partisanship
in-party and out-party legislators‘ vote share. Another alternative measure is the average party loyalty of the
president‘s party without consideration of the party affiliation of legislators. Across these models, the interaction
between party loyalty and income inequality remains statistically significant, while the coefficient for aggregate
party loyalty fails to reach the conventional level of significance (p>.05).
25

I present pooled models here, but the results of random-effect models are not meaningfully different.
41

and the electoral cost of party loyalty are two sides of the same coin.
A corresponding fact is that district partisanship has risen along with income inequality:
when income inequality is higher in a given election year, the average district partisanship is
stronger. The average district partisanship increased from 55.24 % in 1976 to 60.19 % in 2008 as
the Gini Index increased from 35.80 to 44.00. The correlation between average partisanship and
the Gini Index is as high as .90 (p<.001). Given that income inequality is largely driven by
non-political factors such as technological advancement and globalization, it is very likely that
this association is due to income inequality driving the average partisanship up, but not the other
way around.
Since district partisanship could be one of the mechanisms linking income inequality to the
electoral cost of party loyalty (by means of partisan sorting), including district partisanship in the
models would hold constant the factor linking income inequality to the electoral cost of party
loyalty. One can easily imagine that the interaction between party loyalty and income inequality
would become insignificant when district partisanship is included in the models.
To try to account for the effect of district partisanship, I recode it into a dummy variable
denoting the discrete level of partisanship. Partisan District is coded as 1 if a district‘s
partisanship is greater than or equal to the average partisanship of a given election year, and 0
otherwise. Although district partisanship may go hand-in-hand with income inequality, whether a
district is heavily partisan may be less determined by income inequality than is the exact
presidential vote share of the legislator‘s party in that district. Thus, including this dummy
variable may help account for the effect of district partisanship without completely partialling out
the effect of income inequality through district partisanship.
In the pooled model (Model Ⅱ), the coefficients for income inequality and its interaction
with party loyalty are at best significant at the .1 level. However, taking into account the
42

heterogeneity across election years, in the random-effect model in the third column, the
coefficient of the interaction term remains statistically significant. Model Ⅳ and Model Ⅴ show
similar results when income inequality is measured by the 90/10 decile ratio. As these models
show, the inclusion of district partisanship may hold the mechanism linking income inequality
and the electoral cost of party loyalty constant, mitigating the effect of income inequality.
However, it can still be observed that the electoral cost of party loyalty is conditional on income
inequality.

43

Table 1.3. Models with More Control Variables

Loyalty
Inequality
Loyalty×Inequalit
y
Ideology
Freshman
Party Leader
Previous Vote
share
Challenger‘s
Quality
Partisan District
Loyalty×Partisan
District
Aggregate Party
Loyalty
Polarization
Midterm
Presidential
Approval
Unemployment
Rate
Log GDP Per
Capita
Constant
N
N of Clusters
R2

Model Ⅲ

Model I

Model Ⅱ

Β (CRSE)

Β (CRSE)

(RE)
Β (SE)

-.77 (.23)**
-1.51 (.89)

-.64 (.28)*
-1.37 (.66) σ

-.62 (.18)**
1.79 (.92) σ

.02 (.01)**

.01 (.01) σ

.01 (.01)**

Model Ⅳ
Β (CRSE)

Model Ⅴ
(RE)
β (SE)

-.33 (.22)
-.45 (.14)**
11.59 (5.99) σ 11.25 (5.57)*
.05 (.05)

.08 (.04)*

5.11 (1.96)*
1.37 (2.43)
1.83 (1.35)
1.52 (2.24)
1.93 (1.35)
2.44 (.60)** 2.35 (.53)*** 2.31 (.41)*** 2.35 (.52)*** 2.32 (.41)***
-2.12 (1.13) σ -2.14 (1.12) σ -1.86 (1.11) σ -2.08 (1.12) σ -1.85 (1.11) σ
.47 (.03)***

.44 (.03)***

.45 (.01)***

.44 (.03)***

.45 (.01)***

-7.92 (.53)*** -7.49 (.55)*** -7.14 (.39)*** -7.59 (.58)*** -7.15 (.31)***
-3.75 (1.95) σ -4.59 (1.78)*
.09 (.03)**
-.09 (.23)
-2.41 (10.81)
-3.50 (.84)**

-3.13 (2.14)

-4.46 (1.78)*

.10 (.02)***

.09 (.03)**

.10 (.02)***

.10 (.07)

.18 (.23)

.10 (.07)

5.41 (15.50) 5.79 (12.26) 2.35 (11.70)
-3.42 (.92)** -3.60 (.31)*** -3.50 (.98)** -3.60 (.31)***

.13 (.04)**

.08 (.05)

.11 (.02)***

.09 (.05) σ

.11 (.02)***

.10 (.19)

-.05 (.20)

-.05 (.09)

-.03 (.21)

-.05 (.09)

-.00 (.13)

.06 (.13)

.07 (.06)

.07 (.14)

.07 (.06)

108.71
(33.61)**

97.18
(26.41)**

104.72
(27.40)***

75.71 (26.61)*

6548
17

6548
17

6548
17

6548
17

.33

.35

82.91
(17.80)***
6548
17

.35

sd (pcoh)

.03 (.01)
44

.02 (.01)

Table 1.3 (cont‘d)
sd (cons)

1.86 (.83)

1.99 (.75)

sd (residual)
11.26 (.10)
LR test, Chi
261.2***
Square(2)
Note: σ: p<.1; *: p<.05; **:p<.01; ***:p<.001.

11.26 (.10)
239.98***

Monotonically Rising Income Inequality?
Since income inequality increased roughly monotonically along with the changes in party loyalty
and the electoral cost of party loyalty, one might ask whether there is another increasing trend
leading to the rise of income inequality and the reduction in the electoral cost of party loyalty.
For several reasons, I do not consider this issue capable of undermining the empirical
support for my argument. First, as mentioned previously, income inequality did not increase
much until the 1970s, and the correlation between party loyalty and vote share remained negative
before income inequality rose to a high level. Therefore, it is highly likely that the rise of income
inequality in the 1970s was the driving force behind the changes in the electoral cost of party
loyalty. Second, as mentioned, no literature clearly suggests other factors that would
simultaneously affect income inequality and the electoral cost of party loyalty. The previous
models have shown that including aggregate party loyalty and polarization in the House does not
fundamentally change the findings. Furthermore, I have also tested the robustness of the findings
by including a time counter, and the effect of income inequality on the electoral cost of party
loyalty remains significant.26 Therefore, the relationship between income inequality and the
electoral cost of party loyalty is not simply led by other factors or trending. Third, income
inequality did not increase monotonically. It ebbed and flowed in income inequality during the
26

Time counter is a variable that ranges from 1 to 17. It is coded as 1 when the observations are from the first
election year in this dataset (i.e., 1976), 2 when the observations are from the second election year, and so on.
45

time under investigation. The regression analyses have gone beyond the prima facie association
and demonstrated that the electoral cost of party loyalty is lower (higher) when income
inequality is at a higher (lower) level.
Nevertheless, in order to strengthen the empirical support for my conclusion, I conduct a
series of analyses using state-level income inequality. In these analyses, the variation of income
inequality across states is sufficient to demonstrate that the electoral cost of party loyalty changes
across different levels of income inequality.
The data for state-level income inequality is taken from Frank (2009). Frank calculated the
Gini Index for each state from 1959 to 2005, and I take the data for each year in which there
were House elections between 1976 and 2004.27 ModelⅠ in Table 1.4 is the same pooled model
as Model II in Table 2, except that income inequality is measured at the state level. The electoral
cost of party loyalty is found to be smaller in states with a higher level of income inequality.
Model Ⅱ is a random-effect model showing similar findings. Further, Model Ⅲ shows that with
Aggregate Party Loyalty and Polarization included, the interaction term between party loyalty
and state-level income inequality remains statistically significant.28 Since state-level income
inequality varies within each election year, the last model is a fixed-effect model with the
variables that are varying within each election year. This model shows that in any given election
year, party loyalty is less costly in a state with higher income inequality.
These findings further indicate that the effect of income inequality is not simply due to the
prima facie association between the concurrent trends of rising income inequality and falling
27

According to Frank (2009), this measure is calculated using pre-tax income data from the Internal Revenue
Service (IRS).

28

In a pooled model with the same specification, the coefficient for the interaction term between party loyalty and
state income inequality remains the same but has a p-value equal to .50. This suggests that the effect of income
inequality may be mitigated when the heterogeneity across elections years is not taken into account.
46

electoral cost of party loyalty.

Table 1.4. State-level Income Inequality and the Electoral Cost of Party Loyalty

Loyalty
State Inequality
Loyalty×State
Inequality
Ideology
Freshman
Party Leader
Previous Vote share
Challenger‘s Quality
Partisan District

Model I
Β (CRSE)

Model Ⅱ(RE)

Model Ⅲ (RE)

Model Ⅳ (FE)

Β (SE)

Β (SE)

β (SE)

-.47 (.15)**
-.82 (.28)*

-.47 (.12)***
-.74 (.20)***

-.38 (.13)**
-.56 (.21)**

-.39 (.12)**
-.55 (.20)**

.01 (.00)*

.01 (.00)**

.01 (.00)*

.01 (.00)*

6.54 (2.87)*
2.31 (.65)**
-2.35 (1.30) σ
.47 (.03)***
-8.17 (.54)***

5.99 (1.44)***
2.40 (.45)***
-2.00 (1.23)
.48 (.01)***
-7.74 (.43)***

2.17 (1.46)
2.32 (.44)***
-1.90 (1.21)
.44 (.01)***
-7.33 (.42)***
-4.75 (1.87)*

1.72 (1.47)
1.97 (.44)***
1.97 (1.22)
.44 (.01)***
-7.47 (.42)***
-6.35 (1.95)**

.11 (.02)***

.12 (.02)***

Loyalty×Partisan
District
Aggregate Party
Loyalty

.16 (.07)*

Polarization
In-Party
Midterm
Presidential Approval

-5.87 (5.57)
-1.22 (.16)***
-3.04 (.99)**
.11 (.08)

-3.26 (.36)***
.15 (.02)***

-3.03 (.36)***
.08 (.03)**

Unemployment Rate
-.08 (.29)
.01 (.13)
-.23 (.13) σ
Log GDP Per Capita
.11 (.20)
.05 (.08)
.17 (.08)*
Constant
84.56 (15.31)*** 80.06 (10.45)*** 66.40 (11.87)*** 75.85 (10.74)***
N
N of Clusters

5764
15

5764
15

sd (pcoh)

.03 (.01)

.03 (.01)

sd (cons)

1.48 (.95)

1.72 (.93)

sd (residual)

11.61 (.11)

11.45 (.11)

R2

5764
15
.32

47

5764
15

Table 1.4 (cont‘d)
LR test, Chi Square

210.69***

Rho
Note: σ: p<.1; *: p<.05; **:p<.01; ***:p<.001

201.67***
.05

48

Conclusion
Although the rising income inequality is appropriately a cause for concern, ironically, it may lead
to a political consequence that has been desired for years by political scientists. As this paper has
shown, the electoral cost of party loyalty can be reduced by increases in income inequality,
making a responsible party system and a high level of party loyalty in Congress more likely.
Theoretically, based on the literature on the politics of income inequality, I argue in this
paper that increases in income inequality may draw the policy preferences of the rich and poor
apart, sort people into partisan groups and grant party activists more leverage in elections. As a
consequence, voters care more about the success of the party and prefer a loyal legislator to a
dissenting one. The empirical analyses show that the cost of party loyalty is smaller when
income inequality is at a higher level, and the effect of income inequality is robust to various
model specifications.
Due to the limited scope of this research, I have to set aside several interesting research
topics. First, party polarization exists not only in the House of Representatives, but also in the
Senate (Theriault 2008). Garand (2010) has shown that income inequality at the state level leads
to polarization both in the mass and in the Senate. Efforts should be taken to examine the
electoral cost of party loyalty for senators and to investigate whether the cost of party loyalty is
conditional on income inequality for senators as well. Second, it has been shown that individual
voters indeed respond to roll call votes in Congress (Ansolabehere and Jones 2010). Moreover,
an individual voter with strong partisanship has lower confidence in a bi-partisan Congress
where legislators behave as if no party line is separating them (Harbridge and Malhotra 2011).
Combining these studies, my research suggests investigating how income inequality may affect
individual voters‘ attitudes toward legislators‘ party loyalty. Third, the literature lacks a
comparative study covering different countries with various political systems. Party loyalty
49

recently increased in high-ranking personal-vote systems with competing principals, such as
Brazil and Taiwan (Lyne 2008; Sheng 2008). However studying a single-country frees my
analysis from the possible complexity resulting from the differences across political systems, it
would be interesting if a future study could find similar changes in the electoral cost of party
loyalty in those countries.
All in all, if the gap between the rich and poor can not be closed in the near future, we
political scientists should be dedicated to further narrowing the gap in our understanding of the
political impacts of income inequality.

50

CHAPTER 2
CLARITY OF RESPONSIBILITY AND CLARITY OF PARTY LINE: THE IMPACTS OF
PARTY COHESION ON ECONOMIC VOTING IN THE UNITED STATES, 1980-2008

The paradigm of economic voting examines whether ―citizens vote for the government if the
economy is doing all right; otherwise, the vote is against‖ (Lewis-Beck and Stegmaier 2000,
183). In the United States, economy-related factors have been found to be critical conditions that
determine electoral outcomes (Fiorina 1978; Fiorina 1981; Hibbs 2000; Kiewit 1983; Nadeau
and Lewis-Beck 2001; Tufte 1978). Although economic voting is evident, its existence and
strength are found to be contingent (Anderson 2007). Despite a number of studies on the
conditionality of economic voting, it has not been fully investigated how economic voting is
conditional upon party cohesion. This lack of study is particularly important in regards to
contemporary American politics. As intra-party cohesion in Congress has increased to a high
level, it remains unclear how high levels of party cohesion affect the extent to which American
voters hold the incumbent party accountable for economic conditions.
In the literature, party cohesion has been regarded as the key to having a responsible party
system in which voters may be able to hold the incumbent party accountable to the public
interest (APSA 1950; Shugart and Carey 1992).29 More specifically, when it comes to economic
accountability, Powell and Whitten (1993) maintain that voting cohesion of major governing
parties enhances clarity of responsibility and economic voting. However, their research neglects
that party cohesion may vary across time and omits opposition party cohesion. The former
problem makes their findings far from conclusive, and due to the latter, their theory fails to
account for current American politics in which two parties may go hand in hand increasing each
other‘s levels of intra-party cohesion. In addition to these two deficiencies, it is not clear in their
29

In this paper, the term ―incumbent party‖ refers to the party of the president.
51

study how party cohesion affects voting behavior at the micro level. There is lack of knowledge
about whether individual voters hold the incumbent party accountable for economic conditions
as they observe each party acting cohesively in Congress.
The impacts of party cohesion remain puzzling also because several studies indicate that a
high level of party cohesion might instead hinder economic voting. First, at the micro-level, it
has been found that under the influence of partisan cues, accurately assigning responsibility is
not the most important goal to voters (Rudolph 2006). Moreover, partisan cues will have greater
influence, and voters will be more likely to act in a partisan manner when elites from each party
stand firmly with their in-party members (Levendusky 2009, 2-3). Finally, economic conditions
may matter less in determining electoral outcomes when there are more partisan voters (Kayser
and Wlezien 2011). By extension, these studies imply that there exists another mechanism
through which party cohesion affects economic voting. That is, although it might be the case that
voters are able to clearly see the responsibility when each party acts as a unitary actor, a clear
party line separating one party from the other could ignite partisan rivalries that shift voters‘
focus away from economic concerns. The findings from these studies further cast doubt on
Powell and Whitten‘s thesis.
In contemporary American politics, where two cohesive parties continue to engage in party
wars (Sinclair 2006), it is likely that an incumbent party could avoid being held accountable for
the economy if voters focus on partisan rivalries. For example, before the 2012 presidential
election, the Democratic government alleged that the Republicans have waged a ―war on women,‖
and the Republicans responded that it‘s an attempt of the Democrats to dodge the responsibility
for economic downturn due to President Obama‘s policies (Roff 2012). This example shows the
possibility that an incumbent party will not be held accountable for the economy if partisan
rivalries on controversial issues become the focus of the politics. With the clarity of the party line
52

left unstudied, the existing literature on economic voting is insufficient to account for the impact
of party cohesion on economic voting in current American politics
To shed light on the impacts of party cohesion on economic voting, this paper utilizes the
data of time-varying party cohesion in the U.S. House of Representatives and surveys from
American National Election Studies (ANES) between 1980 and 2008. Furthermore, instead of
focusing solely on incumbent party cohesion, I also include opposition party cohesion into my
theory and empirical analysis. I argue that party cohesion is a double-edged sword. On the one
hand, increases in party cohesion may enhance clarity of responsibility and economic voting; on
the other hand, high levels of incumbent and opposition party cohesion may interactively
generate clarity of party line, weakening economic voting. In so doing, this paper may better
account for economic voting in current American politics than the previous literature. Through
the empirical analyses performed in this paper, I discover the mitigating effect of party cohesion
on economic voting that has been concealed in Powell and Whitten‘s study: high levels of
incumbent and opposition party cohesion may interactively contribute to a high clarity of the
party line, decreasing the effects of economic perception on the probability of voting for an
incumbent party. As the findings will imply, a high level of party cohesion is not as desirable as
suggested by the conventional wisdom. Instead, democratic accountability can be hindered by
high levels of party cohesion even if the responsibility might be clear. In particular, incumbent
parties are less likely to be held accountable for economic performance when parties become
cohesive.

53

The Conventional Wisdom: Clarity of Responsibility
Powell and Whitten‘s thesis about the effects of party cohesion on economic voting has three
components. First, diffusion of responsibility may hinder voter from holding the incumbent party
or parties accountable for the economy. Second, a unified control of policymaking can reduce
such diffusion of responsibility. Third, voting cohesion of the major governing party serves as a
stable feature of political systems that affects the control of policy making and hence economic
voting.30 In short, they proclaim that higher levels of incumbent party cohesion enhance clarity
of responsibility, which strengthens economic voting (Powell and Whitten 1993, 399-400).
Although Powell and Whitten enriched the studies of economic voting by linking incumbent
party cohesion to clarity of responsibility, scholars have found problems with their binary
indicator of clarity of responsibility (Anderson 2007; Nadeau, Niemi and Yoshinaka 2002; Royed,
Leyden and Borrelli 2000). These follow-up studies take into account the variation in minority
status or coalition government, however, none of them investigated the variation of party
cohesion and its effects on economic voting. The major problems of the binary indicator are that
it overlooks the variation of party cohesion across time within a country and that it assumes all of
the countries coded as having weak party cohesion have the same level of party cohesion. Due to
the fact that party cohesion varies across time (Canton and Herrnson 1997, 395; Ladewig 2010;
Lyne 2008; Sheng 2008; Kam 2009), the impact of party cohesion on economic voting is far
from conclusive.
Another problem with their study is that they ignored the role of opposition party cohesion.
Incumbent party cohesion alone is not a guarantee for clarity of responsibility. Voters would not

30

The other stable feature is participatory and inclusive committee system in the legislature. In addition, Powell and
Whitten provide three varying conditions: bicameral opposition, minority government, coalition governments
(Powell and Whitten 1993, 399-402).
54

be able to see who is responsible if a large number of legislators from the opposition party work
with the incumbent party. Therefore, if Powell and Whitten‘s thesis is true, a high level of
opposition party cohesion should also contribute to clarity of responsibility and improve
economic voting. Also, if Powell and Whitten‘s thesis fully accounts for the impact of party
cohesion, economic voting should be strongest when both parties have high levels of party
cohesion. Without taking into account opposition party cohesion, it remains uncertain whether
party cohesion improves economic voting through clarity of responsibility. More importantly,
there is a mechanism other than clarity of responsibility linking party cohesion and economic
voting if opposition party cohesion is considered. When the legislators from both parties are
acting cohesively, people observe two unitary actors and a clear party line in between. As such,
even if increases in party cohesion enhance may improve clarity of responsibility, it may
simultaneously enhance clarity of the party line. Clarity of responsibility only affects the ability
of voters to assign responsibility, but voters may not vote on the basis of that assignment
(Hellwig and Samuels 2008). To be clear, while clarity of responsibility enables voters to assign
responsibility for poor economic performance, clarity of party line may create a political
environment where voters‘ voting decisions are not mainly based on their perceptions of the
economy. Powell and Whitten‘s thesis tells us one part of the whole linkage between party
cohesion and economic voting. Yet, their study fails to account for economic voting in
contemporary American politics, where two parties both have high levels of party cohesion.
To fill the gaps in the literature, my research brings the variation of party cohesion across
time and opposition party cohesion into the analysis. The findings demonstrate that concurrent
high levels of incumbent and opposition party cohesion may interactively generate clarity of the
party line, reducing the strength of economic voting.

55

The Missing Story: Clarity of Party Line
Clarity of party line has been missing in the literature on economic voting. Nevertheless, it
should be considered seriously as an important mechanism linking party cohesion and economic
voting. Party cohesion in the United States has increased since the 1970s. The variation of party
cohesion in Table 2.1 shows not only a rising trend, but also that the two parties both increase
their levels of cohesion so as to match each other.31 Lebo, McGlynn and Koger (2007) have
discovered that the two parties were competing with each other by strategically adjusting their
party cohesion. Accordingly, the two parties may concurrently have high levels of cohesion,
resulting in a clear party line that separates the two cohesive parties.

Table 2.1. Party Cohesion in the U.S. House of Representatives
Election
Year

Incumbent's
Cohesion

Opposition's
Cohesion

Incumbent
Party

1980
1984
1988
1992
1996
2000

72.50
76.34
77.12
78.61
80.02
83.33

77.00
75.87
81.15
81.78
87.48
86.89

Democrats
Republicans
Republicans
Republicans
Democrats
Democrats

2004
90.07
86.14
Republicans
2008
85.96
90.54
Republicans
Note: 1. Party cohesion is measured as the average percentage of
a given party‘s legislators voting with the majority of that
party in each roll call vote during a presidential term.
2. The data is taken from voteview.com.

31

In Table 2.1, party cohesion for a given party is calculated as the average percentage of a given party‘s legislators
voting with the majority of that party in each roll call vote during a presidential term. For example, if there are 3
roll call votes during a term in which 60%, 95% and 85% of the legislators from a given party voting in one
direction, then the cohesion for the that party in this term is (60+95+85)/3, namely, 80. The raw data for roll call
votes is taken from the Voteview.com.
56

My theory expects that high clarity of party line may enhance the salience of partisan cues
so as to hinder voters from holding the incumbent party accountable for the economy. 32 A clear
party line may increase the salience of partisan cues through the following mechanisms. First,
high levels of party cohesion enhance the informativeness and the credibility of a party label
(Ashworth and Bueno de Mesquita 2008; Snyder and Ting 2002). Hence, Messages attached to
the party label of a cohesive party has greater influence in determining voting decisions than
those attached to one of a party that are not cohesive. Secondly, a clear party line also means that
there are fewer bi-partisan policies that may draw cross-party support from the members of the
other party.33 The public then observes a ―party war‖ in which parties more frequently vote
down each other‘s bills by intra-party collective effort in the committees, as well as on the floor
(Sinclair 2006). Consequently, the policies and the legislations are sorted along the party line and
become partisan cues. In addition, when there are fewer and fewer dissenting legislators from
each party, parties will dominate policymaking process and become the most important political
actors that voters can identify with and take messages from. Hence, messages that are not
specifically linking to any parties are less available to voters. Finally, due to the proliferation of
partisan agendas in Congress, partisan voters are more likely to be mobilized (Hetherington 2001;
Bernhardt, Duggan and Squintani 2009). These voters might further provide partisan cues in
political discussions.
When a high clarity of party line increases the salience of partisan cues, voters might also
become partisan. It is evident that when both parties increased their cohesion, Americans were
more and more likely to report themselves as party identifiers during the past four decades
32

Partisan cues are messages that ―may reveal party affiliation […] or link a party to a stand on an issue‖ (Bullock

2011, 497).
33

McCarty, Poole and Rothenthal (2006) have pointed out that fewer overlapping of issue positions is a sign of
polarization.
57

(Bartel 2000; Gelmen 2008; Hetherington 2001; McCarty, Poole and Rothenthal 2006). This is
because when the party line is clear, voters are more likely to understand ―the set of policy
positions that accompany being a Democrat or a Republican‖ and ―what it means to be a
Democrat or a Republican.‖ Voters then utilize these clear cues to align their issue positions and
partisanship and become ―more firmly anchored‖ to the party (Levendusky 2009, 2-3). Also, in
Hentherington‘s study, he finds that ―if people are exposed to a heavily partisan stream of
information, which will be more likely if elites are behaving in a partisan manner, then it follows
that respondents will express opinions that reflect the heavily partisan stream‖ (Hetheington
2001, 622). Accordingly, voters may become partisan after perceiving a clear party line and
salient partisan cues.
Evidence shows that voters that become partisan would downplay economic concerns when
evaluating the performance of the incumbent party (Borre 1997; Evans and Anderson 2006;
Pattie and Johnson 2001). Furthermore, salient partisan cues may also affect the way voters
assign responsibility. Rudolgh (2006) draws a comparison between accuracy goals and direction
goals. The former represent the goals to accurately assign the responsibility for the economic
conditions, and the latter are the goals to assign the responsibility in a way that is consistent with
party identification. Rudolgh shows that accuracy goals become less important than direction
goals when partisan cues are present. This study implies that the clarity of responsibility for the
economy may be deemphasized when partisan cues are present and clear.34
To summarize, while any increases in either party‘s cohesion might enhance clarity of
responsibility, they simultaneously improve clarity of the party line. A clear party line that

34

Although Bullock (2010) points out that partisan cues do not necessarily affect opinions better than information
about the policies, my theory stresses that when partisan cues are overwhelmingly spread in politics, partisan cues
will more effectively reduce the weight of economic perception in voters‘ voting decisions.
58

improves the salience of partisan cues may lead voters to downplay economic concerns in
making voting decisions. Accordingly, high levels of party cohesion may weaken economic
voting by enhancing clarity of party line. To test this argument, I conduct a series of regression
analyses with data from the ANES surveys. These analyses are demonstrating that voters‘
economic perceptions have a smaller impact on voting intentions when incumbent and
opposition party cohesion both increase to a high level.

59

Regression Analysis
Economic voting can be depicted by the relationship between voters‘ economic perception and
their intentions to vote for the incumbent (Lewis-Beck and Stegmaier 2000; Anderson 2007;
Duch and Stevenson 2008). Economic voting exists if a negative (positive) perception leads to a
smaller (larger) likelihood to vote for the incumbent. In addition, the strength of economic voting
can be measured as the amount of changes in the likelihood of voting for the incumbent party
that results from the differences in economic perception.
Accordingly, the dependent variable is Voting Intentions, which is a dummy variable
denoting whether a respondent intends to vote for the incumbent party in the coming presidential
election. The time frame is from 1980 to 2008 in order to accommodate the availability of data
for the perception of the economy. In the ANES surveys, respondents are asked, in slightly
different terms, about which party they are going to vote for.35 The respondents who were going
to vote for the incumbent party in a given year are coded as 1, and 0 otherwise. Given that the
dependent variable is binary, the methods of estimation are pooled logit models and
random-effect logit models.36
The key independent variable is Economic Perception. There are two types of economic
perceptions in the literature of economic voting. One is the perception of national economy (a.k.a.
socio-tropic evaluation) and the other is the perception of personal financial situation (a.k.a.
pocketbook voting). While the perception of national economy has been frequently found as a

35

This is the question asked in pre-election surveys. Using pre-election surveys avoids the possibility that
respondents conform their answers to the results of elections.

36

Given the small number of clusters (i.e. 8 election years) in this research, clustered robust standard error is not
appropriate (Primo, Jacobsmeier and Milyo 2007). Clustered stand errors are generally smaller than those without
considering clusters. As a matter of fact, with cluster robust standard errors, the results are even more supportive
to my theory.
60

valid measure of economic evaluation (Lewis-Beck and Paldam 2000; Kiewiet 1983; Hansford
and Gomez 2011), pocketbook voting has also been found in a number of studies (Fiorina 1978;
Sigelman, Sigelman and Bullock 1991). I do not choose socio-tropic evaluation in this study
because the perception of national economy might be more likely to come from a respondent‘s
approval of the incumbent president or partisanship, leading to colinearity and possible
endogeneity. The advantage of using pocketbook evaluation is to get around these
methodological issues since personal financial status might reflect many aspects of life that are
not endogenous to political factors (Borre 1997; Evans and Andersens 2006). As shown in Table
2.2, in this dataset, approval of the president, partisanship and the intentions to vote for the
incumbent party have larger impacts on the perception of national economy than on the
perception of personal financial situation.37 Given that the perception of national economy is
more likely to be endogenous to the dependent variable and other independent variables, it is
better to adopt the perception of personal financial situation in this study.38

37

For example, approval of president only makes about twenty-percent difference (42%-22%) in the percentage of
perceiving financial situation as getting worse, while it makes almost forty-percent difference (68.09%-29.09%)
in the percentage of perceiving national economy as getting worse.

38

As a further test, I replace the pocketbook perception with the perception of national economy in the same
model specifications. The finding is that the conditionality of interest remains when the control variables are not
included. This further demonstrates that the perception of national economy is highly correlated with those
political attitudes and that the pocketbook voting is an option that is more appropriate for this study.
61

Table 2.2. The Associations between Variables
Personal Financial Situation
Not Worse
Worse

National Economy
Not Worse
Worse

Approve of president
Not Approve of
president

78.00

22.00

70.13

29.87

57.78

42.22

31.91

68.09

In-Party
Out-Party

74.78
64.50

25.22
35.50

59.70
46.37

40.30
53.63

Vote for Incumbent
77.10
22.90
67.79
32.21
Not to vote for
60.97
39.03
38.35
61.65
Incumbent
Note: 1.The numbers in the cells are the row percentages. A larger difference between rows in a
given column for each variable
indicates a larger impact of the row variable on the column variable.
2. The Chi-square tests are significant at .001 level for all the associations.

The data for the perception of personal financial situation is also taken from the ANES
surveys. Since 1980, during every presidential election year, the ANES asked questions about
whether respondent‘s financial situation has gotten better or worse over the past year. 39 The
respondents who answered that their personal financial situation has gotten worse are coded as 1,
and 0 otherwise. This way of coding is to accommodate the concern that voters would be more
likely to punish the incumbent party for a bad economy than to reward it for a good economy. 40
Given the way of coding, economic voting exists when this variable has a negative effect—the
respondents who perceive their personal financial situation as getting worse are less likely to
vote for the incumbent party.
The effect of economic perception is expected to be conditional. In my theory, increases in

39

Because my research is designed mainly to investigate the hypothesis regarding clarity of responsibility for past
economic performance, I will focus on retrospective evaluation of economy.

40

Similar arguments have been addressed in the literature about asymmetric effects of economy (Bloom and Price
1975).
62

either party‘s cohesion to a certain extent may enhance clarity of responsibility and economic
voting, while a high level of incumbent party cohesion and opposition party cohesion may
interactively enhance clarity of the party line and hinder voters from holding the incumbent party
accountable for the economy. In other words, the effect of economic perception is conditioned
upon three variables: Incumbent Party Cohesion, Opposition Party Cohesion and the interaction
between these two variables. Such conditionality is being tested by including three interaction
terms: Economic Perception × Incumbent Party Cohesion, Economic Perception × Opposition
Party Cohesion and Economic Perception × Incumbent Party Cohesion × Opposition Party
Cohesion. In my theory, when either party‘s cohesion increases so as to enhance clarity of
responsibility, a negative perception of the economy decreases the likelihood of voting for the
incumbent party to a greater amount. Accordingly, I expect that the first two interaction terms
both have a negative coefficient. The third interaction term is included to capture the idea that
Incumbent Party Cohesion and Opposition Party Cohesion interactively generate a clear party
line and constrain economic voting. If so, the negative effect of economic perception should be
mitigated by the product of the two parties‘ cohesion. Therefore, a positive coefficient is
expected for the third interaction term.
Both Incumbent Party Cohesion and Opposition Party Cohesion are measured as the
average percentage of a given party‘s legislators voting with the majority of that party in each
roll call vote during a presidential term.41 The range of party cohesion is from 50 to 100.
Incumbent Party Cohesion is expected to have a positive effect on an incumbent‘s electoral
fortunes. With a higher level of party cohesion, parties are more likely to keep and improve their
records. Parties that are able to keep a good legislative record in carrying out public policies are
expected to have a better reputation and a higher chance to win the elections (Cox and McCubin
41

Please see the details in footnote 3.
63

1993; Lebo, McGlynn and Koger 2007; Grynaviski 2010). By the same token, Opposition Party
Cohesion may improve the electoral fortune of an opposition party and reduce the likelihood of
voting for the incumbent. Therefore, contrary to Incumbent Party Cohesion, Opposition Party
Cohesion is expected to have a negative effect on the dependent variable.
Several control variables that might also affect the likelihood of voting for the incumbent
party are included. First, In-Party Voter denotes whether a respondent identifies himself/herself
with the incumbent party. It is coded as 1 if a respondent who identifies himself/herself with the
incumbent party and 0 otherwise. All else being equal, an in-party voter is more likely to vote for
the incumbent.
Presidential Approval is included to control for both the effects of voters‘ overall evaluation
about the governing performance and the advantage of incumbency. The data is taken from an
question in the ANES surveys asking whether the respondent approves or disapproves of the
current president‘s handling of his job. The respondents who approve are coded as 1, and 0
otherwise. The respondents who approve of the incumbent president are expected to be more
likely to vote for the incumbent party.
Race is another factor influencing voters‘ voting decisions. Compared to white voters,
Non-white voters are typically prone to supporting the Democratic Party. When the Democratic
Party is the incumbent party, the non-white respondents are coded as 1; when the Republican
Party is the incumbent party, the white voters are coded as 1. All other respondents are coded as
0. This variable is expected to have a positive effect on the likelihood of voting for the incumbent
party.
In addition to individual-level factors, I also control some macro-level variables that might
affect voting decisions. Divided Government is a dummy variable denoting whether an
incumbent party is a minority in the House, in the Senate or in both. When there is a divided
64

government, an incumbent party might shift the blame for bad outcomes to the opposition party.
Therefore, this variable is expected to have a positive effect on incumbents‘ vote shares (Powell
and Whitten 1993; Nicholson, Segura and Woods 2002).
Ideological Polarization is measured as the Polarization Scores used by McCarty, Poole and
Rothenthal (2006). It is simply the difference between two parties‘ median DW-Nominate Scores
during the term closest to each election year. Although ideological polarization might possibly go
hand in hand with high levels of party cohesion, the two variables are essentially different.42
Ideological polarization means that two parties take two opposite positions that are both far from
the median on the entire spectrum. It does not necessarily mean that the members of this party
act cohesively in roll call voting. Ideological polarization among elites might also reinforce the
severity of partisan rivalries. It is necessary to include ideological polarization in order to partial
out its effect from the impact of party cohesion on economic voting.
On-Going War indicates whether there has been any war in which the U.S. engaged during a
presidential term. On-going wars may have two-sided effects. On one hand, wars might reduce
the stability of society because human and material resources are used for warfare. On the other
hand, wars might also increase the popularity of the incumbent president due to the so-called
rally effects (Mueller 1970; Lian and O‘neal 1993). In this research, I hypothesize that wars have
negative effects on respondents‘ intention to vote for the incumbent party because wars overall
impose costs on the societies that are involved. The evidence in the literature also shows that
popular support for a war is reduced as the war continues, and the leaders who are pro-war are
more likely to lose power (Frieden, Lake and Schultz 2010, 141-5). In this research, I define war
as military engagements authorized by Congress. Such military engagements occurred in 1983
(Lebanon), 1991 (Gulf War), 2001 (Afghanistan) and 2003 (Iraq). Accordingly, respondents in
42

Carson et al. (2010) share the same view that party cohesion and ideology are different.
65

1984, 1992 and 2004 are coded as 1, and 0 otherwise.
Reelection is included to control whether a given election features a sitting president up for
reelection. There is a term limit for the U.S. presidents, and hence an incumbent party does not
always have a sitting president as a candidate in an election. Sitting presidents may have
electoral advantages such as nation-wide name recognition and resources from the government.
The successor nominated by the incumbent party may not completely inherit these advantages.
Therefore, voters might be more likely to vote for the incumbent party when the presidential
candidate of the incumbent party is a sitting president than otherwise.

Findings
The results in Table 2.3 show that the effect of economic perception is conditional upon
incumbent party cohesion, opposition party cohesion and the product of the two parties‘ cohesion.
ModelⅠis a pooled model without any control variable included. In this model, all the
interaction terms are statistically significant. The interaction between economic perception and
either party‘s cohesion has a negative coefficient, while the interaction between economic
perception and the product of two parties‘ cohesion has a positive coefficient. This indicates that
the effect of either party‘s cohesion on economic voting is offset by the product of two parties‘
cohesion. In Model Ⅱ, the results are similar when all variables are included, except that the
interaction between economic perception and opposition party cohesion is not significant and
that the triple interaction is significant only at the .1 level. Since the effect of party cohesion in
my theory may take place through partisanship, Model Ⅲ excludes partisanship and the
coefficients for the interaction terms become significant.
This finding seems to be consistent with Powell and Whitten‘s thesis. That is, voters are
more likely to punish the incumbent party for a poor economic condition when incumbent party
66

cohesion increases to a high level. However, such an effect is conditional upon the levels of
opposition party cohesion. Given that the triple interaction has a positive coefficient, as
opposition party cohesion is at a high level, incumbent party cohesion may not enhance
economic voting as Powell and Whitten suggested.
The effect of clarity of responsibility can also be tested by looking into the coefficients that
are involved with opposition party cohesion. As stated, if clarity of responsibility is the only
mechanism linking party cohesion and economic voting, increase in opposition party cohesion
should also enhance clarity of responsibility. Although, the coefficient for the interaction
between opposition party cohesion and economic perception is not significant at the .05 level, its
negative sign suggests that economic voting might be strengthened by opposition party cohesion.
However, such an effect is conditional upon incumbent party cohesion given the triple
interaction.
The simple interpretation of the coefficients shows that Powell and Whitten‘s thesis is not
fully supported by the empirical findings here. First, it is obvious that the effect of either party‘s
cohesion is conditional upon the cohesion of the other party. Increases in incumbent or
opposition party cohesion do not necessarily enhance economic voting. Second, party cohesion is
a double-edged sword to economic voting. Although an increase in one party‘s cohesion may
strengthen the effect of economic perception on voting intentions, it simultaneously weakens the
strengthening effect of the other party‘s cohesion on economic voting. Therefore, when either
party‘s cohesion increases to a high level, economic voting can not be further strengthened and
will diminish.
Speaking of illustration and interpretation, graphs may function better for non-linear models.
Graphically, economic voting can be shown by the differences in probability of voting for an
incumbent party between a respondent who has a bad perception about the economy and a
67

respondent who does not. If economic voting exists, the difference in probability should be
negative and statistically significant. In other words, compared to the voters who do not have a
negative perception about the economy, the voters who perceive the economy as getting worse
should be less likely to vote for the incumbent party. Moreover, such a difference in probability
should be larger as economic voting is strengthened. In Figure 2.1, for both in-party voters (the
top panel) and out-party voters (the bottom panel), I perform three scenarios where the
opposition party cohesion is fixed at the minimum (76), the median (82) and the maximum
(91).43 In each scenario, the incumbent party cohesion increases from the minimum (72) to the
maximum (90).

43

Note that in-party respondents are the respondents who identify with the incumbent party, and out-party
respondents are those who do not identify with the incumbent party.
68

Table 2.3. Regression Analysis

Economic Perception (Worse=1)

Model Ⅰ

Model Ⅱ

Model Ⅲ

β (SE)

β (SE)

β (SE)

53.852 (14.375)*** 34.839 (19.374) σ 49.732 (18.196)**

Incumbent Party Cohesion (IC)

.354 (.108)**

-1.989 (.315)***

-1.183 (.291)***

Opposition Party Cohesion (OC)

.208 (.099)*

-2.592 (.322)***

-1.558 (.297)***

IC×OC

-.004 (.001)**

.030 (.004)***

.016 (.004)***

Economic Perception × IC

-.755 (.182)***

-.487 (.247)*

-.678 (.232)**

Economic Perception × OC

-.587 (.172)**

-.376 (.231)

-.553 (.216)*

Economic Perception × IC × OC

.008 (.002)***

.005 (.003) σ

.008 (.003)**

Approve of President's Job
(Approve=1)

2.457 (.052)***

2.821 (.049)***

In-Party Voter (yes=1)

1.968 (.054)***

Race (White=1)

.898 (.063)***

1.146 (.060)***

-8.093 (2.814)**
-2.794 (.229)***

1.649 (2.587)
-2.106 (.208)***

Reelection

1.905 (.191)***

1.399 (.173)***

Divided Government

2.030 (.236)***

.880 (.216)***

177.432
(24.057)***

109.983
(22.196)***

14571

14571

14571

8

8

8

Ideological Polarization
On-Going War

Constant
Number of Cases
Number of Clusters

2
.030
.398
.322
Pseudo R
Note: 1. σ: p<.1; *: p<.05; **:p<.01; ***:p<.001
2. The dependent variable is a dummy variable indicating whether a respondent is going to
vote for the presidential candidate from the incumbent party in a given election.

As Figure 2.1 shows, for either in-party or out-party voters, the differences in probability
made by economic perception do not consistently increase as the incumbent party cohesion
increases. This means that economic voting is not necessarily strengthened by the increases in
incumbent party cohesion. More specifically, economic voting is enhanced only when incumbent
69

party cohesion increases from a low to mid level; however, the strength of economic voting
decreases as incumbent party cohesion continues to increase from a mid level to a higher one.
Furthermore, when the opposition party cohesion is set at the maximum (91), economic
perceptions do not make significant differences across all levels of incumbent party cohesion.
This result suggests that regardless of the levels of incumbent party cohesion, economic voting is
obscured when a high level of opposition party cohesion makes the product of two parties‘
cohesion large enough to hinder economic voting. These findings do not fully support Powell
and Whitten‘s thesis. Increases in incumbent party cohesion may enhance clarity of responsibility
and strengthen economic voting only when incumbent party cohesion increases from a low to
mid level. However, when incumbent party cohesion is at a high level, economic voting is not
further enhanced by an increase in party cohesion. Instead, voting intentions are less affected or
not affected by economic perceptions when there is a cohesive incumbent party, especially when
the opposition party is also cohesive.
The effect of opposition party cohesion is similar. Since the differences in probability are
not significant when incumbent party cohesion is held at its lowest and highest level, and due to
the limited space, I only present a graph in which incumbent party cohesion is held at its median
(78) as the opposition party cohesion increases from its minimum (76) to maximum (90). In
Figure 2.2, for both in-party voters (the left panel) and out-party voters (the right panel), when
the opposition party cohesion increases from a low to mid level, economic voting is strengthened;
while the strength of economic voting diminishes as the opposition party cohesion continues to
increase from a mid level. This finding is different from what is found in the interpretation of
regression coefficients. However, in regards to a non-linear model, looking into the predicted
probabilities may offer a better understanding of the relationship between variables. Previous
studies of economic voting have long neglected the role of opposition party cohesion, while I
70

find that it matters in improving clarity of responsibility and economic voting as opposition party
cohesion increase from a low to mid level. Furthermore, such an effect is mitigated by the
product of incumbent party cohesion and opposition party cohesion and diminishes as either
party‘s cohesion increases to a high level.

71

72 74 76 78 80 82 84 86 88 90

72 74

76 78 80 82 84 86 88 90

72 74 76 78 80 82 84 86 88 90
Incumbent Party Cohesion

72

74 76 78 80 82 84 86 88 90
Incumbent Party Cohesion

-.14-.12 -.08-.06-.04-.02 .02.04.06.08 .12
-.1
.1
0

Incumbent Party Cohesion

-.14-.12 -.08-.06-.04-.02 .02.04.06.08 .12
-.1
.1
0

-.14-.12 -.08-.06-.04-.02 .02.04.06.08 .12
-.1
.1
0

Incumbent Party Cohesion

Opposition Party Cohesion=91

-.14-.12 -.08-.06-.04-.02 .02.04.06.08 .12
-.1
.1
0

Opposition Party Cohesion=82

-.14-.12 -.08-.06-.04-.02 .02.04.06.08 .12
-.1
.1
0

-.14-.12 -.08-.06-.04-.02 .02.04.06.08 .12
-.1
.1
0

Opposition Party Cohesion=76

72 74

76 78 80 82 84 86 88 90
Incumbent Party Cohesion

72

74 76 78 80 82 84 86 88 90
Incumbent Party Cohesion

Figure 2.1. Incumbent Party Cohesion and Economic Voting
Note: 1.The y-axis is the probability that the respondents who perceive national economy as getting worse vote for the
incumbent party minus the probability that the respondents who do not have the same perception vote for the
incumbent party.
2. The other variables: Approval=1, On-Going War=0, Divided Government=1, Reelection=1, Polarization=mean.
72

.08

.08

-.1

-.08

-.06

-.04

-.02

0

.02

.04

.06

Incumbent Party Cohesion=78

-.12

-.12

-.1

-.08

-.06

-.04

-.02

0

.02

.04

.06

Incumbent Party Cohesion=78

76

78

80

82

84

86

Opposition Party Cohesion
Mean Difference in Probability

88

90

76

78

80

82

84

86

88

90

Opposition Party Cohesion
95% C.I.

Mean Difference in Probability

95% C.I.

Figure 2.2. Opposition Party Cohesion and Economic Voting
Note: 1.The y-axis is the probability that the respondents who perceive national economy as
getting worse vote for the incumbent minus the probability that the respondents who
do not have the same perception.
2. The other variables: Approval=1, On-Going War=0, Divided Government=1,
Reelection=1, Polarization=mean.

In sum, the interpretation of regression coefficients and graphical illustrations offer three
findings about the relationship between party cohesion and economic voting. First, increases in
party cohesion do not necessarily enhance economic voting. The effect of one party‘s cohesion
on economic voting is smaller when the other party‘s cohesion increases. Second, due to the
interaction between the two parties‘ cohesion, party cohesion is a double-edged sword to
economic voting. Although an increase in either party‘s cohesion enhances economic voting, the
strengthening effect of the other party‘s cohesion on economic voting is simultaneously
mitigated. Consequently, economic voting is strengthened when party cohesion increases from a
low to mid level, while it is weakened when party cohesion is at a high level. Third, as an
73

extension of the previous two findings, when both parties have a high level of party cohesion,
economic voting diminishes to none.
In addition to the findings of theoretical interest, the effects of each party‘s cohesion on
voting intentions are worth discussion. The effect of incumbent party cohesion on the log-odds is
the linear combination of the coefficients for Incumbent Party Cohesion and the three interaction
terms. The effect of one-unit increase in incumbent party cohesion on voting intension is
calculated as -1.989 + .030 × Opposition Party Cohesion - .487 × Economic Perception + .005 ×
(Economic Perception × Opposition Party Cohesion). Looking at these coefficients, for the
respondents who do not perceive their financial situation as getting worse (a.k.a. Economic
Perception=0), the linear combination is positive when opposition party cohesion is 70.74 and
above. Meanwhile, for the respondents who perceive financial situation as getting worse (a.k.a.
Economic Perception=1), the linear combination is positive if opposition party cohesion is
greater than 66.31. Thus, as long as opposition party is greater than 74.00, incumbent party
cohesion has a positive effect on the likelihood of voting for the incumbent party, regardless of
voters‘ economic perceptions. Since opposition party cohesion in this dataset is never lower than
70.74, the evidence here shows that incumbent party cohesion improves the electoral fortune of
the incumbent party. Moreover, this finding suggests that incumbent party cohesion benefits the
incumbent party only when opposition party cohesion is high enough. Voters may prefer a
cohesive incumbent party when the opposition party is also cohesive, while increases in party
cohesion are not beneficial to the incumbent party cohesion if the opposition party is not
cohesive.
The effect of opposition party cohesion on voting intensions is conditional upon economic
perception and incumbent party cohesion. The linear combination of the coefficients is calculated
as -2.592+ .030 × Incumbent Party Cohesion - .376 × Economic Perception + .005 × (Economic
74

Perception × Incumbent Party Cohesion). For the respondents who do not perceive their
financial situation as getting worse (a.k.a. Economic Perception=0), the linear combination of
coefficients for opposition party cohesion is negative when incumbent party cohesion is lower
than 84.80; for the respondents who perceive financial situation as getting worse (a.k.a.
Economic Perception=1), the linear combination of coefficients is negative when incumbent
party cohesion is lower than 86.40. Therefore, for most of the cases, increases in opposition party
cohesion may bring troubles to the incumbent‘s electoral fortune, especially when the incumbent
party has weak party cohesion. However, opposition party cohesion could possibly improve the
electoral fortune of an incumbent party when the incumbent party has a high level of cohesion.
Increases in opposition party cohesion might provide more opportunities for a cohesive
incumbent party to shift the blame to the cohesive opposition party and to attribute the
unfavorable governing performance to severe partisan rivalries.
The findings for the effects of incumbent and opposition party cohesion on the electoral
fortunes of an incumbent party are roughly consistent with what has been found by Lebo,
McGlynn and Koger (2007). That is, it is electorally beneficial for a party to increase its
cohesion as the other party becomes cohesive. However, the evidence here suggests that an
opposition party will suffer if it increases cohesion when facing a cohesive incumbent party.
Most of the control variables affect voting intentions in the anticipated way, and their
coefficients are significant at the conventional level (p<.05). The respondents who align with the
incumbent party are more likely to vote for that party. These respondents include those who
identify themselves with, share the same ideology with or are in a racial group traditionally
supporting the given incumbent party, as well as those who approve of the president. Among the
macro-level variables, as expected, on-going war reduces the likelihood of voting for the
incumbent. Additionally, voters in an election year when there is a sitting president or when there
75

is a divided government are more likely to vote for the incumbent party. Finally, ideological
polarization has a negative coefficient. This means that ideological polarization may lead to
voters‘ resentment toward the incumbent party. It is worth further investigating why an
incumbent party would suffer when parties become more ideologically polarized. In my view,
voters might have a general resentment toward ideological extremism, and thus they resent the
current governing party in Washington.
To conclude this section, two findings are worth being stressed again. First, contrary to the
expectation from the conventional theory, the strength of economic voting does not necessarily
increase with incumbent party cohesion. Powell and Whitten‘s argument that a higher level of
incumbent party cohesion enhances economic voting is sustained only when incumbent party
cohesion increases from a low to mid level. Their thesis loses its validity when incumbent party
cohesion is at a high level. Second, increases in incumbent party cohesion may not only enhance
clarity of responsibility but also clarity of party line. The regression analyses and the graphical
illustrations have shown that when incumbent party cohesion and opposition party cohesion are
both at a high level, the clarity of a party line weakens the effect of economic perceptions. This
finding implies that when both parties increase their levels of cohesion, the strength of economic
voting diminishes.

76

Discussion
There has been a consensus that economic voting is contingent, however, it remained unclear
how economic voting changes with party cohesion across time. Unlike the previous literature on
economic voting, this paper takes into account the variation of party cohesion across time and
opposition party cohesion. It demonstrates that a high level of party cohesion does not improve
economic voting and shed new light on the studies of economic voting.
Overall, the findings in this paper do not support Powell and Whitten‘s thesis. If clarity of
responsibility is the only mechanism driving the relationship between party cohesion and
economic voting, increases in either incumbent or opposition party cohesion should make
responsibility clearer and then enhance economic voting. Furthermore, when both parties have
the highest level of party cohesion, and there is no cross-party voting in Congress, the clarity of
responsibility should be also at the highest level. Accordingly, economic voting should be
strongest if both parties‘ cohesion is at their maximum level. However, these theoretical
expectations are not supported by the empirical evidence here. Instead, the evidence suggests
that clarity of responsibility only explains one part of the relationship between party cohesion
and economic voting. Increases in party cohesion not only enhance clarity, but also make the
party line between two parties clearer. When party cohesion is at a high level, the high clarity of
party line increases the salience of partisan cues, hindering voters from holding the incumbent
party accountable for economic conditions.
The findings in this paper have important implications for understanding contemporary
American politics. In contemporary American politics, the Democratic Party and Republican
Party act cohesively in Congress to vote down each other‘s bill (Sinclair 2006). Moreover, as
Lebo, McGlynn and Koger (2007) have pointed out, the two parties adjust their cohesion in order
to compete with each other. My study accounts for these developments and demonstrates how
77

they influence economic accountability in the U.S. As the two parties engage in party wars,
voters are likely to downplay their economic concerns in making their voting decisions.
Incumbent parties may thus have leeway to avoid from being held accountable for economic
conditions. Additionally, the findings for the effect of party cohesion on voting intensions also
suggest that when competing with a cohesive opponent, a higher level of party cohesion makes a
party better off in terms of its electoral fortunes. In other words, it is generally in a party‘s
interest to increase its internal cohesion in order to gain electoral benefits if the other party also
becomes cohesive. Thus, it can be expected that the party wars between the two parties will
continue to be one of the most significant features of American politics.
Additionally, it has been found that economic voting is dependent upon voters‘ partisanship.
The effect of economic perceptions on voting intentions is smaller for voters who identify with
the given incumbent party than that for other voters (Borre 1997; Evans and Andersen 2006).
This paper supports this view. In the graphical illustrations, compared to incumbent party
identifiers, the probability of voting for the incumbent party among those who are not incumbent
party identifiers is less affected by economic perceptions. Additionally, this paper discovers that
the effect of economic perceptions can be further conditioned by clarity of party line, regardless
of voters‘ partisanship. This implies that economic concerns are downplayed by all voters as the
partisan rivalries among elites escalate.
Several potential topics for future research can be extended from this paper. First, although
it is questionable whether voters look into legislators‘ voting record on specific issues, it would
be worth investigating how party cohesion, with respect to economic policies, affects economic
voting. The levels of party cohesion in roll call votes specifically related to economic issues may
have a greater impact on the clarity of responsibility for the economy than by the party cohesion
in all roll call votes. A study that dissects party cohesion in different areas of policies may further
78

discover the effects of party cohesion on economic accountability.
Secondly, a comparative study of the relationship between party cohesion and economic
voting may generalize the findings here to other democracies. Evidence shows that party
cohesion in countries that have had low levels of party cohesion, such as Brazil and Taiwan, has
increased recently (Lyne 2008; Sheng 2008); while others that had a high level of party cohesion,
such as the United Kingdom, have sometimes experienced low levels of party cohesion (Kam
2009). If data for party cohesion is available, a comparative study would enrich the
understanding of the conditioning effect of party cohesion on economic voting.
Methodologically, endogeneity is worth further investigation. The issue of endogeneity
results from a possibility that subjective economic perception might be influenced by voting
decisions. That is, a respondent perceives the economy as getting worse simply because he or she
is going to vote for the opposition party. A traditional solution to endogeneity is to use
instrumental variables. Lewis-Beck, Nadeau and Elias (2008) uses the variables in surveys, such
as race, gender, age, interest in politics and so on, to construct their instrument variable for
economic perception. They find that the effects of economic perception turn out to be higher
when using their instrumental variable. However, some of the variables that they use to construct
the instrument variable have direct relations to voting intention, and the others are not truly
exogenous. Thus, their instrumental variables might not really solve the problem of endogeneity.
Hansford and Gomez (2011) use local economic conditions as instrumental variables, however,
the R-square in their first-stage model is small, suggesting that local economic conditions might
not serve as good instrumental variables for voters‘ perceptions about national economy. In
addition, using local-level variables as an instrument for individual-level perceptions might
induce a concern about ecological fallacy. Ecological fallacy could be induced by using variables
at the macro level to represent variables at the micro level. In my study, with partisanship and
79

presidential approval included in the regression analyses, I have to some extent eliminated the
possibility that voting intentions affect economic perceptions through variables that might reflect
voting intentions. Further, in comparison to socio-tropic perception, pocketbook perception more
involves the aspects of personal life that are exogenous to politics. Adopting the perception of
personal financial situation puts this research in an advantageous position to claim that voting
intentions would not have a large impact on the perception of personal financial situation.

80

Conclusion
Powell and Whitten (1993) assert that a cohesive governing party enhances economic voting by
increasing clarity of responsibility. The doubts surrounding this thesis come from its failures to
take into account the within-country variation of party cohesion and to account for other possible
mechanisms. The former makes the thesis questionable because the salience of economic voting
in a country where the party cohesion is presumably high might occur when the party cohesion is
actually at a low level; the latter theoretically provides a rival hypothesis that higher levels of
party cohesion may result in a clearer party line, which hinders economic voting.
Contrary to Powell and Whitten, I argue that high levels of party cohesion may hinder
voters from holding the incumbent party accountable for the economy. This occurs especially
when both the incumbent party and the opposition party have high levels of cohesion, which
interactively enhance the clarity of the party line. My argument is different from theirs in two
aspects. First, party cohesion conditions economic voting, not only through clarifying the
responsibility for the economy, but also through clarifying the line that separates the two parties.
The former clarity enables voters to know who should be responsible for the poor economic
outcomes and hence makes economic voting more likely. The latter, however, leads to the
salience of partisan cues and draws the focus of voters away from holding the incumbent party
accountable for the economy. Therefore, increases in party cohesion may both enhance and
hinder economic voting. From this perspective, party cohesion is a double-edged sword to
economic voting. Second, while Powell and Whitten only theorize incumbent/governing party
cohesion, I suggest that incumbent and opposition party cohesion should both be considered in
testing the theory of clarity of responsibility. I also theorize that incumbent party cohesion and
opposition party cohesion may interactively condition economic voting. Due to such an
interaction, increases in one party‘s cohesion do not necessarily enhance economic voting.
81

Further, when the two parties both have high levels of party cohesion, the high clarity of the
party line hinders economic voting.
This paper has made two major contributions. First, with the time-varying nature of party
cohesion considered and included into statistical models, this paper enrich our understanding of
how economic voting varies with the ebbs and flows of party cohesion. Economic voting is
strengthened as incumbent party cohesion increases from a low to mid level. However, economic
voting diminishes with incumbent party cohesion when incumbent party cohesion is already at a
high level. Second, the evidence in this paper demonstrates that opposition party cohesion and
the interaction between two parties‘ cohesion affect the strength of economic voting. In particular,
this paper shows that when two parties both become cohesive, voters are less likely to hold
incumbent party accountable for the economic conditions. When the Democrats and Republicans
both have high levels of intra-party cohesion, this paper provides useful information about how
party wars between the two parties influence economic accountability in the United States.
In 1950, the American Political Science Association (APSA) issued a well-known calling
for moving toward a responsible party system which stresses the need for cohesive parties
(American Political Science Association 1950). Since this report was issued, it has been believed
that party cohesion is the key to making government accountable to the public interest (Shugart
and Carey 1992; Carey 2009). Powell and Whitten‘s paper is in line with this belief. However,
my research provides counter-evidence. That is, high levels of party cohesion enhance clarity of
party line, hindering voters from holding the incumbent party accountable for economic
conditions. Rae (2007) has reminded us to ―Be Careful What We Wish For‖ when observing the
severe rivalries between two cohesive parties in contemporary American politics. My research
echoes Rae‘s calling and suggests that as far as economic accountability is concerned,
democracies should be careful in the pursuit of high levels of party cohesion.
82

CHAPTER 3
RESIST BEING HELD ACCOUNTABLE: THE IMPACTS OF PARTY COHESION ON
CONCENTRATION OF VOTES IN THE UNITED STATES

Ideally, ―governments are accountable if voters can discern whether governments are acting in
their interest and sanction them appropriately, so that those incumbents who act in the best
interest of citizens win reelection, and those who do not lose them‖ (Manin, Przeworski and
Stokes 1999, 40). Such accountability would not exist if incumbents are still able to win
reelection without performance approval from a broad range of citizens. In other words, lack of
accountability occurs when incumbents have the ability to resist being held accountable by the
public. In this research, such ability is called ―Resistibility.‖
In studies of accountability, resistibility may come from particularistic vote-interest
exchanges between the incumbents and certain subsets of citizens. When such an exchange
prevails, politicians are held accountable neither for the ―success in delivering collective goods,‖
nor for ―improving overall distributive outcomes along the lines favored by broad categories of
citizens‖ (Kitschelt and Wilkinson 2007, 2). In other words, by serving particularistic goods to
earn sufficient votes from certain subsets of the electorate, politicians may have leeway to avoid
being held accountable for the overall performance in serving the public interest.
It has been argued that the problems led by particularistic exchanges can be solved or
mitigated by high levels of party cohesion. Since a party has to provide programmatic national
policies that are broad enough in order to win national elections, a high level of cohesion not
only indicates that narrow-scope services are less likely to be offered, but also enable a party to
carry out policies that target a broad range of constituencies. Therefore, there has been a belief
that high levels of party cohesion help create programmatic politics in which a universal interest
of the national electorate can be served and a cohesive governing party can be held accountable
83

collectively for their programmatic policies (American Political Science Association 1950; Cox
1987; Shugart and Carey 1992).
It seems logically plausible that party cohesion would constrain particularistic exchanges
and improve accountability. However, empirically, it remains untested whether increases in party
cohesion improve the ability of the public to hold politicians accountable. Franzese, Jusko and
Nooruddin (2007) show that when party cohesion is higher, the number of effective
constituencies shrinks to be less than the number of geographic districts, meaning that local
concerns are reduced by the rise in party cohesion. As a result, spending that is geographically
targeting is also reduced.44 Although their research provided some evidence that party cohesion
may reduce the geographically targeted spending, it is not clear whether the reduction of
spending is due to even narrower targeting of particularistic services, rather than a shift from
particularistic exchanges to broad categories of services.
Theoretically, while a high level of party cohesion could provide the opportunity to serve a
broad constituency, the party may not have the incentive to do so. The incentive of vote
maximization does not necessarily translate into seeking maximization in every sub-unit of the
districts; rather, parties may target disproportionately larger shares of votes from particular
subsets of the electorate. Given that serving the public is difficult, a cohesive party may instead
have the incentive to target particular sets of voters and secure the base of support. McGillivray
(1997; 2004) has pointed out that the incumbent party with stronger party discipline is more
likely to use trade and industrial policies to target certain subsets of the voters. As such,
increasing party cohesion in order to improve accountability would backfire. An incumbent party

44

According to Franzese, Jusko and Nooruddin (2007), the effective number of constituencies is a linear
combination of geographic district and party cohesion. As party cohesion increases, the number of effective
constituencies decreases, and they argue that the particularistic spending will therefore be reduced.
84

with a higher level of party cohesion could be more capable of carrying out policies to target
particular subsets of voters.45 In return, these targeted constituents would be more likely than
others to vote for the incumbent party, leading to a higher concentration of votes. A higher
concentration of votes means that there are disproportionately larger shares of votes may cover
the losses from the entire constituency. Increases in party cohesion may consequently strengthen
the incumbent party‘s resistibility.
Given the two contradictory arguments, a straightforward way to investigate the impact of
party cohesion on resistibility is to directly examine the relationship between party cohesion and
the concentration of votes. In the existing literature, it is not a new idea that the concentration of
votes indicates the prevalence of particularistic exchanges. For example, Batto (2005) considers
that a higher concentration of votes is evidence of rent-seeking or pork-barreling activities. Lynn
(2008) also suggests that ―when votes are won on the basis of policy goods, leading candidates‘
votes will be more evenly distributed across the district‘s units or blocs‖ (Lynn 2008, 88). Lynn‘s
study has even provided some findings regarding the relationship between party cohesion and the
concentration of votes for Brazilian deputies. While she finds that the concentration of votes is
lower in the era when party cohesion is higher, her tests fail to fully capture the relationship
between party cohesion and concentration of votes. The first deficiency is that her comparison
between two periods of time did not show the effect of every one-unit increase in party cohesion
on the concentration of votes. Secondly, a high concentration of votes means that votes are not
spread evenly across the entire constituency. However, her measurement of the concentration of
votes, the Dominance Score, could have a larger value when a candidate in fact receives large but
evenly distributed shares of votes. In other words, the Dominance Score might indicate a high
concentration of votes when the votes are not concentrated. It does not validly distinguish
45

In this paper, an incumbent party is the party of the president of the United States.
85

between a high and low concentration of votes.46 Due to these weaknesses in her tests, the
relationship between party cohesion and the concentration of votes is not conclusive.
With data for party cohesion in the U.S House of Representatives and presidential elections
from 1992 to 2012, this paper attempts to test how concentration of votes varies with party
cohesion across time. I adopt Batto‘s measurement of the concentration votes, which is arguably
better than other measurements adopted in the literature. Additionally I take a step further to test
whether a high concentration of votes improves the chance of winning elections for an
incumbent party.47 In a nutshell, the statistical analyses will provide evidence that a higher level
of party cohesion helps the incumbent party resist being held accountable for the public interest.
Party cohesion is a key element of a responsible party system, and high levels of party
cohesion have been traditionally seen as improving accountability (APSA 1950; Cox 1987).
However, as this paper will show, a high level of party cohesion contributes to the concentration
of votes. Increases in party cohesion may facilitate the exchanges between the parties and
particular subsets of voters, leading to a stronger resistibility of the ruling party.

46

The Dominance Score for a candidate in a district used by Lynn is the weighted average of vote shares in all the
municipalities within that district. The weight for each municipality is the number of votes that a given candidate
receives in a given municipality divided by the number of total votes received by that candidate in all
municipalities of a district. Therefore, a greater dominance score could mean either votes being concentrated or a
strong candidate winning equally large vote shares in all the municipalities.

47

In this paper, the term ―incumbent party‖ refers to the party of the president.
86

Party Cohesion and Concentration of Votes
Party cohesion is the extent to which the legislators of a party vote with the party in Congress. A
high level of party cohesion indicates that most, if not all, legislators of a party vote in the same
manner on policies. Thus, with high levels of party cohesion, incumbent parties may have better
chances to carry out policies in their favor.
These policies can be particularistically targeting. Whenever a party carries out a policy, the
party to some extent knows where the support and opposition will come from, and the party
needs to evaluate the electoral impact of the policy. As such, a cohesive party may carry out
policies that serve particular subsets of voters to optimize their electoral gains and losses. In
return, the benefited voters would give disproportionately larger support to the incumbent parties,
leading to a high concentration of votes.
To be sure, it is not a new idea that policies in favor of particular sets of voters are more
likely to be produced when party cohesion is at a high level. McGillivray (1997; 2004) shared
this view and argued that a strong party discipline may enable an incumbent party to produce
trade and industrial policies that targets certain districts of constituencies. Furthermore, other
studies suggested that a strong party and partisanship may prevent parties from reaching
compromises (Rae 2007; Gerring, Thacker and Moreno 2009). For one reason, a high level of
party cohesion enhances the leverage of the majority party to unilaterally determine the
legislative outcomes; for another, high levels of party cohesion in the U.S. have led to a party
war where policies that might draw cross-party support have diminished in quantities (Sinclair
2006). As a result, policies would be more biased toward some subsets of voters when incumbent
party cohesion is at high levels.
In the U.S., the Republican Party and the Democratic Party traditionally have policies that
serve different subsets of the electorate, leading to different base of support for the two parties.
87

According to Axelrod, the Republicans received support from the non-poor, whites, nonunion
families, Protestants, northerners and voters outside of the central cities; while the Democrats
had been supported by the poor, blacks, unions, Catholics, southerners and the central cites
(Axelrod 1972, 15-6). In addition, the differences in the bases of support can also be
geographically distributed. The north-south division is a typical example. Furthermore,
traditionally the Democrats receive support from urban areas and the areas with larger
percentages of immigrants and blacks (Stonecash, Brewer and Mariani 2003). It is believed that
the Democrats‘ liberal policies on civil rights, immigration, gay marriage, abortion and other
issues have attracted the people living in urban areas and big cities. In sum, the two major parties
have different bases of support, and the social-racial base of support can also be geographically
distributed.
The geographically distributed base of support can vary with party cohesion. With a high
level of party cohesion, an incumbent party may have more leverage to carry out policies in
favor of the voters in some particular areas. As a consequence, the base of support should be
narrower, assuming that those benefited voters would be more likely than other citizens to vote
for the incumbent party in return. Conversely, when the legislators of an incumbent party do not
vote cohesively in Congress, the incumbent party has to negotiate with individual legislators and
the opposition party. Thus, the range of the benefited voters would be broader. In short, a higher
level of party cohesion would lead to a narrower range of support for the incumbent party, and
vice versa.
To test the association between party cohesion and the concentration of votes, I collect data
for the state-level concentration of votes and incumbent party cohesion from 1992 to 2012. The
state-level concentration of votes is calculated with the incumbent party‘s presidential votes

88

disaggregated by congressional districts.48 I adopt Batto (2005)‘s measure of concentration,
which is based on the chi-square statistic and accounts for theoretical range of concentration for
each party given its constituency-wide vote total (Batto 2005, 49). The advantage of Batto‘s
measurement over the standard deviation of vote shares is that it is adjusted by the theoretical
maximum concentration of each state.49 This adjustment avoids the possibility that the variation
in the concentration of votes comes from intrinsic characteristics of the electoral competition in
each state and election. To measure incumbent party cohesion, I take the percentage of the
members of a given incumbent party that forms the majority of the party in each roll call vote,
and I calculate average for each presidential term.50 The record of each roll call vote is taken
from Voteview.com.
As Figure 3.1 shows, when party cohesion started to arise, the average concentration of
votes also increased. Thus, the prima facie association shows that the concentration of votes
increased with party cohesion. The conventional wisdom that party cohesion may constrain the
concentration of votes is not supported by the empirical association here.

48

National concentration of votes means using presidential votes in each congressional districts to calculate a single
value of concentration of votes for each election year; Average state-level concentration of votes means using
presidential votes in each congressional districts to calculate a value of concentration of votes for each state and
then taking the average value of all states for each election year.

49

Please see the appendix for the details of the measurement.

50

For instance, .50 means on average 50 percents of the members from the given incumbent party vote in the same
direction in all the roll call votes during a presidential term.
89

.9

.2

.75

.8

.85

Incunbent Party Cohesion

.05 Concentration of Votes.15
.1
0
1992

1996

2000

2004

2008

2012

Election Year
Average State-Level Concentration of Votes
incumbent Party Cohesion

National Concentration of Votes

Figure 3.1. Party Cohesion and Concentration of Votes, 1992-2012

To further examine the relationship between party cohesion and the concentration of votes, I
conduct a regression analysis. The dependent variable is the concentration of votes for each state,
which is calculated with the presidential popular votes that each incumbent party received in
congressional districts within each state.51 I multiply the raw concentration of votes by 100 to
make it a percentage. The key independent variable, incumbent party cohesion, is measured at
the level of election year, and therefore a multilevel model is adopted as the estimation strategy. 52
51

As mentioned, I exclude some states that have only one congressional district. In this dataset, there are 43 states
for each election year and 258 cases in total.

52

The data is clustered at election years. Clustering it at states does not lead to any meaningful differences in the
results. Conceptually, it makes more sense to see incumbent party cohesion as a second-level factor which is
invariant across states in a given election year, rather than seeing it as a first-level characteristic of each state that
varies across election years. Accordingly, a fixed-effect model is not appropriate in this case since incumbent
party cohesion is not varying within a given election year.
90

I control for some demographic variables that would affect the concentration of votes. First,
South, which is a composite of eleven states in the former Confederate States of America, is
traditionally seen as a conservative area in the U.S. 53 After the 1980s, there have been
widely-spread conservative supporters consistently voting for the Republican Party.54 Therefore,
the votes received by both parties should be less concentrated in these states. Secondly,
Percentage of Rural Population depicts the level of urbanization of a state, which may capture
the density of population and conservative ideology at the same time. In a state with a larger
population living in a rural area, the density of population will be lower compared to the density
in a highly urbanized city (Rodden 2010). It is expected to see a lower concentration of votes
when there is a larger rural population. Accordingly, this variable is expected to have a negative
sign. Third, Percentage of Black People is another factor that might affect the concentration of
votes. It has been noted that black people tend to live together and mostly in urban areas
(Stonecash, Brewer and Mariani 2003). In a state where there is a larger population of black
people, the votes should be more concentrated. Therefore, I take percentages of black people for
each state from the census survey closest to a given election year and expect to see a positive
coefficient for this variable. Further, Percentage of Hispanic people is coded in the same way
with a similar consideration. Because Hispanic people are more likely to live together, in a state
where there is a larger population of Hispanic people, the votes received by an incumbent party
should be relatively more concentrated. Finally, I include Immigrants, which is the four-year

53

The eleven states of the south include Alabama, Arkansas, Florida, Georgia, North Carolina, South Carolina,
Texas, Virginia, Tennessee, Louisiana, and Mississippi.

54

Since the late 1970s, the conservative southerners have stopped supporting the Democratic Party and have begun
Republican identifiers. Moderate and conservative Democrats in Congress were replaced by conservative
Republicans (McCarty, Poole and Rosenthal 2006).
91

average number of people that have been naturalized in a given state. 55 Although this is not the
best measurement of immigration, it captures the proliferation of immigrants in a given state. A
large number of immigrants in a state would increase the concentration of votes since immigrants
tend to live around the people of their own nationality or ethnicity.
In addition to the demographic variables, I also control for political factors. First, In-Party
Governor is included. Since the in-party governor may utilize the resources from the state and
help the campaign of his or her party to expand the base of support, the incumbent party would
receive more widespread votes in states with a in-party governor than it would in other states. As
such, the concentration of votes in the states with an in-party governor can be lower than it is
otherwise. Second, an incumbent party‘s House Seat Share is the total seat share the incumbent
party has in the House of Representatives during the congressional term prior to a given
presidential election. A larger House seat share may indicate a greater diversity within a party
and a broader range of interests that a party may have to target. Hence, a greater share of House
seat may then reduce the possibility of an incumbent party carrying out platforms that target
particularistic subsets of voters. Accordingly, I expect this variable to have a negative coefficient.
Share of House Seats by State is the incumbent party‘s share of the House representatives elected
from a given state during the term prior to the presidential election year. For example, in the
congressional term prior to 1992, two out of the 7 representatives elected from Alabama were
from the incumbent party (i.e. the Republican Party), the Share of House Seats for the incumbent
party in 1992 is coded as 28. 57 %. This variable captures the numbers of congressional districts
in which an incumbent party was advantageous over the opposition party in the most recent
House election. When an incumbent party has won in more congressional districts in a given
state in the previous congressional election, it may also have presidential votes that are more
55

The unit of this variable is one thousand people.
92

widespread within that state. Democrat is coded as 1 if the incumbent party is the Democratic
Party and is coded as 0, otherwise. This variable is included to control for the partisan difference
between the two parties. It has been noted that the Democratic Party has support that is more
concentrated than the Republican Party (Gimpel and Schuknecht 2002; Rodden 2010). Therefore
this variable is expected to have a positive coefficient. The last variable I control for is
Redistricting. It has been argued that redistricting involves gerrymandering or creating
majority-minority districts. 56 Thus, redistricting may lead to more districts with partisan
homogeneity (Carson et.al. 2007; Theriault 2008). If so, redistricting may increase the
concentration of votes. I then include a dummy variable denoting the states that have
experienced changes in congressional districts after the census surveys in 1990, 2000 and 2010.57
In this dataset, states affected by redistricting in the 1992, 2004 and 2012 presidential elections
are coded as 1, and those that did not experience redistricting are coded as 0.
Table 3.1 presents the result of the regression analysis. Incumbent party cohesion has a
positive coefficient across all models, which supports the hypothesis that a higher level of
incumbent party cohesion may boost the concentration of votes for the incumbent party. 58 The
finding shows that a cohesive incumbent party does not necessarily carry out national policies
that target a broad range of the electorate. Rather, an incumbent party with a higher level of
cohesion targets more narrowly.

56

Gerrymandering refers to draw the boundaries of districts in favor of certain parties or politicians. It is named so
because a governor of Massachusett, Elbridge Gerry, made a classic case of Gerrymandering in 1812 (Niemi and
Abramowitz 1994); Majority-minority districts are districts with a minority population that is large enough to
elect a representative (Barreto, Segura and Woods 2004, 65).

57

According to the U.S. Constitution (Article 1, Section 2), redistricting is required for the states with changes in
population after the census surveys in every ten years.

58

There is no meaningful difference between a random effect model and a pooled model. They only slightly differ
in standard errors.
93

Table 3.1. Party Cohesion and Concentration of Votes
Model Ⅰ

Model Ⅱ

Model Ⅲ

β (S.E.)

β (S.E.)

β (S.E.)

.219 (.038)***

.195 (.040)***

.181 (.043)***

South

-.953 (.442)*

-.960 (.438)*

-1.152 (.479)*

Percentage of Rural
Population

-.033 (.011)**

-.029 (.011)*

-.040 (.012)**

Percentage of Blacks

.230 (.020)***

.227 (.020)***

.248 (.022)***

-.004 (.017)

-.003 (.017)

.046 (.017)**

.035 (.005)***

.034 (.005)***

In-Party Governor

-.367 (.246)

-.346 (.157)

-.151 (.266)

House Seat Share

-.049 (.036)

-.022 (.038)

-.006 (.041)

Share of State-Own House
Seats

-.002 (.005)

-.003 (.005)

-.003 (.005)

Democrat

.590 (.246)*

.654 (.245)**

.715 (.268)**

.701 (.312)*

.810 (.341)*

Incumbent Party Cohesion

Percentage of Hispanics
Immigrants

Redistricting
Constant
Number of Cases

-14.527
(2.509)***

-13.996 (2.501)***-13.557 (2.739)***

258

Number of Groups
6
Note: σ: p<.1; *: p<.05; **:p<.01; ***:p<.001

258

258

6

6

Some of the control variables have significant and expected effects on the concentration of
votes, while the other variables do not have coefficients that are significant at the conventional
level (p>.05). In the South, in the states with a larger rural population and in the states with an
in-party governor, incumbent parties have a lower concentration of votes. On the other hand,
incumbent parties have a higher concentration of votes in the states with a larger black
population and the states with a large increase in immigrants during a presidential term. Since
Immigrants may contain a large portion of Hispanic people, I exclude Immigrants, and
Percentage of Hispanic People become significant in the third model. Accordingly, the Hispanic
94

population may still contribute to the concentration of votes. Further, as expected, the
Democratic Party has a higher concentration of votes than the Republicans. Finally, redistricting
leads to a higher concentration of votes. This finding indicates that redistricting indeed create
more districts with partisan homogeneity.
In sum, while most of the literature about particularistic politics sees party cohesion as a
shield that prevents particularistic politics and as a tool that can be used to produce broadly
targeting policies, the evidence here suggests the opposite. When party cohesion increases, the
concentration of votes also increases. This finding supports my theory that high levels of party
cohesion may enable a party to target a narrower range of the electorate in return for
disproportionately larger support from the targeted sets of voters.

95

Concentration of Votes and Electoral Success
The previous section has shown that a higher level of party cohesion leads to a high
concentration of votes, this section is taking a step further to test how the concentration of votes
influences the electoral fortune of an incumbent party.
Assuming that incumbent parties are reelection seekers whose best interest is to stay in
office, serving particular subsets of voters, rather than earning widely-spread support from the
general citizens, may help incumbent parties resume their power.59 Although incumbent parties
may try their best to serve the public interests, it is difficult, if not impossible, to do so without
losing votes. In particular, a policy that may displease a party‘s supporters may not successfully
draw the support for the party from the other voters. Even if it does attract the support from
swing voters or voters who identify with the opposition party, such support might not be reliable.
Negative occurrences or shocks to valence (e.g. scandals, economic shocks, severe accidents or
crises and so on) may push these swing voters or opposition identifiers away. Accordingly,
rational reelection seekers would prioritize securing a base of support.
Furthermore, serving the public interest is not efficient for an incumbent party to secure the
electoral success. The public interest is hard to define, and the reward for serving the public
interest is not guaranteed. To serve the public interest requires efforts to study policies as well as
negotiations with the people affected. Such efforts will not be rewarded by the public if the
policies turn out to be ineffective in solving the problems (e.g. economic depression). Even
worse, the opposition party would oppose and deny the merits of the incumbent‘s policies, even
if the policies are in the best interest of the public. Therefore, it is highly likely that the attempts

59

Parties may have three goals to pursue: policy, office and votes (Müller and Strøm 1999). I assume that the three
goals could be pursued simultaneously as a party may carry out preferred policies to serve particular sets of voters
in return of votes that are sufficient to obtain the office(s).
96

and efforts to serve the public will not be appreciated in the end. In comparison, to identify the
interests of particular voters is easier, and the benefit-vote exchanges between the incumbent
party and these particular voters can be done with a greater certainty (Gimpel and Schuknecht
2002; Lynn 2008).
For the reasons above, securing a base of support by targeting particular subsets of voters
should be in a higher priority for an incumbent party who desires to survive in elections. In
return, disproportionately larger shares of votes from the benefited voters may help compensate
for the loss of votes from failures in serving the public interests. In short, a high concentration of
votes is beneficial to the electoral fortunes of an incumbent party. An incumbent party that
manages to receive votes that are concentrated would be more able to resist being held
accountable to the public interest.
To test the hypothesis that the concentration of votes improves the electoral fortune of an
incumbent party, I regress state-level vote shares of the incumbent parties in presidential
elections (in percentage) on the state-level concentration of votes used in the previous section.60
Although winning the presidency in the U.S. is not absolutely determined by the shares of
popular votes, the shares of popular votes are critical to win the electoral votes from states. Thus,
vote shares in each state may to a large extent indicate an incumbent party‘s success in a
presidential election. As for estimation, I adopt a random-effect model with data clustered at
election years.61
The control variables are slightly different from the previous analysis. I code the
demographic variables with a partisan consideration. Some variables such as Percentage of Black
60

Due to the small number of cases, I am not able to conduct a national-level analysis. Since electoral success in the
states determines the outcomes of presidential elections to a substantial extent, the analysis may also reveal how
concentration of votes affects the electoral fortunes of incumbent parties at the national level.

61

Clustering at states does not lead to any meaningful differences in the results.
97

People, Percentage of Hispanic People, Immigrants, Percentage of Rural Population, and the
South would have different effects on different parties‘ vote shares. Coding these variables
without a partisan consideration would lead to a misleading conclusion about their effects. For
example, a larger black population would help the Democratic Party, while making the
Republican Party inferior. Therefore, its effect should be positive when the incumbent party is
the Democratic Party and negative when the Republican Party is the incumbent. Accordingly,
these five variables are multiplied by -1 when the incumbent party is the Republican Party. As a
result, Percentage of Black People, Percentage of Hispanic People, Immigrants are expected to
have positive coefficients, meaning that these variables have positive effects on the Democratic
Party‘s vote shares and negative effects on the Republican Party‘s. On the contrary, Percentage
of Rural Population and the South are expected to have negative coefficients, meaning that the
two variables have negative effects for the Democratic Party and positive effects for the
Republican Party.
In addition to the demographic variables, I also include Log Median Income and the
Unemployment Rate for each state to control for the impact of the economy on the incumbent
parties‘ electoral fortunes.62 As for political factors, I include the incumbent parties‘ Shares of
House Seats by States and Senate Seats from each state, along with the In-Party Governor from
the previous section. Share of House Seats by States the incumbent party‘s share of the House
representatives elected from a given state during the term prior to the presidential election year.
This variable has been included in the model of the concentration of votes. It captures both the
electoral strength of an incumbent party in a given state and the ability of the party to deliver
―pork‖ to the state that might affect the electoral performance of the incumbent party. By the
same token, Senate Seats could be another important factor. Given that a party may at most have
62

Please see the appendix for the details of the variables.
98

two seats from each state, for the sake of simplicity, this variable is coded as 1 if a given
incumbent party has at least one senator from a given state, and 0 if the incumbent party does not
have a senator from that state. The above two variables are expected to have a positive impact on
the incumbent party‘s vote share. Another political factor, In-Party Governor, is to control for the
impact of a sitting governor from an incumbent party on the incumbent party‘s electoral
performance in a given state. Finally, in one of the models, I include the incumbent party‘s
Previous Vote Shares in each state to control for the support that a given incumbent party has
received in the previous election.
Table 3.2 shows the result of regression analysis. The concentration of votes has a positive
impact on state-level electoral success of the incumbent party, indicating that the concentration
of votes improves the incumbent party‘s electoral fortune. As such, a higher concentration of
votes makes it harder for the public as a whole to hold an incumbent party accountable.

99

Table 3.2. Concentration of Votes and Vote Shares
Model Ⅰ

Model Ⅱ

β (S.E.)

β (S.E.)

.447 (.131)**

.459 (.120)***

-6.498
(1.323)***

-4.622
(1.237)***

-.032 (.020)

.006 (.020)

Percentage of Blacks

.171 (.057)**

.129 (.052)*

Percentage of Hispanics

.089 (.041)*

.104 (.038)**

Immigrants

.019 (.016)

.008 (.015)

Log Median Income

-1.477 (2.599)

-.702 (2.376)

Unemployment Rate

-.634 (.250)*

-.986 (.234)***

.296 (.823)

.015 (.752)

(State-Own) Senate Seat

3.100 (.939)**

1.662 (.881) σ

Share of (State-Own) House
Seats

.176 (.016)***

.117 (.017)***

Concentration of Votes
South
Percentage of Rural
Population

In-Party Governor

Previous Vote Shares
Constant

.436 (.062)***
54.927
(28.647)σ

30.606 (26.382)

258

258

6

6

Number of Cases
Number of Groups

Note: σ: p<.1; *: p<.05; **:p<.01; ***:p<.001

South has a significant and negative coefficient. Because this variable is coded as 1 if the
Democratic Party is the incumbent and -1 if the Republican Party is the incumbent, the negative
coefficient means that the Democratic Party as an incumbent gains smaller vote shares in the
eleven states of the South than in other states, while the Republican Party as an incumbent gains
larger vote shares in the South than in other states. Percentage of Black People and Percentage of
Hispanic People both have a positive and significant coefficient. Due to the way of coding, the
100

positive coefficients mean that the Democratic Party as an incumbent receives larger vote shares
and the Republican Party receive fewer votes in the states where there is a larger black or
Hispanic population. Unemployment Rates has a significant and negative coefficient, indicating
that an incumbent party, on average, has a smaller vote share in the states with a larger
proportion of people unemployed. Senate Seat has a significant and positive coefficient,
suggesting that an incumbent party with at least a senator from a given state can have better
electoral performance in that state. However, after controlling for previous vote shares, the
coefficient is only significant at the .1 level. Finally, the coefficient for Share of House Seats by
State indicates that an incumbent party with more House representatives from a given state can
perform better in the election in that state.
Other variables, such as Percentage of Rural Population, Immigrants, Log Median Income
and In-Party Governor, fail to reach the conventional level of significance (p<.05). A possible
explanation for insignificance would be that they are correlated with one or more variables in the
model. Whereas it is not appropriate to drop any of these variables of theoretical importance, I
decided to include all the variables and leave this issue for future investigations.
In this section, I have shown that the greater the ability of an incumbent party to win strong
support from subsets of voters, the higher the chance for the party to survive the reelection,
regardless of performance in serving public interests. Hence, if a higher level of party cohesion
enhances such ability, a cohesive incumbent party will receive votes that are more concentrated
and therefore will have a better chance of winning elections. Consequently, the incumbent party
can count the electoral success on particular subsets of voters and keep the public in general from
holding the incumbent party accountable for serving public interests.

101

Discussion
Several theoretical implications can be drawn from the findings. To begin with, the findings do
not support the conventional belief in the impact of party cohesion on accountability. The
APSA‘s report and its following studies considered that high levels of party cohesion are required
to have programmatic politics in which the public as a whole may hold the incumbent party
accountable for providing collective or public goods. The rationale is that particularistic
exchanges can be constrained by high levels of party cohesion and the politics may focus on
programmatic party platforms that serve a broad range of citizens. If this was true, it would be
found that a higher level of party cohesion is associated with a higher concentration of votes.
However, the findings in this paper suggest the opposite. The conventional view was not
sufficient to account for the empirical politics in two aspects. For one, while a high level of party
cohesion provides the means for a party to carry out policies that serve national-wide interests, a
cohesive party may not have incentive to serve the public. For another, policy benefits are not
necessarily distributed evenly across the country, and policies can also be produced to
geographically target particular subsets of voters. A high level of party cohesion may instead
enable an incumbent party to carry out policies in order to serve particular subsets of voters. As a
result, the incumbent party may win elections with support from these targeted voters and resist
being held accountable to the public.
To elaborate, it has been argued that parties do not maximize votes, while they only pursue
votes in a sufficient amount to get into office (Riker 1962; Müller and Strøm 1999). If so, a party
may not have incentive to serve as broadly as possible in order to maximize its votes. Also, it is
difficult for a party to serve various interests simultaneously, and the return in terms of votes
from the public, in general, is not certain. Therefore, parties may have incentive to rely on
exchanges with particular subsets of voters to secure their base of support, given that the interests
102

of particular sets of voters are easily identified and the returns from the particularistic exchanges
can be expected with a higher certainty. While a variety of particularistic exchanges in
democracies have been observed, and scholars have believed that higher levels of party cohesion
may constrain such exchanges, the findings in this research suggest that high levels of party
cohesion may facilitate a ―policy-based particularistic exchange‖ between incumbent parties and
voters, hindering voters from holding incumbent parties accountable for public interests.
Secondly, political scientists have discussed the distinctions between policy-based and
particularistic politics or the differences between functional and territorial issues (Caramani
2004). However, these categorizations might not be meaningful, since in reality there is no
clear-cut distinction between each category. Almost no single policy can serve the interest of
every subset of people or every area of a nation. Different sets of people respond to the same
policy differently. When a policy is carried out, some gain and others lose. To some degree, all
policies have a target, and policy-based politics also involves with particularistic exchanges
between parties and voters. By a similar notion, functional issues are not absolutely distinct from
territorial issues because functional issues can be responded to differently across geographical
boundaries. As mentioned, trade and industrial policies may bring benefits to some local
constituencies while they may impose costs to others (McGillivray 2004). For another example,
urban and rural areas may respond differently to environmental and agricultural issues. Likewise,
in regards to energy issues, people living around nuclear power plants would not be in agreement
with those who live far away. Therefore, functional issues can also be territorial, which is why
even in a developed democracy, such as the U.S., regionalism still matters in politics (Gimpel
and Schuknecht 2002). Further, even if parties with high levels of cohesion will compete over
functional issues in a nation-wide constituency, their votes may not be distributed evenly across
the nation. Instead, the geographical concentration of votes in the United States has increased as
103

the partisan rivalries between two parties have escalated. If parties that become cohesive are
more able to serve particular local interests, the nationalization of American parties, which have
been discussed by scholars (Caramani 2004; Brady, D‘Onofrio and Fiorina 2000), might simply
be localization in disguise.
Returning to the main topic of this paper, to put the impact of party cohesion into
perspective, high levels of party cohesion may grant parties with more leverage to serve
particular subsets of voters in pursuit of electoral success. When incumbent parties can rely on
particular subsets of voters and win elections without broadly-spread support from the whole
public, they will not be held accountable for serving public interests. Consequently, the lack of
accountability to the public interest may appear as a problem in American democracy.

104

Conclusion
High levels of party cohesion were desired because political scientists believed that when party
cohesion is at a high level, programmatic and policy-based politics can be expected, and
particularistic exchanges can be constrained. From this perspective, the ruling elites will be held
accountable collectively for serving the best interest of the public when there is a high level of
party cohesion. However, this research has demonstrated three points that shatter the above view.
First, the evidence from the United States shows that the concentration of votes increased
with party cohesion. This would not be observed if high levels of party cohesion really made the
incumbent parties act in the best interest of the whole public, bringing out policies that target
nation-wide constituencies. Second, a higher concentration of votes is associated with a larger
vote share, indicating that the increasing concentration of votes benefits incumbent parties in
elections. This finding not only suggests that incumbent parties may have incentive to make their
votes more concentrated, but also reveals that incumbent parties can further their chances of
winning elections by boosting votes from particular subsets of voters. Third, combining the first
and second piece of evidence, a high level of party cohesion may give an incumbent party
leeway to win elections without general, nation-wide support, which implies that accountability
is diminishing as party cohesion is increasing.
Due to the limited availability of data, this paper only investigates the state-level
concentration of votes through the presidential votes at the level of congressional districts
between 1992 and 2012. Future research may further gather the data that is disaggregated at the
county level for more election years. Also, it would be interesting if one were to study the
relationship between party cohesion and the concentration of votes in other democracies.
Maybe, not all politics is local, and maybe, national issues matter more than local

105

concerns.63 However, neither of these two statements is relevant if the opinions toward national
issues are not distributed evenly across all localities, and if voters from different localities cast
votes differently for the same national policies. When national policies have implications that
vary among localities, what is truly relevant is how incumbent parties utilize higher levels of
party cohesion to target people from certain localities and win elections. Therefore, the impacts
of party cohesion on the concentration of votes, and on accountability, are worth further studies.

63

―All politics is local‖ is a well known political saying in American politics. It emphasizes the importance of local
factors to American politics. The origin of the phrase has been believed to be Senator Tip O‘Neill who used this
phrase in his campaign in 1935.
106

CONCLUSION

This dissertation features party cohesion as a factor that leads to changes in accountability.
Political scientists have believed that a high level of party cohesion may improve accountability.
However, the evidence provided by this dissertation does not sustain such a belief. Rather, by
taking into account the variation of party cohesion across time, the three essays demonstrate that
voters are less likely to hold ruling elites accountable for the public interest when party cohesion
increases.
The first essay revisits a conventional view that a legislator‘s party loyalty is costly to his or
her electoral fortune. I assume that legislators are rational actors who will not commit political
suicide by increasing their loyalty to the party as doing so is electorally costly. Hence, the recent
rise of party loyalty indicates that the electoral cost of party loyalty may have been reduced. This
reduction of electoral cost, I argue, can be attributed to increases in income inequality. Increases
in income inequality may lead voters to care about the success of the party. Hence, voters are less
likely to punish legislators who vote loyally with the party in Congress. With data from the U.S.
House elections from 1976 to 2008, I have demonstrated that the electoral cost of party loyalty
decreases with income inequality. This finding implies that party-oriented accountability and a
responsible party system are likely to exist in the era of high income inequality, despite the
political institutions that may constrain party loyalty.
The second chapter is an attempt to challenge Powell and Whitten (1993)‘s thesis of clarity
of responsibility. They proclaimed that party cohesion may improve clarity of responsibility,
enhancing economic voting, while I find that a high level of party cohesion may instead hinders
voters from holding the incumbent party accountable for economic conditions. My argument is
featuring the interaction between two parties, which has long been neglected by the literature.
107

High levels of incumbent and opposition party cohesion may interactively make the party line
clear. The clear party line proliferates partisan cues that may lead voters to downplay their
concerns about actual economic conditions when making voting decisions. With data from the
American National Election Studies (ANES) between 1980 and 2008, I demonstrate that party
cohesion is a double-edged sword that both improves and weakens economic voting. In particular,
when both parties have a high level of party cohesion, it is less likely that the incumbent party is
held accountable for the economy. This finding not only shatters Powell and Whitten‘s thesis, but
also contributes to the literature of economic voting by analyzing opposition party cohesion and
clarity of party line.
The last essay investigates whether a high level of party cohesion may give the incumbent
party ability to resist being held accountable to the public. It is to directly test the conventional
belief that high levels of party cohesion may make programmatic policies the focus of politics so
that the incumbent party can be held accountable for the public interest. Through the
investigation of the relationship between party cohesion and the concentration of votes received
by the incumbent party, I find that increases in party cohesion may instead lead to a higher
concentration of votes. This implies that an incumbent party may serve a narrower range of the
electorate when it has a higher level of party cohesion. I further demonstrate that a higher
concentration of votes is beneficial to the incumbent party‘s electoral fortune. A high level of
party cohesion enables the incumbent party to target particular sets of the electorate, which gives
the party disproportionately large support that covers the loss of votes from the entire public. In
sum, a high level of party cohesion may enable an incumbent party to win elections on the basis
of exchanges with particular subsets of voters, leading to lack of accountability to the entire
public.
With these three studies, this dissertation has made several contributions. First, this
108

dissertation is one of the few efforts devoted to studying the consequences, rather than the causes,
of party cohesion. Since the APSA‘s report, scholars have devoted efforts to studying why party
cohesion is low and how to increase it in the United States. Theories, including conditional party
government (Aldrich and Rohde 2000a; Aldrich and Rohde 2001; Rohde 1991), procedural cartel
theory (Cox and McCybbins 1993) and strategic party government (Lebo, McGlynn and Koger
2009), have suggested various conditions and factors leading to increases in party cohesion. In
the studies of comparative politics, Carey (2009), Carey and Shugart (1995), Cox (1987), Kim
(2009) and Shugart and Carey (1992) also provide explanations to the variation of party cohesion.
In short, these studies were designed to answer a ―how-so‖ question. However, as party cohesion
in the United States has increased to a high level, my dissertation attempts to answer a ―so-what‖
question by investigating the impacts of party cohesion on accountability. The answer provided
here suggests that a high level of party cohesion may not improve accountability in the way that
many political scientists have wished for.
Second, unlike previous literature, this dissertation takes into account the variation of party
cohesion across time in studying the impacts of party cohesion. In this dissertation, the
conventional wisdom that increases of party cohesion may improve democratic accountability
has been tested with data from the United States. The attempt is to provide empirical evidence
showing how changes in party cohesion affect the way in which voters hold ruling elites
accountable. The findings in the three essays are contradictory to the conventional wisdom. They
demonstrate that when party cohesion is at a high level, voters may hold ruling elites accountable
for the sake of partisan goals, but not the interest of the public. This dissertation provides
theories and empirical evidence that capture the impacts of party cohesion on accountability in
contemporary American politics.
Third, the impacts of party cohesion found in this dissertation have implications on studies of
109

political institutions. The United States does not have the political institutions that create
party-oriented politics in which the role of party is emphasized and parties have high levels of
cohesion. The presidential system and Single Member District system (SMD) with primary
elections have been regarded as reasons why party-oriented politics and a responsible party
government are not likely to exist in the U.S. However, the recent rise of party cohesion shatters
this view. Regardless of these institutions, party cohesion in the United States has still increased,
and the evidence here suggests that income inequality is one of the driving forces behind this
development. Another implication on studies of political institutions is that the effects of political
systems should take the interaction between political actors into consideration. For example,
Powell and Whitten (1993) treat party cohesion as a stable feature that can improve clarity of
responsibility and economic accountability. When the interaction between two cohesive parties is
considered, party cohesion affects economic accountability not only through clarity of
responsibility, but also clarity of party line. Contrary to clarity of responsibility, clarity of party
line may hinder voters from holding the incumbent party accountable for the economy. Another
example is that party cohesion may not improve accountability since parties with a high level of
cohesion may not use it to target the electorate as broadly as possible. When an incumbent party
has incentive to target particular sets of the electorate, institutions that may induce high levels of
party cohesion may instead help the incumbent party to resist being held accountable for the
public interest.
Fourth, while the literature shows that income inequality may lead to lower political
participation (Gaventa 1980; Pateman 1971; Schattscneider 1960; Solt 2008; Solt 2010), political
polarization (Bartels 2004, 2008; Gelman 2008; Garand 2010; McCarty, Poole, and Rosenthal
2006), democratization (Acemoglu and Robinson 2006; Boix 2003) and violence (Nepal, Bohara
and Gawande 2011; Weede 1981), my research demonstrates that income inequality has an
110

impact on accountability. That is, increases in income inequality may lead voters to care about
the success of the party. Hence, loyal representatives who can assist the party to achieve success
are more likely to be rewarded than to be punished when there is a higher level of income
inequality. Such an effect may lead to party-oriented politics and a responsible party government,
which has been desired for years by political scientists.
This dissertation contributes several new paths for future research. First, due to the
availability of data for party cohesion, this dissertation focuses only on the United States.
Although it is easy to hold country-specific factors (e.g. political institutions) constant in a
single-country study, a comparative study may help further understand the impacts of party
cohesion on accountability across different political systems and the interaction between party
cohesion and political institutions. The evidence shows that some countries that have had low
levels of party cohesion, such as the United States, Brazil and Taiwan, have increasing party
cohesion while others that had a high level of party cohesion, such as the United Kingdom, have
sometimes experienced low levels of party cohesion (Canton and Herrnson 1997, 395; Kam 2009;
Ladewig 2010; Lyne 2008; McCarty, Poole and Rosenthal 2006; Sinclair 2006; Sheng 2008).
Comparative studies on the impacts of party cohesion across these political systems will provide
a more thorough view on this topic.
Second, this dissertation is solely based upon party cohesion in the House of Representative,
while an investigation of the party cohesion in the Senate might provide additional findings for
the impacts of party cohesion. Since party polarization and increases in party cohesion also
occurred in the Senate (Garand 2010; Theriault 2008), studies on how senators are held
accountable may be crucial in understanding the effects of party cohesion on accountability.
Third, this dissertation only focuses on election-wise dependent variables, while
accountability can also be studied by investigating attitudes toward the government or
111

democratic politics. This dissertation finds that income inequality is an important factor that
makes party-oriented politics possible in the United States. Numerous studies also relate income
inequality to party polarization (Garand 2010; McCarty, Poole and Rothenthal 2006). A
straightforward extension for future research is to study how income inequality may affect
individual voters‘ attitudes toward legislators‘ party loyalty. It has been shown that individual
voters indeed respond to roll call votes in Congress (Ansolabehere and Jones 2010). Moreover,
an individual voter with strong partisanship has lower confidence in a bi-partisan Congress
where legislators behave as if no party line is separating them (Harbridge and Malhotra 2011).
Combining these studies, my research suggests that income inequality may lead voters to have a
positive attitude toward legislator‘s party loyalty. Another potential topic for research is that
income inequality may enlarge the winner-loser gap in political attitudes. Winner-loser gap is a
phenomenon that has been a recent focus of political science research. It means that voters who
have supported candidates or parties that won the previous election have more supportive
attitudes toward the government or democratic practice than those who have supported the losers
in the election (Anderson et. al. 2005). Given that income inequality may enlarge issue cleavages,
voters who supported the loser in the election would suffer a greater loss when income inequality
increases. Accordingly, the winner-loser gap would be larger in countries with higher income
inequality.
Fourth, this dissertation investigates electoral outcomes, while studies of specific policies
may further discover the effects of party cohesion. For example, although it is questionable
whether voters look into legislators‘ voting record on specific issues, it would be worth
investigating how party cohesion on economic policies affects economic voting. Another
extension for research is to study the impacts of party cohesion on the distribution of policy
benefits. Franzese, Jusko and Nooruddin (2007) examine how party cohesion is related to the
112

amount of locally-targeting spending. However, as this dissertation suggests, policy benefits can
be distributed to narrower sets of the electorate while it decreases in amount. Accordingly, it is
the distribution, rather than the amount, of policy benefits that shows the impacts of party
cohesion on particularistic exchanges between ruling elites and voters. A study examining the
impacts of party cohesion on the distribution of policy benefits would further clarify the
relationship between party cohesion and accountability.
To conclude, this dissertation has made contributions to understanding the impacts of party
cohesion on accountability. Political scientists have once believed that a high level of party
cohesion may lead to a party-oriented politics where democratic accountability can be improved.
However, the evidence from the United States indicates that the way voters hold ruling elites
accountable have become partisan, and the accountability for the interest of the public have
diminished in party-oriented politics. As party cohesion has increased and partisan rivalry has
been getting severe, the trade-off between party-oriented accountability and the ideal
accountability is a new angle to explore and examine democratic politics. Given the benefits and
costs of party cohesion, it is time to reconsider whether a democracy needs to maximize the level
of party cohesion. Maybe, the most important and challenging task for political scientists is to
figure out an optimal level of party cohesion that might really bring us what we have wished
for—a democracy where citizens hold ruling elites accountable for their performance in serving
the best interest of the public.

113

APPENDIX

114

Table A.1. Summary Statistics for Chapter One

Vote Share
Party Cohesion
Gini Index
Decile Ratio
Ideology
Fresh
Party Leader
Challenger‘s Quality
Polarization
Aggregate Party
Loyalty
Log GDP
Unemployment rate
Midterm

Mean

Standard
Deviation

Minimum

Maximum

68.36
84.82
40.65
6.76
.37

14.29
14.73
2.86
1.10
.18

2.35
3.96
35.8
4.13
0

100
100
44.4
12.59
1.239

.17
.02
.17
.72

.17

0
0
0
.49

1
1
1
.99

84.82

6.33

73.42

95.49

10.04
6.18

.50
1.37

9.02
3.97

10.76
9.71

1978, 1982, 1986, 1990, 1994, 1998, 2002, 2006

115

Table A.2. Summary Statistics for Chapter Two
Standard
Deviation

Mean

Minimum Maximum

Vote For Incumbent Party

.40

0

1

Economic Perception (Worse=1)

.33

0

1

Incumbent Cohesion

79.91

5.01

72.5

90.07

Opposition Cohesion

82.92

5.12

75.87

90.54

6647.27

790.27

5582.05

7783.09

.49

0

1

Partisanship (incumbent party=1)

.29

0

1

Ideology (same with incumbent
party=1)

.74

.52

.99

Race

.56

0

1

Incumbent Cohesion × Opposition
Cohesion
Approve of President's Job
(Approve=1)

.17

On-Going War

1984, 1992, 2004

Reelection

1984, 1988, 1992, 1996, 2000, 2008

Divided Government

1980, 1984, 1992, 1996, 2004

Note: The mean values for dummy variables at the micro-level are the percentage of the
respondents who are coded as 1. For the variables at the election level, I simply list the election
years in which the variables are coded as 1.

116

Calculation of Concentration of Votes in Chapter Three
Batto (2005) calculates Concentration Index as following:

Cd
 Cd

2
(
C

C
)
(
(Vd  Cd )  0) 2 
d
d
 ( Ei  Ci )   Vd
V
,
 d
Concentration Index= 
 /
C
C
E
d
d

i


Cd
(Vd  Cd ) 


Vd
Vd
2

Where E i is the expected number of votes that a candidate receives in a precinct, i, and C i is the
actual number of votes received by that candidate. E i is calculated as the number of total votes in
precinct i times the district-wide vote share of the candidate (C d / V d ).
The numerator is equivalent to the formula to calculate chi-square. When a candidate
receives votes from some precincts that are different than what he is expected to receive, the
votes are not distributed evenly within the district. The denominator is the theoretical range of
the numerator. The theoretical maximum of the numerator occurs when a candidate obtains all
his district-wide votes solely from a precinct, while the minimum is 0 as all the observed votes
are equal to the expected votes.64 Accordingly, the value of concentration index ranges from 0 to
1.
To calculate national concentration of votes for a given party, C d is the national-wide votes
received by that party, and V d is the total votes cast in the whole country. Since the lower level
64

In Batto‘s explanation, ―This value is at a minimum when the candidate gets exactly the same vote share in all
precincts. If C i = E i for all i, then this formula will yield a value of 0. The value reaches a maximum when, for a
given district-wide vote total, C d , the district can be divided into two sets of precincts. One set has a sum of
exactly C d valid votes, and the candidate gets all C d of those votes. The other set has a sum of V d - C d valid
votes, and the candidate gets none of those votes. Mathematically, the number or size of the precincts in each of
the sets is unimportant, and it is convenient to represent the district as having only two precincts, one of size C d
and one of size V d - C d ‖ (Batto 2005, 58-9).
117

unit here is each congressional district, E i is the expected number of votes that a party receives
in a congressional district i, and C i is the actual number of district votes received by that party.
For the case of state-level concentration, C d is the state-wide votes received by a party in a
given state and V d is the total votes cast in that state. E i is the expected number of votes that a
party receives in a congressional district i of that given state, and C i is the actual number of
district votes received by that party.
Regarding the sources of data, the presidential votes in congressional districts comes from
three sources. For the elections from 1992 to 2004, I take the data from the website of Polidata;
for the election in 2008, the data is drawn from the Swing State Project; and for the election in
2012 is from the website of Daily Kos.

118

Table A.3. State-level Concentration of votes, 1992-2012
1992
State
Incumbent Opposition
Alabama
0.047
0.060
(R)
(D)
Arizona
0.006
0.111
Arkansas
0.009
0.017
California
0.034
0.066
Colorado
0.023
0.030
Connecticut
0.010
0.008
Florida
0.031
0.045
Georgia
0.038
0.060
Hiwai
0.002
0.000
Idaho
0.001
0.001
Illinois
0.050
0.093
Indiana
0.019
0.030
Iowa
0.004
0.003
Kansas
0.002
0.006
Kentucky
0.003
0.007
Louisiana
0.054
0.078
Maine
0.001
0.000
Maryland
0.049
0.091
Massachusetts
0.005
0.016
Michigan
0.032
0.064
Minnesota
0.010
0.021
Mississippi
0.021
0.033
Missouri
0.013
0.012
Nebraska
0.003
0.009
Nevada
0.006
0.012
New
0.001
0.001
New
Jersey
0.020
0.030
Hampshire
New Mexico
0.003
0.007
New York
0.042
0.094
North
0.026
0.040
Ohio
0.027
0.036
Carolina
Oklahoma
0.018
0.018
Oregon
0.007
0.014
Pennsylvania
0.036
0.059
Rohde Island
0.001
0.002
South
0.025
0.043
Tennessee
0.017
0.027
Carolina
Texas
0.036
0.064
Utah
0.009
0.010
Virginia
0.021
0.035
Washington
0.017
0.031

1996
Incumbent Opposition
0.067
0.059
(D)
(R)
0.014
0.012
0.017
0.013
0.061
0.060
0.028
0.030
0.006
0.008
0.049
0.042
0.056
0.044
0.000
0.002
0.001
0.000
0.087
0.067
0.030
0.029
0.004
0.004
0.012
0.011
0.007
0.004
0.048
0.040
0.000
0.001
0.088
0.067
0.010
0.007
0.056
0.045
0.015
0.012
0.036
0.028
0.040
0.030
0.009
0.006
0.013
0.009
0.000
0.000
0.037
0.031
0.003
0.003
0.084
0.059
0.043
0.036
0.045
0.042
0.019
0.020
0.016
0.016
0.061
0.048
0.001
0.000
0.039
0.031
0.034
0.026
0.064
0.058
0.014
0.012
0.033
0.028
0.024
0.022
119

2000
Incumbent Opposition
0.065
0.060
(D)
(R)
0.012
0.013
0.013
0.011
0.073
0.079
0.032
0.038
0.007
0.008
0.057
0.056
0.075
0.072
0.000
0.001
0.000
0.000
0.094
0.093
0.036
0.035
0.006
0.006
0.012
0.012
0.011
0.011
0.061
0.059
0.001
0.001
0.100
0.099
0.008
0.010
0.060
0.057
0.020
0.026
0.038
0.038
0.056
0.055
0.015
0.017
0.016
0.015
0.000
0.000
0.040
0.037
0.005
0.007
0.089
0.087
0.023
0.023
0.043
0.044
0.010
0.010
0.030
0.033
0.071
0.068
0.001
0.001
0.033
0.030
0.042
0.041
0.071
0.073
0.015
0.023
0.027
0.027
0.039
0.047

Table A.3.(cont‘d)
0.007
0.005
0.017
0.017
0.030
0.025
2004
2008
State
Incumbent Opposition Incumbent Opposition
Alabama
0.066
0.067
0.089
0.156
(R)
(D)
(R)
(D)
Arizona
0.021
0.021
0.023
0.023
Arkansas
0.009
0.009
0.005
0.006
California
0.094
0.092
0.079
0.078
Colorado
0.049
0.049
0.048
0.047
Connecticut
0.005
0.005
0.005
0.005
Florida
0.056
0.058
0.079
0.082
Georgia
0.125
0.125
0.118
0.120
Hiwai
0.001
0.001
0.001
0.001
Idaho
0.000
0.000
0.000
0.000
Illinois
0.093
0.093
0.081
0.084
Indiana
0.040
0.041
0.034
0.035
Iowa
0.011
0.011
0.012
0.012
Kansas
0.018
0.019
0.022
0.023
Kentucky
0.013
0.014
0.023
0.025
Louisiana
0.070
0.071
0.060
0.063
Maine
0.001
0.001
0.003
0.004
Maryland
0.093
0.094
0.102
0.105
Massachusetts
0.016
0.015
0.023
0.024
Michigan
0.058
0.060
0.048
0.050
Minnesota
0.035
0.034
0.034
0.034
Mississippi
0.047
0.047
0.065
0.067
Missouri
0.069
0.071
0.075
0.077
Nebraska
0.018
0.018
0.028
0.029
Nevada
0.014
0.015
0.013
0.013
New
0.001
0.001
0.001
0.001
New
Jersey
0.042
0.043
0.048
0.049
Hampshire
New Mexico
0.010
0.010
0.011
0.012
New York
0.084
0.086
0.094
0.097
North
0.042
0.042
0.046
0.048
Ohio
0.048
0.048
0.043
0.045
Carolina
Oklahoma
0.007
0.007
0.008
0.008
Oregon
0.033
0.036
0.035
0.035
Pennsylvania
0.087
0.092
0.075
0.077
Rohde Island
0.002
0.002
0.002
0.002
South
0.031
0.033
0.032
0.033
Tennessee
0.048
0.049
0.074
0.077
Carolina
Texas
0.072
0.072
0.083
0.084
Utah
0.010
0.011
0.007
0.008
West Virginia
Wisconsin
Mean

0.004
0.012
0.018

0.008
0.024
0.033

120

0.005
0.028
0.033

0.004
0.031
0.034

2012
Incumbent Opposition
0.099
0.096
(D)
(R)
0.049
0.049
0.007
0.006
0.080
0.084
0.041
0.041
0.007
0.012
0.067
0.065
0.134
0.131
0.000
0.001
0.000
0.000
0.095
0.093
0.037
0.036
0.008
0.008
0.017
0.017
0.043
0.042
0.109
0.105
0.004
0.004
0.060
0.059
0.024
0.025
0.068
0.067
0.038
0.040
0.075
0.072
0.101
0.097
0.024
0.022
0.021
0.021
0.002
0.002
0.068
0.067
0.011
0.013
0.104
0.102
0.077
0.076
0.060
0.058
0.010
0.010
0.044
0.046
0.103
0.090
0.004
0.004
0.052
0.050
0.102
0.099
0.106
0.104
0.013
0.013

Virginia
Washington
West Virginia
Wisconsin
Mean

0.035
0.054
0.001
0.040
0.039

0.036
0.054
0.002
0.040
0.039

Table A.3.(cont‘d)
0.040
0.052
0.000
0.042
0.041

121

0.041
0.053
0.000
0.041
0.044

0.061
0.054
0.002
0.056
0.050

0.060
0.055
0.002
0.056
0.049

Table A.4. Control Variables in Chapter Three
Variables

Description
Model of Concentration

Source

Model of Vote Shares
1: The eleven
Confederate States and
the incumbent party
(a.k.a the president‘s
party) is the

South

1: The eleven Confederate Democratic Party.
states.
-1: The eleven
Collected by Author
0: otherwise.
Confederate States and
the incumbent party
(a.k.a the president‘s
party) is the
Republican Party.
0: The other 32 states

The percentage of
people living in rural
areas in a given state, if
the incumbent party is
the Democratic Party.
Percentage of Rural
Population

The percentage of people
living in rural areas.

122

-(The percentage of
people living in rural
areas in a given state) ,
if the incumbent party
is the Republican
Party.

The census survey
nearest to a given
election year from
Census of Bureau
(Access:
12/12/2013).

Table A.4.(cont‘d)

Percentage of Blacks

The percentage of people
that are black.

The percentage of
people that are black, if
the incumbent party is
the Democratic Party.
-( The percentage of
people that are black),
if the incumbent party
is the Republican
Party.

Percentage of
Hispanics

The percentage of people
that are Hispanic.

The percentage of
people that are
Hispanic, if the
incumbent party is the
Democratic Party.
-(The percentage of
people that are
Hispanic) , if the
incumbent party is the
Republican Party.

The census survey
nearest to a given
election year from
Census of Bureau
(Access:
12/12/2013).

The census survey
nearest to a given
election year from
Census of Bureau
(Access:
12/12/2013).

Four-year average
number of people
naturalized in a state, if
the incumbent party is
Four-year average number the Democratic Party.
The Department of
Immigrants

of people naturalized in a
state. (unit: thousand
people)

Log Median Income

123

-( Four-year average
Homeland Security
number of people
(Access:
naturalized in a state) ,
12/12/2013).
if the incumbent party
is the Republican
Party.
(unit: thousand people)
Log of median income

Census of Bureau

in each state during the
election year.

(Access:
12/12/2013).

Table A.4.(cont‘d)
Unemployment Rate in
each state during the
election year.

Unemployment Rate

In-Party Governor

The governor of the
The governor of the state is
state is from the
from the incumbent party
incumbent party during
during the election year.
the election year.

Bureau of Labor
Statistics (Access:
12/12/2013).
The US50.com
(Access:
12/12/2013).

1, if at least one of the
senators from the state
during the
congressional term
nearest to the election
year is affiliated with
the incumbent party.

(State-Own) Senate
Seat

Share of State-Own
House Seats

House Seat Share

0, none of the senators
from the state during
the congressional term
nearest to the election
year is affiliated with
the incumbent party.

Voteview.com
(Access:
12/12/2013).

The percentage of House
representatives elected

The percentage of
House representatives

The Office of Clerk

from the state that are

elected from the state

of House of

affiliated with the
incumbent party.

that are affiliated with
the incumbent party.

Representatives

The incumbent party‘s
share of total House seats
during the term nearest to
the election year.

124

The Office of Clerk
of House of
Representatives

Table A.4.(cont‘d)

Democrat

1, if the incumbent party is
the Democratic Party;
0, otherwise.

Collected by Author

States affected by
redistricting in the
presidential elections in
Redistricting

1992, 2004 and 2012 are
coded as 1,
and those that did not
experience redistricting are
coded as 0.

125

Census of Bureau
(7/28/2014)

Table A.5. Web Links in Chapter Three
Variables

Percentage of Rural
Population

Source

link

The census survey
nearest to a given
election year from
Census of Bureau
(Access: 12/12/2013).

1990 and 2000:
http://www.census.gov/compendia/statab/
cats/population.html.
2010:
http://www.census.gov/geo/reference/ua/u
rban-rural-2010.html

The census survey
nearest to a given
http://www.census.gov/population/www/d
Percentage of Blacks
election year from
ocumentation/twps0056/twps0056.html.
Census of Bureau
(Access: 12/12/2013).
The census survey
nearest to a given
http://www.census.gov/population/www/d
Percentage of Hispanics
election year from
ocumentation/twps0056/twps0056.html.
Census of Bureau
(Access: 12/12/2013).
The Department of
https://www.dhs.gov/yearbook-immigrati
Immigrants
Homeland Security
on-statistics.
(Access: 12/12/2013).
Census of Bureau
http://www.census.gov/hhes/www/income
Log Median Income
(Access: 12/12/2013). /data/statemedian/.
Bureau of Labor
http://www.bls.gov/schedule/archives/all_
Unemployment Rate
Statistics (Access:
nr.htm#SRGUNE.
12/12/2013).
Partisanship of
governors from The
In-Party Governor
http://www.theus50.com/.
US50.com (Access:
12/12/2013).
Voteview.com
(State-Own) Senate Seat
http://voteview.com/downloads.asp.
(Access: 12/12/2013).
The Office of Clerk
Share of (State-Own) House
http://clerk.house.gov/member_info/electi
of House of
Seats
onInfo/.
Representatives
126

Table A.5.(cont‘d)

House Seat Share

Redistricting

The Office of Clerk
of House of
Representatives.
Census of Bureau
(7/28/2014)

127

http://history.house.gov/Institution/PartyDivisions/Party-Divisions/.
https://www.census.gov/geo/maps-data/m
aps/reference.html.

Table A.6. Summary Statistics for Chapter Three

Incumbent Concentration of
Votes
Incumbent State-Level Vote
Shares
Incumbent Party Cohesion
South
In-Party Governor
Percentage of Hispanic
People
Percentage of Black People
Rural Population
Immigrant (thousand people)
Democrat
House Seat Share
Share of (State-Own) House
Seats
Redistricting
Senate Seats
State Log Median Income
State Unemployment Rate
Previous Vote Shares

Mean

Standard
Deviation

Minimum

Maximum

3.53

3.08

9.80e-07

13.44

47.03

8.89

24.75

71.61

84.36

4.15

78.62

90.07

0
0

1
1

.26
.40
9.06

9.67

0

1

10.97
25.74
13.32
.5

9.47
13.89
29.91

0
5.05
.19
0

1
61.34
231.74
1

.17

.52

.99

26.04

0

100

.16
1.72
8.43

0
0
10.39
2.2
24.08

1
1
11.19
11.3
71.85

45.14
.22
.66
10.86
5.71
49.56

128

REFERENCES

129

REFERENCES

Abramowitz, Alan I., and Kyle L. Saunders. 1998. ‗‗Ideological Realignment in the U.S.
Electorate.‘‘ Journal of Politics 60 (3): 634–52.
Abramowitz, Alan I., and Kyle L. Saunders. 2008. ‗‗Is Polarization a Myth?‘‘ Journal of Politics
70 (2): 542–55.
Acemoglu, Daron, and James Robinson. 2006. Economic Origins of Dictatorship and
Democracy. Cambridge, NY: Cambridge University Press.
Adams, James F., Samuel Merrill Ⅲ, and Bernard Grofman. 2005. A Unified Theory of Party
Competition: A Cross-National Analysis Integrating Spatial and Behavioral Factors. New
York, NY: Cambridge University Press.
Alderson, Arthur S., and François Nielsen. 2002. ―Globalization and the Great U-Turn: Income
Inequality Trends in 16 OECD Countries.‖ The American Journal of Sociology 107 (5):
1244-99.
Aldrich, John H. 1983. "A Downsian Spatial Model with Party Activism." American Political
Science Review 77: 974-90.
Aldrich, John. 1995. Why Parties? The Origin and Transformation of Political Parties in
America. Chicago, IL: The University of Chicago Press.
Aldrich, John H., and David W. Rohde. 2000a. "The Consequences of Party Organization in the
House: The Role of the Majority and Minority Parties in Conditional Party Government" In
Polarized Politics: Congress and the President in a Partisan Era, eds. Jon R. Bond and
Richard Fleisher. Washington, DC: CQ Press, 31-72.
Aldrich, John H., and David W. Rohde. 2000b. "The Republican Revolution and the House
Appropriation Committee." The Journal of Politics 62 (1): 1-33.
Aldrich, John. H., and David W Rohde. 2001. ―The logic of conditional party government.‖ In
Congress Reconsidered (7th ed), eds. L. C. Dodd and B. I. Oppenheimer. Washington, DC:
CQ Press, 269-92.
American Political Science Association. 1950. ―Toward a More Responsible Two-Party Systems:
A Report of the Committee on Political Parties.‖ American Political Science Review 44 (3):
Part 2, Supplement.
Anderson, Christopher J. 2007. ―The End of Economic Voting? Contingency Dilemma and the
Limits of Democratic Accountability.‖ Annual Review of Political Science 10: 271-96.
Anderson Christopher J, Silvia M. Mendes, and Yuliya V. Tverdova. 2004. ―Endogenous
economic voting: evidence from the 1997 British election.‖ Electoral Studies. 23: 683–708.
Ansolabehere, Stephen, James M. Snyder, Jr., Charles Stewart III. 2001. ―Candidate Positioning
in U.S. House Elections.‖ American Journal of Political Science 45 (1): 136-59.
Ansolabehere, Stephen, and Philip Edward Jones. 2010. ―Constituents' Responses to
Congressional Roll Call Voting.‖ American Journal of Political Science 54 (3): 583-97.

130

Aranson, Peter H., and Peter C. Ordeshook. 1972. "Spatial Strategies for Sequential Elections."
In Probability Models of Collective Decision Making, edited by R. G. Niemi and H. F.
Weisberg, 298-331. Columbus, OH: Charles E. Merrill.
Arnold, R. Douglas. 1990. The Logic of Congressional Action. New Haven, CT: Yale University
Press.
Ashworth, Scott, and Ethan Bueno de Mesquita. 2008. ―Informative Party Label with
Institutional and Electoral Variation.‖ Journal of Theoretical Politics 20 (3): 251-73.
Axelrod, Robert. 1972. ―Where the Votes Come From: An Analysis of Electoral Coalition,
1952-1968.‖ American Political Science Review 65 (1): 11-20.
Barreto, Matt A., Gary M. Segura, and Nathan D. Woods. 2004. ―The Mobilizing Effect of
Majority-Minority Districts on Latino Turnout.‖ The American Political Science Review 98
(1): 65-75.
Bartel, Larry. 2000. ―Partisanship and Voting Behavior, 1952-1996.‖ American Journal of
Political Science 44 (1): 35-50.
Bartels, Larry. 2004. ―Partisan Politics and the U.S. Income Distribution.‖ Russell Sage
Foundation Working Paper Series.
Bartels, Larry. 2008. Unequal Democracy: The Political Economy of the New Gilded Age. New
York: Russell Sage Foundation, and Princeton, NJ: Princeton University Press.
Batto, Nathan F. 2005. ―Electoral Strategy, Committee Membership, and Rent Seeking in the
Taiwanese Legislature, 1992-2001.‖ Legislative Studies Quarterly 30 (1): 43-62.
Bernhardt, Dan, John Duggan, and Francesco Squintani. 2009. "The Case for Responsible
Parties." American Political Science Review 103 (4): 570-87.
Blanchflower, David G., and Matthew J. Slaughter. 1999. ―The Causes and Consequences of
Changing Income Inequality‖ In Growing Apart: The Causes and Consequences of Global
Wage Inequality, eds. Albert Fishlow and Karen Parker. New York, NY: Council on Foreign
Relations Press, 67-94.
Bloom, Howard S., and H. Douglas Price. 1975. Voter Response to Short-Run Economic
Conditions: The Asymmetric Effect of Prosperity and Recession.‖ American Political
Science Review 69 (4): 1240-54.
Boix, C. 2003. Democracy and Redistribution. Cambridge, NY: Cambridge University Press.
Borre, Ole. 1997. ―Economic Voting in Danish Electoral Surveys.‖ Scandinavian Political
Studies 20 (4): 347-66.
Brady, David W., Hahrie Han, and Jeremy C. Pope. 2007. "Primary Elections and Candidate
Ideology: Out of Step with the Primary Electorate?" Legislative Studies Quarterly
32:79-105.
Brady, David W., Robert D‘Onfrio, and Morris P. Fiorina. 2000. ―The Nationalization of
Electoral Forces Revisted.‖ In Continuity and Change in House Elections, edited by David
W. Brady, John F. Cogan, and Morris P. Fiorina. Standford, CA: Standford University Press.

131

Brewer, Mark D. 2005. ―The Rise of Partisanship and the Expansion of Partisan Conflict within
the American Electorate.‖ Political Research Quarterly 58 (2): 219-29.
Brody, Richard A. 1991. Assessing the President: The Media, Elite Opinion, and Public Support.
Stanford: Stanford University Press.
Bullock, John G. 2011. Elite Influence on Public Opinion in an Informed Electorate. American
Political Science Review 105 (3): 496-515.
Bureau of
Labor
Statistics, United States
Department
http://www.bls.gov/cps/cpsaat01.htm (October 22, 2013).

of

Labor.

2012,

Canes-Wrone, Brandice, David W. Brady, and John F. Cogan. 2002. ―Out of Step, Out of Office:
Electoral Accountability and House Members‘ Voting.‖ American Political Science Review
96: 127–40.
Cantor, David M., and Paul S. Herrnson. 1997. ―Party Campaign Activity and Party Unity in the
U. S. House of Representatives.‖ Legislative Studies Quarterly 22 (3): 393-415.
Caramani, Daniele. 2004. The Nationalization of Politics. New York, NY: Cambridge University
Press.
Carey, John M., and Matthew Shugart. 1995. ―Incentives to Cultivate a Personal Vote: A Rank
Ordering of Electoral Systems.‖ Electoral Studies 14: 417-39.
Carey, John M. 2007. ―Competing Principals, Political Institutions, and Party Unity in
Legislative Voting.‖ American Journal of Political Science 51 (1): 92-107.
Carey, John M. 2009. Legislative Voting and Accountability. New York, NY: Cambridge
University Press.
Carroll, Royce, and Henry Kim. 2010. ―Party Government and the Cohesive Powerof Public
Plunder.‖ American Journal of Political Science 54 (1): 34-44.
Carson, Jamie L. 2005. ―Selection, and Candidate Competition in U.S. House and Senate
Elections.‖ The Journal of Politics 67 (1): 1-28.
Carson, Jamie, Michael H. Crespin, Charles J. Finocchiaro, and David W. Rohde. 2007.
"Redistricting and Party Polarization in the U.S. House of Representatives." American
Politics Research 35(6): 878-904.
Carson, Jamie L., Gregory Koger, Matthew J. Lebo, and Everett Young. 2010. ―The Electoral
Costs of Party Loyalty in Congress.‖ American Journal of Political Science 54 (3):
598-616.
Charles-Coll, Jorge A. 2011. ―Understanding Income Inequality: Concept, Causes and
Measurement.‖ International Journal of Economics and Management Sciences 1 (3): 17-28.
Cheibub, Jose Antonio, and Fernando Limongi. 2002. ―Democratic Institutions and Regime
Survival: Paliamentary and Presidential Democracies Reconsidered.‖ Annual Review of
Political Science 5: 151-79.
132

Coleman, James S. 1971. "Internal Processes Governing Party Positions in Elections." Public
Choice 11: 35-60.
Cox, Gary. 1987. The Efficient Secret: The Cabinet and the Development of Political Parties in
Victorian England. Cambridge, NY: Cambridge University Press.
Cox, Gary W., and Mathew D. McCubbins .1993. Legislative Leviathan: Party Government in
the House. Berkley, CA: University of California Press.
Cox, Gary W., and Mathew D. McCubbins. 2005. Setting the Agenda: Responsible Party
Government in the U.S .House of Representatives. Cambridge, NY: Cambridge University
Press.
Crespin, Michael, David W. Rohde, and Ryan J. Vander Wielen. 2011. ―Measuring Variations in
Party Unity Voting: An Assessment of Agenda Effects.‖ Party Politics (July): 1-26.
Desposato, Scott W. 2005. ―Correcting for Small Group Inflation of Roll-Call Cohesion Scores.‖
British Journal of Political Science 35(4): 731–44.
Duch, Raymond M., and Randolph T. Stevenson. 2008. The Economic Vote: How Political and
Economic Institutions Condition Election Results. New York: Cambridge University Press.
Erikson, Robert S. 1988. ―The Puzzle of Midterm Loss.‖ The Journal of Politics 50 (4): 1011-29.
Evans, Geoffrey, and Robert Andersen. 2006. ―The Political Conditioning of Economic
Perceptions.‖ Journal of Politics 68 (1): 194-207.
Fenno. 1978. Home style: House members in their districts. Boston, MA: Little Brown.
Ferejohn, John. 2007. ―Accountability in Global Context.‖ International Law and Justice
Working
Papers:
Global
Administrative
Law
Series.
http://iilj.org/publications/documents/2007-5.GAL.Ferejohn.web.pdf (July 15, 2014).
Fiorina, Morris P. 1977. Congress: Keystone of the Washington Establishment. New Haven: Yale
University Press.
Fiorina, Morris P. 1978. ―Economic Retrospective Voting in American National Elections: A
Micro-Analysis.‖ American Journal of Political Science 22: 426-43.
Fiorina, Morris P. 1981. Retrospective Voting in American National Elections. New Haven: Yale
University Press.
Fiorina, Morris P., Samuel J. Abrams and Jeremy C. Pope. 2006. Culture War? The Myth of a
Polarized America. 2nd ed. New York, NY: Pearson Longman.
Frank, Mark W. 2009. ―Inequality And Growth In The United States: Evidence From A New
State-Level Panel Of Income Inequality Measures.‖ Economic Inquiry 47 (1): 55-68.
Franzese, Robert, Karen Long Jusko, and Irfan Nooruddin. 2007. ―Party Unity and The Effective
Constituency in Distributive Politics.‖ Working Paper.
133

Frieden, Jeffry A., David A. Lake, and Kenneth A. Schultz. 2010. World Politics: Interests,
Interactions, Institutions. New York: W. W. Norton & Company.
Garand, James C. 2010. ―Income Inequality, Party Polarization, and Roll-Call Voting in the U.S.
Senate‖ The Journal of Politics 72 (4): 1009-28.
Gaventa, John. 1980. Power and Powerlessness: Quiescence and Rebellion in an Appalachian
Valley. Urbana, IL: University of Illinois Press.
Gelman, Andrew. 2008. Red State, Blue State, Rich State, Poor State: Why Americans Vote the
Way They Do? Princeton, NJ: Princeton University Press.
Gerring, John, Strom C. Thacker, and Carola Moreno. 2009. ―Are Parlimentary Systems Better?‖
Comparative Politics 42 (3): 329-59.
Gimpel, James G., and Jason E. Schuknecht 2002. ―Reconsidering Political Regionalism in the
American States.‖ State Politics & Policy Quarterly 2 (4):325-52.
Grynaviski, Jeffrey D. 2010.Partisan Bonds: Political Reputation and Legislative Accountability.
Cambridge, NY: Cambridge University Press.
Hansford, Thomas G., and Brad T. Gomez. 2011. ―Reevaluating the Sociotropic Economic
Voting Hypothesis.‖ paper presented at Annual Meetings of the American Political Science
Association, Seattle.
Harbridge, Laurel, and Neil Malhotra. 2011. ―Electoral Incentives and Partisan Conflict in
Congress: Evidence from Survey Experiments.‖ American Journal of Political Science 53
(3): 494-510.
Hellwig, Timothy, and David Samuels. 2008. ―Electoral Accountability and the Variety of
Democratic Regimes.‖ British Journal of Political Science 38 (1): 65-90.
Hetherington, Marc J. 2001. ―Resurgent Mass Partisanship: The Role of Elite Polarization.‖
American Political Science Review 95 (3): 619-31.
Hibbs, Douglas A. Jr. 2000. ―Bread and Peace Voting in U.S. Presidential Elections.‖ Public
Choice 104 (2):149-80.
Hicken, Allen, and Joel Simmons. 2008. ―The Personal Vote and the Efficacy of Education
Spending,‖ American Journal of Political Science 52 (1): 109-24.
Jacobson, Gary C. 2004. The politics of Congressional elections. 6th edition. New York, NY:
Longman.
Jacobson, Gary C., and Gregory L. Bovitz. 1998. ―The Electoral Politics of Budgets and Deficits,
1980-1996.‖ Paper Presented at the Annual Meeting of American Political Science
Association, Boston, MA, 1998.
Jones, Mark, and Scott Mainwaring. 2003. "The Nationalization of Parties and Party Systems:
An Empirical Measure and an Application to the Americas" Party Politics 9 (2): 139-66.

134

Kam, Cristopher J. 2009. Party Discipline and Parliamentary Politics. Cambridge, NY:
Cambridge University Press.
Kayser, Mark M., and Christopher Wlezien. 2011. ―Performance Pressure: Patterns of
Partisanship and the Economic Vote.‖ European Journal of Political Research 50 (3):
365-94.
Keefer, Philip, and Stuti Khemani. 2009. ―When Do Legislators Pass on Pork? The Role of
Political Parties in Determining Legislator Effort,‖ American Political Science Review 103
(1): 99-112.
Kiewiet, D. Roderick, 1983. Macroeconomics and Micropolitics: The Electoral Effects of
Economic Issues. Chicago: University of Chicago Press.
Kitschelt, Herbert. 2007. ―The Demise of Clientelism in Affluent Capitalist Democracies.‖ In
Patrons, Clients, and Policies: Patterns of Democratic Accountability and Political
Competition, edited by Herbert Kitschelt and Steven Wilkinson. Cambridge, NY:
Cambridge University Press.
Kitschelt, Herbert and Steven Wilkinson (eds). 2007. Patrons, Clients, and Policies: Patterns of
Democratic Accountability and Political Competition. Cambridge, NY: Cambridge
University Press.
Krehbiel, Keith. 1998. Pivotal Politics: A Theory of U.S. Lawmaking. Chicago, IL: University of
Chicago Press.
Krehbiel, Keith. 2000. ―Party Discipline and Measures of Partisanship." American Journal of
Political Science 44 (2): 212-27.
Krueger, Alan B. 2012. ―The Rise and Consequences of Inequality in the United States,‖ Speech
Delivered at Center for American Progress, January 12, 2012.
Ladewig, Jeffery W. 2005. ―Conditional Party Government and the Homogeneity of Constituent
Interests‖ The Journal of Politics 67 (4): 1006-29.
Ladewig, Jeffery W. 2010. ―Ideological Polarization and the Vanishing of Marginals‖ The
Journal of Politics 72 (2): 499-512.
Lampman, Robert J. 1973. ―Measured Inequality of Income: What Does It Mean and What Can
It Tell Us?‖ Annals of the American Academy of Political and Social Science 409 (Income
Inequality): 81-91.
Lebo, Matthew J., Adam J. McGlynn, and Gregory Koger. 2007. ―Strategic Party Government:
Party Influence in Congress, 1789–2000.‖ American Journal of Political Science 51 (3):
464-81.
Levendusky, Matthew S. 2009. ―Clear Cues, More Consistent Voters: A Benefit of Elite
Polarization.‖ Political Behavior 32 (1): 111-31.
Levendusky, Matthew S. 2009. The Partisan Sort: How Liberals Became Democrats and
Conservatives became Republican. Chicago, IL: University of Chicago Press.
135

Leyden, Kevin M., and Stephen A. Borrelli. 1990. ―Party Contributions and Party Unity: Can
Loyalty Be Bought?‖ The Western Political Quarterly 43 (2):343-65.
Lewis-Beck, Michael S., and Martin Paldam. 2000. ―Economic Voting: An Introduction.‖
Electoral Studies 19 (2): 113-21.
Lewis-Beck, Michael S., Richard Nadeau, and Angelo Elias. 2008. ―Economics, Party, and the
Vote: Causality Issues and Panel Data.‖ American Journal of Political Science 52 (1):
84-95.
Lyne, Mona M. 2007. ―Rethinking Economics and Institutions: The Voter‘s Dilemma and
Democratic Accountability.‖ In Patrons, Clients, and Policies: Patterns of Democratic
Accountability and Political Competition, eds. Herbert Kitschelt and Steven Wilkinson.
Cambridge: Cambridge University Press, 159-81.
Lyne, Mona M. 2008. The Voter’s Dilemma and Democratic Accountability: Latin America and
Beyond. University Park, PA: The Pennsylvania State University Press.
Manin, Bernard, Adam Przeworski, and Susan Stokes. 1999. ―Elections and representation.‖ In
Democracy, Accountability, and Representation, ed. Przeworski, Adam, Susan Stokes, and
Bernard Manin. Cambridge, NY: Cambridge University Press.
Mayhew, David. 1974. Congress: The Electoral Connection. New Haven, CT: Yale University
Press.
McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 2006. Polarized America: The Dance of
Ideology and Unequal Riches. Cambridge, MA: The MIT Press.
McGillivray, Fiona. 1997. ―Party Discipline as a Determinant of the Endogenous Formation of
Tariffs.‖ American Journal of Political Science 41 (2): 584-607.
McGillivray, Fiona. 2004. Privileging Industry: The Comparative Politics of Trade and
Industrial Policy. Princeton, NJ: Princeton University Press.
Mueller, John E. 1970. ―Presidential Popularity from Truman to Johnson.‖ The American
Political Science Review 64 (1): 18-34.
Müller, Wolfgang C., and Kaare Strøm. 1999. Policy, Office, or Votes? How Political Parties in
Western Europe Make Hard Decisions. Cambridge, NY: Cambridge University Press.
Nadeau, Richard, and Michael Lewis-Beck. 2001. ―National Economic Voting in U.S.
Presidential Elections,‖ Journal of Politics 63: 159-81.
Nadeau, Richard, Richard G. Niemi, and Antoine Yoshinaka. 2002. ―A cross-national analysis of
economic voting: Taking account of the political context across time and nations.‖ Electoral
Studies 21: 403–23.
Nannestad, Peter, and Martin Paldam. 1994. ―The VP-function: A Survey of the Literature on
Vote and Popularity Functions After 25 Years.‖ Public Choice 79: 213-45.
NationMaster.com. 2012. ―Economy Statistics/GDP per capita, PPP/current international $ (most
136

recent)
by
country.‖
http://www.nationmaster.com
/graph/eco_gdp_per_cap_ppp_cur_int-per-capita-ppp-current-international&date (October
22, 2013).
Neckerman, Kathryn M., and Florencia Torche. 2007. ―Inequality: Causes and Consequences.‖
Annual Review of Sociology 33: 335-57.
Niemi, Richard G., and Alan I. Abramowitz. 1994. ―Partisan Redistricting and the 1992
Congressional Elections.‖ The Journal of Politics 56 (3): 811-14.
Nepal, Mani, Alok K. Bohara, and Kishore Gawande. 2011. ―More Inequality, More Killings:
The Maoist Insurgency in Nepal.‖ American Journal of Political Science 55 (4): 886-906.
Nicholson, Stephen P., Gary M. Segura, and Nathan D. Woods. 2002. ―Presidential Approval and
the Mixed Blessing of Divided Government.‖ Journal of Politics 64: 701-20.
OECD. 2013. ―Decile Ratios of Gross Earnings.‖ http://stats.oecd.org/ Index.aspx?
DatasetCode=DEC_I (October 22, 2013).
OECD iLibrary. 2013. ―Factbook 2011-2012: Economic, Environmental and Social
Statistics-Income Inequality.‖ http://www.oecd-ilibrary.org/sites/factbook-2011-en/03/05/01
/index.html?itemId=/content/chapter/factbook-2011-31-en (October 22, 2013).
Office of The Clerk, U.S. House of Representatives. 2012. http://clerk.house.gov/member_info/
electionInfo/index.aspx. (October 22, 2013).
Page, Benjamin I., Larry M. Bartels, and Jason Seawright. 2013. ―Democracy and the Policy
Preferences of Wealthy Americans.‖ Perspectives on Politics 11 (1): 51-73.
Pateman, Carole. 1971. "Political Culture, Political Structure and Political Change." British
Journal of Political Science 1 (3): 291-305.
Pattie, Charles, and Ron Johnston. 2001. ―Routes to party choice: ideology, economic
evaluations and voting at the 1997 British general election.‖ European Journal Political
Research 39: 373–89.
Powell, G. Bingham, and Guy D. Whitten. 1993. ―A cross-national analysis of economic voting:
Taking account of the political context.‖ American Journal of Political Science 37:
391–414.
Primo, David M., Matthew L. Jacobsmeier, and Jeffrey Milyo. 2007. ―Estimating the Impacts of
State Policies and Institutions with Mixed-Level Data.‖ State Politics and Policy Quarterly
7 (4): 446-59.
Rae, Nicol. 2007. "Be Careful What You Wish for: The Rise of Responsible Parties in American
National Politics." Annual Review of Political Science 10:169-91.
Ranney, Austin. 1962. The Doctrine of Responsible Party Government:
Present State. Urbana, IL: University of Illinois Press.

137

It’s Origins and

Riker, William H. 1962. The Theory of Political Coalitions. New Haven, CT: Yale University
Press.
Rodden, Jonathan. 2010. ―The Geographic Distribution of Political Preferences.‖ Annual Review
of Political Science 13: 321-40.
Roff, Peter. 2012. ―Obama's Economy and Hilary Rosen Lead the War on Women‖ U.S. News,
April 12th. http://www.usnews.com/opinion/blogs/peter-roff/2012/04/12/obamas-economyand-hilary-rosen-lead-the-war-on-women (July 27, 2014).
Rohde, David W. 1991. Parties and Leaders in the Post-reform House. Chicago, IL: University
of Chicago Press.
Royed, Terry J., Kevin M. Leyden, and Stephan A. Borrelli. 2000. ―Is Clarity of Responsibility
important for Economic Voting? Revisiting Powell and Whitten‘s Hypothesis.‖ British
Journal of Political Science 30 (4): 669-98.
Rudolph, Thomas J. 2006. ―Triangulating Political Responsibility: The Motivated Formation of
Responsibility Judgments.‖ Political Psychology 27 (1): 99-122.
Sanders, David and Sean Carey. 2002. ―Temporal Variations in Economic Voting: A Comparative
Cross-national Analysis.‖ In Economic Voting, eds. Han Dorussen and Michaell Taylor.
London: Routledge.
Schattschneider, E. E. 1960. The Semisovereign People: A Realist's View of Democracy in
America. New York: Holt, Reinhart, and Winston.
Sheng, Shing-Yuan. 2008. ―Party Leadership and Cohesion in the Legislative Yuan: Before and
After the First Party Turnover in the Executive Branch‖ Taiwan Journal of Democracy 5 (4):
1-46.
Shepsle, Kenneth A. and Barry R. Weingast. 1995. Positive Theories of Congressional
Institutions. Ann Arbor, MI: University of Michigan Press.
Shugart, Matthew S., and John M. Carey. 1992. Presidents and Assemblies: Constitutional
Design and Electoral Dynamics. New York, NY: Cambridge University Press.
Sigelman, Lee, Carol K. Sigelman, David Bullock. 1991. "Reconsidering pocketbook voting: An
experimental approach." Political Behavior 13 (2): 129-49.
Sinclair, Barbara. 2006. Party Wars: Polarization and the Politics of National Policy Making.
Norman: University of Oklahoma Press.
Smith, Steven S. 2007. Party Influence in Congress. Cambridge, NY: Cambridge University
Press.
Snyder, James M. and Michael M. Ting. 2002. ―An Informational Rationale for Political Parties‖
American Journal of Political Science 46 (1): 90–110.
Solt, Frederick. 2008. ―Economic Inequality and Democratic Political Engagement.‖ American
Journal of Political Science 52: 48–60.
138

Solt. Frederick. 2010. ―Does Economic Inequality Depress Electoral Participation? Testing the
Schattschneider Hypothesis.‖ Political Behavior 32: 285–301.
Stonecash, Jeffrey M. 2006. ―The Income Gap.‖ PS: Political Science and Politics July: 461-5.
Stonecash, Jeffrey M., Mark D. Brewer, and Mack D. Mariani. 2003. Diverging Parties: Social
Changes, Realignment, and Party Polarization. Boulder, CO: Westview Press.
Swing State Project. 2010. http://www.swingstateproject.com/.
The World Bank. 2012. ―World Development Indicators.‖ http://databank.worldbank.org/data/
views/variableSelection/selectvariables.aspx?source=world-development-indicators#s_g
(October 22, 2013).
Theriault, Sean M. 2008. Party Polarization in Congress. New York, NY: Cambridge University
Press.
Tuftle, Edward R. 1978. Political Control of the Economy. Princeton: Princeton University Press.
Voteview.com. 2012. http://voteview.com/.
Weede, Eric. 1981. ‖Income Inequality, Average Income, and Domestic Violence.‖ The Journal
of Conflict Resolution 25 (4): 639-54.
Zaller, John R. 1992. The Nature and Origins of Mass Opinion. Cambridge: Cambridge
University Press.

139