A gw
.. z ... m

Ev. Lx
0.9.1.?“

I.

.. 22p.

 

nu... mwﬂa

5! 15.x.» .n-aa

‘11:.‘1‘ it.

3 ﬂag“...
5.1.: 2: ..

«In!
«nix...
.15. ﬁnal... .
. .2!
.1‘ .
.(flx‘r

.H

”z... .. ...X..W:..E.uu..~, . 7.51.5. .n‘

 

 

LIBRARY
Michigan State 3
University

This is to certify that the

dissertation entitled

The Electoral Consequences of Position Taking
in Congress: The Impact of Roll Call Behavior
on Electoral Fortunes

presented by

Jamie L. Carson

has been accepted towards fulﬁllment
of the requirements for

 

doctorate degreein Political Science

 

ﬁmﬂ 6d. W

Major professor

Date May 12, 2003

 

MS U is an Afﬁrmative Action/Equal Opportunity Institution 0-12771

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

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/ClRC/DateDue.p65-p. 15

THE ELECTORAL CONSEQUENCES OF POSITION TAKING IN CONGRESS:
THE IMPACT OF ROLL CALL BEHAVIOR ON ELECTORAL FORTUNES

By

Jamie L. Carson

A DISSERTATION
Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY

Department of Political Science

2003

ABSTRACT

THE ELECTORAL CONSEQUENCES OF POSITION TAKING IN CONGRESS:
THE IMPACT OF ROLL CALL BEHAVIOR ON ELECTORAL FORTUNES

By

Jamie L. Carson

How does the voting behavior of representatives in Congress affect their electoral
fortunes? While much has been written to date on congressional elections, we lack a
comprehensive study that predicts when, and under what conditions, positions taken on
roll call votes produce electoral consequences. This dissertation seeks to ﬁll this void in
the congressional elections literature. In Chapter 1, I provide a brief introduction to the
central research questions employed in this dissertation. In Chapters 2 and 3, I analyze
key roll call votes from both the House and the Senate since the early 19703 to assess
how and when legislators are rewarded or punished at the polls for their position taking
behavior on roll call votes. I ﬁnd that a wide-range of roll call decisions yield electoral
consequences at the margins for members of Congress. I also show that the extent to
which a given roll call is controversial, salient, and a catalyst for intra-party disagreement
affects whether it has electoral implications. In sum, my analysis explains both when and
why legislators are rewarded or punished at the polls for their position taking behavior in
Congress.

In Chapter 4, I test an assumption that is implicit in much of the work on US.
congressional elections—that strategic politicians regularly monitor the roll call voting
behavior of members of Congress. In the context of candidate competition in House and

Senate elections, I employ a strategic choice analysis to examine the incidence of quality

challenger emergence and incumbent retirements for the 1976-2000 election cycles.
Most notably, I ﬁnd that experienced candidates are more likely to run against legislators
who regularly vote with their party on key votes. I also ﬁnd that the amount of money
raised and spent by the challenger does not always have a deterrent effect on challenger
entry. In terms of incumbent career decisions, I ﬁnd that the emergence of a quality
candidate can deter an incumbent from seeking another term of ofﬁce, irrespective of
their political position in their chamber. In the concluding chapter, I summarize the
central ﬁndings of the dissertation and raise potential issues for consideration in future

research.

Cepyright by
JAMIE 1L. CARSON
2003

For my Mother and Father

ACKNOWLEDGMENTS

It is hard to identify exactly when and where the idea underlying this dissertation
originated, but it was strongly inﬂuenced by a number of different people. Most notably,
I owe a tremendous debt of gratitude to David Rohde, my mentor and major professor at
Michigan State University. Over the past few years, Dave has played a substantial role in
my development and training as a political scientist and has been nothing short of
supportive in terms of my endless questions and varied research interests. As a fellow in
the Political Institutions and Public Choice Program under Dave’s direction and
leadership, I beneﬁted tremendously from the conducive learning environment, hands-on
training with congressional data and excellent resources, and interaction with other highly
motivated graduate students. Most importantly, Dave has set a standard as a political
science scholar that I will always try to live up to throughout my academic career.

I am also especially thankful to Jeffery Jenkins and Greg Bovitz, both of whom
played an important role in my development as a political scientist. Beyond his valuable
comments and suggestions with respect to this dissertation, Jeff has stimulated my
interest in historical work on Congress, elections, and political parties. Through the
various research projects we have worked on together, Jeff has inﬂuenced my way of
thinking about political science, particularly in terms of the emphasis he places on theory
development. Greg has been a valuable colleague and ﬁiend, who largely stimulated my
thinking about many of the ideas presented in this dissertation during his year as a post-
doc in the PIPC program. Although I do not get the chance to talk with Greg as much
these days (especially about college basketball), I would be remiss not to mention how

inﬂuential he was in shaping my research interests.

vi

Beyond those already mentioned, I owe thanks to Bill Reed and Reggie Sheehan,
both members of my dissertation committee who offered a number of helpful comments
and suggestions as I was writing my dissertation. I would also like to acknowledge a
number of professors within the department who have inﬂuenced my development as a
political scientist, including Paul Abramson, Darren Davis, Ada Finifter, Scott Gates, Jim
Granato, Tom Hammond, Mark Jones, Burt Monroe, Brian Silver, and Ken Williams.
On the whole, my experience as a graduate student at Michigan State University was a
very rewarding experience, one that I will always look back upon with fond memories.

In terms of classmates, I am especially thankful to have attended graduate school
with Charles Finocchiaro and Mark Souva, two of the best ﬁiends and colleagues that a
person could ever have. During our time in graduate school, we have had a number of
lively debates and discussions, and have freely shared our ideas about life, politics, and
religion. I would also like to thank Mark Hurwitz, Bryan Marshall and Brandon Prins
who set a solid example for the “second” generation of PIPC students, as well as Kirk
Randazzo and Mike Crespin who have become good friends and colleagues in the past
few years.

Finally, I would like to give a special thanks to my Mom, Dad, and sister Tammy
who have been especially supportive of my interest in graduate school. It is because of

their encouragement that I am dedicating this dissertation to my Mom and Dad.

vii

TABLE OF CONTENTS

LIST OF TABLES ........................................... x-xi
LIST OF FIGURES ............................................................................................................... xii
CHAPTER 1
INTRODUCTION ................................................................................................................... 1
CHAPTER 2
ELECTORAL ACCOUNTABILITY AND ROLL CALL VOTING IN THE US. HOUSE ................. 10
Literature Review: Deﬁning the Scope of Current Evidence ................................. 11
When and Why Do Roll Calls Matter? ................................................................... 19
Data and Methods ................................................................................................... 24
Results ..................................................................................................................... 27
Explaining Variation in the Electoral Signiﬁcance of Key Roll Calls ................... 38
Understanding Directionality of Roll Call Signiﬁcance ......................................... 45
Conclusion .............................................................................................................. 47
CHAPTER 3
ELECTORAL ACCOUNTABILITY AND ROLL CALL VOTING IN THE US. SENATE ................ 49
Prior Findings .......................................................................................................... 52
Extending the Analysis to the US. Senate ............................................................. 54
Data and Methods ............................ 61
Exploring the Frequency of Electorally Consequential Roll Calls
in the Senate ....................................................................................................... 63
Explaining Variation in the Electoral Signiﬁcance Across Key
Senate Roll Calls ................................................................................................ 80
Conclusion .............................................................................................................. 86
Chapter 4
THE IMPACT OF LEGISLATIVE BEHAVIOR ON CANDIDATE COMPETITION
IN HOUSE AND SENATE ELECTIONS: A STRATEGIC CHOICE ANALYSIS ............................ 89
Theories of Competition and Candidate Emergence in
Congressional Elections ..................................................................................... 92
A Uniﬁed Strategic Interaction Model ................................................................. 100
Data, Hypotheses, and Methods ............................................................................ 108
Empirical Results .................................................................................................. 1 13
US. Senate Elections ....................................................................................... 120
Comparing Results From the Strategic Choice Model to a
Standard Probit Model ...................................................................................... 125
Conclusion ............................................................................................................ 128

viii

Chapter 5

CONCLUSION ................................................................................................................... 131
Summary of Findings ............................................................................................ 133
Extensions and Future Directions for this Research ............................................. 137

RFERENCES ........................................... 139

ix

Table 1:

Table 2:

Table 3:

Table 4:

Table 5:

Table 6:

Table 7:

Table 8:

Table 9:

Table 10:

Table 1 1:

Table 12:

Table 13:

Table 14:

LIST OF TABLES

Percentage of CQ House Key Votes With Signiﬁcant

Electoral Impact, 93'd-106th Congresses ........................................................ 28
Key Votes in the House Yielding Signiﬁcant Electoral

Effects on Incumbent Margins .................................................................. 29-36
Probit Model of Electorally Signiﬁcant House Roll Call Votes .................... 42

Percentage of CO Senate Key Votes With Signiﬁcant
Electoral Impact, 93'rd-106th Congresses ........................................................ 64

Key Votes Yielding Signiﬁcant Electoral Effects in 1974
Senate Elections ............................................................................................. 66

Key Votes Yielding Signiﬁcant Electoral Effects in 1976
Senate Elections ........................................................................................ 66-67

Key Votes Yielding Signiﬁcant Electoral Effects in 1978
Senate Elections ............................................................................................. 67

Key Votes Yielding Signiﬁcant Electoral Effects in 1980
Senate Elections ........................................................................................ 67-68

Key Votes Yielding Signiﬁcant Electoral Effects in 1982
Senate Elections ............................................................................................. 68

Key Votes Yielding Signiﬁcant Electoral Effects in 1984
Senate Elections ............................................................................................. 69

Key Votes Yielding Signiﬁcant Electoral Effects in 1986
Senate Elections ........................................................................................ 69-70

Key Votes Yielding Signiﬁcant Electoral Effects in 1988
Senate Elections ............................................................................................. 71

Key Votes Yielding Signiﬁcant Electoral Effects in 1990
Senate Elections ............................................................................................. 72

Key Votes Yielding Signiﬁcant Electoral Effects in 1992
Senate Elections ............................................................................................. 73

Table 15:

Table 16:

Table 17:

Table 18:

Table 19:

Table 20:

Table 21:

Table 22:

Key Votes Yielding Signiﬁcant Electoral Effects in 1994
Senate Elections ........................................................................................ 73-74

Key Votes Yielding Signiﬁcant Electoral Effects in 1996
Senate Elections ............................................. 74

Key Votes Yielding Signiﬁcant Electoral Effects in 1998
Senate Elections ............................................................................................. 75

Key Votes Yielding Signiﬁcant Electoral Effects in 2000
Senate Elections ............................................................................................. 76

Probit Model of Electorally Signiﬁcant Roll Call Votes in
1974-2000 Senate Elections ........................................................................... 83

Strategic Choice Model of Competition in House Elections,
1976-2000 .................................................................................................... 115

Strategic Choice Model of Competition in Senate Elections,
1976-2000 .................................................................................................... 121

Probit Model of Competition in House & Senate Elections,
1976-2000 .................................................................................................... 126

xi

LIST OF FIGURES

Figure 1a: Strategic Interaction Between Challengers and Incumbents .................... 102

Figure 1b: Equilibrium Outcome Probabilities .......................................................... 106

xii

CHAPTER 1
INTRODUCTION

During the past 30 years, students of congressional elections have documented a
fairly stable pattern in the context of House and Senate elections: the relative ease with
which incumbents seem to get reelected. Beginning with the work of scholars who ﬁrst
recognized the underlying advantages of incumbency (see, e.g., Erikson 1971a; Mayhew
1974b; F erejohn 1977; Cover 1977; F iorina 1977) and continuing with more recent and
innovative attempts to estimate the extent of that advantage in congressional elections
(see, e.g., Gelman and King 1990; Cox and Katz 1996, 1999; Erikson and Palfrey 1998;
Ansolabehere, Snyder, and Stewart 2000; Brady, Cogan and Fiorina 2000), political
scientists have widely examined both the causes and consequences of this recurring
phenomenon. To some, the incumbency advantage is indicative of an electorate that is,
by and large, content with the performance of its elected representatives. To others, it
raises potential concerns about the issue of democratic accountability, oﬂen resulting in
increased demands for campaign ﬁnance reform or enactment of term limits for
incumbents—issues that have received widespread attention since the early 19903.

On the whole, concern about democratic accountability has sparked an interest
among students of congressional elections to investigate the extent to which legislative
behavior has discernible electoral consequences. While some scholars have focused on
the electoral rewards associated with distributive beneﬁts (see, e.g., Bickers and Stein
1994, 1996; Alvarez and Saving 1997; Sellers 1997), others have examined the potential
advantages linked to casework (Alford and Hibbing 1981; Serra and Moon 1994) and

incumbents’ roll call voting records (Erikson 1971b; Wright 1978; Jacobson 1993).

Though far from conclusive, these studies offer evidence that legislators typically receive
marginal beneﬁts from these and other activities. This, of course, has been a source of
concern among journalists, pundits, and scholars alike who question whether potential
advantages that accrue to incumbents makes it impractical for quality candidates to
mount an effective campaign. One student of congressional elections summed up the
state of affairs in the following statement:

Elections are supposed to be the means by which the public exercises

control over its government. If elections are competitive this system

works well. People are faced with Viable options and make their

choices. But if the deck is somehow stacked so that one candidate is

virtually guaranteed victory, then public accountability is undermined

(Krasno 1994: 5).
Aﬁer all, how can we expect incumbent legislators to be responsive to their constituents
if they have little reason to believe that they will be defeated in the upcoming election?

Clearly, congressional incumbents spend a substantial amount of time engaging in

activities that they anticipate will cultivate a sense of loyalty with their constituents and
improve their chances of reelection. Indeed, Mayhew (1974a) ﬁnds that legislators split
their time between advertising, credit claiming, and position taking to facilitate their goal
of getting reelected. F enno (1978) observes that incumbents cultivate a personal “home
style” with their constituents in an attempt to foster a sense of trust and loyalty among
prospective voters. Fiorina (1989) argues that Congress is responsible for the creation of
a Washington establishment that allows individual legislators to frequently engage in

pork barrel and casework activities, which curries favor with constituents. Incumbents

also raise large sums of money with the expectation that it will enhance their prospects

for reelection by discouraging the emergence of experienced, prospective challengers
(Hersch and McDougall 1994; Box-Steffensmeier 1996).

AS the preceding studies suggest, there is little doubt that legislators believe that
their actions and behavior impact their electoral fortunes. Because legislators are directly
elected by the people, most (if not all) recognize the importance and necessity of being
responsive to their constituents. Given that they also possess a ﬁnite amount of time and
scarce resources, we should not expect them to act fn'volously in terms of the allocation
of these limited resources. Moreover, the extent to which we evaluate an incumbent as
“marginal” or “safe” is largely contingent upon the activities that the legislator engages in
while in ofﬁce. Incumbents who appear “out of touch” with their constituencies will
generally have a much harder time getting reelected than those legislators who attempt to
maintain a connection with their voters. As Mayhew (1974a: 37) contends, “what
characterizes ‘safe’ representatives is not that they are beyond electoral reach, but that
their efforts are very likely to bring them uninterrupted electoral success.”

Despite the perceptions of legislators, students of elections remain surprisingly
agnostic about the degree to which legislative behavior affects incumbents’ electoral
fortunes. While involvement in a scandal could spell disaster for an incumbent’s career
in either the House or Senate (see, e. g., Welch and Hibbing 1997), there is less evidence
supporting the notion that an incumbent’s day-to-day activities will consistently yield
observable effects (either positive or negative) during the subsequent election. Indeed,
reconciling this apparent puzzle between legislators’ beliefs and the ﬁndings of the extant
literature has preoccupied the attention of a number of scholars interested in better

understanding the dynamics of electoral politics (see, e.g., Fiorina 1977; Johannes and

McAdams 1981; Alford and Hibbing 1981; Jacobson 1987; Abramowitz 1991; Cox and
Katz 1996; Bianco, Spence, and Wilkerson 1996; Gaines and Nokken 2001).

In continuing in this same tradition, this dissertation examines an issue that has
received comparatively less attention in the congressional elections literature—electoral
accountability in the context of roll call voting. Students of congressional behavior have
long recognized that legislators act strategically when expressing a public position on the
House or Senate ﬂoor when voting on legislation. More speciﬁcally, scholars agree that
legislators seeking to be reelected express positions that can be considered political
commodities in and of themselves (Mayhew 1974a; Fiorina 1974; Kingdon 1989; Arnold
1990). Nevertheless, these same scholars have noted that it can be difﬁcult at times to
ﬁnd evidence that election results are impacted by legislators’ position taking behavior.

One congressional scholar who has written about this relationship, R. Douglas
Arnold (1990: 37), describes the challenges associated with studying the electoral effects

of position taking behavior in the following illustrative example:

If legislators consulted the scholarly literature on congressional elections,
they might conclude that they need not worry much about either the
positions that they take or the effects that they produce because these are
not the major determinants of electoral outcomes. This would be a
dangerous conclusion. It would be equivalent to concluding that one need
not ﬁle a tax return because the Internal Revenue Service prosecutes only
a few thousand individuals each year for tax evasion. The problem is that
legislators as a group have not offered congressional scholars much
variance to analyze. No legislators have offered to take positions directly
opposite to their electoral interests so that we may measure the full impact
of positions on electoral margins. None have acknowledged selecting
their positions by ﬂipping a coin. Legislators as a group have been quite
uncooperative in providing scholars with the kind of evidence we would
need to assess the potential importance of issues in congressional
elections.

Nevertheless, Arnold maintains that it is a worthwhile venture to continue to investigate
the potential electoral consequences of position taking behavior in Congress.

The preceding discussion raises a number of potentially interesting questions for
students of congressional elections. To what extent does roll call behavior of incumbents
affect their electoral fortunes? Are incumbent legislators routinely held accountable for
the positions they express on roll call votes? Do certain types of votes have a greater
impact on electoral fortunes than others? Are senators more or less insulated electorally
from position taking behavior than their counterparts in the House? Are experienced
challengers attuned to the position taking behavior of incumbent legislators as students of
congressional elections typically assume? The answers to these questions have notable
implications for issues of representation and accountability. Indeed, if representative
democracy is working, then constituents should have the opportunity to reward or punish
incumbents, depending on whether legislative decisions reﬂect their interests.

Extant research examining the linkage between legislative behavior and election
outcomes provides some preliminary evidence suggesting that patterns of behavior on
highly salient roll calls may impact incumbents’ electoral fortunes. To date, however, we
lack a comprehensive study that predicts when, and under what conditions, positions
expressed on roll call votes will yield electoral consequences. This dissertation attempts
to ﬁll this void in the congressional elections literature with a large-scale investigation of
the relationship between roll calls and electoral fortunes. In particular, I examine the
electoral implications of hundreds of roll call votes in both the House and the Senate
from the 19705, 19805, and 19905 and arrive at new insights regarding when, and under

what conditions, position taking affects incumbents’ electoral fortunes.

This dissertation suggests that the positions legislators take on individual roll call
votes can, and do, have observable effects in terms of their electoral fortunes. Generally
speaking, we should expect constituents to evaluate a legislator based partly upon his or
her voting record in Congress. Just as issues are of notable consequence in presidential
contests (Page 1978; F iorina 1981), it should come as no surprise that issues produce at
least marginal, observable effects for incumbent legislators since there is considerable
overlap in the voters who select both sets of candidates. AS Arnold (1990: 38) maintains,
“The more one ﬁnds evidence that issues matter in presidential elections, the more one
must accept the possibility that, under the proper conditions, issues could become equally
important in congressional elections.” In the context of Senate elections, Kahn and
Kenney (1999) conclude that issues often play a prominent role during the campaign due
to the greater competitiveness and intensity associated with Senate elections.

The amount of attention afforded a particular roll call should also be a signiﬁcant
factor in determining whether or not positions taken have observable electoral effects.
When a representative takes a speciﬁc position by casting a vote on a given issue, he or
she is engaging in highly Visible behavior; one cannot avoid going on the record with an
announced roll call vote (Kingdon 1989). Since we should theoretically expect roll call
votes to reach different levels of salience based largely on the nature of the issue under
consideration and the degree of consensus/controversy among the members, the electoral
impact of roll call votes will vary. What remains unclear is the degree to which, and the
conditions under which, roll call votes become electorally salient. Accordingly, this
dissertation seeks to address long-standing questions regarding how and when legislators

are rewarded or punished at the polls for their position taking behavior in Congress.

In departing from the traditional format, this dissertation combines three separate,
but interrelated essays addressing issues of electoral accountability and position taking in
both chambers of Congress. In the ﬁrst essay, I focus on the electoral consequences of
position taking on roll call votes in the context of US. House elections from 1974 to
2000. The analysis includes a much larger number of roll call votes than previous studies
and I examine their effects individually to understand which, if any, yield observable
electoral consequences. In particular, I consider the electoral effects of all Congressional
Quarterly (CQ) “key votes” in the House from the early 19705 to the present. I focus
speciﬁcally on CQ’s key votes in this analysis to gain a better understanding of when
prominent roll call votes that are reasonably well known to both political elites and voters
will have observable electoral effects. As I argue in the dissertation, this subset of votes
provides a useful starting point for the analysis since these votes represent a fairly broad
picture of the diverse agenda dealt with by the House across congressional sessions.

Beyond speaking to the frequency of electorally consequential roll calls, the ﬁrst
essay also attempts to offer insights into the conditions under which position taking
yields observable electoral implications. While the preliminary analysis is useﬁil for
determining which votes are and are not electorally consequential, it does not reveal why
some votes yield discernible electoral effects while others do not. Nevertheless, these
initial ﬁndings play a pivotal role in helping us to begin to identify factors impacting
whether or not positions on key votes will impact incumbents’ vote margins. Using these
results as the dependent variable for the second part of this study, I am able to test a
model where the roll call vote is the unit of analysis. This, in turn, provides a unique

opportunity to examine factors such as the saliency and level of consensus on the vote to

determine if they are Signiﬁcantly related to the likelihood of yielding observable effects
for incumbent legislators.

The second essay also focuses on the issue of electoral accountability and roll call
voting, but does so in the context of Senate elections held since the early 19705. While
the electoral effects of roll call voting in the House of Representatives has been given
limited consideration in the extant literature, virtually no attention has been devoted to
studying the same relationship in the US. Senate. This essay seeks to remedy this
oversight by examining both the frequency of and the conditions under which positions
on key roll calls exert electoral implications for incumbent senators. Given that this
analysis parallels the one performed on the House, I am able to make comparisons and
draw distinctions between the two chambers. Thus, I can Show how and why positions
on key votes impact incumbents in both chambers while also highlighting differences
resulting from electoral, institutional, and political factors unique to each chamber.

The third essay departs from the initial analyses by testing an assumption implicit
in the ﬁrst two essays—that experienced candidates routinely scrutinize position taking
behavior of legislators to determine whether or not they should emerge to challenge the
incumbent. Building upon the work of previous scholars who have explored the issue of
candidate competition in congressional elections, I examine the impact of roll call voting
behavior on candidate emergence from the mid-19705 to the present. Unlike previous
work, however, I control for the effects of both selection bias and strategic interaction
through the use of a strategic choice model developed by Curtis Signorino (2001). This
model allows me to investigate strategic factors inﬂuencing candidate emergence while

simultaneously controlling for events affecting an incumbent’s decision to seek reelection

or retire. From this, I offer new insights on candidate competition that calls into question
many previously accepted conventions regarding electoral politics.

In the concluding section of the dissertation, I draw out the larger implications of
the results reported in each of the three separate essays. While many of the ﬁndings
presented in this dissertation challenge the conventional wisdom concerning position
taking on roll call votes and electoral accountability, other results serve to independently
conﬁrm conclusions that are commonly held in the congressional elections literature.
Overall, I believe that the ﬁndings from each of the three essays add signiﬁcantly to our
understanding of electoral politics, particularly with regard to the electoral impact of
legislative behavior. I hope the results reported in this dissertation will encourage others
to continue to advance our understanding of House and Senate elections with respect to

both institutional and political behavior.

CHAPTER 2

ELECTORAL ACCOUNTABILITY AND ROLL CALL VOTING IN THE U.S. HOUSE

Students of legislative behavior have long recognized that members of the U.S.
Congress are strategic when they take public positions (e.g., when casting roll call votes
on the House or Senate ﬂoor). As legislators pursue reelection, the stances they take on
issues can be political commodities in and of themselves. As Mayhew (1974a: 69-70)
and others have noted, however, it is difﬁcult to ﬁnd systematic evidence of legislators’
“position taking” behavior affecting election results. To what extent does roll call voting
affect congressional elections? The answer to this question has important implications
for issues of representation and accountability. Indeed, if representative democracy is
working, then we should observe constituents rewarding or punishing incumbents based
on whether legislative decisions reﬂect their interests.

Case study research exploring the relationship between speciﬁc important roll
calls and election outcomes offers us some preliminary evidence suggesting that patterns
of behavior on certain key roll calls affect incumbents’ electoral margins. Yet, to date we
are left without a comprehensive study that tells us about the frequency of electorally
consequential roll calls and lends insight into the conditions under which position taking
has electoral implications. This chapter takes a step toward ﬁlling that void through a
large-scale investigation of the relationship between roll calls and election results in the
context of elections to the House of Representatives. More speciﬁcally, I examine the
electoral implications of “key” roll call votes from the 19705, 19805, and 19905 to arrive

at new generalizations regarding how and when position taking affects electoral fortunes.

10

My results demonstrate that, after controlling for the usual factors, a wide range
of roll calls have signiﬁcant effects on incumbents’ electoral margins. I also ﬁnd that the
extent to which a particular roll call is controversial, salient, and a catalyst for intra-party
disagreement affects whether it has observable electoral implications. In sum, my central
ﬁndings begin to address long-standing questions regarding how and when legislators are
rewarded or punished at the polls for their position taking behavior in Congress.

The organization of this chapter is as follows. I begin by reviewing the existing
literature linking position taking behavior with electoral outcome and clarify how the
empirical work presented in this paper is different. Then, I present theoretical evidence
detailing both when and why roll call votes should become electorally consequential.
Next, I formulate and conduct an empirical test of the electoral effects of roll call voting
in the House from the early 19705 to the present. Building on these ﬁndings, I then
examine why some roll calls, and not others, have observable electoral consequences.
Finally, a concluding discussion summarizes the principal implications of the results and

suggests avenues for further research.

Literature Review: Deﬁning the Scope of Current Evidence
Mayhew (1974a) views legislators as single-minded seekers of reelection and, in
turn, draws some general conclusions about expected legislative behavior. According to
Mayhew’s theory, legislators employ some combination of three ﬁrndamental activities in
their attempt to garner electoral support: advertising, credit claiming, and position taking.
While the electoral effects of advertising (see, e.g., Mayhew 1974b; Cover and Brumberg

1982) and credit claiming (see, e.g., Rivers and Fiorina 1989; Stein and Bickers 1994;

ll

Alvarez and Saving 1997; Sellers 1997) have been widely explored in the congressional
literature, substantially less attention has been given to the electoral consequences of
position taking behavior in Congress.

Extant research on the relationship between legislative behavior and election
outcomes offers an interesting puzzle to students of congressional elections. On the one
hand, surveys of individual members consistently Show that legislators believe the
positions they take on roll call votes impact their electoral fortunes (see, e.g., Froman
1963; Clausen 1973; Matthews and Stimson 1975; Kingdon 1989; Sullivan, Shaw,
McAvoy, and Barnum 1993). Yet, some recent literature ﬁnds little to no evidence of
legislative behavior affecting elections, suggesting that legislators may not need to be
very concerned about the potential electoral implications of position taking (see, e. g.,
Bianco, Spence, and Wilkerson 1996; Ansolabehere, Snyder, and Stewart 2001; Gaines
and Nokken 2001).1 If this is indeed the case, why might legislators worry about the
potential electoral effects of their position taking behavior? Mayhew (1974a), Fiorina
(1974), Jacobson (1987), Kingdon (1989), and Arnold (1990) all emphasize that
incumbents are extremely sensitive to the potential electoral implications of their votes,

and as a result, behave strategically when announcing a position on a roll call vote.2

 

1 While not focusing explicitly on roll call voting, Jacobson and Dirnock (1994) and Groseclose and
Krehbiel (1994) present an alternate View on the electoral effects of member behavior in connection with
the House banking scandal, showing that members may face consequences even if they do not expect their
actions to become widely publicized.

2 Representatives who occasionally take roll call positions that are “out of touch” with their constituencies
will not necessarily be removed from ofﬁce. As a result of comfortable electoral margins, incumbents
often have a considerable degree of leeway that allows them to take some controversial positions. In any
case, because so few incumbents are defeated from one election to the next, we cannot expect to ﬁnd any
large-scale evidence of roll call behavior impacting electoral wins V5. losses. When position taking does
underlie winning vs. losing, it presumably involves a “lagged” effect. As electoral margins narrow as a
result of more and more “wrong” positions, experienced challengers are more likely to emerge, which
gradually translates into narrower margins and eventually seat turnover. For a more detailed discussion of
this phenomenon, see Froman (1966).

12

Early research on the effects of position taking questioned the possibility of a
relationship between legislators’ behavior and election outcomes because of voters’
tendencies of apathy and ambivalence toward politics. Stokes and Miller (1962), among
the ﬁrst to explore this question, used survey evidence to Show that voters are indeed
apathetic toward politics and therefore few take the time to gather information regarding
speciﬁc issue positions of congressional candidates. Lacking information about their
representatives, voters are portrayed as incapable of holding incumbents accountable for
decisions in ofﬁce (Stokes and Miller 1962: 545). One implication is that incumbents do
not need to be too concerned with the electoral consequences of position taking since so
few constituents pay attention.

Schoenberger (1969), on the other hand, highlights an important link between
legislative position taking and electoral outcomes in his study of the 1964 congressional
elections. Challenging the established notion that campaign platform positions play a
minOr role in determining electoral outcomes, Schoenberger ﬁnds that those Republican
incumbents who took a favorable public position toward Goldwater did substantially
worse in the 1964 elections than those who did not. Thus, in contrast to Stokes and
Miller, Schoenberger concludes that the relationship between position-taking and election
returns is indeed signiﬁcant.

Since the early 19705, the effects of position taking have generally been analyzed
in terms of ﬂoor roll call voting behavior (and almost exclusively with respect to the

House).3 Moreover, literature examining the relationship between position taking and

 

3 While Mayhew (1974a) describes other activities beyond roll call voting in his discussion of legislative
position taking (e.g., ﬂoor addresses, speeches, television appearances, press releases), I focus solely on the
effects of position taking behavior on House roll calls. In future work, I plan to focus on other aspects of
position taking to determine if they yield related consequences for incumbent legislators.

13

election outcomes generally falls under one of two alternative empirical approaches. The
ﬁrst approach focuses on how aggregate voting records affect electoral outcomes. These
analyses typically examine differences between representatives and their constituents by
either exploring ideological extremeness on all roll calls or distinctiveness as measured
by a limited, subset of roll calls (e.g., those on which ADA scores are based). Generally,
the more legislators’ roll call choices depart from constituent ideology, the more their
vote margins in subsequent elections are expected to diminish.

The second approach focuses on the electoral effects of a single roll call vote or a
particular subset of votes in Congress. These studies explore whether the positions that
legislators express on roll calls signiﬁcantly affect the size of the vote margin by which
they are reelected (or, defeated). Scholars examine whether controversial positions taken
on highly Visible or salient votes in Congress have a signiﬁcant and measurable impact
on the percentage of the vote that incumbent legislators receive.

In connection with the former empirical approach, Erikson (1971b) was among
the ﬁrst to systematically explore the relationship between roll call voting in the House
and electoral returns. The central question of his analysis is whether legislators are
rewarded or punished at the polls on the basis of their aggregate roll call record even
though relatively few constituents regularly monitor their behavior. Using electoral and
survey data from 1952 to 1968, Erikson tests the hypothesis that voting behavior in
congressional elections is inﬂuenced by the degree to which the incumbent’s roll call
behavior reﬂects his or her party’s ideological extreme. He ﬁnds that the evidence is
inconclusive for Democratic legislators; however, the results support the hypothesis that

the degree of conservatism among Republican legislators had a considerable negative

14

effect on their vote margins (i.e., the more conservative a Republican’s voting record
was, the more his vote margin was reduced). Although Erikson cautions that the small
nature of the sample necessitated that his ﬁndings be taken with caution, he suggests that
they are “sufﬁciently provocative” to warrant further analysis (1031).

In his analysis of the 1974 congressional elections, Bumham (1975) identiﬁes a
general pattern in legislative behavior that echoed the ﬁndings of both Schoenberger and
Erikson. Based on aggregate voting patterns, Bumham ﬁnds that electoral losses among
House Republicans in the 1974 elections were greatest among the conservative members
of the party. As had been the case in the 1964 congressional elections, a relationship
emerged between the degree of policy liberalism among Republican incumbents and their
relative success on election day (Bumham, 418). Although Bumham acknowledges that
it is possible only to draw limited inferences from the aggregate analysis, his ﬁndings
suggest that Stokes and Miller may have underestimated the policy attentiveness of
voters, at least during the 1974 elections.

Following up on the work of Erikson and Bumham, Wright (1978) challenges the
argument advanced by scholars such as Stokes and Miller in attempting to determine
whether candidates’ issue positions actually do have an effect on their electoral margins.
Using the 1966 NBC survey of candidates for the House and the SRC national election
survey, Wright examines the electoral impact of policy positions of candidates. He ﬁnds
that, contrary to much of the existing literature, constituents often behave as though they
are attentive to individual candidates and that positions taken on policy issues affect the
manner in which voters make up their minds on election day. Interestingly, Wright

explains that it is not necessary for issues to account for much of the variance in election

15

outcomes Since electorally insecure legislators only require minimal incentives to take
note of the policy interests of their constituents (446). Moreover, he concludes that,
“. . .since the candidate’s issue stance is one of the few factors relevant to his reelection
that is also within his control, the representative is well advised to bring his issue
positions into line with those of his constituency. Not to do so could be the determining
factor in electoral defeat” (45 9).

To examine the impact of issue positions on electoral outcomes, Serra and Moon
(1994) employ a district-level analysis to determine whether constituent service and
relative issue proximity affect vote choice. Prior to their study, Serra and Moon note that
the criteria by which voters reward legislators for casework and issue positions had not
been adequately addressed. To explore this dynamic more systematically, they conduct
telephone interviews with two groups of constituents ﬁom one congressional district to
determine the effects of casework separate from incumbent positions. Their results
suggest that while casework does seem to make a difference in terms of increasing the
legislator’s evaluation and recognition, the effects of the incumbent’s issue positions
cannot be discounted (205). Indeed, Serra and Moon suggest that legislators seem to do
better the more attention they devote to both casework and issue considerations (211).

In a series of collaborative essays, Erikson and Wright (2001) have taken their
respective analyses of voters, candidates, and issues in congressional elections one step
further by focusing on long-term electoral trends. From their analysis of election returns
over time, they argue that not only are incumbents required to be fairly attentive to their
districts, but they must also take into consideration the policy interests of constituents

since failing to do so can result in electoral defeat. Although Erikson and Wright do not

16

look at the effects of roll call voting on electoral returns, they do suggest that legislators
who stray from the party’s ideological norm may be more at risk in upcoming elections
(145).4 Also, Erikson and Wright maintain that incumbents who represent “safe” districts
are more likely to diverge in their roll call behavior than non-safe incumbents (147). This
ﬁnding may suggest that voters are more attentive to certain types of votes in Congress,
especially those that result in legislators taking controversial positions.

In a recent study, Canes-Wrone, Brady, and Cogan (2002) explore the connection
between members’ electoral margins and overall support for their party and demonstrate
that legislators are indeed accountable for their roll call voting behavior. Using data from
the 1956-1996 elections, they ﬁnd that incumbents receive smaller electoral margins on
average the more they vote in support of their party across roll calls. They also show that
this effect is comparable to other determinants of electoral margins including challenger
quality and campaign Spending. Finally, they argue that the distinction between “safe”
and “marginal” members is tenuous at best, since this phenomenon affects all members
equally, regardless of their previous electoral performance.

Unlike studies based on aggregate voting patterns or behavior, other research has
explored the electoral implications of a single vote or certain types of votes in the House.5
Jacobson (1993), for instance, offers a compelling case study of budget deﬁcit reduction
legislation in 1990 that includes evidence of how roll call positions inﬂuence election
outcomes. Jacobson’s fundamental argument is that electoral circumstances strongly

shape the positions taken by members of Congress—and Vice versa. Those legislators in

 

4 In a similar study, Ansolabehere, Snyder, and Stewart (2001) ﬁnd that candidates’ vote share is inversely
related to support for the party’s ideological extreme.

5 For a related study focusing on the potential electoral consequences of abstentions (Shirking) on roll calls,
see Rothenberg and Sanders (2000).

17

a position to absorb the costs associated with supporting the 1990 deﬁcit reduction bill
could vote for it while those members who could suffer at the polls as a result had to vote
against it (Jacobson 1993: 376; for a related discussion, see Mann and Wolﬁnger 1980:
629). The prediction, then, is that House members who supported the deﬁcit reduction
package should not incur any particularly serious electoral damage despite its tax hikes
and cuts in popular spending.6 Indeed, Jacobson’s analysis estimates the impact of roll
call votes regarding deﬁcit reduction on the incumbent’s share of the two-party vote in
the 1990 House elections and shows that supporting Bush’s budget deal only hurt those
incumbents in vulnerable districts.

Following up on the previous study, Jacobson and Bovitz (1998) investigate more
generally the electoral conditions shaping House incumbents’ decisions on budget-related
roll calls as well as the effects these votes had on subsequent election returns. Jacobson
and Bovitz examine 43 key budget-related votes from 1981-1996 to explore the linkage
betWeen electoral politics and budgets. Jacobson and Bovitz conclude, “loyalty to the
party’s positions on difﬁcult budget votes was regularly penalized on election day” (20).
Even though their ﬁndings only speak to the relationship between budget votes and
electoral results, they suggest a broader implication that legislative position taking
regularly has consequences on election day.

While research to date has provided us with some insights regarding the effects of
legislative behavior on election outcomes, in some ways the two approaches described
above can only give us a limited understanding of how roll call voting impacts election

returns. In particular, studies that rely on aggregate measures of legislative behavior are

 

6 Jacobson also ﬁnds that the risk to legislators supporting the bill was reduced since the number of quality
challengers running in the 1990 election was relatively low.

18

unable to distinguish the electoral relevance of one vote from another. Since all or even a
subset of votes are considered together, we do not have the opportunity to determine
which speciﬁc votes, if any, are having a signiﬁcant effect on an incumbent’s margins.

In terms of the second approach, analyses examining the electoral implications of
one or even a few votes in Congress are valuable case studies, but they do not allow for
broader generalizations. In particular, these types of analyses do not tell us whether other
roll call votes will have signiﬁcant electoral effects. Without looking at a broader subset
of roll calls that legislators take positions on, we cannot begin to understand the degree to
which Congress’s diverse legislative agenda impacts members’ electoral margins.

I investigate the extent to which roll call behavior affects the electoral fortunes of
incumbents by attempting to combine the best features of the previous approaches. In
particular, this analysis includes a much larger number of roll calls than prior studies and
I explore their effects individually to understand which, if any, have observable electoral
effects. As a result, I expect to be able to offer more comprehensive ﬁndings that speak
to the frequency of electorally consequential roll calls as well as offer insights into the

conditions under which legislative position taking will yield electoral implications.

When and Why Do Roll Calls Matter?

Because legislators cast literally hundreds of votes in a congressional session, one
should not expect the vast majority of roll calls to have signiﬁcant electoral effects. First,
as Carson, F inocchiaro, and Rohde (2001) indicate, as much as two-thirds of legislation
considered in the House is highly consensual. It is rare for behavior on low variance roll

calls to have any noticeable electoral effects since they do little to help voters or special

19

interest groups distinguish legislators from one another. Moreover, only a fraction of roll
calls with variance in legislative behavior may become important enough to play a central
role in an election since voters generally tend to be unaware of how their representatives
vote on much of the legislation considered in Congress (Kingdon 1989). While voters
may recognize the signiﬁcance of a few important roll calls in Congress each year, they
are less likely to learn or recall exactly how their member voted with respect to those
bills. In nearly all cases, the public’s information about positions taken by legislators is
often limited to a few extremely salient issues (Aldrich 1994).

Even with prominent, controversial votes, the experience of the challenger and
campaign spending may often overshadow particular positions taken (see Mann and
Wolﬁnger 1980: 629; Jacobson 1987: 139). As Jacobson and Bovitz (1998: 20) contend,
if roll call “. . .decisions reﬂected only members’ personal electoral circumstances, and
members sought only to maximize their vote in the next election, then in equilibrium, no
roll call vote should have any signiﬁcant effect on district-level election results. That is,
if members simply adapted their votes to their constituents’ preferences as accurately as
they could, mistakes would cancel out, leaving no systematic effect.” Overall, then, the
deck is stacked against individual roll calls producing observable election-day effects.

Given the relative obscurity of most House roll calls and the fact that voters are
generally unaware of legislators’ voting behavior, why then might legislators worry about
which positions they take on the House ﬂoor? Indeed, chances are that any Single roll
call will remain outside of the electoral arena. However, studies of legislative behavior
suggest that members of Congress are risk averse and worry about the positions they take

because they suspect some roll calls may become electorally salient (see, e.g., Mayhew

20

1974a; Jacobson 1987; Arnold 1990), and they can never be absolutely sure about which
ones will ﬁgure prominently in the next election. Therefore, legislators have Signiﬁcant
incentive to make choices as if every roll call could become a factor in their reelection
bid—and nearly thirty years of Congress scholarship, since Mayhew (1974a), contends
emphatically that they generally act accordingly.

To date, various explanations have been offered to address the hypothesized link
between House ﬂoor roll call voting and citizens’ VOting behavior at the polls. In some
instances, voter awareness implies a thorough and sophisticated understanding of an
incumbent’s speciﬁc roll call choice; however, there is a second level of awareness that is
sufﬁcient to link legislative behavior to electoral outcomes. Indeed, some voters gain a
more general sense of awareness of a speciﬁc roll call decision. For example, proposed
legislation may garner signiﬁcant attention inside the beltway and this stimulates elite
action, which in turn provides voters with some (relatively generic) understanding of
their legislators’ decisions on hot-button issues. These voters may not necessarily be
cognizant of all the details of the vote, but they are at least somewhat familiar with the
nature of a specific position their legislator has taken on an issue that is salient to them.
And this can be sufﬁcient for the roll call-—which is the ofﬁcial articulation of the
position taken—to have an impact on election day.

Arnold (1990: 46) argues that legislators must be careful when casting roll calls
because citizens may use any number of the incumbents’ actions when engaging in
retrospective voting. Additionally, Wright (1978: 446) claims, “Issues do not need to
explain most of the variance in election outcomes; they just need to have enough of an

impact to give congressmen—who are surprisingly insecure about their own re-elections

21

(F enno, 1977; Mayhew, 1974b)—an incentive to heed the general policy wishes of their
constituents.” Moreover, incumbent legislators must develop voting records that prevent
challengers from being able to expose inconsistencies in their position taking. As Arnold
(1990: 272-273) contends, “the fear is not simply that citizens will notice on their own
when a legislator errs, but that challengers will investigate fully a legislator’s voting
record and then share their interpretations of how he or she has gone wrong.” Indeed,

“. . .a prominent position on the wrong side of a major issue [can]. . .galvanize potential
opponents” (Jacobson 1987: 139).

Media coverage surrounding prominent votes in Congress suggests that legislators
have sufﬁcient reason to worry about their roll calls on these issues being politicized. As
suggested above, media coverage of key votes may be all that is necessary to expose
“attentive publics” (Arnold 1990: 64-65) to an incumbent’s roll call choices. Thus, those
most capable of using legislators’ behavior as a political device—namely, party activists,
prospective challengers for House seats, and political elites—have ample opportunities to
become aware of key votes and, in turn, transform them into electorally salient political
issues (i.e., raise voter awareness). Legislators, then, have every reason to be cautious
when taking public positions on such matters, largely because the degree to which votes
will become electorally consequential is to some extent beyond their control.

In sum, positions on roll calls can have signiﬁcant electoral effects not because
voters are always aware of legislators’ speciﬁc roll call choices but because legislators’
roll call behavior affects the actions of political elites (e.g., the media, special interest
groups, and challengers), which, in turn, affects voters’ choices. This conceptualization

of the accountability process, involving various potential links between roll call decisions

22

and voters as well as different levels of voter “awareness,” leads to the prediction that a
signiﬁcant proportion of key votes should have observable effects on electoral margins.
What factors impact whether political elites (including the legislators themselves)
take sufﬁcient follow-up action for positions on roll calls to have measurable electoral
implications? A number of hypothesized factors come to mind. One important factor
may be the degree to which legislators in the same party take opposing sides on the issue.
With an increase in intra-party defections on a given bill, the degree of controversy may
become sufﬁcient to motivate elites to take notice and bring decisions to the attention of a
broad set of voters. This may be especially true when defection from the party’s
positions results in a more attractive outcome for defectors’ constituents. A related factor
in determining the electoral implications of a given roll call position may be the political
salience of the legislation under consideration. Many bills considered in Congress are
brought to the ﬂoor and routinely passed into law with little or no notice paid to them by
constituents or interest groups. However, a considerable set of votes relate to highly
Visible issues that receive widespread attention and are controversial. The relative
salience of a roll call should be expected to drive the extent of electoral consequence.
Aside from factors such as the degree of intra-party controversy and saliency, one
must also consider the broader political context under which issues are brought up in the
House. To a certain degree, whether or not legislation will be viewed by elites or mass
voters as controversial or salient will depend upon the political environment and the
nature of the issue consider consideration. A vote on deﬁcit reduction legislation taken
during the late 19805 or early 19905 would be much more salient, for instance, than one

taken during the early 19705.

23

Data and Methods

To address the question regarding the frequency of electorally consequential roll
calls, this analysis goes beyond the case study and policy-speciﬁc approaches employed
in earlier studies by examining the election-day implications of a relatively large set of
roll call votes. In particular, I consider the effects of all Congressional Quarterly (CQ)
“key votes” from 1973-2000 in the House of Representatives.7 Each year Since 1945, CO
has selected a series of key votes based on what the editors determine to be the “major”
issues of the year. Speciﬁcally, an issue is evaluated in terms of whether it involves a
degree of controversy, is a matter of political or presidential power, and has the potential
to have an enormous impact on the lives of the American people.

My analysis begins with an examination of the 1974 elections since this is the ﬁrst
year in which campaign ﬁnance data for individual House elections is available. I focus
on CQ’S key votes in order to develop a sense of how often prominent roll call votes that
are relatively well known to political elites and voters have observable effects on election
day. In so doing, I can begin to paint a fairly complete picture regarding the frequency of
electorally consequential roll calls since those votes that do not make the CO key vote list
should be less likely to register on the electoral radar screen.

I examine the electoral effects of legislators’ roll call votes by utilizing the same

regression model used by Jacobson (1993).8 This model incorporates the usual predictors

 

7 From 1973-2000, CQ classiﬁed 435 individual roll calls in the House as “key votes.” The total number of
roll calls in this analysis is reduced to 339 for Republicans and 387 for Democrats because I omit all cases
(e.g., the Public Debt Limit vote in November of 1985) where there is negligible intra-party variance (i.e.,
less than 10 defectors from the majority of the party's viewpoint) in terms of roll call choices. For a more
explicit discussion of the rationale for this approach, see Jacobson and Bovitz (1998).

8 While Jacobson includes a dummy variable for the legislator’s party afﬁliation to control for partisan
effects, I opted instead to analyze Republican and Democratic legislators separately to account for any
behavioral differences attributable to partisan attachments. Since I also include members’ NOMINATE

24

of congressional election outcomes while adding an additional explanatory variable—
namely, legislators’ roll call vote choice on a key vote. The statistical method facilitates
empirical estimation of the impact of incumbents’ key vote roll call decisions (the central
explanatory variables) on election returns (the dependent variable), while controlling for
the effects of other factors that are known to affect House election outcomes. Regression
coefﬁcients for each statistically signiﬁcant roll call, as reported in Table 2, represent the
estimated change in the incumbent’s share of the two-party vote that results from a “yea”
vote regarding a given key vote considered on the House ﬂoor.

The dependent variable is the incumbent’s share of the two-party vote in legislator
i ’5 district during year t.9 The central explanatory variables are the legislator’s choices on
key vote roll calls from year t and year t-I (i.e., the roll call votes from the two sessions
of Congress preceding the election). In order to isolate the effects of the legislators’ roll
call votes on their electoral fortunes, I control for a series of factors that are known to
affect incumbent vote share (see, e.g., Jacobson 1993). In particular, the control variables
in the model include: the legislator’s previous vote margin, the district’s partisanship
measured independently of the congressional vote, incumbent and challenger spending,
and challenger quality. In addition, I control for the legislator’s overall voting record in
order to isolate the effects of the particular key votes under examination.

In sum, I posit the following model of legislators’ electoral fortunes:

INCUMBENT Two-PARTY VOTE SHARE = or + Bllncumbent Vote ShareH+

BzDistrict Partisanship + Bgln(lncumbent Spending) + B4ln(Challenger Spending) +
BsChallenger Quality + B6Voting Record + [37KEY VOTE + e

 

score (as a proxy for their ideology) in the model, this seemed like an acceptable alternative that avoids
including both partisan afﬁliation and member ideology on the right-hand side of the equation.

9 Given the general focus on the electoral impact of incumbents’ voting behavior, I exclude open seat
contests, incumbents who previously ran and won in a special election, and the few races with non-
Democratic/Republican incumbents.

25

KEY VOTE represents a legislator’s decision of whether or not to support a roll call vote
considered on the House ﬂoor. KEY VOTE is a dummy variable where a position taken
on the vote in favor of the roll call is coded 1, 0 otherwise.10 I test the model with each
key roll call separately because, if all key votes are included in a single model, a number
of issues (e. g., multicollinearity, listwise deletion) can adversely affect the estimates.11
Incumbent vote share“ is operationalized as the legislator’s share of the two-party
vote in the previous election.12 District Partisanship is measured as the district-level two-
party vote share for the presidential candidate from the incumbent representative’s party
(subtracted ﬁom the national vote) in the most recent presidential election (see Brady,
Canes-Wrone, and Cogan 2000; Jacobson 2000; Ansolabehere, Snyder, and Stewart
2001). Following Jacobson’s (1980) seminal work on money in elections, the model also
controls for the effects of incumbent spending and challenger spending. Speciﬁcally, I
include the natural logarithm of both dollar amounts as separate independent variables in

the model (Jacobson 1980: 40).13 Challenger Quality is operationalized as a dummy

 

'0 Abstentions on roll calls votes are treated as missing in the dataset, as are votes that are paired for or
against in some of the earlier Congresses. While it would be interesting to examine the observed effects of
abstentions more systematically in this analysis, one cannot readily determine whether votes were avoided
for strategic reasons or because a member was unable to be present at the time votes were taken.

1' Including all the roll calls from a given congressional session in a single equation makes it nearly
impossible to interpret results Since legislative voting behavior across roll calls tends to be highly inter-
correlated. Additionally, including all the votes in the same model leads to the serious issue of listwise
deletion resulting from varying patterns of member abstentions across all key votes. By including only one
vote per model, I avoid these estimation problems. More importantly, however, I am in a better position to
offer insights regarding the electoral impact of individual key votes on members’ vote margins. While this
analysis is not without potential limitations, I see it as an appropriate ﬁrst step toward establishing an
empirical understanding of when, and under what conditions, roll calls impact electoral fortunes.

'2 For those Congresses following a decennial reapportionment (97th and 102'”) and occasionally when
other districts were redrawn (e.g., North Carolina in the 19905), I attempted to estimate what a member’s
two-party vote share in the new district would have been in the previous election. To assist with this
process, I collected county-level data and attempted to match up former and current district-level data as
accurately as possible. In many cases, congressional districts were only slightly redrawn, thereby
minimizing obstacles to the analysis.

'3 As Jacobson argues, the advantage of using the natural logarithm of expenditures for empirical analysis
of elections is that doing so avoids the assumption of a linear relationship between money and votes,
thereby accounting for diminishing returns from campaign spending.

26

variable coded 1 if the candidate has previously held elected ofﬁce, 0 otherwise. This
coding also follows J acobson’s classic study that views a successful campaign for another
public ofﬁce as a proxy for candidate quality.14

For the operationalization of the Voting Record variable, I use Poole-Rosenthal
ﬁrst dimension DW-NOMINATE scores (see, e. g., Poole and Rosenthal 1997; McCarty,
Poole, and Rosenthal 1997). These scores, constructed Via scaling techniques that use
legislators’ entire roll call records to measure ideology, provide summary statistics of
legislators’ overall voting tendencies on the traditional liberal-conservative economic
dimension. DW-NOMINATE scores for legislators range from -1 to 1, with low scores
indicating policy liberalism and high scores indicating policy conservatism. By including
this variable in the model, I seek to control for the legislator’s overall ideological
reputation and, in turn, isolate the impact of a particular roll call decision that may or may

not be consistent with the broader voting record.15

Results
A summary of the estimated electoral effects of C Q key votes for the 1974-2000
House elections are displayed in Table 1 where I report the percentage of key votes that
yielded observable electoral effects for incumbent legislators. As summarized in Table 1,

a total of 66 out of 339 CQ key votes (20%) have a signiﬁcant impact on Republican

 

'4 See Jacobson (1980: 106-107; 1990). For a dissenting View, see Green and Krasno (1988: 884-907).

'5 By controlling for members’ voting record as reﬂected by their NOMINATE score, I conduct a relatively
stringent test of the electoral irrrpact of individual roll calls. Including both the individual-level vote and
the NOMINATE score in the model does not necessarily make it more difﬁcult to infer statistical signiﬁcance
since a single roll call represents a fairly negligible component of a member’s voting record. In fact, I ran a
series of correlation analyses which show that individual key votes are ahnost never strongly correlated
with members’ NOMINATE scores. Additionally, I re-ran the OLS models for the ﬁrst part of our analysis
without the NOMINATE score and found that the substantive results rarely change in any notable fashion. I
nevertheless chose to include legislators’ NOMINATE scores because I believe it is theoretically important
and it allows me to include some degree of control for the impact of legislators’ ideological tendencies.

27

incumbents’ electoral fortunes, while 62 of 387 key votes (16%) yield a statistically
signiﬁcant impact on Democratic legislators’ margins. Additionally, Table 2 displays the
actual estimates from the 1974-2000 House elections where I report the vote-speciﬁc
information for each roll call that has a signiﬁcant impact on incumbent vote share for
either Democrats or Republicans (or both).16 Each coefﬁcient reported in the tables

represents the estimated impact of voting “yea” on the corresponding roll call vote.17

Table 1 - Percentage of CQ House Key Votes With Signiﬁcant
Electoral Impact, 93'd-106‘h Congresses

 

% 01' “Key” Roll Total Number of % Of “Key” Roll Total Number of

 

Congress Calls Signiﬁcant Key Votes During Calls Signiﬁcant Key Votes
for Republicans Congress for Democrats During Congress
93 18.5 27 16.0 25
94 15.4 26 6.5 31
95 4.0 25 18.8 32
96 18.2 22 23.3 30
97 10.7 28 12.9 ‘ 31
98 7.4 27 12.9 31
99 44.4 27 13.8 29
100 8.0 25 14.8 27
101 48.3 29 22.6 31
102 12.5 24 26.7 30
103 15.8 19 13.3 30
104 22.7 22 15.4 26
105 33.3 21 15.8 19
106 5.9 17 6.7 15
Total 19.5 339 16.0 387

 

 

'6 All relevant information concerning bill descriptions, the date the roll call was taken, and the partisan
breakdown of the vote was collected from Congressional Quarterly’s annual series, Congressional Roll
Call.

17 I do not report the coefﬁcient results for the control variables in our model due to space limitations.
Overall, the pattern of results for these variables is consistent with extant work on House elections.

28

Table 2 — Key Votes Yielding Signiﬁcant Electoral Effects on Incumbent Margins

 

 

Roll Call Congress Date Party & Coefficient Robust
Vote Result St. Error
Foreign Military and 93 7/26/73 D 119-94 2.36 1.54
Economic Aid R 69-89 2.45* 1.35
Defense Procurement 93 7/31/73 D 160-63 -2.60* 1.57
Authorization R 27-121 1.60 1.10
Alaskan Pipeline 93 8/2/73 D 133-101 -1.24 1.81
R 65-120 5.14" 1.20
Gerald R. Ford 93 12/6/73 D 201-35 -4.66* 2.58
Conﬁrmation R 186-0 --- ---
National Energy 93 12/14/73 D 78-137 1.71 1.63
Emergency Act R 102-73 -2.00"' 1.10
Minimum Wage Increase 93 3/20/74 D 220-11 0.94 4.46
R 155-26 -3.41"‘ 1.98
Elementary and Secondary 93 3/26/74 D 145-88 4.37" 1.82
Education Act R 148-29 -0.83 1.69
Amendments
Defense Supplemental 93 4/4/74 D 55-127 271* 1.52
Authorization, Fiscal 1974 R 99-50 0.45 1.17
Public Service Jobs 93 12/12/74 D 203-11 3.32 4.30
R 119-42 -3.03** 1.39
South Vietnam Assistance 94 5/1/75 D 72-200 0.82 1.24
R 90-46 -3.67** 1.35
Executive Level Pay Raises 94 7/30/75 D 178-105 -2.35** 1.09
R 36-108 -2.58 1.58
Aid to New York City 94 12/2/75 D 175-103 -0.94 1.55
R38-100 -3.l8** 1.53
Fiscal 1976 Congressional 94 12/12/75 D 186-61 -2.86* 1.50
Budget Resolution R 3-126 --- ---
Defense Department 94 1/27/76 D 251-30 -0.43 2.22
Appropriations, Fiscal R 72-69 2.16* 1.12
1976
Lobbying Disclosure 94 9/28/76 D 54-173 -1 .79 1.53
R 94-29 -2.90* 1.62
Stimulus Tax Cuts 95 3/8/77 D 54-218 2.77" 1.38
R 140-1 --- ---

 

29

Table 2 — Key Votes Yielding Signiﬁcant Electoral Effects on Incumbent Margins

 

 

(cont.)
Roll Call Congress Date Party & Coeﬂicient Robust
Vote Result St. Error
Obey Commission House 95 10/12/77 D 160-113 -2.46""" 1.05
Reforms R 0-139 --- ---
National Energy Act 95 10/13/78 D 199-79 -2.03** 0.92
R 8-127 -1.69 2.37
Fiscal 1979 Budget Targets 95 5/3/78 D 58-202 3.16" 1.21
R 139-1 --- ---
Labor-HEW 95 6/13/78 D 98-163 3.22" 1.16
Appropriations, Fiscal R 122-18 -4.01** 2.03
1979
Full Employment Act 95 3/9/78 D 56-221 4.67" 1.20
R 142-2 --- ---
Energy Mobilization Board 96 11/1/79 D 126-129 3.42" 1.19
R 27-121 1.03 1.28
Nuclear Regulatory 96 1 1/29/79 D 1 12-133 2.01 * 1.17
Commission R 23-121 0.62 1.39
Panama Canal Treaties 96 6/20/79 D 195-68 0.14 1.48
Implementation R 25-132 -3.68** 1.19
Chrysler Loan Guarantees 96 12/18/79 D 209-48 -2.43*"' 1.08
R 62-88 -2.10** 1.03
Defense Department 96 5/15/80 1) 123-130 3.62" 1.14
Authorization R 29-120 -0.26 1.33
Draft Registration Funding 96 4/22/80 D 135-122 -2.13* 1.10
R 83-66 0.35 1.04
Labor-HHS-Education 96 8/27/80 D 134-124 1.03 0.97
Appropriations, Fiscal R 65-91 2.07* 1.09
1981
Windfall Proﬁts Tax 96 3/12/80 D 178-85 2.27" 1.06
R 23-130 3.79" 1.60
Michael Myers Expulsion 96 10/2/80 D 225-29 4.65" 1.52
R 151-1 --- ---
Legal Services Corporation 97 6/18/81 D 186-21 -2.44 2.80
R59-116 -2.12* 1.14
Tax Cuts 97 7/29/81 D 48-194 2.98" 1.37
R 190-1 --- ---

 

30

Table 2 - Key Votes Yielding Significant Electoral Effects on Incumbent Margins

 

 

(cont.)
Roll Call Congress Date Party & Coefficient Robust
Vote Result St. Error
Urgent Supplemental 97 5/12/82 D 215-15 438* 2.64
Appropriations, Fiscal 1982 R 128-52 0.62 1.16
Budget Reconciliation Tax 97 7/28/82 D 164-60 0.12 1.17
Increases/Spending Cuts R 44-137 3.92" 1.18
Soviet Economic Sanctions 97 9/29/82 D 82-146 -2.03* 1.09
R 124-57 0.41 1.14
Continuing Appropriations, 97 12/14/82 D 20-208 7.07" 3.18
Fiscal 1983/Jobs R 171-7 0.90 2.51
Automobile Domestic 97 12/15/82 D 171-58 2.15 1.27
Content Requirements R 44-130 -2.63** 1.14
Emergency Housing 98 5/11/83 D 210-41 -2.89* 1.50
Assistance Act R 6-155 -1.70 1.99
Department of Housing and 98 6/2/83 D 138-86 1.73* 0.91
Urban Development R 89-50 -0.44 1.05
Approp.
Defense Department 98 11/1/83 D 190-72 -0.32 1.24
Appropriations, Fiscal 1984 R 18-145 ”2.53" 1.13
Defense Department 98 11/15/83 D 198-63 2.68" 1.10
Appropriations, Fiscal 1984 R 60-103 -0.80 0.90
Tax Reform Act 98 11/17/83 D 191-65 -0.85 1.08
R 13-149 3.54" 1.07
Equal Access Act 98 5/15/84 D 123-134 -1.77* 1.03
R 147-17 -0.62 1.26
Supplemental 99 6/6/85 D 95-133 -0.01 0.89
Appropriations, Fiscal 1985 R 108-69 2.15" 0.82
Department of Defense 99 6/19/85 D 86-162 -0.87 1.11
Authorization, Fiscal 1986 R 143-34 -2.59** 1.12
Farm Programs Reauthor., 99 10/3/85 D 82-164 182* 0.96
Fiscal 1986-90 R 169-10 3.39" 0.89
Plant Closing Notiﬁcation 99 1 1/21/85 D 183-54 -2.33* 1.30
R 20-154 252* 1.37
Superfund Reauthorization, 99 12/10/85 D 147-101 2.98" 0.92
Fiscal 1986-90 R 73-105 0.80 0.89

 

31

Table 2 — Key Votes Yielding Signiﬁcant Electoral Effects on Incumbent Margins

 

 

(cont.)
Roll Call Congress Date Party & Coeﬁ'icient Robust
Vote Result St. Error
Tax Overhaul 99 12/11/85 D 188-59 -l.44 1.01
R 14-164 -2.92** 1.37
Omnibus Highway Bill 99 8/6/86 D 81-161 -0.45 0.94
R 117-57 -2.59""" 0.97
Defense Authorization, 99 8/8/86 D 208-26 -0.09 1.73
Fiscal 1987 R 37-130 2.40" 1.13
Defense Authorization, 99 8/8/86 D 200-31 -0.22 1.74
Fiscal 1987 R 34-124 2.73" 1.02
Defense Authorization, 99 8/ 12/ 86 D 206-34 -0.06 1.67
Fiscal 1987 R 33-142 2.13" 0.98
Defense Authorization, 99 8/13/86 D 172-72 0.80 1.21
Fiscal 1987 R 38-137 1.62" 0.98
Pesticide Control 99 9/19/86 D 135-65 -0.45 1.24
Reauthorization R 22-1 18 3.14" 1.26
Tax Overhaul 99 9/25/86 D 176-74 -1.56* 0.92
R 116-62 _ 1.29 0.89
South African Sanctions 99 9/29/86 D 232-4 --- ---
R 81-79 2.57" 1.06
Speed Limit/Omnibus 100 3/18/87 D 92-156 2.29" 0.97
Highway Reauthorization R 125-50 -1.52"' 0.87
FSLIC Rescue 100 5/5/87 D 81-160 -2.77** 0.97
R 72-98 -0.88 0.78
Coast Guard Authorization/ 100 7/8/ 87 D 200-38 -1.87 1.50
Re-ﬂagging Kuwaiti Ships R 22-146 3.01'MI 1.00
Fiscal 1988 Transportation 100 7/13/87 D 124-102 1.86" 0.86
Approp./In-Flight Smoking R 74-91 -0.44 0.81
Fiscal 1988 Budget 100 10/29/87 D 205-41 -l.82* 1.08
Reconciliation R 1-164 --- ---
Fiscal 1990 Budget 101 5/4/89 D 157-96 -O.22 1.01
Resolution/ Adoption R 106-61 -3.00** 1.22
Minimum-Wage 101 6/14/89 D 227-28 -3.55* 1.96
Increase/Veto Override R 20-150 6.92" 1.88

 

32

Table 2 — Key Votes Yielding Signiﬁcant Electoral Effects on Incumbent Margins

 

 

(cont.)
Roll Call Congress Date Party & Coeﬂicient Robust
Vote Result St. Error

Fiscal 1990-91 Defense 101 7/25/89 D 214-38 -O.50 1.75
Department Auth./SDI R 34-137 339* 1.74
Funding
Fiscal 1990-91 Defense 101 7/26/89 D 116-135 1.27 1.18
Department Auth./’Stealth’ R 28-144 3.73" 1.86
Fiscal 1990 Budget Reconc./ 101 10/5/89 D 36-214 4.24" 1.86
Childcare Substitute R 159-16 -4.58* 2.62
Fiscal 1990 Labor, HHS and 101 10/11/89 D 175-72 2.49* 1.30
Education Approp./Abortion R 41- 1 34 1.66 1.35
Oil-Spill Liability/State 101 11/8/89 D 204-44 -1.67 1.67
Pre-cmption R 75-99 2.77” 1.09
Government Pay-and-Ethics 101 11/16/89 D 168-85 -3.54** 0.98
Package/Passage R 84-89 -0.55 1.26
Fiscal 1990 Foreign Aid 101 5/22/90 D 219-28 -1.45 1.99
Supplemental Authorizations R 3 1-135 3 .54* 1.81
Clean Air Act 101 5/23/90 D 231-20 -0.77 2.80
Reauthorization/ Transition R 43-126 2.65" 1.32
Aid '
Family and Medical Leave 101 7/25/90 D 194-57 -2.96* 1.51
Act/Veto Override R 38-138 305* 1.79
Frank Reprimand/Censure 101 7/26/90 D 12-241 3.55 2.74

R 129-46 3.51M 1.55
Fiscal 1991 Defense 101 9/18/90 D 205-39 -2.09 1.84
Authorization/ SDI Funding R 20-150 3.65”“ 2.07
Legal Immigration 101 10/3/90 D 186-65 -2.57* 1.32
Revision/Passage R 45-127 0.13 1.44
Fiscal 1991 Budget 101 10/4/90 D 108-149 -0.75 1.14
Resolution/ Conference R 71-105 -2.4l * 140
Report
Civil Rights Act of 101 10/17/90 D 239-15 -3.28 2.27
1990/Conference Report R 34-139 4.77" 1.58
Fiscal 1991 Intelligence 101 10/17/90 D 195-50 -2.75* 1.56
Appropriations/Aid to R 12-156 4.73" 2.59
UNITA
Civil Rights Act of 1991/ 102 6/5/91 D 250-15 5.75” 1.93
Passage R 22-143 1.30 1.72

 

33

Table 2 — Key Votes Yielding Signiﬁcant Electoral Effects on Incumbent Margins

 

 

(cont.)
Roll Call Congress Date Party & Coeﬂicient Robust
Vote Result St. Error

Fiscal 1992 Interior 102 6/25/91 D 184-78 -3.45** 1.27
Appropriations/ Grazing Fees R 47-1 14 0.14 1.35
Fiscal 1992-93 Foreign Aid 102 10/30/91 D 131-127 1.08 1.28
Authoriz./Conference Report R 28-134 3.92“ 1.5 8
Fiscal 1992 Labor, HHS, and 102 11/19/91 D 222-43 4.05“ 1.88
Education Approp./Passage R 53-113 0.74 1.39
Campaign Finance/Passage 102 1 1/25/91 D 251-12 -3.80** 1.28

R21-144 0.28 1.95
1992 Tax Bill/Democratic 102 2/27/92 D 219-46 -2.29* 1.23
Substitute R 1-164 -0.72 5.42
Campaign Finance Reform/ 102 4/9/92 D 239-20 -2.77 2.97
Conference Report R 19-145 499* 2.60
Family Flaming 102 4/30/92 D 212-45 4.10" 1.73
Reauthorization/ Passage R 55-105 1.28 1.32
Fiscal 1993 Defense 102 6/5/92 D 149-77 -2.83** 1.33
Authoriz./ Strategic Defense R 11-134 1.91 2.74
Initiative
Cable Television 102 7/23/92 D 221-23 -2.34* 1.19
Reregulation/ Program R 116-45 -0.15 1.31
Access
Western Water Bill/Central 102 10/6/92 D 42-202 0.53 1.90
Valley Project R 117-41 -3.20** 1.27
Fiscal 1994 Budget 103 3/18/93 D 242-11 -3.35"‘ 1.86
Resolution/ Adoption R 0-172 --- ---
Fiscal 1994 Defense 103 9/28/93 D 157-101 0.02 1.08
Authorization] Gay Ban R 11-163 -3.03* 1.75
NAFTA 103 11/17/93 D 102-156 -2.31** 0.70
Implementation/Passage R 132-43 -0.90 0.93
Mining Law 103 11/18/93 D 245-6 -O.53 2.70
Overhaul/Passage R 70-102 -2.56"”" 1.05
Abortion Clinic 103 5/5/94 D 200-43 -1.89* 1.00
Access/Conference Report R 40-131 -0.84 1.10
Fiscal 1995 Defense 103 6/9/94 D 117-132 1.36" 0.68
Authoriz./ Bosnia Arms R 127-45 -0.15 1.00

Embargo Term.

34

Table 2 — Key Votes Yielding Signiﬁcant Electoral Effects on Incumbent Margins

 

 

(cont.)
Roll Call Congress Date Party & Coeﬂicient Robust
Vote Result St. Error

California Desert 103 7/12/94 D 92-157 -0.19 0.98
Protection/Hunting Exception R 146-26 2.56”“ 1.47
Clean Water Act 104 5/16/95 D 45-150 -2.52** 1.11
Revisions/Passage R 195-34 -1.88** 0.78
1995 Budget- 104 10/26/95 D 4-192 --- ---
Reconciliation/Passage R 223-10 -2.14** 1.00
Gift Rules/Full Disclosure 104 1 1/16/95 D 46-150 -2.69** 0.96
Alternative R 108-125 1.21 0.64
Fiscal 1996 Defense 104 1/3/96 D 34-139 -1.37 1.27
Authorization/ Veto Override R 206-16 -1.66"‘* 0.82
Farm Bill/Passage 104 2/29/96 D 54-135 -0.26 1.04

R216-19 -l.45* 0.88
Anti-Terrorism and Death 104 3/13/96 D 67-117 -1.70 1.01
Penalty Enforcement R 178-54 1.44* 0.74
Authority
Anti-Terrorism & Death 104 3/14/96 D 122-65 2.56" 1.02
Penalty Enforcement/Habeas R 12-218 -0.21 1.26
Corpus ‘
Fiscal 1997 Omnibus 104 9/28/96 D 167-13 -3.66** 1.34
Approp./Conference Report R 202-24 -0.14 1.14
Reprimand of 105 1/21/97 D 198-2 --- ---
Rep. Newt Gingrich R 196-26 1.94* 0.98
Public Housing System 105 5/14/97 D 71-130 1.77* 1.05
Overhaul R 222-1 --- ---
Fiscal 1998 Budget 105 5/21/97 D 155-48 -l.91** 0.83
Resolution R 58-168 1.36 0.93
Fiscal 1998 Budget 105 5/21/97 D 132-72 -1.96*"‘ 0.87
Resolution R 201-26 0.44 0.88
Fiscal 1998 Foreign 105 9/4/97 D 171-33 -0.58 1.43
Operations R 38-185 -3.67** 1.20
Appropriations/ Abortion
Funds
Private School Vouchers 105 11/4/97 D 4-192 --- ---

R 187-35 -3.09** 1.16
Special Transportation 105 4/1/98 D 12-184 0.13 l .64
Projects R 67- 1 52 -2.02** 0.72

 

35

Table 2 — Key Votes Yielding Signiﬁcant Electoral Effects on Incumbent Margins

 

(cont.)
Financial Services Overhaul 105 5/13/98 D 6.1-139 0.63 0.76
R 153-73 -2.19** 0.74
Food Stamps for Legal 105 5/22/98 D 2-190 --- ---
Immigrants R 118-98 -1.80** 0.77
Contraceptive Coverage 105 7/16/98 D 175-21 ’ -1.91 1.53
R48-177 -3.53"'* 1.10
Physician-Assisted Suicide 106 10/27/99 D 71-135 1.60“ 0.83
R 200-20 0.82 1.01

Campaign Finance Disclosure 106 6/28/00 D 205-0 --- ---
R 178-39 1.83" 0.78

 

Mp s .05, *p s .10

A cursory glance at the results in Table 2 reveals a number of notable ﬁndings.
First, there is considerable variation in the types of issues yielding observable electoral
consequences. In particular, one can detect Signiﬁcant electoral effects on roll calls
dealing with budget and tax policy, foreign policy and national defense, environmental,
transportation, civil rights, and social issues. This is an especially interesting result as it
suggests that the relative inﬂuence of “attentive publics” (e.g., interest groups and
political elites) in the election is not limited to a narrow category of issue areas.

I also observe that many of the votes in Table 2 are signiﬁcant in both a statistical
and substantive sense. While some of the votes appear to have a minimal effect on
incumbents vote margins in the subsequent elections, many other roll calls exert what is
considered to be a sizeable impact on vote share, especially in light of the other factors
being controlled for in the model (e.g., campaign spending, challenger quality). Indeed,

some roll calls appear to impact legislators’ electoral margins by as many as six or seven

36

percentage points on average. Consider, for instance, the electoral impact of voting “yea”
on the veto override for the bill to increase the minimum wage (6/14/89). The 20
Republicans who voted in favor of this override received, two-party votes shares that were
on average approximately 7 percent higher than other Republican incumbents in the 1990
House elections (assuming all else equal).18

The results displayed in Table 2 also reveal considerable variation across House
elections in terms of the number of key votes yielding signiﬁcant electoral effects. While
we observe as many as 14 and 17 key votes yielding Signiﬁcant electoral effects in the
1986 and 1990 House elections respectively, other election years (e.g., 2000) have as few
as 2 key votes prove to be electorally consequential. (The average number of signiﬁcant
roll calls between 1974 and 2000 is 8.) This suggests (although it is not directly tested)
that the political context under which a vote occurs in the House may play an important

role in determining whether or not the vote will produce observable electoral effects.19

 

'8 Why did these 20 Republicans beneﬁt so much from their position-taking on the Minimum Wage
Increase veto override? There are at least two key reasons. First, the override vote was very high proﬁle,
including national news coverage (e.g., in the New York Times) as well as highly publicized reactions to
the vote by some of the most powerful special interest groups in America (e.g., the AFL-CIO) who have a
direct line of communication to large constituencies. These conditions facilitated voter awareness of the
issue—and, of representatives’ positions. Second, the Republicans who defected from the party to vote
against the president’s veto came from districts where they had a signiﬁcant opportunity to beneﬁt
electorally as a result. These Republicans hailed from relatively moderate districts (as compared with other
Republican incumbents), where the minimum wage increase proposal was more popular. Indeed, these 20
Republicans—many of whom were Northeasterners—generally had a relatively strong Democratic
presence in their districts (based on presidential vote shares, which is a frequently used measure of district
partisanship). In these 20 districts, the 1988 Democratic presidential candidate (Dukakis) received, on
average, 45 percent of the vote (range: 42-57 percent). In contrast, nearly all of the 150 Republicans who
voted against the veto override came from districts where Bush generally received relatively large shares of
the district-level presidential vote. AS I discuss later in the paper, the pattern of electoral signiﬁcance of
roll calls suggests that strategic party defections in favor of specialized district preferences can often be
rewarded.

‘9 Since the possibility exists that some of the signiﬁcant ﬁndings in the regression models could be due to
“false positives” resulting from similar alignments of voting coalitions among members, it is important to
address this issue brieﬂy. As it now stands, the positions that legislators take on a Speciﬁc key vote could
appear electorally consequential (even though they actually may not be) simply because the same members
that voted together on a previously signiﬁcant vote also did so on a subsequent vote. Nonetheless, I do not
believe that this potential issue undermines the reported ﬁndings for two main reasons. First, from a purely

37

The degree of attention given to a particular vote by attentive publics, the media, special
interest groups, or experienced challengers may inﬂuence the likelihood that incumbents’
roll call decisions will be revisited during a subsequent election campaign.

Given the sizable percentage of electorally signiﬁcant votes for the Republicans
and Democrats (20% and 16% respectively) from 1973-2000, it appears that more than
chance is at work when we observe some key roll calls—and not others—having a
signiﬁcant electoral effect. This, of course, raises one additional question. How can one
systematically explain the signiﬁcant effects of some roll calls and not others, especially
given the large volume of roll calls legislators consider in one year and given citizens’

general lack of awareness of legislators’ decisions? It is to this question that I now turn.

Explaining Variation in the Electoral Signiﬁcance of Key Roll Calls

Many of the roll call votes under investigation yielded no observable signiﬁcant
effect on electoral outcomes. This is not surprising given that legislators cast hundreds of
roll call votes every session of Congress and only a fraction of those votes can become
large enough issues to play a role in the election. However, a sizable number of key roll
calls from 1973-2000 were controversial, became political issues and, according to the
earlier results, appear to have had signiﬁcant effects on congressional elections.

Given that some roll calls have observable electoral consequences and others do

not, a theoretical explanation is needed to explain why certain roll calls emerge from the

 

theoretical perspective, there is little reason to suspect that the same coalitions of legislators routinely join
together and vote the same way except in those circumstances when their positions are likely to set them
apart electorally from other members. Second, a related analysis (results not Shown) that combined
legislators’ voting patterns across all key votes in a party unity score (to reﬂect the proportion of times
members voted with a majority of their party) revealed considerable variability in signiﬁcance both across
legislators and over time, thus appearing to minimize the likelihood of false positives.

38

vast number of House ﬂoor votes to inﬂuence elections. Toward that end, I draw on the
regression analysis presented in the previous section to test empirically what drives roll
calls’ electoral effects (or lack thereof). How does this analysis move us down the path
toward a theoretical framework for the conditions under which roll calls are electorally
consequential? The results of the previous section provide data on which votes are—and
which votes are not—signiﬁcant. Using these data as the dependent variable, I can test a
model designed to predict whether votes will have electoral implications and then use the
results to inform more comprehensive theory development.

The results of the analysis summarized in Table 2 provide 435 observations of roll
calls that either do or do not have observable electoral effects for Democrats and/or
Republicans.20 These observations provide the dependent variable—Roll Call Electoral
Signiﬁcance—for this aspect of the analysis. This variable is coded 1 if the vote on the
roll call has a signiﬁcant electoral effect according to the regressions presented earlier,
and 0 otherwise. Since Roll Call Electoral Signiﬁcance is a binary variable, I utilize
probit to estimate the model, which can accommodate the complexities of estimating a
model with a limited dependent variable.

What causes a roll call to have observable electoral implications? Drawing on the
earlier theoretical discussion, many factors come to mind. For example, a large number
of intra-party defections on a roll call vote may instigate electoral relevance. As the
proportion of defections within the party increases, political elites may take notice of this
level of disagreement and draw it to the attention of the voters (especially if it results in a

more favorable outcome for a representatives’ constituents). Alternatively, if a roll call

 

2° Roll calls where there is negligible intra-party variation were coded as not electorally signiﬁcant in this
model. The results reported in Table 3 do not differ substantively when these votes are excluded from the
model.

39

vote is a major news story (e.g., it makes the front page of the New York Times), then it
may have strong political salience and, in turn, affect elections. Yet another possible
factor is the roll call vote’s particular issue area, since some may seep into the electoral
arena more regularly than others. A ﬁnal consideration is that electoral effects may hinge
on whether there was consensus surrounding a roll call (e.g., a vote result of 425-10) vs.
being conﬂictual (e. g., 218-217). All of these factors are incorporated into the model as
independent variables.

Since I pool data from across elections, it is important to control for ﬁxed effects
(see Kennedy 1998: 226-228). Thus, I include dummy variables for all but the 93rd
Congress as independent variables to control for the possibility that certain congresses
may produce more electorally consequential roll calls than others. I also control for
whether or not a particular roll call took place during a congressional election year. By
doing this, I account for the possibility that roll calls cast during election years are better
poSitioned to have electoral implications due to their proximity to the upcoming election.

In sum, I posit the following model to explain whether a roll call is electorally
Signiﬁcant or not:

’ Pr(Roll Call Electoral Signiﬁcance) = or + BlElection Year + BzParty Defections +

B3COUﬂICt Level + [34 NYTimeS + BsAppropriations Vote + BsDefense Vote +
B7Welfare Vote + BgCongress-Speciﬁc Dummies + e

As before, I run the model separately for Democrats and Republicans since some factors
may determine whether or not a roll call is electorally consequential for one party’s
incumbents while having no effect on the other party’s reelection-seeking legislators.

Each variable in the model is measured as follows.

40

Election year is a dummy variable coded 1 for an election year, 0 otherwise.
Party Defections is coded as the proportion of a party’s incumbents who voted with the
other party on the key vote. Conﬂict Level is measured as the proportion of members
dissenting from the majority View on the roll call. NYTimes is a dummy variable
indicating whether or not the key vote was a front-page story in the New York Times
immediately following the vote in the House (thus, tapping a measure of saliency for the
vote). The issue codes (e.g., appropriations, defense, and welfare votes) are dummy
variables taken from Rohde’s (2003) roll call database for the House of Representatives
that spans from 1953-2000. Finally, I include a dummy variable for each Congress
(except for the 93rd).

The results of the model for Democrats and Republicans are presented in Table
3.21 The probit estimates indicate that, for both parties, the level of conﬂict has a
signiﬁcant positive effect on whether a roll call will be electorally signiﬁcant. The
estimates suggest that the higher the level of conﬂict on a particular roll call, the higher
the probability that a roll call vote will be electorally consequential. I ﬁnd this result to
be particularly appealing and quite informative as it suggests that the more embattled
Congress is over a particular key roll call, the more likely the roll call is to help voters
distinguish among candidates and then cast votes accordingly. This conclusion is quite
intuitive, providing a straightforward explanatory factor that can be used to build a more
comprehensive theory and to make predictions regarding whether or not positions taken

on an alleged “key” vote will actually be critical come election time.

 

2' As is common in the literature that presents similar analyses with time-period controls, I do not report the
coefﬁcients and standard errors of the Congress-speciﬁc dummy variables in Table 3.

41

Table 3 - Probit Model of Electorally Signiﬁcant House Roll Call Votes

 

 

Variable Democrats A P Republicans A P

Election Year 0156 0.133
(0.152) (0.162)

Party Defections 0.869 1.556" .03
(0.622) (0.564)

Level of Conﬂict 0.788" .03 1.574M .07
(0.400) (0458)

NY Times Front-Page Story 0.425" .09 0.346" .05
(0.159) (0.172)

Appropriations Vote -0.040 -0.249
(0.223) (0.256)

Defense Vote -0.107 0.478* .09
(0.261) (0.248)

Welfare Vote -0.020 0.084
(0285) (0.325)

Constant -2.226** -3.012**
(0.439) (0.564)

Pseudo R2 0.080 0.179

Log-Likelihood 32.97 59.5 1

% Correctly Predicted 85.75 85.98

Reduction of Error 26.25 53.48

Sample Size (N) 435 435

 

Note: the coefﬁcients are probit estimates; the dependent variable is 1 if the roll cmote is electorally
signiﬁcant according to our analysis, 0 otherwise; “p S .05, *p s .10

The change in predicted probability (A P) is calculated from a baseline model using Tomz, Wittenberg, and
King’s (1999) CLARIF Y program where interval-level variables are held at their mean and dichotomous
variables are held at their modal value. For interval-level variables, changes reﬂect an increase of one
standard deviation. For dichotomous variables, the values reported reﬂect changes from one category to
the other. For more information about the CLARIFY program, see King, Tomz, and Wittenberg (2000).

The substantive effect of the covariate(s) can be illustrated by calculating the
change in predicted probability of a roll call yielding electoral effects as the explanatory

variable of interest increases one standard deviation (or moves from zero to one) while

42

holding all other variables at their mean or modal values. These estimates are displayed
in the table (A P) for the explanatory variables yielding signiﬁcant effects. For instance,
roll calls with higher levels of conﬂict were 3 percent more likely to result in electoral
consequences among Democrats. Among Republicans, higher levels of conﬂict on roll
calls dramatically increased the likelihood of observable electoral consequences by as
much as 7 percent.

Additionally, the results support my theoretical expectation that roll calls with
high levels of salience as measured by their appearance on the front page of the New York
Times are signiﬁcantly more likely to yield observable electoral consequences. Among
Democrats, roll call votes that made the front page of the Times were 9 percent more
likely to result in observable effects. While the effect among Republicans was smaller,
votes appearing on the front page of the Times still increased the probability of electorally
signiﬁcant effects by as much as 5 percent. This ﬁnding is noteworthy as it suggests that
when voters are better informed about highly salient votes in the House, they are more
likely to reward or punish their representative for their position taking behavior.

Aside from these main results, I also ﬁnd that two additional factors appear to
play a Si gniﬁcant role in determining whether or not a roll call has electoral effects for
Republicans only. First, I observe that the number of party defections on roll call votes
among Republicans had a signiﬁcant and positive impact on those roll calls becoming
electorally consequential. While the variable for party defections among Democrats was
also in a positive direction, it failed to reach conventional levels of statistical signiﬁcance
(p<.162). One possible explanation for this result is that Republicans in relatively

moderate districts may occasionally be rewarded for joining with Democrats on votes in

43

certain key policy areas, while the same cannot be said for Democrats who defect to vote
with Republicans.

I also ﬁnd that roll calls on national defense issues were signiﬁcantly more likely
than other types of votes to produce observable electoral consequences (p<.065) for
Republican incumbents. This is also an intuitive result, since defense is generally a
Republican issue and therefore we might expect GOP incumbents to be held accountable
by constituents for legislative decisions in this policy area (on this point, see Abramson,
Aldrich, and Rohde 2003). Indeed, the predicted probabilities indicate that roll call votes
on defense issues were 9 percent more likely to yield observable electoral consequences
among Republicans.

In sum, this second part of the analysis offers important evidence that can be used
to build a more comprehensive theory of the conditions under which roll call votes on the
House ﬂoor will transcend intra-chamber politics and become important political issues
in congressional election races. Indeed, I now have some clear evidence suggesting that
both the level of conﬂict and salience of the roll call vote impacts its electoral relevance
for both Democrats and Republicans. Moreover, one can observe that the rate of intra-
party defections and the particular issue area of a given roll call vote impacts whether or
not it would produce observable electoral consequences among Republicans. Thus, these
results have illustrated a number of potential determinants of electorally signiﬁcant roll
calls.

Additional empirical work is necessary to provide more intuition regarding the
dynamic in question, so that reasonable assumptions can be made in the theory-building

process. Moreover, while this approach has broadened the scope of analysis regarding

when roll calls relate to elections, it is important to return to the case study methodology
as a complement in order to gain insights as to why some roll calls and not others are
electorally consequential. By developing a more complete understanding of why some
roll calls and not others have observable electoral effects, I hope to be able to add critical

nuances to how we think about Mayhew’s seminal position-taking argument.

Understanding Directionality Of Roll Call Signiﬁcance

Although it is beyond the scope of the research presented here, I wish to comment
at least brieﬂy on the issue of directionality of the roll call effects. In particular, what are
some of the key insights related to the signs of the coefﬁcients in the main analysis
investigating the frequency of electorally signiﬁcant roll calls? Are members regularly
rewarded or punished for loyalty (or lack thereof) to their party?

The direction of electoral signiﬁcance can go in one of two ways: either in favor
of intra-party loyalists or in favor of intra-party defectors. One hypothesis regarding
directionality is that intra-party defectors (rather than the loyalists) should generally be
the ones who are rewarded at the polls due to signiﬁcant roll calls. The logic behind this
hypothesis is that votes against the party’s position on key roll calls must be highly
strategic decisions by incumbents who anticipate Special reward ﬁ'om constituents for
putting the district ﬁrst on important issues where the party position and the district-need
are disparate. This of course does not suggest that incumbents should regularly defect
from their party in anticipation of reward—alter all, constituents presumably want and
expect a relatively high degree of party loyalty, since the party is a key reason why most

supported the candidate in the ﬁrst place. Rather, the hypothesis suggests that most

45

defections on important roll calls are highly strategic, presumably motivated at least in
part by notable constituency-demand; and therefore they garner special attention from
elites and voters.

When signiﬁcant votes occur, do they generally reward strategic intra-party
defections as hypothesized? A review of the evidence presented in Table 2 suggests that
most signiﬁcant roll calls are in fact cases where intra-party defection is being rewarded.
Indeed, for both Democrats and Republicans, more than two-thirds (68%) of the
electorally signiﬁcant votes are cases of intra-party defectors receiving higher average
two-party vote shares than loyalists, all else equal. This is an important ﬁnding as it
strongly suggests that roll call votes are usually signiﬁcant because party defectors are
rewarded for taking strategic positions against the party’s point of View.

What might explain the remaining 32% of roll calls yielding a Signiﬁcant positive
impact of party loyalty? One hypothesis is that party loyalists are sometimes rewarded
for sticking with the party when a core party issue is facing a relatively large number of
intra-party defections. The results provide some initial evidence to support this assertion
(although I think more detailed investigation of this complicated question—including
case-study analysis—is required). For both Democrats and Republicans, the median
percentage of intra-party defections is higher for roll calls that have a positive impact on
loyalists, as compared with the median percentage of intra-party defectors for roll calls
that have a positive impact on defectors.22 This suggests that when there is a relatively

large degree of intra-party defection, the electoral impact often is not a reward for

 

22 The median percentage of intra-party defectors for roll calls that have a positive impact on Democratic
loyalists is 32%, as compared with 24% for roll calls that have a positive impact on Democratic defectors.
The median percentage of intra-party defectors for roll calls that have a positive impact on Republican
loyalists is 25%, as compared with 19% for roll calls that have a positive impact on Republican defectors.

46

defectors but rather a reward for loyalists who provide the counterbalancing force that is
working to sustain the party’s position on an issue that is important to the party’s agenda.
Many other questions relate to the directionality of the effects uncovered here.
For example, do defections in some policy areas tend to bring electoral rewards while
defections in other areas tend to generate losses at the polls? Do directionality patterns
change over time due to various circumstances? Answering these types of questions in
future research can help us move beyond identifying the frequency and conditions of
electorally signiﬁcant rolls calls to a more thorough understanding of how electorally

consequential roll calls affect incumbents' reelection bids.

Conclusion

The analysis in this paper has been motivated by two central questions pertaining
to the connection between legislative position taking behavior and electoral fortunes.
First, I sought to determine whether positions expressed by incumbent legislators on key
votes yielded observable electoral effects. Second, I attempted to better understand why
some votes, and not others, produced such effects. In both instances, I found evidence to
support my expectations. In particular, the results strongly support the contention that a
wide range of prominent roll call votes have electoral implications for incumbent House
members. The analysis also enables one to begin to identify the conditions (e.g., the level
of conﬂict and salience of the roll call vote) under which position-taking behavior has
electoral implications.

By developing an understanding of the theoretical relationship between roll calls

and elections, we can enhance our capacity to assess how democratic accountability

47

manifests itself in the American political process. In particular, we can identify precisely
how, when, and why legislators are rewarded or punished for position taking. Moreover,
by specifying electoral implications of roll calls, we can gain insight into the underlying
logic of legislative behavior (see, e.g., Mayhew 1974a: 67-69; F iorina 1974; Kingdon
1989; Arnold 1990). Indeed, if incumbents’ roll call choices are driven in part by
whether they anticipate electoral effects, then a comprehensive View of how legislators
formulate these expectations can help us predict when they will be conservative in their
roll call decisions (e. g., vote like they always have or based on what the party wants)

versus when they will engage in “entrepreneurial position taking” (Mayhew 1974a: 68).

48

CHAPTER 3

ELECTORAL ACCOUNTABILITY AND ROLL CALL VOTING IN THE U.S. SENATE

To what extent are members of Congress held accountable for the positions they
express on roll call votes? Despite varying degrees of attention to this important issue in
the scholarly literature (see, e.g., Stokes and Miller 1962; Erikson 1971b; Wright 1978;
Mann and Wolﬁnger 1980; Kingdon 1989; Jacobson 1993), this remains a persistent and
largely unanswered question for students of legislative behavior. While it is clear that
legislators behave strategically when expressing public positions on roll call votes, it is
less clear what effects such position-taking behavior has on legislators’ electoral fortunes.
AS a result, it is now fairly common for scholars to conclude that roll call voting has little
or no observable electoral consequences for legislators, especially Since we rarely see
incumbents in the contemporary era get defeated.

In thinking about the potential impact of roll call voting on electoral fortunes, the
extant literature exploring this relationship has largely been characterized by two
alternative empirical approaches. The ﬁrst approach focuses exclusively on aggregate
voting patterns in Congress to determine the extent to which ideological extremeness on
all or a subset of roll calls produces observable electoral effects (see, e.g., Erikson 1971b;
Bumham 1975; Wright 1978; Erikson and Wright 2000, 2001; Ansolabehere, Snyder,
and Stewart 2001). The more extreme legislators are perceived to be in their overall
voting behavior when compared with their constituencies, the greater impact this
extremeness should have on their vote margins in subsequent elections, all else being

equal.

49

A more recent and alternative empirical approach has focused exclusively on the
electoral effects of a single vote or certain types of votes in Congress. Studies of this
nature have sought to explore whether the positions legislators express on individual roll
call votes signiﬁcantly affect the size of the vote margin by which they are reelected or
defeated (see, e.g., Jacobson 1993; Bianco, Spence, and Wilkerson 1996; Jacobson and
Bovitz 1998). Scholars following this approach assume that controversial positions taken
on highly visible or salient roll call votes in Congress (i.e., those dealing with budgetary
or ﬁscal issues) have a signiﬁcant and measurable impact on the percentage of the vote
incumbent legislators receive.

While the extant literature offers us some preliminary insights concerning the
relationship between legislative behavior and election outcomes, shortcomings with these
approaches limit our understanding of how roll call voting impacts electoral fortunes. In
particular, studies that rely on aggregate measures of legislative behavior or ideology
miss important individual effects since they are unable to distinguish the electoral effect
of one vote from another. Because all or even a subset of votes are combined together, it
becomes impossible to evaluate which speciﬁc votes, if any, are producing Signiﬁcant
effects at the margins for incumbents. At the other extreme, analyses examining the
electoral consequences of one or even a few votes in Congress can at best be described as
case studies from which it is difﬁcult to generalize. More importantly, these latter studies
offer us little or no predictive power concerning whether or not other votes will yield
signiﬁcant electoral effects since they are not taken into account in existing models.

One solution for overcoming these problems in an attempt to more systematically

evaluate the extent to which roll call behavior affects the percentage of the vote received

50

by incumbents involves combining the best features of the previous two approaches. In
fact, recent work by Bovitz and Carson (2000) and Carson (2003) has done just this in the
context of House elections. In particular, their analyses (include a much larger number of
roll calls than previous studies and they explore their effects individually to understand
which, if any, have observable electoral effects. By doing this, they have sought to offer
a more comprehensive account that speaks to the frequency of electorally consequential
roll calls and offers insights into the conditions under which legislative position taking
has electoral implications.

Despite the different approaches adopted in the literature examining the linkage
between position taking on roll call votes and electoral outcomes, previous analyses share
one common factor: each focuses exclusively on the House of Representatives. As such,
we know considerably less about the electoral consequences of roll call voting behavior
in the U.S. Senate. In this chapter, I move beyond the House-centric approaches adopted
in previous scholarship to more systematically examine the representational connection
between substantive positions on roll calls and electoral fortunes in the context of Senate
elections. Building upon and extending previous work focusing exclusively on the
House, I examine the electoral implications of hundreds of “key” roll call votes in the
Senate from the early 19705 to the present. From this, I hope to present new insights
regarding the observable electoral effects of roll call voting in the Senate as well as offer
comparative differences between House and Senate elections.

The organization of this chapter proceeds as follows. I brieﬂy review the ﬁndings
from Bovitz and Carson (2000) and Carson (2003) before turning to a general discussion

on the U.S. Senate. From there, I discuss the data and methodology utilized and present

51

results detailing the frequency of electorally consequential roll call votes in the Senate.
Based on these results, I offer a comparative assessment of the differences in electoral
consequences of roll call voting for the House and Senate. Next, I examine why some
roll calls, and not others, have observable electoral effects. I conclude by discussing the

implications of the results and explore possible avenues for future research.

Prior Findings

Bovitz and Carson (2000) and Carson (2003) explore the electoral consequences
of position taking on roll call votes in the context of House elections. Using “key votes”
as classiﬁed by Congressional Quarterly across multiple congresses,23 they examine the
extent to which positions expressed on a relatively large set of highly visible and salient
roll call votes yield observable electoral effects at the margins for House incumbents.
They focus on key votes in their analyses to assess the extent to which prominent roll call
votes that are reasonably well known to both political elites and constituents will have
observable effects on election day. Although the authors do not claim that key votes
represent the only subset of roll calls that can produce observable electoral effects, they
do suggest that these votes offer a useful starting point for an analysis of electoral
accountability and roll call voting.

Bovitz and Carson (2000) and Carson (2003) ﬁnd that a wide range of roll call
votes in the House produce observable effects at the margins. More speciﬁcally, they

ﬁnd considerable variation in the types of issues yielding discernible electoral effects for

 

23 Each year since 1945, Congressional Quarterly has selected a series of key votes based on what the CO
editors determine to be the “major” issues of the year. In particular, issues are evaluated in terms of
whether they involve a substantial degree of controversy, are a matter of political or presidential power, or
have the potential to have an enormous impact on the lives of the American people.

52

incumbent legislators. For instance, they ﬁnd that not only do budget and appropriations
votes yield observable electoral effects, but that many other issues such as environmental
and foreign policy, transportation, civil rights, and social issues do as well. Indeed, their
results suggest that many votes on these types of issue areas have both a strong statistical
and substantive effect on members’ electoral margins.

Building upon their preliminary results, Bovitz and Carson (2000) and Carson
(2003) also seek to understand why certain roll calls, and not others, yield observable
electoral effects. Toward this end, they draw upon the ﬁndings in their initial analyses to
test empirically whether certain vote-speciﬁc factors inﬂuence the likelihood of a roll call
vote yielding electoral consequences. Using votes that both are and are not electorally
signiﬁcant as their dependent variable, they test a probit model intended to predict
whether roll call votes will yield electoral implications. They ultimately ﬁnd that the
extent to which a particular roll call is controversial, prominently covered by the media,
and subject to intra-party disagreement plays an important role in determining whether or
not it has electoral consequences.

Beyond attempting to address a persistent question in the congressional literature,
extant work on the House facilitates our understanding of both how and when positions
taken on roll calls can yield observable electoral effects. With these results, we now have
a better idea why certain roll calls in the House can play such a pivotal role in the
subsequent election. Although many of the substantive effects of the signiﬁcant House
votes were somewhat marginal, this should not suggest that positions on roll call votes
are without potential consequences. Indeed, these ﬁndings offer empirical support for a

commonly held perception in the context of legislative voting behavior:

53

The question facing a legislator is seldom whether a speciﬁc roll-call
vote might cost him the next election. . .Each issue may have only a
slight impact on a legislator’s electoral margin; yet small effects can
quickly add up to become large effects when summed over the many
issues that Congress considers each year (Amold, 1990, 61-62).

While the ﬁndings in Bovitz and Carson (2000) and Carson (2003) are instructive,
their exclusive focus on the House leaves us with a number of unanswered questions. For
instance, to what extent are their results generalizable to the U.S. Senate? Are senators
more insulated electorally from position taking behavior than their counterparts in the
House? Do roll calls with higher levels of saliency have a similar impact on the electoral
fortunes of senators as they do on representatives? In an attempt to address each of these
questions systematically, this chapter builds upon prior work on the House to extend the
discussion to the Senate. From this, I hope to be able to offer insights into mechanisms

affecting electoral accountability in the Senate as well as offer a comparative assessment

of the differences between the two legislative chambers.

Extending the Analysis to the U.S. Senate

Moving beyond the House-centric approach employed in past research, I examine
the issue of electoral accountability and roll call voting in Senate elections to determine
whether senators are rewarded or punished for their position taking behavior. From a
comparative perspective, Senate elections have received far less scholarly attention than
elections to the U.S. House of Representatives. With notable exceptions in the work by
Mann (1978), Mann and Wolﬁnger (1980), Hinckley (1980a), and Jacobson (1980), most
early analyses of congressional elections focused almost exclusively on the House.

While this emphasis helped illuminate a number of important ﬁndings in the context of

54

House elections, it did little to identify signiﬁcant patterns or trends in association with
Senate elections or allow comparisons to be drawn between House and Senate elections.
Over time, and especially in recent years, Senate elections have begun to receive
slightly more attention by students of congressional elections. For instance, Wright and
Berkman (1986) examined the 1982 U.S. Senate elections to determine the extent to
which policy issues played a role in the selection of senators. Using data from the CBS
News/ New York Times 1982 congressional poll and from 23 statewide exit polls, they
found that Senate candidates behave as though policy issues matter to their constituents
and that voters are systematically responsive to the policy positions of U.S. senators.
This ﬁnding challenged much of the previous research, which had suggested that policy
considerations were of relatively minor importance in congressional elections.

In the early 19905, further emphasis on the Senate helped illuminate a number of
relevant ﬁndings in the context of Senate elections. Most notably, students of elections
began to recognize that Senate elections are considerably more competitive than House
elections, due largely to the greater ntunber of experienced candidates challenging Senate
incumbents (see, e. g., Westlye 1991; Abramowitz and Segal 1992; Krasno 1994). Unlike
most House incumbents who generally face lawyers or other candidates lacking electoral
experience, incumbent senators typically face experienced challengers or well-known
amateurs (Krasno 1994), which tends to increase the challengers’ visibility and makes for

a harder fought and more intense electoral campaign.24

 

2‘ This is especially true given the greater amount of media coverage afforded Senate races since there are
fewer of them in a given electoral cycle (on this point, see Abramowitz 1988; Kuklinski and Sigelrnan
1992). With more media attention devoted to the challenger (and thus more campaign visibility),
incumbent senators have to work substantially harder to get reelected.

55

In addition to campaign intensity, the inﬂuence of money in Senate elections has
also received attention in the context of congressional elections research. In discussing
the impact of campaign spending on Senate elections, Gerber (1998) challenges the
prevailing notion that spending by the incumbent has little or no effect on the election
outcome (see, e.g., Jacobson 1980, 1985; Abramowitz 1988; Grier 1989). Arguing that
much of the previous ﬁndings concerning the impact of incumbent spending are biased
statistically, Gerber adopts an instrumental variables approach in his analysis of Senate
elections. As a result, when the endogeneity issues associated with campaign spending
are statistically controlled for, Gerber (1998: 409) contends that the marginal effects of
incumbent and challenger Spending are approximately the same.25 Thus, he concludes
that balancing the spending levels of Senate candidates could fundamentally alter the
electoral environment for both incumbents and challengers.

Recent work by Kahn and Kenney (1999) has also helped to further illustrate the
implications of the greater competitiveness of Senate elections. In focusing on Senate
races between 1988 and 1992, Kahn and Kenney examine the extent to which campaign
intensity inﬂuences the behavior of the voting population. AS the news media tends to
place a greater emphasis on Senate races (due to their greater competitiveness) than those
held in the House, they ﬁnd that the central themes of both of the candidates come to the
forefront of the political campaign. Consequently, voters become more aware of the
positions advanced by competing candidates and become more knowledgeable when

heading to the voting booth. In the end, more knowledgeable citizens make informed

 

25 Although Gerber (1998) utilizes a two-stage least squares model in his analysis of campaign spending in
Senate elections, I do not adopt his approach in this paper since the relative inﬂuence of campaign Spending
by incumbent senators is not the central focus of this analysis. Whereas incumbent spending was the
central explanatory variable included in Gerber’s analysis, I simply include it as a control variable to
evaluate the degree to which spending by the incumbent offsets position taking behavior.

56

decisions when selecting between the competing candidates, thus increasing the relative
degree of accountability in Senate races.

In thinking about the mechanisms by which roll callvotes translate into vote gain
or loss for incumbents, there are at least four ways in which positions on roll calls might
have a signiﬁcant impact on a member’s vote share. First, constituents who previously
supported an incumbent could decide to withhold support ﬁ'om that legislator based on
their position taking behavior on a series of roll calls. Second, positions expressed on roll
call votes could stimulate turnout among those constituents who are strongly opposed to
the views expressed by the incumbent, thus reducing his or her electoral margin during
the next election (the reverse may also be true—positions expressed on key roll calls
could stimulate turnout among constituents strongly in favor of the Views expressed by
the incumbent, thus enhancing turnout). Third, voters who had not previously supported
an incumbent could change their minds and decide to support a legislator based on his/her
position taking behavior. Lastly, experienced challengers could emerge in response to
the positions taken by the incumbent, thereby providing the voters with a Viable
alternative(s) to the incumbent (thus implying vulnerability) during the next election.

The preceding discussion provides a basis for thinking about issues of electoral
accountability and roll call voting in the U.S. Senate from a purely theoretical Viewpoint.
Based on similarities between the House and the Senate in terms of roll call voting, the
earlier ﬁndings concerning House elections give us related expectations about electoral
accountability in the context of Senate elections. For instance, we should expect to ﬁnd
considerable variation in the types of Senate key votes that yield electoral implications.

While the issues taken up in the Senate may be slightly different than those considered in

57

the House, they should still deal with a diverse array of issues, many of which will garner
sufﬁcient interest from attentive publics (Arnold 1990: 64). Additionally, those persons
most capable of using legislators’ behavior as a political device—namely, party activists,
prospective challengers, and political elites—have ample opportunities to become aware
of key votes and, in turn, transform them into electorally salient political issues.

Given the variation in House and Senate elections presented earlier, we might also
expect differences in electoral implications of roll call voting in the Senate. To begin, the
greater competitiveness of Senate races may result in an increased number of electorally
consequential roll calls for incumbent senators. With a greater proportion of experienced
candidates vying for a limited number of Senate seats, the election campaign may focus
more heavily upon speciﬁc policy issues or positions (see, c. g, Wright and Berkman
1986; Kahn and Kenney 1999). In particular, the positions Senate incumbents express on
prominent roll calls may become salient issues in an upcoming election campaign,
especially if they diverge from the overriding interests of a legislator’s constituency.

We might also expect variation in the electoral consequences of roll call voting in
the Senate due to certain inherent differences between the House and Senate. As
Mayhew (1974a: 66) argues, controversial positions taken on one or even a few isolated
roll calls should not normally have an enormous impact on an incumbent’s chances of
getting reelected. However, Mayhew also asserts that

On rare occasions single roll calls achieve a rather high salience among

the public generally. This seems especially true of the Senate, which

every now and then winds up for what might be called a “showdown

vote,” with pressures on all Sides, presidential involvement, media

attention given to individual senators’ positions, and suspense about the
outcome (Mayhew 1974: 67).

58

Thus, it would seem that if the stakes are high enough, the positions expressed on even a
Single roll call vote in the Senate could spell electoral disaster for an incumbent senator
given the greater scrutiny afforded roll call voting behavior in the Senate. Moreover,
given the smaller size of the Senate, single-vote margins of Victory (or defeat) are more
likely to occur, thus making it easier for voters or political elites to assign credit or blame
for positions adopted on key votes in the Senate.

The nature of the legislative agenda in the Senate might also inﬂuence the extent
to which roll call votes yield observable electoral consequences. Unlike the House, the
Senate routinely deals with treaty ratiﬁcations and presidential nominations (e.g., judicial
and cabinet posts). Given the highly Visible and potentially contentious nature of such
decisions, expressing a controversial position on a roll call could severely undermine a
senator’s chances of getting reelected. In particular, an incumbent senator with a track
record of one or a series of controversial positions could attract an experienced and well-
ﬁnanced challenger in the next election, making it extremely difﬁcult for the incumbent
to retain his or her seat.

Additionally, the differences in members’ constituencies may also inﬂuence the
relative frequency of electorally consequential roll calls in the Senate. Since members of
the House represent roughly the same number of persons per district within a Speciﬁc
geographic region, their constituencies tend to be relatively homogeneous. As a result,
the positions House members express on roll calls may be less likely to stimulate conﬂict
among their constituents due to greater similarities in voter preferences. Senators, on the
other hand, often represent large and more heterogeneous constituencies, making it more

difﬁcult to avoid alienating segments of the population with their roll call behavior. For

59

this reason, we should expect positions expressed by senators on key roll call votes to be
of greater electoral consequence than those recorded by their colleagues in the House.

Moving beyond discussion of the frequency of electorally consequential roll calls,
what types of factors might impact whether roll call positions have observable electoral
implications? One relevant factor may be the underlying degree of conﬂict on the roll
call as measured by the closeness of the vote. Roll calls that elicit controversy in the
form of a close vote may be more likely to stimulate attention on the part of attentive
publics. A related factor in determining the electoral implications of a given roll call
position is the political salience of the legislation under consideration. Many bills
considered in the Senate are brought to the ﬂoor and routinely passed into law with little
notice paid to them by political elites, constituents, or interest groups. However, those
involving highly Visible issues that receive widespread coverage should be the most
likely to produce Signiﬁcant electoral effects.

2 Aside from factors such as the degree of controversy associated with or saliency
of votes in Congress, one must also consider the broader political context under which
issues are brought up in the Senate. To a certain degree, whether or not legislation will
be Viewed as controversial or salient will depend upon the characteristics of the political
environment and the nature of the issue consider consideration. For instance, votes that
occur during an election year may be more likely to yield signiﬁcant electoral effects than
those taken during non-election years. Also, factors such as the occurrence of divided
government or the type of vote under consideration may also play a role in translating

particular roll calls into electorally salient issues.

60

Data and Methods

To a certain extent, students of congressional elections have focused primarily on
House elections since they are easier to study systematically than Senate elections. To
examine the electoral consequences of roll call voting in a given House election, for
instance, one need only examine the voting behavior of representatives for the preceding
Congress since House members come up for reelection every two years. Since senators
are elected for six-year terms, in contrast, it is necessary to examine their voting behavior
for the three preceding congresses if one hopes to determine whether their roll call voting
behavior on key votes impacts their electoral fortunes. Additionally, given that only one-
third of the Senate is up for reelection at any particular time, one needs to recognize that
roll call positions may impact senators differently due to variation in the proximity to the
election.

To address the question regarding the frequency of electorally consequential roll
calls in the Senate, I utilize an Ordinary Least Squares (OLS) multivariate regression
analysis similar to those adopted in the elections literature (see, e.g., Jacobson 1993,
Sellers 1997) and in the previous analyses on the House (see Bovitz and Carson 2000;
Carson 2003).26 The model I employ incorporates the usual predictors of congressional
election outcomes while adding an additional explanatory variable—namely, legislators’
positions on individual key votes. This method facilitates empirical estimation of the
impact of incumbents’ positions on key votes (i.e., the central explanatory variables) on

election returns (i.e., the dependent variable to be explained), while controlling for the

 

26 Unlike previous and related research on the House, I elected not to run the analysis separately for
Democratic and Republican incumbents due to the limited number of Senate elections in a given year (only
one-third of the Senate is up for reelection during any election cycle).

61

effects of other factors known to affect election outcomes (see, e. g., Jacobson 2001). As
modeled, the regression coefﬁcients for each roll call will reﬂect the estimated change in
the incumbent’s share of the two-party vote that results from an afﬁrmative position on
key votes taken in the Senate.

The dependent variable is the incumbent’s Share of the two-party vote in legislator
i ’5 state during year t. The principal explanatory variables are the legislators’ positions
on each of the key roll call votes from year t and t-1 for each of the preceding congresses
(i.e., the roll call votes from both sessions of the three congresses prior to the election).
To isolate the effects of legislative position taking on electoral fortunes, I control for a
variety of factors known to affect incumbent vote share. In particular, I include control
variables for legislators’ electoral security, the underlying partisanship of the state,
incumbent and challenger spending, and challenger quality.

In sum, I posit the following model of legislators’ electoral vote margins:
INCUMBENT VOTE SHARE = 01 + BlElectoral Security + BZState Partisanship + Bglncumbent
Spending + B4Challenger Spending + BsChallenger Quality + [36Voting Record + ByKey Vote + 8

Key Vote represents the position incumbents took with respect to each key roll call vote
and is coded 1 for an afﬁrmative position and 0 otherwise. Electoral Security reﬂects the
percentage of the two-party vote the incumbent senator received in the previous election.
State Partisanship is measured in terms of the popular vote for president received in each
state.

To control for the effects of both incumbent spending and challenger spending,
following J acobson’s seminal work on money in elections, I include the natural logarithm

of both dollar amounts as separate independent variables in the model (Jacobson 1980:

62

40).27 The variable for Challenger Quality is dichotomously coded as a 1 if the candidate
has previously held elected ofﬁce, 0 otherwise. This coding also follows J acobson’s
classic study that views having run a successful elective campaign as a proxy for
candidate quality.28

Additionally, I use Voting Record to control for the legislators’ general voting
tendencies as reﬂected by their ﬁrst dimension DW-NOMINATE scores (see, e. g., Poole
and Rosenthal 1997). These scores provide summary statistics of legislators’ overall
voting tendencies on the traditional liberal-conservative economic dimension. DW-
NOMINATE scores for legislators range from —1 to 1, with low scores indicating policy
liberalism and high scores indicating policy conservatism. By including this variable in
the model, I seek to control for the legislator’s general ideological reputation and, in turn,
isolate the impact of a particular roll call decision that may or may not be consistent with

the incumbents’ broader voting record.29

Exploring the Frequency of Electorally Consequential Roll Calls in the Senate
A summary of the estimated electoral consequences of CQ key votes for the

1974-2000 Senate elections are displayed in Table 4 where I report the percentage of key

 

27 As Jacobson (1980) argues, the advantage of using the natural logarithm of campaign expenditures for
empirical analysis of elections is that doing so avoids the assumption of a linear relationship between
money and votes, thereby accounting for diminishing marginal returns from campaign spending.

28 See Gary C. Jacobson, Money in Congressional Elections (1980): 106-107. For an alternative View on
how to measure challenger quality, see Donald Philip Green and Jonathan S. Krasno (1988).

29 One might argue that controlling for members’ overall voting record as reﬂected by their NOMINATE
score results in a fairly stringent test of the electoral impact of individual roll calls given that roll calls are
in the model twice. However, including both the individual-level vote and the NOMINATE score in the
model does not make it more difﬁcult to ﬁnd signiﬁcant electoral effects Since a single roll call represents a
fairly negligible component of a member’s voting record. To verify this claim, I reran the OLS models for
the ﬁrst part of the analysis without the NOMINATE score and found that the results did not substantively
change. I nevertheless included the legislators’ NOMINATE scores in the model to determine if individual
votes served as a proxy for defections from general voting tendencies.

63

votes that yielded observable electoral effects for incumbent senators.3O As shown in the
table, there is considerable variation in the proportion of key votes that are electorally
consequential across the elections included in the analysis. Nearly one-fourth of the key
votes in the 97th-99th Congresses appear to be signiﬁcant for those senators running in the
1986 election, for instance, compared with only two percent of those prior to the 1978
election showing up as electorally consequential. On the whole, it appears that routinely
between 10-20 percent of Senate key votes yield electoral effects for incumbent senators,

somewhat more than we Should expect by chance alone.

Table 4 - Percentage of CQ Senate Key Votes With Signiﬁcant
Electoral Impact, 93"'-106"I Congresses

 

% Of “Key” Roll Total Number of
Senate Election Calls Signiﬁcant Key Votes Prior to

 

for Senators the Election
1974 Election 11.1 90
1976 Election 14.6 96
1978 Election 2.1 95
1980 Election 13.7 95
1982 Election 9.6 94
1984 Election 15.6 90
1986 Election 24.1 87
1988 Election 18.6 86
1990 Election 18.7 91
1992 Election 3.2 93
1994 Election 13.7 95
1996 Election 12.8 94
1998 Election 14.0 86
2000 Election 21.5 79

 

 

3° I begin with an examination of the 1974 Senate elections since this is the ﬁrst year that comprehensive
campaign ﬁnance for individual Senate races is available.

64

Also of interest from Table 4 is the suggestive pattern that may be a result of
different cohorts of members seeking reelection from one election to the next. Not only
do the results indicate that roll call votes reach varying degrees of electoral salience based
on the underlying political context, but there may also be considerable variation among
incumbent senators in terms of which votes will yield observable electoral effects. This
is a particularly interesting ﬁnding (especially when compared to the House) as it would
appear to conﬁrm the expectation that saliency of roll calls vary both across time and
across individual members.

Turning to more speciﬁc vote information, Tables 5-18 display estimates from the
1974-2000 Senate elections where I report the vote-speciﬁc information for each roll
call.31 Each coefﬁcient reported in the tables represents the estimated impact of voting
“yea” on the corresponding roll call vote. A total of 173 out of 476 (36%) of CO Senate
key votes yielded a statistically signiﬁcant impact on incumbent senators’ electoral
fortunes.32 Due to limited space considerations, I only report coefﬁcients for key votes in

the table that are statistically signiﬁcant in the Senate elections.33

 

3 ' All relevant information concerning bill descriptions, the date the roll call was voted upon, and the vote
tally was taken from Congressional Quarterly’s Annual Series, Congressional Roll Call from 1969-2000.

32 Since the possibility exists that some of the signiﬁcant ﬁndings reported in Tables 5-18 could be due to
“false positives” resulting from similar alignments of voting coalitions among members (particularly in the
Senate as a result of its smaller membership), it is important to touch upon this issue brieﬂy. As it now
stands, the positions that legislators take on a speciﬁc key vote could appear electorally consequential (even
though they actually may not be) simply because the same members that voted together on a previously
signiﬁcant vote also did so on a subsequent vote. Nonetheless, I do not believe that this potential issue
undermines the reported ﬁndings for two main reasons. First, from a purely theoretical perspective, there is
little reason to suspect that the same coalitions of legislators routinely join together and vote the same way
except in those circumstances when their positions are likely to set them apart electorally from other
members. Second, when I conducted an analysis (results not shown) combining senators’ voting patterns
across all Senate key votes into a party unity score (to reﬂect the proportion of times members voted with a
majority of their party) and included it in the original models, the results revealed considerable variability
in signiﬁcance both across legislators and over time. Given that the pattern exhibited sufﬁcient variation
(instead of appearing monotonic), it would appear that the likelihood of obtaining false positives in the
results is minimal.

33 I chose not to report the coefﬁcient results for the control variables in the model due to space limitations.
Overall, the pattern of results for these variables is consistent with extant work on Senate elections.

65

Table 5 — Key Votes Yielding Signiﬁcant Electoral Effects in 1974 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬁicient Robust
St. Error

Tax Reform 91 12/1/69 30-62' -7.35** 3.17
Labor-HEW 91 12/17/69 52-37 -6.66* 3.57
Appropriations
Drug Control 91 1/27/70 35-50 4.79“ 2.61
Voting Rights Act 91 3/12/70 64-17 -9.13** 3.34
Amendments
Revenue Act of 1971 92 1 1/22/71 52-47 6.83“ 3.87
Equal Rights Amendment 92 3/22/72 84-8 -20.10** 3.95
Federal Aid Highway 92 9/19/72 48-26 -5.74* 2.74
Program
Debt Ceiling 92 10/17/72 27-39 -6.48"'* 2.96
Farm Program Extension 93 6/8/73 42-44 -5.82** 1.78
Capital Punishment 93 3/13/74 54-33 -5 .44* 3 .02
Standards

 

Table 6 — Key Votes Yielding Signiﬁcant Electoral Effects in 1976 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬂicient Robust
St. Error

Public Works Acceleration 92 7/14/71 57-36 1019* 5.59
Rehnquist Nomination 92 12/10/71 68-26 10.76" 5.05
Equal Employment 92 2/15/72 45-39 10.88" 3.18
Opportunities Enforcement
Omnibus Education 92 2/29/72 63-34 -9.31** 1.96
Amendments of 1972
Foreign Military Aid 92 7/24/72 42-48 5.70* 3.08
Authorizations
Debt Ceiling 92 10/17/72 27-39 -5.85* 2.70
Second Supplemental 93 5/29/73 55-21 1199* 5.47
Appropriations, 1973
National Energy 93 1 1/15/73 40-48 7.04M 2.78

Emergency Act

66

 

Northeast Rail
Reorganization
Sinai Agreement

Defense Department
Appropriations

Food Stamp Reform Act

Clean Air Act
Amendments

Tax Revision

93
94

94

94

94

12/11/73
10/9/75
12/19/75

4/8/76
8/3/76

8/5/76

37-59
70-18
54-22

52-22
3 1-63

48-42

-9.94**
-4.74*
11.03"

13.77"
6.58"

-6.40*

5.47
2.57

3.44

2.52
2.27

3.32

 

Table 7 — Key Votes Yielding Signiﬁcant Electoral Effects in 1978 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬁicient Robust
St. Error
ERDA Authorizations 94 6/25/76 31-50 -1 7.1 3" 1.69
Waterway User Fees 95 5/3/78 43-47 -3.67* 1.75

 

Table 8 - Key Votes Yielding Signiﬁcant Electoral Effects in 1980 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬂicient Robust
St. Error

Amend Cloture Rule 94 3/5/75 73-21 15.74" 7.38
Open Senate Committee 94 11/5/75 16-77 -18.25** 4.70
Meetings
Defense Department 94 11/18/75 52-47 9.96“ 5.32
Appropriations, 1976
Federal Election 94 3/ 1 6/ 76 46-47 - l 3.90" 6.1 1
Commission
Intelligence Committee 94 5/19/76 31-63 -10.27* 5.18
Defense Procurement 94 5/20/ 76 44-37 -8.63** 4.06
Authorization
Social Security Financing 95 11/4/77 42/41 -12.05"‘* 5.28

 

67

 

Airline Deregulation
Emergency Farm Bill

Fiscal 1980 Binding
Budget Levels

Abolishment of the
Electoral College (Const.
Amendment)

State, Justice, Commerce,
Judiciary Appropriations,
198 1

Fiscal 1981 Continuing
Appropriations

95

96

96

96

4/19/78
3/21/78
9/18/79

7/10/79

9/26/ 80

9/29/ 80

69-23
67-26
55-42

51-48

43-39

47-37

14.14“

-1 1.14”
-10.00**

7.84*

-lO.99*

-10.04**

4.71
2.22
4.06

4.35

5.66

4.69

 

Table 9 — Key Votes Yielding Signiﬁcant Electoral Effects in 1982 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬂicient Robust
St. Error

Public Ofﬁcials’ Integrity 95 6/27/77 74-5 6.92“ 2.26
Act
ERA Deadline Extension 95 10/4/78 44-54 6.26” 2.24
Elementary and Secondary 95 8/23/78 49-47 -5.74** 2.26
Education Act
Amendments
Foreign Aid 96 10/9/79 49-46 10.30" 3.43
Appropriations, Fiscal
1980
Energy Mobilization Board 96 10/3/79 37-56 8.15“ 1.66
Fiscal 1981 Budget Targets 96 5/7/80 64-30 6.60" 2.59
Windfall Proﬁts Tax 96 3/27/80 66—31 -7.27** 3.31
Budget Reconciliation Tax 97 7/22/82 50-47 -9.88** 2.95
Increases
Transportation Assistance 97 12/20/82 56-34 -6.41 ** 2.48

Act of 1982

 

68

Table 10 — Key Votes Yielding Signiﬁcant Electoral Effects in 1984 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬂicient Robust
St. Error

Sen. Herman E. Talmadge 96 10/11/79 81-15‘ 9.43" 3.87
Investigation
Windfall Proﬁts Tax 96 12/12/79 44-53 9.73" 4.32
Trucking Deregulation 96 4/ 1 5/ 80 70-20 -9. l 8" 1.56
Domestic Violence 96 9/4/80 46-41 -7.94* 4.11
Oil Import Fee/Debt Limit 96 6/6/ 80 68-10 -9.01** 2.41
Tax Cuts (Motion to Kill) 96 9/25/80 54-38 -1 1.36" 3.79
Energy and Water Develop. 97 1 1/4/ 81 46-48 7.53" 2.49
Appropriations, Fiscal 1982
Debt Limit Increase 97 2/6/81 73-18 6.77“ 3.17
Urgent Supplemental 97 5/27/ 82 69-23 -10.24* * 2.17
Appropriations, Fiscal
1982
Balanced Budget/T ax 97 8/4/ 82 69-31 -7.44** 3.08
Limitation Amendment
Outside Income of Senators 97 12/14/82 54-38 -5.50** 2.40
Supplemental 98 6/ l 6/ 83 49-47 -5.81** 2.47
Appropriations, (Senator
Salary Increase)
Deﬁcit Reduction 98 4/ l 2/ 84 62-19 -5 .87" 2.23
Deﬁcit Reduction 98 5/17/84 65-32 -10.35*"' 3.87

 

Table 11 - Key Votes Yielding Signiﬁcant Electoral Effects in 1986 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬂicient Robust
St. Error
Energy and Water Develop. 97 11/4/81 46-48 4.22" 1.57
Appropriations, Fiscal 1982
Energy and Water Develop. 97 1 1/4/81 48-46 -4.98** 2.39

Appropriations, Fiscal 1982

 

69

 

Fiscal 1981 Supplemental
Appropriations

Agriculture and Food Act

First Budget Resolution,
1983

Continuing Appropriations,
Fiscal l983/MX Missile

Continuing Appropriations,
Fiscal 1983/Clinch River

Continuing Appropriations,
Fiscal 1983/FT C

Continuing Appropriations,
Fiscal 1983

Emergency Supplemental
Appropriations, Fiscal 1983

Supplemental
Appropriations, Salary
Increase for Senators

Multinational Force in
Lebanon

Labor, Health, and Human
Services, Education

Approp.

Defense Department
Appropriations, Fiscal 1984

Omnibus Defense
Authorization

Motor Vehicle Safety/
Minimum Drinking Age

Continuing Appropriations,
Fiscal 1985

First Budget Resolution,
1986

Superfund Reauthorization,
Fiscal 1986-90

Daniel A. Manion
Nomination

South Aﬁica Sanctions

97

97
97
97

97

97

97

98

98

98

98

98

98

98

98

99

99

99

99

5/20/81

9/18/81
5/21/82
12/16/82

12/16/82

12/16/82

12/16/82

3/11/83

6/16/83

9/29/ 83

10/4/ 83

11/7/83

6/ 14/ 84

6/26/ 84

9/27/ 84

5/9/85

9/24/ 85

6/26/ 86

10/2/86

49-48

44.40
4943
56-42

49-48

59-37

46-50

34-53

49-47

54-46

50—45

37-56

49-48

81-16

51-48

50-49

49-45

48-46

78-21

40.23”

-4.3l**
-10.51**
-7.09**

-5.9l**
-4.11*
-6.87*

13.05"

-4.96**

-10.51**

-1192"

7.17"
-7.67**
-5.81**

6.88”“
-7.99**
-6.15**

-12.43**

-6.l7**

4.39

1.83
4.18
3.54

2.07

2.18

3.66

3.45

2.00

4.18

2.91

2.80

1.89

2.24

3.59

3.71

1.91

3.11

2.32

 

70

Table 12 — Key Votes Yielding Signiﬁcant Electoral Effects in 1988 Senate Elections

 

 

Roll Call Congress Date Vote Result Coefficient Robust
St. Error

First Budget Resolution, 98 5/19/83 50-49 6.14“ 3.34
1984
Debt Limit Increase 98 10/31/83 39-56 4.72* 2.53
Agricultural Programs 98 3/22/84 78-10 -10.20** 3.87
Adjustment Act
Department of Agriculture 98 4/4/84 59-36 -5.33** 1.73
Urgent Supplemental
Approp.
Motor Vehicle Safety/ 98 6/26/ 84 81-16 9.74" 3.62
Minimum Drinking Age
Continuing Appropriations, 98 9/27/84 51-48 -9.68** 4.30
Fiscal 1985
State Department 99 6/7/85 71-13 7.93" 3.75
Authorizations, Fiscal 1986-
87
Aid to Nicaraguan Rebels 99 3/27/86 53-47 8.75“ 4.62
Department of Defense 99 8/5/86 50-49 6.15* 3.20
Authorization, Fiscal 1987 .
Fiscal 1988 Budget 100 5/6/87 56-42 -10.22** 4.37
Resolution
Defense 100 9/17/87 58-38 -8.50* 4.81
Authorization/Missile Tests
Bork Nomination 100 10/23/ 87 42-58 10.22" 4.37
Housing and Community 100 11/17/87 57-43 -6.60* 3.60
Development/Budget
Waiver
Fiscal 1989 Defense 100 5/11/88 48-50 -7.56** 3.23
Authorization/SDI Funding
INF Treaty/Ground- 100 5/25/88 28-69 -7.08*"' 2.78
Launched Cruise Missile
Deﬁnition
Catastrophic Health 100 6/8/ 88 86-11 -l3.50** 1.92

Insurance/ Conference
Report

 

71

Table 13 — Key Votes Yielding Signiﬁcant Electoral Effects in 1990 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬂicient Robust
St. Error

First Budget Resolution 99 5/9/ 85 50-49 -15.07* 7.97
Department of Defense 99 6/4/85 38-57 11.30" 3.06
Authorization, Fiscal 1986
Public Debt Limit 99 10/9/85 75-24 5.58* 2.94
Public Debt Limit/Anti- 99 7/30/86 63-36 6.37" 2.69
Deﬁcit Act
Department of Defense 99 8/5/ 86 50-49 -10.39** 4.71
Authorization, Fiscal 1987
Fiscal 1988 Budget 100 5/6/87 56-42 15.22“ 7.46
Resolution
Temporary Debt-Limit 100 7/31/87 41-52 14.14" 4.14
Increase/Deﬁcit Targets
Senate Campaign Finance 100 9/15/87 51-44 10.41 * 5.42
Housing and Community 100 11/17/87 57-43 10.21 ** 4.97
Development/Budget
Waiver
Catastrophic Health 100 6/8/88 86-11 -7.34* 3.86
Insurance/Conference
Report
Fiscal 1989 Defense 100 8/10/88 49-47 1455* 7.90
Appropriations/Contra Aid
John Tower Nomination 101 3/9/89 47-53 -9.35* 5.09
Fiscal 1989 Supplemental 101 6/1/89 77-18 6.55" 2.37
Appropriations/Budget Act
Child Care/Dole Substitute 101 6/22/89 44-56 -14.50* 7.93
Government Pay-and-Ethics 101 11/17/89 56-43 -4.85* 2.71
Package/Pay Raise
Clean Air Act 101 3/29/90 49-50 -5.59** 2.24
Reauthorization/ Coal Miner
Beneﬁts
Campaign Finance Overhaul 101 7/30/90 46-49 -17.92** 5.48

 

72

Table 14 — Key Votes Yielding Signiﬁcant Electoral Effects in 1992 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬁ’icient Robust
St. Error

Civil Rights Restoration 100 3/22/ 88 73-24 -10.34** 4.09
Act/ Veto Override
Americans w/ Disabilities 101 7/13/90 91-6 15.28M 1.54
Act
Fiscal 1993 Defense 102 8/7/92 43-49 -4.89* 2.70
Authorization/ SDI

 

Table 15 — Key Votes Yielding Signiﬁcant Electoral Effects in 1994 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬂicient Robust
St. Error

Fiscal 1989 Supplemental 101 6/1/89 77-18 -4.l6** 2.09
Appropriations/Budget Act
Catastrophic Revision/ 101 10/6/89 37-62 4.60" 2.07
Durenberger Substitute
Clean Air Act 101 3/29/90 49-50 -7.75** 2.98
Reauthorization/ Coal Miner
Beneﬁts
Omnibus Crime Package/ 101 5/23/90 48-52 -4.48"‘ 2.34
Assault-Style Weapons
Farm Programs 101 7/24/90 54-44 6.01 ** 2.47
Reauth./ Sugar Price
Supports
Fiscal 1991 Defense 101 10/15/90 46-50 -5.50** 2.39
Approp./ Troop Cuts
Crime Bill/Handgun Waiting 102 6/28/91 67-32 -7.58** 3.48
Period
Conditional MFN for China 102 7/23/91 55-44 4.69* 2.42
Clarence Thomas 102 10/15/91 52-48 -12.55** 2.37
Conﬁrmation Vote
Campaign Finance 102 5/13/92 57-42 -1 1.45" 4.58
Family Planning 102 10/1/92 73-26 -7.58** 3.48
Amendments

 

73

 

Cable Television 102 10/5/92 74-25 -6.24** 2.19
Reregulation

Lift U.S. Trade Embargo 103 1/27/94 62-38 -10.15** 2.70
with Vietnam .

 

Table 16 — Key Votes Yielding Signiﬁcant Electoral Effects in 1996 Senate Elections

 

 

Roll Call Congress Date Vote Result Coefﬁcient Robust
St. Error

Use of Force Against Iraq 102 1/12/91 52-47 -4.35** 1.78
Conditional MFN for China 102 7/23/91 55-44 3.97* 1.77
CFE Treaty 102 11/25/91 86-8 4.28“ 1.70
Implementation/ Soviet
weapons dismantlement
Aid for Former Soviet 102 7/1/92 35-60 -3.87** 1.55
Republics/Russian Troops
Fiscal 1993 Defense 102 9/18/92 55-40 -5.22** 2.65
Authoriz./ Nuclear Testing
Moratorium
Campaign Finance 103 6/17/93 60-38 -6.73** 2.51
Abortion Clinic Access 103 5/12/94 69-30 4.24" 2.12
Fiscal 1995 Defense 103 7/1/94 50-50 3.72* 2.02
Authoriz./ Unilateral
Termination
General Agreement on 103 12/1/94 68-32 4.06" 1.85
Tariffs and Trade (GATT)
Shareholder Lawsuits 104 12/22/95 68-30 -4.51** 2.12
Line-Item Veto 104 3/27/96 69-31 -4.76** 2.28
Fiscal 1997 Budget 104 5/23/96 46-53 4.72" 1.51

Resolution/ Centrist
Coalition Alternative

 

74

Table 17 — Key Votes Yielding Signiﬁcant Electoral Effects in 1998 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬂicient Robust
St. Error

Ruth Bader Ginsburg 103 8/3/93 96-3 8.17" 3.32
Conﬁrmation
Fiscal 1994 Defense 103 9/9/93 50-48 -3.90* 2.02
Authoriz./ Strategic
Weapons
Omnibus Crime/Assault 103 11/17/93 56-43 5.41 *"‘ 2.70
Weapons Ban
Fiscal 1994-95 State Dept. 103 1/27/94 62-38 4.59” 1.79
Authorizati on/Vietnam
Abortion Procedures 104 12/7/95 54-44 -4.56** 1.73
Budget Reconciliation 104 8/1/96 78-21 -4.87"‘ 2.55
Welfare Overhaul
Abortion Procedure Ban 104 9/26/96 57-41 -5.26** 1.40
Abortion Procedure Ban 105 5/20/97 64-36 -4.25** 1.88
Fast Track Procedures 105 1 1/4/97 69-31 3.41 "' 1.82
NATO Expansion 105 4/30/98 41-59 -3.55* 1.88
Skilled Worker Visas 105 5/18/98 78-20 6.25" 2.53
“Partial Birth” Abortion 105 9/18/98 64—36 -4.25** 1.88

 

75

Table 18 - Key Votes Yielding Signiﬁcant Electoral Effects in 2000 Senate Elections

 

 

Roll Call Congress Date Vote Result Coeﬂicient Robust
St. Error

Product Liability Overhaul 104 5/4/95 47-52 -2.62** 0.86
Abortion Procedures 104 12/7/95 54-44 -1 1.81" 4.16
Shareholder Lawsuits 104 12/22/95 68-30 7.32" 3.14
Cuba Sanctions 104 3/5/96 72-22 -8.08** 2.79
Fiscal 1997 Budget 104 5/23/96 46-53 6.41" 2.33
Resolution/ Centrist
Coalition Alternative
Campaign Finance 104 6/25/96 54-46 14.03" 3.97
Overhaul
Sexual Orientation Non- 104 9/10/96 49-50 10.37" 3.55
Discrimination
Chemical Weapons Treaty 105 4/24/97 74-26 5.16” 2.56
Abortion Procedure Ban 105 5/20/97 64-36 -10.63** 3.63
Campaign Finance 105 10/7/97 53-47 14.03" 3.97
Overhaul _
Campaign Finance 105 2/26/98 51-48 14.03M 3.97
Overhaul
Same-Sex Military 105 6/25/98 39-53 -4.68** 2.20
Training
“Partial-Birth” Abortion 105 9/18/98 64-36 -10.63** 3.63
Impeachment of President 106 2/12/99 50-50 -14.03** 3.97
Clinton-Obstruction of
Justice
Fiscal 2000 Transportation 106 9/15/99 40-55 9.61 ** 3.91
Appropriations (Fuel Use)
Abortion Procedure Ban 106 10/21/99 51-47 14.03" 3.97
(Afﬁrm Roe v. Wade)
Missile Defense System 106 7/13/00 52-48 -14.03** 3.79

Testing

 

Mp s .05, *p s .10

76

As was the case with the House, there is considerable variation in the types of
issues yielding observable electoral consequences in the Senate. In addition to signiﬁcant
electoral effects associated with votes on appropriations and budgetary issues, we also
observe Si gniﬁcant effects on votes involving agriculture policy, defense and military
issues, educational and environmental policy, health issues, and civil rights legislation.
In contrast to the House (given the differential nature of the legislative process in the
lower chamber), votes on both treaty ratiﬁcations and executive nominations also proved
to be electorally consequential for incumbent senators. Most notably, senators who voted
in favor of conﬁrming Clarence Thomas’s appointment to the U.S. Supreme Court in
1991 and ran for reelection in 1994 suffered nearly a 13% decline in their subsequent
vote margin (compared to members who voted against conﬁrmation) as a result of their
public position on this key vote.

Also of interest from the results is the sizeable electoral impact of many of the
key votes. As hypothesized by Mayhew (1974a), the substantive effect of several of the
key votes reported in the tables can be, at times, quite large. To illustrate this point,
consider the effects of a ‘yea” vote on the Americans with Disabilities Act in the context
of the 1992 Senate elections. Senators who voted in support of this legislation received
an electoral “gain” of slightly more than 15% in the subsequent election when compared
with members voting against this measure. In contrast, senators who supported a ban on
an abortion procedure prior to the 1996 elections experienced an 11% loss in electoral
support, even controlling for factors such as campaign spending and the underlying

support in their constituencies.

77

While these votes initially appear to be of enormous electoral consequence, one
needs to place them in the appropriate perspective to fully appreciate their impact on
incumbent senators’ chances of reelection. In the ﬁrst place, incumbent senators seeking
reelection do not run on the basis of one or even a few votes in Congress, even when
those votes are especially prominent or salient. Instead, they run based on their entire
record, which may involve a considerable number of controversial positions on important
legislation. As such, it is important to keep in mind the combined effect of electorally
consequential roll calls. Just because a particular vote impacts an incumbent’s vote
margin by 10 or 12 percent does not mean that specific effect will be felt at the election
since a member’s chances of reelection are inﬂuenced by his or her combined roll call
voting record. Accordingly, a number of recorded roll call positions could produce
observable individual effects in both a positive and negative direction, yet cancel each
other out when the election rolls around.34

One additional pattern of interest in Tables 5-18 pertains to the proximity of key
votes to the upcoming election and their related electoral signiﬁcance. While some might
argue that the proximity of a vote to the upcoming election should be an important factor
in predicting whether or not votes yield signiﬁcant electoral effects, the results show that
electorally signiﬁcant votes are not conﬁned to the session immediately preceding the
election. Indeed, a considerable proportion of electorally consequential roll calls (74%)
are voted upon in the ﬁrst four years of individual senators’ terms of ofﬁce. This should

not come as a surprise, however, given the cumulative nature of a senator’s voting record.

 

3" At the same time, however, an extended pattern of controversial positions on key votes could have
combined negative consequences in the subsequent election, especially if these effects are not offset by
more electorally beneﬁcial roll call votes. Furthermore, it is still possible for a controversial vote in the
Senate to have an enormous impact on an incumbent’s margin, especially if that vote becomes the focus of
attention in the next campaign.

78

During a span of Six years, a senator may be called upon to express thousands of roll call
positions, many of which may be questioned in his or her home state depending on the
underlying political context. Over time, a senator’s roll call record may become the focus
on his or her next reelection campaign, especially if an experienced challenger emerges to
exploit positions that are “out of step” with the constituency.

Substantively, the preceding results have implications for our understanding of
electoral accountability in the context of House and Senate elections. First, when the
results for the Senate are compared with previous analyses of the House, it appears that
there are a greater proportion of electorally consequential roll calls in the former chamber
than in the latter. This is not entirely unexpected, however, given that Mayhew (1974a)
has argued that we should see more observable effects from position taking in the Senate
than in the House. Additionally, it appears that many of the issues taken up in the Senate
are more likely to attract the attention of political elites, experienced challengers, and the
general electorate due to more extensive coverage of the Senate chamber (Abramowitz
1988; Kuklinski and Sigelrnan 1992). To a considerable extent, senators may be held to a
greater level of accountability than their colleagues in the House for the positions they
express on prominent issues such as key votes.35

Moreover, the results suggest that a sizeable proportion of the roll calls on key
votes in the Senate appear to have rather large marginal effects for incumbent senators

seeking reelection. While previous analyses found evidence of electorally consequential

 

35 To be clear, greater accountability does not imply that senators who take roll call positions that are “out
of touch” with their constituencies will necessarily be removed from ofﬁce. All else equal, incumbents
often have a certain degree of leeway in expressing controversial positions as a result of their comfortable
electoral margins. Thus, while position taking behavior on roll calls can and does produce effects at the
margins, it is somewhat impractical to expect it to exert a direct effect on whether senators are reelected
(especially since so few incumbents are defeated in a given electoral cycle).

79

roll calls in the House, the magnitude of the electoral implications of these votes were
considerably less than those voted upon in the Senate. This ﬁnding strongly suggests that
senators are more likely to be rewarded or punished for their individual roll call behavior
than their counterparts are in the House. As before, this could be a function of increased
media attention of the upper chamber, greater awareness of Senate roll call behavior by
“attentive publics,” or the nature of the issues under consideration in the Senate.
Additionally, these ﬁndings lend support to the contention that “attentive publics”
are not simply attuned to position taking behavior on key votes in the Senate immediately
preceding an election. As was often the case in the context of this analysis, key votes
from early in a senator’s tenure can prove to be electorally consequential. Thus, even
though senators serve four years longer than their colleagues in the House, they do not
appear to be better insulated ﬁ'om electoral considerations as a result of their longer terms
of ofﬁce. In fact, the evidence marshaled in this analysis seems to suggest just the
opposite: senators have every reason to be concerned about the positions they express on

roll call votes since they often do yield signiﬁcant electoral effects.

Explaining Variation in the Electoral Signiﬁcance Across Key Senate Roll Calls
The preceding analysis has illustrated that while a signiﬁcant proportion of key
roll call votes yield observable effects at the margins for incumbent senators, many others
fail to produce such effects. This should not come as a surprise given the vast number of
roll calls voted upon in the Senate in a given year, of which only a limited proportion can
become prominent enough to impact senators’ electoral fortunes. Nonetheless, the earlier

ﬁndings can begin to help us explore an intriguing theoretical puzzle for those seeking to

80

understand why certain roll call votes yield electoral consequences for legislators. After
all, the preceding results indicate that a sizeable number of roll calls from 1969 to 2000
elicited controversy, became salient political issues, and appear to have had signiﬁcant
electoral effects for incumbent legislators.

To begin to understand why certain roll calls—and not others—yield signiﬁcant
electoral effects at the margins for incumbent senators, I draw upon the results from the
previous analysis to test whether certain factors systematically impact the likelihood of a
vote yielding observable effects. How can this analysis move us down the path toward a
theoretical understanding for the conditions under which roll calls will be electorally
consequential? The results from the earlier regression models provide 474 observations
of roll call votes that either do or do not have observable electoral effects for incumbent
senators. Using these observations as the dependent variable, I can test an empirical
model designed to predict whether votes will have electoral implications.

What factors might affect whether a roll call has electoral effects? Drawing on
the earlier theoretical discussion, a number of explanatory variables come to mind. First,
if a particular roll call occurred during an election year it may be better positioned to have
electoral implications due to its proximity to the upcoming congressional election. Also,
if a roll call is a major news story (e. g., it makes the front page of the New York Times),
then it may be especially salient and, in turn, impact electoral fortunes. The type of vote
on which a roll call occurs (e.g., amendment, ﬁnal passage) may also affect its electoral
relevance due to the varying degrees of support on a bill at various stages of the process.
Finally, whether or not a roll call vote is conﬂictual (e.g., 51-49) versus consensual (e.g.,

92-8) may also affect the likelihood of it being electorally consequential.

81

Since the data for this analysis is pooled across multiple elections, it is important
to account for ﬁxed effects (see Kennedy 1998: 226-228). To accomplish this, I include
dummy variables for all but the 9lst Congress as independent variables to control for the
possibility that certain congresses may produce more electorally consequential roll calls
than others. Along the same lines of thought, I also include a control for the occurrence
of divided government in order to better understand what impact, if any, divided
government has on the likelihood of a speciﬁc vote becoming electorally consequential.

In sum, 1 posit the following model to explain whether a key roll call vote is
electorally signiﬁcant or not:

Pr(Roll Call Electoral Signiﬁcance) = or + B.Election Year + BzNew York Times +
B3Conﬂict Level + B4Amendment Vote + BsFinal Passage Vote +
BéDivided Government + B7Congress-Speciﬁc Dummies + e
The dependent variable—roll call electoral signiﬁcance—is coded 1 if the vote on the
roll call has a signiﬁcant electoral effect according to the earlier regression estimates, and
0 otherwise. Given the binary nature of this variable, I estimate the model using probit,
which can accommodate the complexities of estimating a model with a limited dependent
variable.

Each variable in the model is measured as follows. Election year is coded 1 for
years in which a Senate election occurred, 0 otherwise. New York Times is a dummy
variable indicating whether or not the key vote was a front-page story in the New York
Times immediately following the vote in the Senate (thus, tapping a measure of political
saliency for the vote). Conﬂict Level is measured as the proportion of members who
dissent from the majority View on the roll call vote. Votes on Amendments and Final

Passage are dichotomous variables and are coded from the annual series, Congressional

82

Roll Call. Divided government is coded 1 if the Senate and the president were controlled

by opposite political parties, 0 otherwise. Finally, I include a dummy variable for each

Congress, with the exception of the 9lst.

Table 19 — Probit Model of Electorally Signiﬁcant Roll Call Votes
1974-2000 Senate Elections

 

 

Variable Democrats A P"5

Election Year 0.377" .07
(0.144)

New York Times Front-Page Story 1.811" .56
(0.166)

Divided Government -0.855** -.08
(0.43 8)

Level of Conﬂict 0.692" .03
(0342)

Vote on Amendment 0.468M .09
(0.181)

Vote on Final Passage -0.225
(0.204)

Constant -l .881"
(0.443)

Pseudo R2 0.3183

Log-Likelihood 1 56.00

% Correctly Predicted 78.57

Reduction of Error 49.32

Sample Size (N) 474

 

Note: the coefﬁcients are probit estimates; the dependent variable is 1 if the roll call vote is electorally
Signiﬁcant, 0 otherwise; ”p S .05

 

3" The change in predicted probability (A P) is calculated from a baseline model using CLARIF Y (Tomz,
Wittenberg, and King’s 1999) where interval-level variables are held at their mean and dichotomous
variables are held at their modal value. For interval-level variables, changes reﬂect an increase of one
standard deviation. For dichotomous variables, the values reported reﬂect changes from one category to
the other. For more information about the CLARIF Y program, see King, Tomz, and Wittenberg (2000).

83

The results of the model for electorally consequential roll calls in the Senate are
presented in Table 19.37 The probit estimates indicate that roll call votes taken during an
election year were more likely to be electorally consequential. The substantive effect of
this (and other) variable (5) can be estimated via the change in predicted probability of a
roll call yielding electoral effects as the explanatory variable of interest increases one
standard deviation (or moves from zero to one) while holding all other variables at their
mean or modal values. These estimates are displayed in the table (A P) for the covariates
yielding signiﬁcant effects. In this case, roll calls occurring during an election year were
7 percent more likely to yield electoral consequences for incumbent senators.

The results also indicate that votes with high levels of salience as measured by
their appearance on the front page of the New York Times were signiﬁcantly more likely
to yield observable electoral effects for senators. Indeed, the level of saliency as tapped
by this variable exerts the largest substantive effect on the likelihood‘of a vote producing
electoral consequences. All else equal, votes that made the front-page of the New York
Times were 56 percent more likely to result in observable effects. This is an especially
appealing ﬁnding as it conﬁrms the original theoretical expectation that votes with higher
levels of attention and saliency are more likely to be electorally consequential.

The probit estimates in Table 19 indicate that the level of conﬂict has a signiﬁcant
positive effect on whether a roll call will be electorally consequential. More speciﬁcally,
as the level of conﬂict increases on a given roll call, the higher the probability that a roll
call vote will be electorally consequential. While the substantive effect of this variable (3

percent) is rather small, this ﬁnding is nonetheless informative. In particular, it suggests

 

37 To simplify the presentation, I do not report the coefﬁcients and standard errors of the Congress-speciﬁc
dummy variables in Table 19.

84

that the more embattled the Senate is over a particular key roll call, the more likely that
roll call is to help voters distinguish among candidates in the next election. A5 a result,
those votes that stimulate the most controversy among incumbent legislators should be
more likely to become the focal point of a reelection campaign.

In considering the impact of vote type on roll call electoral signiﬁcance, while
votes on amendments are positive and signiﬁcant, votes on ﬁnal passage are negative and
insigniﬁcant. This result is neither surprising nor unexpected, however, given the nature
of the legislative process. As the earlier results suggest, roll calls with higher levels of
conﬂict are more likely to elicit controversy than those that are more one-sided. Since
amendments tend to be more conﬂictual and attract greater opposition from legislators, it
is not surprising that they Should be more likely to be electorally consequential. Votes on
ﬁnal passage, on the other hand, tend to be more lopsided and consensual since many of
the differences in opinion have already been worked out prior to the ﬁnal stages of the
legislative process.

Additionally, it appears that the occurrence of divided government signiﬁcantly
affects the likelihood of a roll call yielding electoral consequences. That being said, the
incidence of divided government actually reduces the chance of a roll call vote becoming
electorally signiﬁcant. This is an intriguing ﬁnding as it suggests that increased conﬂict
between the president and the Senate does not manifest itself in a greater likelihood of
electoral Signiﬁcance of roll calls voted upon in this environment. Perhaps this is best
explained by increased efforts on the part of senators to downplay those votes that attract

greater amounts of opposition from the chief executive.

85

In sum, this second part of the analysis provides a crucial understanding of the
conditions under which one should expect to observe roll calls yielding Signiﬁcant
electoral effects. More importantly, it offers a critical ﬁrst. step in the development of a
broader theoretical framework for understanding why certain roll calls, and not others,
transcend intra-chamber politics and become important political issues in congressional
election races. Indeed, the results presented here suggest that factors such as the saliency
of the vote and the level of conﬂict on a roll call determine its electoral relevance for
incumbent senators. Additionally, votes occurring during an election year and those
taken on amendments are also more likely to become electorally salient during the next
election cycle. Additional empirical work building upon these ﬁndings should allow us
to further reﬁne a theory of the conditions under which roll calls are predicted to yield

observable and signiﬁcant effects for incumbent legislators.

Conclusion

While students of legislative politics have studied the representational connection
between substantive positions and electoral outcomes in U.S. House elections, Virtually
no systematic attention has been given to the same relationship in the context of Senate
elections. This paper takes a ﬁrst step in ﬁlling that void in the congressional elections
literature by examining the electoral consequences of position taking on key votes ﬁ'om
the 1974-2000 Senate elections. In particular, I ﬁnd that considerable variation exists in
the types of issues dealt with in the Senate that yield observable electoral effects. Beyond
votes on appropriations or budgetary issues where one would typically expect special

interest groups (e. g., Citizens Against Government Waste) to hold members accountable

86

for their roll call behavior, I ﬁnd compelling evidence of electoral effects on a range of
issues such as agriculture policy, civil rights legislation, and environmental issues. These
results are especially interesting as they suggest that senators, like their colleagues in the
House, are held accountable for their roll call behavior on a much wider range of issues
than previously believed.

Additionally, the results indicate that the substantive effects of many key votes in
the Senate can often be quite large. In comparison to the marginal effects of key votes in
the House, those in the Senate appear to be substantially larger and more signiﬁcant in
affecting senators’ electoral fortunes. Mayhew (1974a) speculated that this might indeed
be the case in his discussion of the differential effects of position taking in the House and
Senate. As suggested earlier, this ﬁnding is most likely a result of more extensive media
coverage of the Senate where key votes become considerably more salient to “attentive
publics” (Arnold 1990).

The results also reveal substantial variation in the number of Senate votes yielding
observable effects at the margins across cohorts of senators. This ﬁnding is of particular
interest as it suggests roll calls reach different levels of saliency based on which group of
senators is up for reelection. On a related issue, the results imply that incumbent senators
are not only held accountable for the positions they express on roll calls in the two years
prior to the upcoming election. Instead, positions expressed by incumbent senators early
in their tenure often yield signiﬁcant electoral effects, suggesting that their longer terms
of ofﬁce do not safeguard them from electoral concerns.

In examining why certain roll call votes and not others yield signiﬁcant electoral

effects for senators, the results ﬁ'om the ﬁrst part of the analysis provide valuable insights

87

into the mechanisms affecting the electoral signiﬁcance of roll calls. Using these results
as the dependent variable, I determined that factors such as political saliency and the
level of conﬂict on a roll call signiﬁcantly impact its electoral Signiﬁcance. Additionally,
votes that occur during election years and those that deal principally with amendments
are more likely to yield signiﬁcant electoral effects. In sum, these ﬁndings help us begin
to understand why certain key roll calls become electorally salient for incumbent senators
while others do not.

While the emphasis on the Senate has broadened the scope of analysis regarding
the relationship between roll call votes and electoral fortunes, additional empirical work
in this mode is necessary to provide more intuition so that reasonable assumptions can be
made in the theory-building process. For instance, when, and under what conditions, do
incumbent legislators experience tangible beneﬁts or costs with respect to other aspects
of position taking (e.g., ﬂoor addresses, public speeches, press releases)? How might
legislators attempt to explain away controversial positions to their constituents when they
are campaigning for reelection? Also, how effective are experienced challengers at using
incumbents’ position taking behavior against them in subsequent electoral contests? By
focusing on these and related questions, I hope to further enhance our understanding of

the representational connection between incumbent legislators and their constituents.

88

CHAPTER 4
THE IMPACT OF LEGISLATIVE BEHAVIOR ON CANDIDATE COMPETITION IN HOUSE AND
SENATE ELECTIONS: A STRATEGIC CHOICE ANALYSIS

On August 5, 1993, the U.S. House of Representatives voted upon one of the most
important proposals offered by President Bill Clinton in his ﬁrst year of ofﬁce—a ﬁve-
year, $492-billion deﬁcit reduction package. With time running out on the vote and a
216-216 tie displayed on the House scoreboard, the Democratic leadership turned to ﬁrst-
term Representative Marjorie Margolies-Mezvinsky (D-Pa.), one of the only remaining
legislators who had yet to vote. In the previous election, Margolies-Mezvinsky had won
by a narrow margin of 1,373 votes from her suburban Philadelphia district, the ﬁrst
Democrat to be elected in that seat during the past 76 years. As a staunch opponent to tax
increases, she had initially voted against Clinton’s proposal due to the precariousness of
her electoral circumstances, but ﬁnally acquiesced on ﬁnal passage given the pressure
placed upon her and the pivotal nature of the vote. While her support on the vote was
crucial to President Clinton and the party leadership, in the end it cost her reelection in
1994 when an experienced Republican candidate emerged to challenge her resulting in
her defeat by a margin of 8,000 votes (Loomis 1999: 2-3).

The preceding example illustrates how a vote taken on the House or Senate ﬂoor
can have serious ramiﬁcations for an incumbent seeking reelection. While it is rare for a
single vote to have such noticeable consequences for an incumbent legislator in the
contemporary era, the outcome is not a complete surprise given the enormous degree of
Visibility afforded the vote and the underlying competitiveness of the district. Generally

speaking, incumbent legislators face little difﬁculty in getting reelected since they have

89

access to considerable resources, can run on their legislative accomplishments, and often
do not face a quality challenger (Erikson 1971a; Mann and Wolﬁnger 1980; Cox and
Katz 1996; Erikson and Wright 2000, 2001). Nevertheless, if incumbent legislators begin
expressing positions on roll calls out of touch with their constituencies, they may be more
likely to attract the attention of an experienced challenger. Moreover, if an experienced
candidate decides to enter the race, it becomes considerably more difﬁcult for the
incumbent to get reelected (see Jacobson 2001).

Extant research tells uS that candidates with prior electoral experience are more
likely to run when an incumbent retires, is involved in a political scandal, or won the
previous election by a narrow electoral margin (see Jacobson and Kemell 1983, Jacobson
1989, Banks and Kiewiet 1989, Kazee 1994, Maisel and Stone 1997, Welch and Hibbing
1997, Gaddie and Bullock 2000). Building upon these ﬁndings, I examine the extent to
which incumbents support their party’s majority on key roll call votes affects the strategic
emergence of quality challengers in a given election. In other words, are experienced
challengers more likely to run in response to the positions incumbents take on prominent
or salient roll calls? Using House and Senate elections data ﬁ'om the mid-19705 to the
present, I systematically examine this question in this chapter.

Prior research suggests that experienced candidates act strategically in response to
anticipated behavior on the part of incumbent legislators. At the same time, the choice by
a candidate to challenge an incumbent could impact the likelihood that a legislator seeks
reelection (Gates and Humes 1997). Thus, it is important to model this behavior as a
sequential process, with the challenger moving ﬁrst and the incumbent responding to this

initial move. Given this consideration, it is reasonable to assume that some of the same

90

factors inﬂuencing whether a quality candidate enters might also indirectly inﬂuence the
decision of an incumbent to run again or voluntarily retire. That being the case, it is
necessary to consider how factors inﬂuencing these processes may be interrelated. If we
examine these processes separately, we run the risk of model misspeciﬁcation resulting
from a possible selection eﬁ’ect.38 Analyzing this relationship jointly with a selection
model provides an opportunity to statistically control for sample selection bias, thus
avoiding potential threats to statistical inference.

The issue of candidate competition is ﬁrrther complicated by the fact that there is
more than one actor involved in this sequential process and each is operating within a
strategic context. When two or more actors are engaged in a strategic interaction under
uncertainty, it is important and necessary to model such strategic behavior appropriately.
If we ignore the strategic dimension to the interaction between the challenger and the
incumbent, then we increase the probability of introducing bias into our results. Indeed,
Signorino (2002) has demonstrated that parameter estimates ﬁ'om a standard selection
model will be biased if the dependent variable we are trying to explain is generated by
strategic interaction among the players. To fully capture the strategic nature of the
decision calculus involved with challenging an incumbent, this analysis tests hypotheses
derived from a game theoretic model and a test of the model using STRAT, a program for
analyzing statistical strategic models developed by Si gnorino (2001).

This chapter is organized as follows. After reviewing the theoretical literature on

candidate competition in congressional elections, I discuss the motivations for employing

 

38 A selection effect can occur when actors select or are selected into a particular situation by a strategic
process, but the interaction between the actors is assumed to be random. Failing to control for a potential
selection effect can lead to selection bias in the estimation of an empirical model. For a more extensive
discussion of the impact of selection effects, see Heckman 1979; Achen 1986; Signorino 1999; Reed 2000.

91

a strategic choice model in the context of this research question. From there, I formulate
hypotheses and empirical tests of factors inﬂuencing the strategic interaction between
challengers and incumbents. Next, I compare the results generated from the strategic
probit model to those obtained from a standard probit model to show how the latter are
affected by selection bias. The concluding section summarizes the principal implications

of the ﬁndings.

Theories of Competition and Candidate Emergence in Congressional Elections

Scholars interested in candidate competition in congressional elections have spent
considerable energy examining the issue in question and have approached the topic from
a variety of perspectives. While several early studies redirected attention away from an
exclusive emphasis on the incumbent to focus on the role of the challenger in explaining
election outcomes (see, e.g., Mann 197 8; Mann and Wolﬁnger 1980;Hinckley 1980a,
1980b), Kazee (1980, 1983) was among the ﬁrst to study various conditions under which
candidates chose to emerge in House races.39 Drawing upon interviews with individual
candidates, Kazee (1980) found that many candidates ran for Congress Simply for the
experience or joy of running for ofﬁce. In his subsequent work, Kazee (1983) conﬁrmed
that incumbency regularly served as an effective deterrent to potential challengers ﬁ'om
emerging in congressional races.

In their now classic study of challenger emergence, Jacobson and Kemell (1983)
examine whether political candidates exhibit strategic behavior in deciding whether or

not to seek ofﬁce. Through an examination of aggregate patterns of candidates’ career

 

39 For a general discussion of ambition theory that examines the motivations for why individuals choose to
run for ofﬁce or make a career out of politics, see Schlesinger (1966), Rohde (1979), and Abramson,
Aldrich, and Rohde (1987).

92

decisions, they speculate as to the underlying motivations for politicians’ behavior. As
their theory is premised on rational calculations, they argue that experienced candidates
are more likely to run for the House when national and partisan conditions are more
favorable in terms of their likelihood of success. Jacobson and Kemell test their theory
of strategic behavior on data from the 1974, 1980, and 1982 congressional elections and
ﬁnd convincing evidence in support of their hypotheses concerning strategic politicians.
Not only do they conclude that experienced challengers wait until circumstances are
optimal before they decide to run, they also ﬁnd that strategic politicians play a pivotal
role in determining the results of both district-level elections and the overall partisan
composition of Congress.40
Jacobson (1989) offers additional support for the strategic politicians theory by
testing it against elections data from 1946 to 1986. Through his examination of elections
data during this 40-year period, he ﬁnds that experienced challengers do not emerge
arbitrarily. Rather, their likelihood of running varies with their perceived chance of
winning (775). Indeed, Jacobson concludes that a greater proportion of experienced or
quality candidates emerge when prospects appear favorable to their party. As a result, he
argues that strategic decisions by congressional candidates, based on factors such as
likelihood of victory, value of the seat, and opportunity costs, both reﬂect and enhance
national partisan tides. In support of his contention that experienced politicians act
strategically, he recognizes that quality challengers are more likely to emerge when a seat
is uncontested and they rely increasingly on an incumbent’s prior margin of victory as an

important cue in deciding whether or not to run (Jacobson 1989: 778).

 

4° For critiques and extensions of the Jacobson-Kemell strategic politicians hypothesis in the context of
House races, see Bianco (1984), Bond, Covington, and Fleisher (1985), and Born (1986).

93

In an attempt to further discern challengers’ motivations in running for Congress,
Banks and Kiewiet (1989) examine an interesting puzzle regarding the behavior of non-
experienced or weak candidates who emerge to challenge incumbents. While they agree
with Jacobson and Kemell (1983) concerning the deterrent effects of incumbency with
respect to the emergence of experienced challengers, they seek to understand why
incumbency does not have the same effect on weak challengers; as they point out, nearly
all incumbents are challenged from one election to the next, usually by candidates lacking
electoral or political experience. Through their analysis of congressional primary data
from 1980 through 1984, Banks and Kiewiet conclude that weak challengers run against
incumbents for the same reason that strong challengers are more likely to run in open seat
contests—to maximize their probability of getting elected to Congress (1002)."1 Even
though their chances of defeat in the fall election are high, political amateurs recognize
that running against an incumbent affords them the best opportunity to win their party’s
nomination, especially since more experienced challengers are likely to stay out of the
race absent favorable national or partisan conditions.

Beyond these comprehensive and inﬂuential studies of challenger emergence in
House races, other scholars have employed case study methodology to better understand
candidates’ motivations in running for Congress. Fowler and McClure (1989) examine
the behavior of candidates that chose to run and those that did not emerge in New York’s

30th district in the 1984 election to draw a distinction between individual motivations. In

 

4' For an alternative perspective regarding political amateurs’ motivation for running for Congress, see
Canon (1990, 1993). In brief, Canon argues that amateurs would be better off running in open seat contests
since they actually do not maximize their chances of winning by running against incumbents. Canon
(1993: 1130-1138) adds that to fully understand why inexperienced candidates challenge “safe”
incumbents, it is necessary for scholars to distinguish between experience-seeking amateurs and ambitious
amateurs.

94

a series of collected essays, Kazee (1994) presents the work of authors who studied nine
congressional districts during the 1991-1992 election cycle in an attempt to identify why
some candidates chose to run for Congress while others preferred to stay on the sidelines.
More recently, Maisel and Stone (1997) have employed an innovative analysis that seeks
to identify potential candidates across the country in an attempt to ascertain what factors
inﬂuence their decision calculus concerning whether or not to emerge in congressional
races. AS with earlier studies, Maisel and Stone ﬁnd that potential candidates’ decisions
about emergence are most directly inﬂuenced by their perceived chance of success.

What factors do potential candidates consider when evaluating the likelihood of
running a successful campaign? Jacobson and Kemell (1983) and Jacobson (1989) have
systematically Shown that the incumbent’s margin of Victory in the previous election and
the underlying political preferences of the district are important factors in inﬂuencing a
potential challenger’s decision calculus. The decision by an incumbent to forgo another
term in ofﬁce is also an issue of consequence for candidates as evidence suggests that
experienced or quality candidates are more likely to emerge in open seat contests, thus
increasing the level of competitiveness of these races (Bianco 1984; Jacobson 1989;
Banks and Kiewiet 1989; Squire 1989a; Wrighton and Squire 1997; Gaddie and Bullock
2000). The involvement of an incumbent in a scandal could also persuade experienced
candidates to run given their perceptions of Victory would be higher against a legislator
viewed as vulnerable or even prone to retire (Peters and Welch 1980; Groseclose and
Krehbiel 1994; Jacobson and Dimock 1994; Welch and Hibbing 1997).

An additional factor that has received widespread attention in the literature on

candidate emergence is the inﬂuence of campaign money on challengers’ entry decisions.

95

As is often the case, however, there is considerable disagreement over the precise effects
of campaign spending on candidate emergence. On one side of the debate, Goldenberg,
Traugott, and Baumgartner (1986) found evidence that preemptive fund-raising by the
incumbent may deter strong challengers from entering a race. Similarly, Goidel and
Gross (1994), Hersch and McDougall (1994) and Box-Steffensmeier (1996) concluded
that campaign war chests affect the entry decisions of experienced challengers. More
speciﬁcally, they found that incumbents are usually successful in “scaring off” potential
challengers by raising large war chests early in their election campaign.

The previous results notwithstanding, other scholars argue that fundraising efforts
and the accumulation of campaign war chests have virtually no effect on entry decisions.
In their study of House races from 1972 to 1980, Krasno and Green (1988) conclude that
preemptive spending by the incumbent has little effect on candidate emergence when
compared with local conditions in the district. Epstein and Zemsky (1995) employ a
signaling model to evaluate the extent to which incumbent fundraising deters quality
challengers from entering a race. While they ﬁnd that the amount of money raised by the
incumbent can convey information to a potential challenger, they also argue that it does
not deter quality challengers from emerging since the two factors are virtually unrelated
(296). Recently, Goodliffe (2001) has considered the impact of campaign fundraising in
House elections while controlling for selection effects and reached the conclusion that
legislators’ war chests do not deter quality candidates from challenging incmnbents (see
also Milyo 1998; Ansolabehere and Snyder 2000).

While the preceding discussion has focused exclusively on candidate emergence

in U.S. House elections, this is not meant to suggest that this issue has been ignored in the

96

context of Senate elections. Many scholars have noted the increased competitiveness of
Senate races, due largely to the greater proportion of experienced challengers who run for
seats in a given election year (see, e. g., Abramowitz 1988; Squire 1989b; Westlye 1991;
Abramowitz and Segal 1992; Krasno 1994). To test the applicability of Jacobson and
Kemell’s theory outside of its original laboratory, Lublin (1994) evaluates the strategic
politicians hypothesis in the context of the Senate and ﬁnds that potential challengers
consider both local and national conditions when deciding whether or not to emerge in
Senate contests. In exploring the impact of fundraising on candidate emergence in Senate
elections, Squire (1991) found that money raised by the incumbent does not discourage
experienced challengers ﬁom emerging. Moreover, research on the Senate conducted by
Oppenheimer (1996) and Lee and Oppenheimer (1999) has shown that the size of a
state’s population directly impacts the nature of senator-constituency linkages and the
relative competitiveness of the Senate election.42

Since experienced candidates are more likely to run when an incumbent decides
not to seek reelection, it is important to consider factors that inﬂuence an incumbent’s
decision calculus in exploring the central issue of candidate competition. If we assume
that reelection is an important objective for legislators (Mayhew 1974a), then the reasons
for voluntary retirement must be sufﬁcient to deter incumbents from seeking another term
in ofﬁce. Frantzich (1978) examined the issue of voluntary retirements among House
members and found that many willingly chose to leave because they became disaffected
by their service in Congress. Following up on this study, Hibbing (1982a) found that a

greater percentage of incumbents chose to retire in the 19705 as a result of institutional

 

’2 This ﬁnding is related to that of Hibbing and Brandes (1983) who found that senators from heavily
populated states enjoy less of an incumbency advantage than those coming from sparsely populated states.

97

reforms implemented in the House.43 Hibbing (1982b) later discovered that monetary
compensation and advancement within the committee system also played a central role in
inﬂuencing incumbents’ decisions concerning retirement (see also Hibbing 1991).

In investigating the impact of political scandal on incumbent behavior, Groseclose
and Krehbiel (1994) and Jacobson and Dimock (1994) found that legislators involved in
the House banking scandal were signiﬁcantly more likely to retire than seek reelection in
1992. While the opportunity to personally proﬁt from their accumulated war chests also
inﬂuenced their decision calculus, both groups of scholars concluded that the risks of
losing the next election far outweighed the potential beneﬁts of serving another term in
Congress. By and large, this pattern of activity corresponds with other studies of
congressional behavior, which concluded that members involved in personal or political
scandals suffered a signiﬁcant decline in their vote margin in the subsequent election if
they chose not to retire (Peters and Welch 1980; Welch and Hibbing 1997).

A few additional studies of incumbent behavior have enhanced our understanding
of factors inﬂuencing legislators’ decisions to seek reelection or retire from Congress. In
their analysis of career decisions from 1947 to 1986, Kiewiet and Zeng (1993) found that
members facing the prospect of running in a redrawn district were signiﬁcantly more
likely to run for higher ofﬁce if an opportunity presented itself.44 In fact, they concluded
that incumbents only chose to retire if they lacked such an opportunity. Hall and Van

Houweling (1995) examine the 1991-1992 election cycle and conclude that legislators

 

43 Hibbing (1982a) also found evidence that incumbents with ideologically extreme positions on legislation
were more likely to retire voluntarily from the House than were more moderate members.

4’ Box-Steffensmeier and Jones (1997) also consider the role of redistricting in incumbents’ decisions to
retire from the House as well as the impact of serving in the minority party. AS they argue, members of the
disadvantaged party should be more likely to retire given their limited opportunities to inﬂuence policy
outcomes in the House.

98

positioned to exert inﬂuence in the legislative process (e.g., committee chairs and party
leaders) are less likely to retire from the House. They also found that electoral insecurity
plays a critical role in inﬂuencing members of Congress to retire. In a recent study of
voluntary departures from Congress, Moore and Hibbing (1998: 1105) concluded that
incumbents who feel constrained in their opportunity to achieve their objectives are
signiﬁcantly more likely to turn down a chance to serve another term in the House.

With few exceptions, prior congressional research has focused on questions of
candidate emergence or incumbents’ career decisions independently, without considering
the interrelationship between the actors’ decision calculus. While we are beginning to
recognize that experienced candidates are more likely to run when an incumbent decides
not to seek reelection, less attention has been given to the impact of challengers’ strategic
behavior on incumbents’ career decisions. Jacobson and Kemell (1983: 49), for instance,
conclude that “given the remarkable ease with which most [incumbents] win reelection,
few retirees during any given Congress are likely to be driven from ofﬁce by the specter
of defeat.” In their analysis of relationships between House incumbents and challengers,
Ragsdale and Cook (1987) contend that incumbents and challengers behave in ways
relatively independent of one another, rather than basing their decisions on the actions or
behavior of their opponents.

One important consideration that previous studies overlook, however, is that
incumbents may also behave strategically in response to either anticipated or observed
behavior on the part of an experienced candidate. Rather than risk losing in a tough
reelection campaign, incumbents may opt out of the legislative game in response to the

emergence of an experienced challenger. Indeed, Gates and Humes (1997) consider this

99

very scenario when they analyze strategic behavior on the part of both challengers and
incumbents through the lens of a game theoretic model. While they conﬁrm the belief
that experienced challengers may decide not to emerge due to the apparent invulnerability
of an incumbent, they also conclude that “incumbents may choose to leave ofﬁce because
of a potential challenger” (Gates and Humes 1997: 77). AS ﬁrrther evidence in support of
this claim, Jacobson (2001: 154) offers the following disclaimer based on his analysis of
congressional elections: “Incumbents of the disadvantaged party are marginally more
likely to retire when facing the prospect of a tougher-than-usual campaign; the struggle

for one more term may not be worth the effort.”

A Uniﬁed Strategic Interaction Model
The above discussion lends additional support to the View that the decision by an

experienced candidate to challenge an incumbent may impact the choice of a legislator to
run for reelection or retire. As a result, it is important to consider how factors inﬂuencing
these processes may be related. If we suspect that the two processes are interrelated, but
fail to control for it in our analysis, we risk introducing selection bias into our estimates
(Heckman 1979). When two processes are related but estimated separately, correlation
between the disturbance terms can produce inconsistent or inefﬁcient estimates resulting
from a selection effect. To statistically control for selection bias, it becomes necessary to
incorporate the sample selection into the statistical model to avoid biasing the inferences.

Since experienced candidates decide to challenge incumbents in certain instances
and not others, case selection is an important factor to consider statistically in analyses

that examine candidate emergence in congressional elections. In the terminology of the

100

selection literature, an experienced candidate’s decision to enter a race or stay out selects
the incumbent into the sub-sample where the incumbent must decide whether or not to
seek reelection. If candidates choose not to emerge in a given race, their actions can be
considered as nonevents. However, if we fail to systematically examine why prospective
challengers decide not to run in certain races, we risk missing the underlying motivations
for their behavior. Perhaps experienced candidates choose not to emerge because they
suspect their likelihood of winning is relatively small. Or, they may ﬁnd it difﬁcult to
raise the amount of money necessary to unseat an incumbent and hence, decide to stay
out of the race. Regardless of their reasoning, one can only begin to understand
candidate behavior by examining both events and nonevents.

If there are two or more actors involved in a decision calculus, then it is important
to understand the role of strategic interaction in the underlying selection process. Figure
1a depicts a typical situation in which an experienced challenger makes the ﬁrst move by
deciding whether or not to run against the incumbent. In considering this binary choice,
likely challengers condition their behavior on the anticipated reaction of the incumbent
since legislators generally have a large advantage in winning reelection. If they suspect
that an incumbent may opt for retirement instead of face a tough reelection campaign,
however, they may be more likely to emerge in a given race. Likewise, incumbents do
not need to make a choice about whether to seek reelection absent a challenge from an
experienced candidate, unless they have already decided to retire for altogether different
reasons. AS a result, decisions by both an experienced challenger and an incumbent in

this scenario are predicated upon strategic considerations.45

 

‘5 One should note that regardless of which actor moves ﬁrst in the game described above, the same four
outcomes occur when the theoretical model is estimated statistically: (Enter, Run), (Enter, Not Run), (Not

101

Figure la — Strategic Interaction Between Challengers and Incumbents

Challenger

   
   
  

Incumbent Incumbent

Retire Retire

(ES) (ER) (NS) (NR)

As Figure 1a illustrates, experienced candidates have two available choices: they
can enter the race (E) or they can stay out of the race (N) and make it Signiﬁcantly easier
for an incumbent to get reelected. In deciding whether or not to run, strategic politicians
obviously consider a number of relevant factors in their utility calculation. If experienced
challengers anticipate that an incumbent will decide to seek reelection, they will consider
the incumbent’s previous vote margin, the partisan characteristics of the constituency, the
voting behavior of the incumbent, and the legislator’s fundraising ability before making a
decision. If they believe or strongly suspect that an incumbent will not seek reelection,
they should be more likely to emerge since this Situation would Signiﬁcantly enhance

their prospects of success.

 

Enter, Run), (Not Enter, Not Run). Nevertheless, 1 model the strategic interaction with the challenger
moving ﬁrst without loss of generality since the challenger is often looking ahead on the game tree in an
attempt to predict what the incumbent will do. I recognize that the real world Situation is much more
complex than is depicted in Figure 1. However, I utilize this Simpliﬁed model to begin to identify and
understand the factors that inﬂuence the strategic calculus of incumbents and challengers in congressional
races.

102

Once the challenger decides whether to enter a race, incumbents must choose to
seek reelection (S) or retire (R). Absent a challenge from an experienced candidate,
incumbents are in a better position to get reelected since theywill be facing a political
novice in the general election (or no one at all). If an experienced candidate does decide
to run, however, incumbents must weigh a number of different factors in their utility
calculations. In particular, they must consider their position in the chamber, their tenure
in ofﬁce, involvement in a personal or political scandal, their relationship with their
constituency, and their overall voting record before making a decision. If the potential
costs associated with running are too high, then we should expect incumbents to be more
likely to voluntarily exit the chamber.

While the model of candidate emergence above implies strategic interaction on
the part of both the challenger and the incumbent, questions remain concerning how to
capture empirically the strategic interdependence illustrated by the theoretical model. In
an attempt to address this issue methodologically, Signorino (1999, 2002) has developed
a statistical discrete choice model derived from McKelvey and Palfrey’s (1995, 1996,
1998) work on statistical equilibrium concepts for extensive form games that integrates
the underlying strategic dimension among the various actors. This strategic probit model,
as Signorino refers to it, combines elements of the strategic interaction in the selection
process by translating theoretical choice models into statistical tests.

In calculating the utilities for the actors involved in the model, the payoff for each
actor is comprised of two parts: an observable component and a private or unobservable
component. As Signorino (2002) contends, including a private component in the analysis

accounts for limitations encountered by the analyst in specifying the players’ utilities or

103

by the actors themselves in evaluating the player’s preferences over outcomes. Since the
possibility for error exists in each of these situations, inclusion of the private component
minimizes the impact of this error in estimating actors’ utilities. Likewise, “if we assume
that the distribution of the private component is known to all players and is independently
and identically distributed, N (0,62), then the theoretical model can be translated into a
statistical model” (Leblang 2001: 14).

Returning to the previous example, the players’ actions lead to the four outcomes
displayed in Figure la: quality challenger enters, incumbent seeks reelection (ES); quality
challenger enters, incumbent retires (ER); quality challenger stays out, incumbent seeks
reelection (NS); quality challenger stays out, incumbent retires (NR). Now, the players’
actual utilities for these outcomes are represented by UC(ES), Uc(ER), UC(N S), and
Uc(NR) for the challenger and U1(ES), U.(ER), U1(N S), and Ui(NR) for the incumbent.
Since neither the actors involved nor the analyst can assume perfectly observable utilities
as mentioned above, I assmne that private information enters the players’ choice
probabilities presented below. This private information is represented by 1: (the random
error component) for each actor’s utilities. I

In ﬁn'ther assuming the random error component is independently and identically
distributed normal (and this distribution is known to each player), then we need only
calculate the equilibrium choice probabilities for each actor (since the interaction in
Figure 1a is also probabilistic) as well as the strategic equilibrium outcome probabilities
before estimating the statistical model. To begin, the equilibrium choice probability b
that the incumbent decides to retire when an experienced challenger emerges would be

represented as follows (where (D is the standard Normal cumulative distribution):

104

b = Pr[U, (ER) + 7:, > U, (ES) + as]
= P472, —7r,, < U, (ER) — U, (ES)]

U,(ER)—U,(ES)

b: (1) $03,, +0.3...

It can also be shown that the probability that the incumbent retires when an

experienced challenger does not enter, d, is as follows:

d = Pr[U, (NR) + 7:, > U, (NS) + as]

= Pr[7rs — 7:, < U, (NR) — U, (NS)]

U1(NR)-U,(NS)

’ 2 2
(0” NS + 0.3M?

d=<l>

In a similar fashion, we can find the equilibrium choice probabilities for an
experienced challenger. As before, the challenger’s true utility for either entering or not
entering a race has both an observable and a private component. Assuming that this
random error component is IID Normal, then the probability m of a challenger not

entering would be as follows:

m = Pr[UC(N)+;rN > UC(E) +ftE]
= Pr[7rE - 7:, < UC(N) - UC(E)]
= Pr[7r5 — 7:, < dUC(NR) + (1 — d)UC(NS) -bUC(ER) + (1 -—b)UC(ES)]

dUC(NR)+(1 —d)UC(NS)—bUC(ER)+(1 —b)UC (ES)

m = (I) 2 2
J03,” +03%

 

 

105

When we multiply the equilibrium choice probabilities for all combinations of the
two actors, we obtain the requisite equilibrium outcome probabilities as illustrated in
Figure lb. Speciﬁcally, the probability of a challenger entering and an incumbent
seeking reelection (ES) is (l-m)*(1-b). The probability of a challenger entering and an
incumbent retiring (ER) is (1-m)*b. The probability of a challenger not entering and an
incumbent seeking reelection (NS) is m*(1-d). Lastly, the probability of a challenger not
entering and an incumbent retiring (NR) is m*d. Since these outcome probabilities are
functions of a set of explanatory variables and their parameters, it is possible to obtain

maximum likelihood estimates with respect to the parameters using the STRAT program.

Figure lb —Equilibrium Outcome Probabilities

Challenger

   
    
 

Incumbent Incumbent

Retire Retire

(l-m)*(l-b) (l-m)*b m*(l-d) m*d

What advantages does using a strategic choice technique have over a traditional
selection model in analyses of candidate competition? First, this model recognizes the

sequential nature of the choices offered to the actors involved and incorporates strategic

106

interaction into the selection process. Once a candidate decides to emerge, the decision
self-selects the incumbent into the subsarnple where a strategic choice must be made to
seek reelection or retire. Also, a strategic choice model accommodates the complexities
of estimating a selection model where there are two or more actors involved (Signorino
2002). While strategic models are inherently selection models, the latter cannot account
for the strategic interdependence among players, which can result in biased estimates if
improperly applied. Indeed, failing to model the strategic interaction between two or
more actors can lead to a form of misspeciﬁcation far worse than failing to account for
selection bias resulting from correlated disturbances (Signorino 2002).

An alternative strategy for examining candidate emergence in congressional
elections is to employ a maximum likelihood estimation technique such as multinomial
logit where three or more outcomes are estimated simultaneously. This technique is
inappropriate for examining this particular question, however, both because it imposes
the property of independence of irrelevant alternatives (IIA) on the choices available to
the actors (Alvarez and Nagler 1998) and it ignores the sequential nature of the choice
made by the challenger and the incumbent. Multinomial probit, which does not suffer
from the restrictiveness of the IIA assumption, also is unsuitable, as it does not allow for
the sequential nature of the decision process. Only an estimation technique that both
controls for sample selection bias and recognizes the strategic dimension to the choices
available to the actors will minimize the risk of model misspeciﬁcation. In sum, then, a
strategic choice model addresses these methodological concerns by accounting for the
strategic interaction between actors and treating the decision calculus of challengers and

incumbents as a sequential process.

107

Data, Hypotheses, and Methods

To evaluate the extent to which position taking on key votes in the House and the
Senate impacts the emergence of experienced challengers, I go beyond existing models in
the extant literature by including a large sample of roll call votes. In particular, I examine
the effects of Congressional Quarterly (CQ) “key votes” ﬁ‘om 1975-2000 in the House
and Senate to ascertain whether the positions incumbents express on prominent roll call
votes affect the decision calculus of experienced challengers. These roll calls are selected
on the basis of whether the issue at question involves a degree of controversy, is a matter
of political or presidential power, and has the potential to signiﬁcantly impact the lives of
Americans. Ultimately, if experienced candidates are attuned to the position taking
behavior of incumbent legislators on these votes, then we should expect candidates to
behave strategically by emerging in those congressional races where incumbents take
positions that diverge from the interests of their constituencies.

In departing from traditional approaches employed in earlier studies of candidate
emergence, I employ a strategic probit model that captures the strategic interdependence
in both challenger and incumbent behavior. By utilizing a strategic choice model, I can
avoid potential problems associated with selection bias given the sequential nature of the
decision calculus as well as control for the strategic element associated with interaction
by two or more actors under uncertainty. This model also allows me to incorporate
factors inﬂuencing challenger emergence into the model while simultaneously controlling
for variables affecting an incumbent’s decision to seek reelection or retire. Maximmn
likelihood estimates for the parameters, standard errors, and predicted probabilities as

reported in Tables 20 and 21 are calculated using STRAT.

108

The dependent variables for the strategic choice analysis are the four outcomes for
the model presented in Figure la: (ES), (ER), (NS) and (NR). Given the emphasis in this
paper on the inﬂuence of incumbents’ roll call behavior on candidate emergence, the
central explanatory variable is essentially each legislator’s party unity score on selected
key votes. In other words, this is the percentage of time incumbents support the position
taken by a majority of their party on key roll call votes (see, e. g., Canes-Wrone, Brady,
and Cogan 2002).46 This variable is continuous as it is computed as a proportion ranging
from 0 (no support for the party) to 1 (absolute support for the party). To isolate the
effects of legislators’ voting across key votes on challenger emergence, I control for a
number of factors hypothesized to inﬂuence candidates’ entry decisions.

Jacobson and Kemell (1983) and Jacobson (1989) demonstrate that a legislator’s
electoral security in the previous election and the partisanship of the district are central
factors in determining whether or not an experienced candidate emerges to challenge an
incumbent.47 Accordingly, I control for each of these variables in this analysis. The
incumbents’ electoral security is coded as the percentage of the two—party vote each
legislator received in the previous election. To assess the underlying partisan leanings of
the constituency, I include the percentage of the two-party vote the Democratic candidate
received in each district/state in the most recent presidential election (subtracted from the

national vote) as a proxy for political preferences (see, e. g., Brady, Canes-Wrone, and

 

4" I focus speciﬁcally on positions taken on key votes during the ﬁrst session of each Congress since these
are the ones most likely to inﬂuence a challenger’s decision calculus concerning whether or not to emerge.
For the most part, challengers have to make a decision about whether or not to run in a congressional
primary early in an election year, preventing votes that take place in the second session from having little,
if any impact, on their entry decisions.

’7 For a variety of reasons (e. g., insufﬁcient or incomplete data, the failure of the STRAT model to
converge with the inclusion of additional variables), I do not include economic indicators or presidential
approval in the models. In future work, I plan to extend the ﬁndings presented here by incorporating such
variables in related analyses.

109

Cogan 2000; Jacobson 2000; Ansolabehere, Snyder, and Stewart 2001).48 This variable
taps long-term forces such as the partisan composition of the district/state independent of
party support for the congressional candidates.

Although it is important to control for the effect of money to ascertain whether it
has a deterrent effect on challenger emergence in the context of a strategic choice model,
questions remain about how to measure this variable. In examining the deterrent effect of
campaign money on candidate emergence, Goodliffe (2001) collects data on the size of
an incumbent’s war chest on January lst of odd years and includes this variable in his
model. While this operationalization represents one of several alternatives for measuring
the impact of money on a challenger’s decision calculus, it would perhaps make more
sense to collect data on the amount of money in an incumbent’s war chest on January lst
of an election year (to reﬂect fundraising efforts during the previous year). Indeed, it
would be ideal to have a record of the amount of money an incumbent cmrently has on
hand prior to the ﬁling deadline when a challenger has to make a decision about whether
or not to run in a congressional primary.

As an alternative to the above speciﬁcations, 1 control for the amount of money
spent by the incumbent in the previous election to determine whether strong challengers
are discouraged from entering a race by an incumbent’s past record of spending.49 In
general, one should expect that a quality candidate would be less likely to enter a race if

an incumbent has demonstrated an ability to raise substantially large sums of money to

 

’8 For alternative operationalizations of this measure of district/state political preferences, see Abramowitz
(1991); Levitt and Snyder (1995); and Ansolabehere, Snyder, and Stewart (2000).

49 More speciﬁcally, I include the natural logarithm of previous incumbent expenditures to account for
potential “scare off” effects. As Jacobson (1980) argues, the advantage of using the natural logarithm of
expenditures for empirical analyses of elections is that doing so avoids the assumption of a linear
relationship between money and votes, thereby accounting for diminishing returns from campaign
spending.

110

ensure reelection. While such behavior may be perceived as a signal of vulnerability,
only ambitious, experienced, and well-ﬁnanced challengers will be in a position to defeat
an incumbent who has veriﬁed ﬁmdraising prowess (Jacobson 2001). Accordingly, I
utilize incumbents’ past record of spending in the previous electoral cycle as a proxy for
their ability to scare off potential strong challengers.

Given that prior research has shown that open seats are strongly competitive due
to the greater emergence of experienced challengers (Bianco 1984; Squire 1989a; Gaddie
and Bullock 2000), I incorporate this variable into the analysis. Additionally, based on
the work of Lee and Oppenheimer (1999), I control for the size of the state population in
the analysis of candidate emergence in Senate races. In their discussion of the effects of
equal apportionment on Senate elections, Lee and Oppenheimer ﬁnd that the larger the
state’s population, the more competitive the election is likely to be. Since this effect
seems to imply a higher probability of entry by an experienced challenger(s), I examine
whether there is a related connection between population size and candidate emergence
resulting from the greater heterogeneity of more populous states.

Turning to factors inﬂuencing an incumbent’s decision to seek reelection or retire,
Hall and Van Houweling (1995) have demonstrated that legislators who are positioned to
exert inﬂuence in the chamber are signiﬁcantly less likely to retire from Congress. They
also found that as the prospect of electoral security declines, incumbents become more
likely to choose retirement over possible defeat.50 Following their lead, I include control

variables for both status as a party leader and that of committee chair since incumbents

 

5° The inclusion of the incumbents’ previous vote margin and the democratic presidential vote share at the
district/state level will allow me to test the veracity of this claim in conjunction with incumbents’ strategic
career choices.

111

holding such positions should be less likely to retire, all else being equal.51 To examine
possible career effects while also accounting for diminishing marginal rettuns associated
with length of service in Congress, I include a squared term for seniority. Incorporating
this variable into the model tests for the occurrence of “incumbency fatigue,” whereby
constituents become less supportive of legislators the longer they remain in ofﬁce (Lin
and Guillen 1999).

Following up on the work of Kiewiet and Zeng (1993) and Box-Steffensmeier and
Jones (1997), I also include a variable indicating whether a representative’s district in the
House was redrawn following the decennial census and subsequent reapportionment of
House seats. Districts that were substantially redrawn after the census were coded as a 1
while others were coded 0.52 Since Box-Steffensmeier and Jones (1997) and Jacobson
(2001) conclude that members of the minority party are also more likely to retire from
Congress given their limited input in the legislative process, I also control for this
variable in the analysis. Prior to 1995, Republican members in the House are coded as a
1 while all others are coded as a 0; after 1995, Democrats are coded as a l with all others
coded as a 0. For the Senate analysis, I control for Republican control of the chamber
from 1981-1986 and 1995-2001 (in all other cases, they are coded as the minority party).

Additionally, I control for incumbents’ general voting tendencies as reﬂected by

their ﬁrst dimension DW-NOMINATE scores (see, e.g., Poole and Rosenthal 1997). These

 

5 ' Most of this information was coded from Nelson’s (1993) volume outlining committee jurisdictions and
member rosters and supplemented with information from the Ofﬁce of the Clerk for the U.S. House of
Representatives for more recent Congresses. The web site for the House Clerk can be accessed at:
lng/clerkwebhousegov. While it would also be useful to explore the effects of subcommittee chair
status or the prospect of becoming a committee chair in the subsequent Congress on incumbent retirement
decisions, I lack suitable data to examine the effects of these factors more systematically. Such issues are
ripe for future analysis, however, and will be examined when such data become more readily available.

52 This coding includes districts that were redrawn later in the decade as a result of judicial decisions
striking down existing district boundaries as unconstitutional.

 

112

scores, constructed via scaling techniques that use legislators’ entire roll call records to
measure ideology, provide summary statistics of legislators’ overall voting tendencies on
the traditional liberal-conservative economic dimension. DW—NOMINATE scores range
from -1 to 1, with low scores indicating policy liberalism and high scores indicating
policy conservatism. By including the absolute value of this variable in the model, I seek
to control for the legislator’s general ideological reputation and, in turn, evaluate the
extent to which members who adopt ideologically extreme positions on legislation are
more likely to retire.

One ﬁnal variable that I attempt to control for is the involvement of the incumbent
in a personal or political scandal and its observable effects in terms of challenger entry
decisions and legislative retirements. While I lack suitable data on political scandals in
the House, I have been able to incorporate data on political scandals in the Senate. All
else being equal, we should expect an incumbent mired in a political scandal to be more
likely to retire, especially as the seriousness of the charges increase (Welch and Hibbing
1997). An incumbent’s involvement in a scandal may also encourage more experienced
challengers to emerge in a given race, thus further increasing the likelihood of defeat for

an incumbent who chooses to seek reelection.

Empirical Results
Table 20 presents estimates for the strategic probit model examining challenger

and incumbent behavior in U.S. House elections.53 The columns in Table 20 correspond

 

’3 Since the samples in this analysis consist of pooled, cross-sectional data, it is important to consider the
possibility of serial correlation within the estimates. To control for this potential issue, 1 included a series
of ﬁxed effects (i.e., congress-speciﬁc dummies) with the estimation of each of the models. At no time

113

to each of the outcomes for the two players (challenger and incumbent) as identiﬁed in
Figure 1a. Cell entries are maximum likelihood estimates with robust standard errors in
parentheses. Predicted probabilities are denoted in the last column (A P) of the table and
are interpreted in the same manner as with any maximum likelihood model. One should
note that the challenger’s utility for not emerging and the incumbent’s utility for seeking
reelection are represented by constants to avoid problems associated with identiﬁcation.54
As anticipated, the model yields a number of substantively interesting ﬁndings.
First, the hypothesis regarding the impact of party support on the subset of votes included
in the analysis is supported by the ﬁndings, as the coefﬁcient on this variable is both
positive and signiﬁcant. The higher the proportion among incumbents on the party
support index for key votes, the greater likelihood an experienced candidate emerges to
challenge a legislator. In probabilistic terms, for every percentage increase in the
incumbent’s support of the majority of their party’s position, an experienced candidate’s
likelihood of entering increases by nearly two percent. While this effect is small, this
ﬁnding is important as it conﬁrms the expectation that strategic politicians do take notice
of and run in response to the positions expressed by legislators on prominent legislative
issues. It also suggests that members may ﬁnd it in their best interests at times to defect
from their party on salient votes to discourage quality candidates from running against

them.55

 

were these variables individually or jointly signiﬁcant. As a result, the results reported in Tables 1 and 2 do
not include ﬁxed effects.

5’ The STRAT model fails to converge due to problems of identiﬁcation when the same set of variables are
included in more than one set of utilities for a single actor.

55 These conclusions are most clearly supported by the tactics employed by Republican challengers in the
1994 congressional elections. In an attempt to associate Democratic legislators with an unpopular

114

Table 20 — Strategic Choice Model of Competition in House Elections, 1976-2000

 

Challenger’s Utilities

Incumbent’s Utilities

 

 

 

Variable Emerge Not Emerge Retire Seek Reelection A P
Constant 5.633 0.678*
(3.331) (0059)

Party Index Score on 0388* 1.9%

Key Votes (0.148)

Incumbent -0.048* -2. 1%

Expenditures (0.019)

Previous Incumbent -0.004* -2.2%

Vote (0.002)

Democratic Pres. Vote -0.019* -5.8%

Share (0.004)

Open Seat 3.533* 50.7%

(0.142)

Minority Party Status 0112* 3.0%
(0.05 1)

Redistricting 0.206“ 5.3%
(0.098)

Committee Chair 0.130 3.4%
(0.100)

Party Leader -0.083 -1 .9%
(0.147)

Ideology 0.005 0.2%

(DW-Nominate) (0.014)

Seniorityz 0003* 19.6%
(0.001)

N=5666 Mean log-likelihood = -0.683 * Signiﬁcant at p < 0.05

Estimates are calculated via the strategic probit model described in the text and implemented in
STRAT. Standard errors are in parentheses.

 

Democratic president, the Republicans ran a series of television advertisements that had the faces of
incumbents morphing into the face of President Clinton. Additionally, the advertisements linked members’
roll call voting behavior to a number of policy initiatives favored by Clinton in an attempt to publicize their

liberal voting tendencies.

115

The results in Table 20 also indicate that an incumbent’s past record of spending
does play a signiﬁcant role in inﬂuencing candidates’ strategic behavior. The negative
sign on the variable suggests that money spent by the incumbent in the previous electoral
cycle deters experienced candidates from entering a race in the subsequent election. In
particular, the larger the amount of money raised and spent by the incumbent in the
previous election, the less likely an experienced challenger is to emerge in the current
electoral cycle. This ﬁnding is entirely consistent with earlier expectations as it suggests
that strategic politicians use an incumbents’ ﬁrndraising ability in the previous election as
a cue in determining whether or not to emerge. 56 While this result does not directly
refute the ﬁndings of Goodliffe (2001) and others who fail to ﬁnd that incumbents’ war
chests deter quality candidates from entering, it does conﬁrm the existence of a potential
“scare of’ effect associated with incumbent spending within the context of a strategic
choice analysis.

Also noteworthy from the results are the impact of electoral considerations on
candidates’ entry decisions. In particular, the percentage of the two-party vote the
incumbent received in the previous election strongly inﬂuences strategic behavior on the
part of a potential challenger. The larger an incumbent’s previous vote margin, the less
likely an experienced candidate is to enter a race, all else equal. Indeed, for every one-
percentage increase in an incumbent’s previous vote margin, an experienced challenger’s

likelihood of challenging an incumbent declines by slightly more than two percent. This

 

5" In an alternative speciﬁcation of this model, I included a control for whether or not the incumbent ran
against a quality challenger in the previous election, but this variable was not statistically signiﬁcant. To a
certain extent, this result may explain why experienced challengers are less likely to enter a race when an
incumbent spent a large sum of money in the previous election. Even if they had sufﬁcient resources to
mount an effective campaign, they rrright be discouraged from entering the race by the fact that an
experienced challenger was unable to defeat the incumbent in the previous election.

116

is an intuitive ﬁnding that serves to conﬁrm Jacobson and Kemell’s (1983) argument that
strategic politicians use incumbents’ electoral security as a signal when deciding whether
or not to run. As they conclude, strategic politicians are much less likely to run against a
safe incumbent as they realize that their chances of victory are relatively slim against an
otherwise popular legislator.

The underlying competitiveness of the district also appears to inﬂuence a decision
by challenger to enter or stay out of a race. As the share of the Democratic presidential
vote margin (subtracted ﬁom the national total) in the district increases, experienced
candidates become signiﬁcantly less likely to emerge. Probabilistically, a one percent
increase in the Democratic presidential vote share in the congressional district decreases
the likelihood of challenger emergence by nearly six percent. Since this variable taps the
relative degree of competitiveness of the district, potential challengers become less likely
to risk their political futures by challenging an incumbent in a “safe” seat. All else being
equal, we should expect to see more experienced candidates emerge in districts that are
slightly more competitive as reﬂected by close presidential contests at the district level.

When it is announced in advance that an incumbent is not'seeking reelection, the
results indicate that experienced candidates are signiﬁcantly more likely to decide to run.
Indeed, the change in predicted probability reveals that quality challengers are 50 percent
more likely to run when there is an open seat than when an incumbent seeks reelection.
This ﬁnding conﬁrms the conventional wisdom that open seat races tend to be the most
competitive (Gaddie and Bullock 2000). Without an incumbent running, experienced
politicians realize that this is their best opportunity to get elected to Congress. Not only

are they more likely to capture the nomination from their respective party, but they are

117

also in a better position to raise the money necessary to run an effective campaign against
experienced challengers from the opposing party.

Turning to the incumbents’ utility for choosing between running for reelection
and retiring, the estimates tell an interesting story about strategic behavior. As expected,
members of the minority party were more likely to retire from Congress, especially when
they faced the prospect of running against an experienced challenger.57 This result is
quite intuitive as it suggests that members of the disadvantaged party have less incentive
to remain in Congress as their opportunities to inﬂuence legislation are reduced by their
party status. More importantly, the prospect of running against an experienced challenger
from the majority party would make the decision to retire even easier since members
would rather leave ofﬁce voluntarily than risk the potential costs associated with defeat.

The results in Table 20 also conﬁrm that redistricting affects incumbent career
decisions. The positive sign on this variable indicates that representatives forced to run
in a redrawn district are signiﬁcantly more likely to retire, all else equal. As reﬂected in
probabilistic terms, incumbents are about 5 percent more likely to voluntarily leave ofﬁce
when their district boundaries are altered. By and large, this suggests that incumbents are
reluctant to run for reelection when the level of uncertainty increases as to their prospects
for success. Not only must they compete in a “new” district, they also run the risk of
facing an experienced challenger seeking to capitalize on this turn of events. Moreover,
two incumbents may face off against one another in certain instances (when a state loses
a congressional seat), thus ﬁrrther increasing the probability that at least one legislator

opts for retirement instead of facing a tough reelection campaign.

 

57 Indeed, serving in the minority party increases the probability of retirement among incumbent legislators
by approximately 3 percent.

118

In contrast to the ﬁndings reported by Hall and Van Houweling (1995), the party
leadership and committee chair variables do not exert an independent and signiﬁcant
effect in the strategic choice model when controlling for other factors. Although the sign
on the party leadership variable leads one to believe that members serving in this capacity
are less likely to retire from the House, it is impossible to conclude that the effect of this
variable is statistically different from zero. Additionally, the sign on the committee chair
variable is positive, suggesting that members serving in this capacity are actually more
likely to retire when faced with the prospect of running against an experienced candidate.
Nevertheless, it is difﬁcult to draw deﬁnitive conclusions about this variable since it fails
to reach conventional levels of statistical signiﬁcance. What these results do seem to
suggest, however, is that members holding leadership positions may cut their losses and
decide to retire when the prospect of facing an experienced challenger in the next election
becomes reality.

The lack of statistical signiﬁcance on the ﬁrst dimension DW-NOMINATE variable
also makes it impossible to conclude that the ideology of legislators plays a role in their
decision to retire from the House. Although Hibbing (19823) found that ideologically
extreme members were more likely to voluntarily exit the House, this conclusion cannot
be conﬁrmed in the strategic choice model when controlling for other factors impacting
incumbents’ decision calculus. In evaluating the impact of seniority on incumbents’
career decisions, the results do support the argument of diminishing marginal returns for
legislators the longer they remain in ofﬁce. While there is a tangible incentive for
members of the House to accrue seniority, resulting in better placement in the committee

and party leadership structure, incumbents may eventually reach a ceiling whereby they

119

choose to retire. Whether or not this effect is the result of incumbency “fatigue” remains
unclear, but there does appear to be a point when incumbents recognize the beneﬁts of

leaving ofﬁce voluntarily.

U.S. Senate Elections

To facilitate comparisons between House and Senate elections, Table 21 presents
the results from a strategic choice model for Senate contests. As we saw with the House,
the estimates for the Senate provide several substantively interesting ﬁndings regarding
candidate competition. Most notably, the hypothesis concerning the impact of party
support on key votes in the Senate is also sustained by the ﬁndings. Incumbents who
frequently vote with a majority of their party on key votes are more likely to face an
experienced challenger in the next election, all else equal. For every percentage increase
in the incumbent’s support of the majority of their party’s position, the likelihood of
strategic challenger emergence increases by 1.9 percent. Thus, even though Senate
elections are usually more competitive than House elections, strategic politicians in the
Senate do not appear to differ substantially in their propensity to run in response to the
positions incumbents express on key votes. Moreover, senators may be advantaged from
time to time for defecting ﬁom their party on such votes, as it may reduce the likelihood

of facing a quality challenger in the future.

120

Table 21 — Strategic Choice Model of Competition in Senate Elections, 1976-2000

 

 

 

 

Challenger’s Utilities Incumbent’s Utilities
Variable Emerge Not Emerge Retire Seek Reelection A P
Constant 0.5226 1.136*
(2.2824) (0.185)
Party Index Score on 0023* 1.8%
Key Votes (0.009)
Incumbent 0. l 70 2.6%
Expenditures (0.087)
Previous Incumbent -0.066* -4.1%
Vote (0.011)
Democratic Pres. -0.006 -0. 1%
Vote Share (0.010)
Open Seat 2.280* 5.7%
(0.550)
State Population 0004* 5.6%
(in thousands) (0.002)
Minority Party 0411* 2.5%
Status (0.201)
Ethics Violation 1.158* 4.4%
(0.586)
Committee Chair 0.230 1.8%
(0.272)
Party Leader 0.351 2.3%
(0.5 13)
Ideology (DW- 0.642 1.3%
Nominate) (0.762)
Seniorityz 0.001 0.1%
(0.001)
N=465 Mean log-likelihood = -1.039 * Signiﬁcant at p < 0.05

Estimates are calculated via the strategic probit model described in the text and implemented in
STRAT. Standard errors are in parentheses.

121

Whereas the amount of money spent by the incumbent legislator in the previous
electoral cycle deters the emergence of experienced challengers in House races, the same
does not hold for the Senate. Indeed, the results in Table 21 suggest that experienced
candidates in Senate races decide to enter irrespective of the amount of money spent by
the incumbent in the most recent election. In particular, the variable measuring previous
incumbent spending is statistically signiﬁcant, but the sign on the coefﬁcient is positive
rather than negative. This ﬁnding conﬁrms the results presented by Squire (1991) and is
perhaps best explained by the recognition that experienced candidates who challenge
Senate incumbents are typically well connected ﬁnancially, which helps to increase their
chances of winning. As a result, these individuals are less likely to be deterred by the
fundraising prowess of an incumbent since they generally have signiﬁcant resources to
draw upon to organize a competitive election campaign (Krasno 1994).

In considering the impact of state population on the propensity for experienced
challenger emergence in Senate contests, the estimates indicate a strong relationship
between these variables. More speciﬁcally, the results suggest that the likelihood of a
quality candidate entering a race is directly related to the population of the state. The
larger the state’s population, the more likely an experienced candidate is to emerge and
challenge an incumbent (partly as a function of the greater pool of potential quality
challengers in the larger, more populous states). From a candidate’s perspective, running
in a more populous state is preferred since these states tend to be more heterogeneous,
and hence, more competitive. Absent a more homogeneous constituent base, incumbents
face a tougher reelection battle in larger and more popular states, especially if they must

run against a well-ﬁnanced and experienced challenger.

122

As was the case with candidate competition in the House, we also observe that
electoral considerations play a role in shaping strategic politicians’ behavior in Senate
races. In particular, the proportion of the two-party vote incumbent senators received in
the previous election is inversely related to the likelihood of an experienced challenger
emerging. For every percentage increase in the incumbent’s previous vote margin, the
likelihood of strategic challenger entry declines by 4 percent. Once again, incumbent
popularity and strategic considerations on the part of prospective challengers in Senate
races appear to be interrelated. When considering whether or not to challenge a sitting
incumbent, prospective challengers rely on incumbents’ previous vote margins as an
important indicator of vulnerability to evaluate their prospects for success.

While comparatively less attention has been given to the issue of candidate
competition in open seats in the context of Senate elections, it should come as no surprise
that strategic politicians are more likely to run when an incumbent senator voluntarily
exits the chamber. Indeed, the results in Table 21 clearly support this conclusion (albeit
somewhat less so than was the case for the House). In most instances, the experienced
candidates running for open Senate seats are probably current or former House members
who recognize the rarity of the opportunity to win a seat in the upper chamber. Even
though these individuals have to give up their seats in the House to run for the Senate,
progressively ambitious House members most likely conclude that the risks associated
with losing their seats are worth the potential payoff for a position in the Senate.

In contrast to the House, the Democratic share of the presidential vote at the state
level does not appear to be statistically related to the emergence of quality candidates in

Senate races. While the sign on the variable is in the expected negative direction, it fails

123

to reach conventional levels of statistical signiﬁcance when controlling for other factors
in the strategic model. This ﬁnding is most likely attributable to greater heterogeneity
within the states resulting from larger, more diverse populations.

Shifting our attention to the incumbents’ utility in opting for retirement versus
seeking reelection, the results indicate that senators are more likely to voluntarily exit the
chamber when they are in the minority party. Just as we saw with their counterparts in
the House, senators in the minority almost certainly feel disadvantaged by their party’s
status, making the decision to leave the Senate easier—especially when they recognize
that they will be facing an experienced challenger in the next election.58 Moreover, the
results in the table show that senators involved in an ethical scandal or violation are
signiﬁcantly more likely to retire from the chamber. More speciﬁcally, senators involved
in an ethical violation are approximately 4 percent more likely to retire from the Senate
than are other members. Given that senators mired in scandal are more likely to face
defeat at the hands of a quality challenger, they recognize the utility of voluntarily
leaving the Senate to avoid the uphill battle associated with their reelection campaign.

Similar to the pattern in the House, serving as a party leader or committee chair in
the Senate does not deter incumbents from retiring from the Senate. In both instances,
the effects of these variables are not statistically different from zero when we control for
other factors in the strategic model. More speciﬁcally, committee and party leaders do
not appear less likely to retire than rank-and-ﬁle members when faced with the prospect
of a quality challenge in the next election. Furthermore, the variable tapping senators’

ideology fails to reach conventional levels of statistical signiﬁcance, suggesting as before

 

’8 While relatively small, senators serving in the minority are 2 percent more likely to retire from the
chamber than are their colleagues in the majority.

124

that ideologically extreme members of the Senate are not necessarily more like to retire
than their more moderate colleagues. Finally, we observe that seniority does not exert an
independent or signiﬁcant impact on incumbents’ strategic career choices, making it

difﬁcult to draw deﬁnitive conclusions about the effect of this variable.

Comparing Results From the Strategic Choice Model to a Standard Probit Model
One of the central arguments in this paper is that failing to model the underlying
strategic relationship between a challenger and an incumbent leads to biased or inaccurate
estimates. To illustrate the validity of this argument, I reran each of the models presented
in Tables 20 and 21 with a standard probit technique that is not designed to capture the
strategic dimension in each of the actors’ utilities. Through a comparison of these results
to the original STRAT estimates, I seek to demonstrate exactly the type of bias that can
occur when failing to control for strategic behavior on the part of one or both actors. The
results from the probit models for both the House and the Senate are reported in Table 22.
A cursory glance at the results in Table 22 reveals several noteworthy ﬁndings.
First, and perhaps most importantly, the party index score variable is insigniﬁcant in the
estimates for both the House and the Senate models. In other words, simply estimating
the decision by a challenger to emerge in a given congressional race without controlling
for strategic interaction would lead one to believe that experienced challengers do not pay
attention to an incumbents’ voting pattern on key votes. To some extent, this ﬁnding
may help illustrate why other scholars have failed to identify the effect of legislative roll
call behavior on candidate emergence in congressional races. Only by estimating this
relationship in the context of a strategic choice model, however, do we obtain results that

are consistent with the underlying theoretical expectations presented earlier in the paper.

125

Table 22 — Probit Model of Competition in House & Senate Elections, 1976-2000

 

 

 

House Elections Senate Elections
Variable Challenger Incumbent Challenger Incumbent
Party Index Score 0.049 ' 0.009
on Key Votes (0.145) (0.006)
Incumbent 0.053* 0.088
Expenditures (0.022) (0.069)
Previous Incumbent -0.006* -0.037*
Vote (0.001) (0.006)
Democratic Pres. -0.009* -0.006
Vote Share (0.002) (0.008)
Open Seat 1.969* 1.330*
(0.074) (0.422)
State Population 0.001
(in thousands) (0.001)
Minority Party 0.022 0.084
Status (0.048) (0.154)
Redistricting 0.025
(0.059)
Committee Chair -0.094 -0.160
(0.1 16) (0.206)
Party Leader -0.529* -0.036
(0.217) (0.265)
Ideology -0.387* 0.003
(DW-Nominate) (0.129) (0.368)
Seniorityz 0001* 0001*
(0.000) (0.000)
Ethics Violation 1.290*
(0.419)
Constant -0.892"‘ -1.293 * 1.018 -1.072*
(0.298) (0.053) (1.124) (0.177)
Log-Likelihood -1878.01 -1780.32 -262.81 ~229.9l
N 4304 5666 456 465

 

Estimates are calculated using probit. Standard errors are in parentheses. * Signiﬁcant at p <0.05

126

Additionally, the results in Table 22 suggest that the amount of money raised and
spent by the House incumbent in the previous electoral cycle does not deter experienced
candidates from entering. While the expenditures variable is statistically signiﬁcant, the
sign on the variable is positive instead of negative, leading one to believe that the more
money spent by the incumbent in the previous election, the greater chance that he or she
would face an experienced challenger in the subsequent election. As before, this result is
opposite that reported in the strategic choice model (for House races) and contrary to the
underlying theoretical story in this paper. What makes this ﬁnding of particular interest
is that it may help shed light on the ongoing debate over the effects of campaign spending
on deterring quality challengers from emerging. Without the appropriate control for
strategic interaction, one might be falsely led to believe that money raised and spent by
the incumbent is not a sufﬁcient condition for deterring challenger entry, even though it
actually may be.

Beyond the differences in the probit estimates for the challengers’ utilities, one
can also observe a number of dissimilarities between the two models when focusing on
the incumbents’ decisions to seek reelection or retire. The variables indicating minority
party status and redistricting are not signiﬁcant in the probit model for House elections,
leading one to infer that these variables do not have a positive effect on an incumbent’s
decision to retire from the House when forced to run against an experienced challenger.
In general, these results are not consistent with the estimates obtained from the strategic
choice model nor do they coincide with the theory and evidence presented earlier. Thus,
these ﬁndings provide further evidence that failing to control for strategic interaction

between challengers and incumbents can signiﬁcantly bias our results.

127

The estimates in the standard probit model also reveal differences in connection
with the variables tapping leadership status in the party and member ideology. Most
notably, the variable measuring party leadership status is both negative and signiﬁcant in
the probit model, suggesting that party leaders are less likely to retire, even when they
may face an experienced candidate in the upcoming election. While this result may have
a certain intuitive appeal, it is not supported by the estimates generated in the context of
the strategic choice model. Additionally, the results from the probit model seem to
suggest that more ideologically extreme members are less likely to retire when facing a
quality challenge. From a purely substantive perspective, this ﬁnding is somewhat
puzzling since we should generally expect the opposite to be true—the more extreme a
member is ideologically, the more likely that incumbent may be to invoke a challenge

from an experienced candidate, all else being equal.59

Conclusion
The results in this chapter serve to both challenge and conﬁrm several insights in
the extant literature on candidate competition in congressional elections. First, I have
shown that experienced challengers decide to run in situations other than when legislators
exhibit signs of vulnerability or decide to opt out of a tough election campaign. Strategic
politicians run against incumbents who increasingly vote with their party on prominent
roll calls. This is an important ﬁnding as it reveals that there are electoral consequences

associated with toeing the party line on key votes.6O Moreover, this result conﬁrms the

 

’9 This presumes, of course, that the incumbent does not represent an ideologically extreme constituency.
In all other cases, we should expect this general principle to hold.

6° Mayhew (1974: 99) reaches a similar conclusion in discussing the electoral beneﬁts of position taking.
“What is important to each congressman, and vitally so, is that he be free to take positions that serve his

128

assumption implicit in congressional elections research that candidates are attuned to the
position taking behavior of incumbents in both the House and the Senate.

In addition to showing that electoral security and the. partisan characteristics of the
constituency inﬂuence strategic challenger emergence, I also demonstrate that the amount
of money raised and spent by the incumbent in the prior election can deter experienced
challengers from running. While I am not able to speak directly to the debate concerning
deterrent effects of campaign war chests, the results do suggest that strategic politicians
can be scared off by an incumbent’s ability to raise and spend large sums of campaign
money. Moreover, from the perspective of rational choice theory, this ﬁnding has a
certain intuitive appeal. If we assume that legislators are rational actors and we know
that they dislike raising large sums of money, why would they continue to repeat this
cycle each election if money had little or no measurable impact on strategic challenger
entry?“ On the other hand, if we assume that legislators do not habitually squander their
limited time or scarce resources, then it is reasonable to conclude that they are making
rational calculations when raising money for their reelection campaigns.

The results in the preceding models also demonstrate that incumbents’ career
decisions are inﬂuenced by a variety of external and institutional factors. Indeed, serving
in the minority party, running in a radically redrawn district, involvement in a political

scandal, and length of service in the chamber can all encourage incumbents to retire from

 

advantage. There is no member of either house who would not be politically injured—or at least who
would not think he would be injured—by being made to toe a party line on all policies (unless of course he
could deterrrrine the party line). There is no congressional bloc whose members have identical position
needs across all issues.”

6' To argue that incumbents raise and spend large sums of money because it has an impact on winning
would not be inconsistent with my argument. Indeed, incumbents often win because they face an
inexperienced challenger or no challenger at all. If raising large sums of money has any deterrent effect on
challenger entry, however, then we face an observational equivalence problem of determining why
incumbents are winning.

129

their respective chambers. In contrast, serving as a committee or party leader does not
appear to be sufﬁcient to deter members from retiring, all else equal. At the same time,
legislators representing the extreme wing of either party arenot more likely to retire
when compared with their more moderate colleagues in the chamber. Equally important,
the results conﬁrm my expectations that strategic behavior on the part of experienced
candidates can inﬂuence the decision calculus of incumbents when choosing between
seeking another term in ofﬁce and retiring.

The ﬁndings in this chapter also have broader methodological implications for
analyses examining strategic behavior between multiple actors. As illustrated by the
model comparisons in the preceding section, many of the results reported in this paper
almost certainly differ from those in the extant literature as a result of including effective
controls for selection bias and strategic interaction between challengers and incumbents.
As we investigate additional and related questions involving strategic interaction between
multiple actors (e. g., members of Congress and the president, party leaders and rank-and-
ﬁle members), we should be wary of the methodological issues raised here. Research
continuing in the application of the strategic choice model demonstrated in this paper
should be able to offer new insights that seek to further enrich our understanding of

electoral and political behavior.

130

CHAPTER 5

CONCLUSION

In his discussion of the usefulness of position taking as an electoral strategy for
members of Congress, Mayhew (1974a: 62) underscores the fact that “the congressman
as [a] position taker is a speaker rather than a doer.” Legislators are not always required
to pass laws or write legislation that necessarily results in favorable policy outcomes.
Instead, legislators can reap electoral rewards merely by creating the appearance of
maintaining a connection with their constituencies through an effective use of position
taking strategies. In other words, the most effective electoral strategy for legislators may
simply involve staking out a position on an issue that voters ﬁnd agreeable. As Mayhew
(1974a: 62) maintains, “[t]he position itself is the political commodity.”

This dissertation has explored the electoral consequences of one aspect of position
taking in Congress—namely, that of roll call voting. 1 focus speciﬁcally on the electoral
effects of roll calls in the House and Senate since this is the one area of position taking in
which legislators are unable to avoid expressing positions that may be divisive to voters.
“Outside the roll call process the congressman is usually able to tailor his positions to suit
his audiences” (Mayhew 1974a: 63-64). However, with roll call votes, “there is no way
for a member to avoid making a record on hundreds of issues, some of which are
controversial in the home constituencies” (Mayhew 1974a: 65-66). I also examine the
electoral impact of positions expressed on “key votes” as identiﬁed by Congressional
Quarterly since these votes often deal with prominent or salient issues that attract the

interest of political elites and “attentive publics” (Arnold 1990).

131

Since the positions expressed on prominent or salient roll calls may yield effects
at the margins for incumbents, I argue that is important to consider the mechanism by
which these positions generate observable electoral effects. _While it is difﬁcult to argue
convincingly that voters regularly go to the polls with complete knowledge of how their
representatives voted on even the most prominent legislative issues, it is reasonable to
conclude that political elites draw attention to incumbents’ voting behavior on individual
roll calls. In discussing the strategic calculations associated with position taking, Arnold
(1990) equates the behavior of challengers and incumbents to a game played between the
competing actors:

Challengers are perhaps the most diligent players in this game. Few

challengers fail to sift through incumbents’ records in search of the

smoking gun. They then employ their newly discovered evidence to

persuade citizens how poorly their current representatives have served

their interests (49).
Moreover, Arnold (1990: 49-50) contends that other disenchanted groups may highlight
speciﬁc “inconsistencies” in legislators’ position taking behavior, particularly if they
suffered as a result of a member’s pattern of voting.

As each of the essays in this dissertation suggests, legislators behave strategically
in an attempt to avoid “mistakes” on roll call positions that may have detrimental effects
in the next election. Whether or not this strategic behavior will have observable electoral
effects is to some extent beyond the incumbents’ control, however. While legislators are
partly responsible themselves for translating positions into viable political commodities,

the ultimate impact of their positions is determined to a considerable extent by decisions

in the media, among special interests, and other political elites (e. g., quality challengers)

132

who are able to transform roll call positions into electorally consequential issues. Thus,
even though voters themselves may not be aware of speciﬁcally how legislators voted on
individual roll calls, there are still opportunities for awareness manifested by “attentive
publics” who attempt to draw attention to the roll call voting patterns of incumbents in

both the House and Senate.

Summary of Findings

In terms of my ﬁndings, the ﬁrst essay examines the effects of position taking on
key roll calls in the House of Representatives. In considering the frequency of electorally
consequential roll calls in the House, I found considerable variation in the number of key
votes that yielded observable electoral effects for incumbent legislators. Additionally,
there were sufﬁcient differences in the types of issues yielding electoral effects as well as
notable variation across congresses in the number of statistically signiﬁcant key votes.

At times, the signiﬁcance of these votes was notable in both a statistical and substantive
sense, particularly on key votes where members appeared to be earning electoral rewards
for defecting from a majority of their fellow partisans.

In examining why certain key roll calls yielded observable electoral effects while
others did not, the second part of the analysis explored this issue systematically. Using
the ﬁndings from the ﬁrst part of the analysis as the dependent variable, I tested a probit
model that examined whether positions on key votes would yield electoral consequences
for incumbent House members. As expected, factors such as the degree of conﬂict on the
original vote, the overall saliency of the issue, and the presence of intra-party conﬂict all

strongly predicted which key votes would be electorally consequential. Also, differences

133

among Republicans and Democrats manifested themselves in this aspect of the analysis,
suggesting that the factors that affected the electorally saliency of key votes is not always
the same for members of the rival political parties.

The analysis in the second essay largely mirrors that conducted in the ﬁrst, but
shifts the central focus to the U.S. Senate. While the representational connection between
position taking behavior and electoral outcomes has been given limited attention in the
context of U.S. House elections, a related examination has not been carried out in terms
of Senate elections. Thus, my analysis of electoral accountability in Senate elections in
the second essay seeks to ﬁll this void in the congressional elections literature. In terms
of the results, many of the reported ﬁndings parallel those for the House. For instance,
considerable variation exists in the types of Senate issues that yield observable electoral
effects as well as the number of electorally signiﬁcant key votes across cohorts of
senators. In contrast, the results also suggest that the substantive effects of many key
Senate votes can often be quite large and many votes that occur early (versus later) in a
senator’s tenure eventually prove to be electorally consequential.

Perhaps the most interesting comparison to be drawn between the analysis of
electorally consequential votes in the House and the Senate involves the distinction in
electoral safety of the membership in both chambers. Contrary to the conventional
wisdom, senators do not appear to be more insulated ﬁom electoral considerations as a
result of their longer terms of ofﬁce. In certain respects, senators are more vulnerable
than their counterparts, due partly to the greater competitiveness of Senate elections
(fewer uncontested seats and more experienced challengers rtuming in a given election

year). This is, in part, a ﬁrnction of the greater attention afforded legislative behavior in

134

the Senate, a process that Mayhew (1974a) alludes to in his discussion of the potential
electoral effects of position taking in the Senate versus the House. While senators were
surely more insulated from popular sentiment prior to the passage of the Seventeenth
Amendment (direct election of senators), this no longer appears to be the case under the
existing rules for elections in the upper chamber.

In examining why certain votes and not others yield observable electoral effects
for incumbent senators, many of the results are consistent with those of the House. As
we saw with House votes, factors such as political saliency and the level of conﬂict on a
roll call signiﬁcantly impact its electoral signiﬁcance. Beyond these factors, however,
we also observe that key votes occurring in election years and ones on amendments are
more likely to produce signiﬁcant electoral effects. In conjunction with the ﬁndings for
the House, this part of the analysis helps us better understand theoretically important
factors that explain why a number of prominent roll calls yield effects at the margins for
incumbent senators.

The third and ﬁnal essay systematically examines whether experienced candidates
regularly scrutinize the position taking behavior of legislators in the House and Senate
when deciding if they should emerge in a given race. Unlike much of the extant literature
focusing on issues of candidate emergence, I focus on the degree of strategic interaction
between the challenger and the incmnbent in this process. I employ a strategic choice
model to analyze the extent to which the behavior of both incumbents and challengers are
premised on strategic considerations. While we are aware that challengers may elect to
stay out of race if conditions are perceived as less than favorable for them to win (see,

e. g., Jacobson and Kemell 1983), we have less information about the strategic behavior

135

of incumbents when experienced candidates decide to emerge and challenge them. In
this essay, I systematically test a number of hypotheses related to candidate competition
in congressional races using STRAT, a program designed by Curtis Signorino to evaluate
and test strategic selection models.

The results from the strategic model yielded a number of substantively interesting
ﬁndings. First, experienced challengers’ entry decisions are signiﬁcantly related to the
position taking behavior of legislators in both chambers. The more often members vote
with their party on key votes, the more likely they are to face an experienced challenger
in the subsequent election. Furthermore, the amount of money raised and spent by the
incumbent in the previous appears to deter challengers in House races, but not in Senate
races. When we shift to the incumbent side of the equation, the emergence of a strong
challenger can effectively deter an incumbent from seeking reelection, particularly if the
incumbent was involved in a scandal or won the previous election by a rather slim
electoral margin. Moreover, holding a position of prestige (committee or party leader)
does not necessarily deter incumbents in either the House or the Senate ﬁ'om retiring if
they face a tough reelection battle against an experienced challenger.

In sum, the results reported in this dissertation shed new light on a ﬁrndarnentally
important issue in the context of congressional elections research—whether legislative
behavior directly impacts legislators’ electoral fortunes. Through an analysis of positions
taken on key votes in both the House and the Senate, I have conﬁrmed that incumbents’
voting behavior on prominent and salient issues can be electorally consequential. Not
only can the positions that legislators express on key votes cost them at the margins, they

may also invite a challenge from an experienced candidate in the subsequent election,

136

making it more difﬁcult for an incumbent to get reelected. Although incumbents today
appear to enjoy a modest degree of success in avoiding electoral “mistakes” that may cost
them the next election, the results in this dissertation suggest that many of the principles

of democratic accountability are still intact in the electoral arena.62

Extensions and Future Directions for this Research

Given the encouraging nature of the ﬁndings reported here, I doubt that this will
be the last word on the subject of electoral accountability and roll call voting. For one,
my exclusive focus on the electoral consequences associated with key roll calls precludes
me from speaking to the relevance of other types of votes taken in Congress. While the
evidence presented in this dissertation is largely consistent with my expectations, I cannot
necessarily generalize to other types of votes. I focused exclusively on key roll call votes
since they typically attract sufﬁcient attention to result in observable electoral effects. To
generalize beyond the analysis in this dissertation, it would be beneﬁcial to consider a
broader subset of votes in future work.

Additionally, the analysis in this dissertation deals principally with position taking
on roll call votes, instead of considering position taking activities more broadly. For my
analysis, I was interested in examining whether or not the positions legislators expressed
on highly salient or prominent roll calls would produce observable electoral effects. To

consider broader electoral effects associated with position taking in general, it would be

 

62 Even when one considers the large number of “safe” seats in a given election year, one should not forget
the invaluable insight offered by Mayhew (1974a: 37): “When we say ‘Congressmen Smith is unbeatable,’
we do not mean that there is nothing he could do that would cost him his seat. Rather we mean,
‘Congressman Smith is unbeatable as long as he continues to do the things he is doing. . .norrnal political
activity among politicians with healthy electoral margins should not be confused with inactivity. What
characterizes ‘safe’ congressmen is not that they are beyond electoral reach, but that their efforts are very
likely to bring them uninterrupted electoral success.”

137

necessary to look at other aspects of legislative behavior such as positions expressed on
amendments to legislation, press releases or public statements, or even speeches on the
House or Senate ﬂoor. While such an analysis is beyond the scope of this dissertation, an
examination of these issues could further enhance our understanding of the effects of
position taking behavior of elected representatives and build upon many of the important
ideas presented by Mayhew in his classic text on Congress.

Finally, those seeking to build upon the ﬁndings reported in this dissertation could
further examine the relationship between issues taken up in roll call votes and the impact
of those votes in terms of legislators’ constituencies. For instance, are constituents more
likely to reward or punish incumbents for the positions they express on appropriations
legislation given that the amount of money allocated to speciﬁc districts can be more
easily tracked? Additionally, to what extent do members attempt to offset their position
taking behavior with other potentially advantageous activities such as advertising and
credit claiming? Further examination of these and other issues should continue to enrich
our understanding of legislators’ behavior, electoral accountability, and the underlying

representativeness of democratic institutions.

138

References

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

Abramowitz, Alan I. 1991. “Incumbency, Campaign Spending, and the Decline of
Competition in U.S. House Elections.” The Journal of Politics 53: 34-56.

Abramowitz, Alan I. And Jeffrey A. Segal. 1992. Senate Elections. Ann Arbor, MI:
University of Michigan Press.

Abramson, Paul R., John H. Aldrich, and David W. Rohde. 1987. “Progressive
Ambition among United States Senators, 1972-1988.” The Journal of Politics 49:
3-35.

Abramson, Paul R., John H. Aldrich, and David W. Rohde. 2003. Change and
Continuity in the 2000 and 2002 Elections. Washington, DC: CQ Press.

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

Aldrich, John H. 1994. “Rational Choice Theory and the Study of American Politics.”
In Dynamics of American Politics, ed. C. Lawrence Dodd and-Calvin J illson.
Boulder: Westview Press.

Alford, John R., and John H. Hibbing. 1981. “Increased Incumbency Advantage in the
House.” Journal of Politics 43: 1042-1061.

Alvarez, R. Michael and Jason L. Saving. 1997. “Deﬁcits, Democrats, and Distributive

Beneﬁts: Congressional Elections and the Pork Barrel in the 19808.” Political
Research Quarterly 50: 809-831.

Alvarez, R. Michael and Jonathan Nagler. 1998. “When Politics and Models Collide:
Estimating Models of Multiparty Elections.” American Journal of Political
Science 45: 55-96.

Ansolabehere, Stephen and J arnes M. Snyder. 2000. “Campaign War Chests in
Congressional Elections.” Business and Politics 2: 9-33.

Ansolabehere, Stephen, James M. Snyder, Jr., and Charles Stewart, III. 2000. “Old
Voters, New Voters, and the Personal Vote: Using Redistricting to Measure the
Incumbency Advantage.” American Journal of Political Science 44: 17-34.

139

Ansolabehere, Stephen, James M. Snyder, Jr., and Charles Stewart, III. 2001.
“Candidate Positioning in U.S. House Elections.” American Journal of Political
Science 45: 136-159.

Arnold, R. Douglas. 1990. The Logic of Congressional Action. New Haven: Yale
University Press.

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

Bianco, William T. 1984. “Strategic Decisions on Candidacy in U.S. Congressional
Districts.” Legislative Studies Quarterly 9: 351-364.

Bianco, William T., David B. Spence, and John D. Wilkerson. 1996. “The Electoral
Connection in the Early Congress: The Case of the Compensation Act of 1816.”
American Journal of Political Science 40: 145-171.

Bickers, Kenneth N., and Robert M. Stein. 1994. “Congressional Elections and the Pork
Barrel.” Journal of Politics 56: 377-399.

Bickers, Kenneth N., and Robert M. Stein. 1996. “The Electoral Dynamics of the Federal
Pork Barrel.” American Journal of Political Science 40: 1300-1326.

Bond, Jon R., Cary Covington, and Richard Fleisher. 1985. “Explaining Challenger
Quality in Congressional Elections.” The Journal of Politics 47: 510-529.

Born, Richard. 1986. “Strategic Politicians and Unresponsive Voters.” American
Political Science Review 80: 599-612.

Bovitz, Gregory L. and Jamie L. Carson. 2000. “The Electoral Consequences of
Position-Taking in Congress: Exploring the Relationship Between Roll Call
Behavior and House Election Results.” Paper delivered at the American Political
Science Association Annual Meeting.

Box-Steffensmeier, Janet M. 1996. “A Dynamic Analysis of the Role of War Chests in
Campaign Strategy.” American Journal of Political Science 40: 352-371.

Box-Steffensmeier, Janet M. and Bradford S. Jones. 1997. “Time is of the Essence:
Event History Models in Political Science.” American Journal of Political
Science 41: 1414-1461.

140

Brady, David W., Brandice Canes-Wrone, and John F. Cogan. 2000. “Differences in
Legislative Voting Behavior Between Winning and Losing House Incumbents.”
In Continuity and Change in House Elections. David W. Brady, John F. Cogan,
and Morris P. Fiorina, Editors. Stanford: Stanford University Press.

Brady, David W., John F. Cogan, and Morris P. Fiorina. 2000. Continuity and Change
in House Elections. Stanford: Stanford University Press.

Bumham, Walter Dean. 1975. “Insulation and Responsiveness in Congressional
Elections.” Political Science Quarterly 90: 411-435.

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: 127-140.

Canon, David T. 1990. Actors, Athletes, and Astronauts: Political Amateurs in the
United States Congress. Chicago: University of Chicago Press.

Canon, David T. 1993. “Sacriﬁcial Lambs or Strategic Politicians? Political Amateurs
in U.S. House Elections.” American Journal of Political Science 37: 1119-1141.

Carson, Jamie L. 2003. “Electoral Accountability and Roll Call Voting in U.S. House
Elections.” Typescript. Michigan State University.

Carson, Jamie L., Charles J. F inocchiaro, and David W. Rohde. 2001. “Consensus and
Conﬂict in House Decision Making: A Bill-Level Examination of Committee and

Floor Behavior” Paper delivered at the Midwest Political Science Association
Annual Meeting.

Clausen, Aage. 1973. How Congressmen Decide: A Policy Focus. New York: St.
Martin’s Press.

Cover, Albert D. 1977. “One Good Term Deserves Another: The Advantage of
Incumbency in Congressional Elections.” American Journal of Political Science
21: 523-541.

Cover, Albert D. and Bruce S. Brumberg. 1982. “Baby Books and Ballots: The Impact
of Congressional Mail on Constituency Opinion.” American Political Science
Review 76: 347-359.

Cox, Gary W. and Jonathan N. Katz. 1996. “Why Did the Incumbency Advantage in
U.S. House Elections Grow?” American Journal of Political Science, 40: 478-
497.

141

Cox, Gary W. and Jonathan N. Katz. 1999. “The Reapportionment Revolution and Bias
in U.S. Congressional Elections.” American Journal of Political Science 43:
812-841.

Epstein, David and Peter Zemsky. 1995. “Money Talks: Deterring Quality Challengers
in Congressional Elections.” American Political Science Review 89: 295-308.

Erikson, Robert S. 1971a. “The Advantage of Incumbency in Congressional Elections.”
Polity 3: 395-405.

Erikson, Robert S. 1971b. “The Electoral Impact of Congressional Roll Call Voting.”
American Political Science Review 65: 1018-1032.

Erikson, Robert S. and Gerald C. Wright. 2000. “Representation of Constituency
Ideology in Congress.” In Change and Continuity in House Elections, ed. David
W. Brady, John F. Cogan, and John F. Ferejohn. Stanford, California: Stanford
University Press.

Erikson, Robert S. and Gerald C. Wright. 2001. “Voters, Candidates, and Issues in
Congressional Elections.” In Congress Reconsidered, 7’” edition. Lawrence C.
Dodd and Bruce I. Oppenheimer, editors. Washington, DC: CQ Press.

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

Fenno, Richard F ., Jr. 1977. “U.S. House Members in Their Constituencies: An
Exploration.” American Political Science Review 71: 883-917.

Fenno, Richard F. 1978. Home Style: House Members in Their Districts. New York:
Harper Collins Publishers.

Ferejohn, John A. 1977. “On the Decline of Competition in Congressional Elections.”
American Political Science Review 28: 127-146.

F iorina, Morris P. 1974. Representatives, Roll Calls, and Constituencies. Lexington,
Massachusetts: Lexington Books.

Fiorina, Morris P. 1977. “The Case of the Vanishing Marginals: The Bureaucracy Did
It.” American Political Science Review 71: 177-181.

Fiorina, Morris P. 1981. Retrospective Voting in American National Elections. New
Haven: Yale University Press.

142

Fiorina, Morris P. 1989. Congress: Keystone of the Washington Establishment, 2nd Ed.
New Haven: Yale University Press.

Fowler, Linda L. and Robert D. McClure. 1989. Political Ambition: Who Decides to
Run for Congress. New Haven: Yale University Press.

F rantzich, Stephen. 1978. “Opting Out: Retirement from the House of Representatives.”
American Politics Quarterly 6: 251-273.

Froman, Lewis A., Jr. 1963. Congressmen and Their Constituencies. Chicago: Rand
McNally.

Froman, Lewis A., Jr. 1966. “A Realistic Approach to Campaign Strategies and
Tactics.” In The Electoral Process, M. Kent Jennings and L. Harmon Zeigler,
editors. Englewood Cliffs, NJ: Prentice-Hall, Inc.

Gaddie, Ronald Keith and Charles S. Bullock, III. 2000. Elections to Open Seats in the
U.S. House: Where the Action Is. New York: Rowman & Littleﬁeld Publishers,
Inc.

Gaines, Brian J. and Timothy P. Nokken. 2001. “The Presidential Shadow on Midterm
House Elections: Presidential Support, Presidential Agendas, and Seat Loss.”
Paper delivered at the Midwest Political Science Association Annual Meeting.

Gates, Scott and Brian D. Humes. 1997. Games, Information, and Politics: Applying
Game Theoretic Models to Political Science. Ann Arbor: University of Michigan
Press.

Gelman, Andrew and Gary King. 1990. “Estimating Incumbency Advantage Without
Bias.” American Journal of Political Science 34: 1 142-1164.

Gerber, Alan. 1998. “Estimating the Effect of Campaign Spending on Senate Election
Outcomes Using Instrumental Variables” American Political Science Review 92:
401 -41 1.

Goidel, Robert K. and Donald A. Gross. 1994. “A Systems Approach to Campaign
Finance in U.S. House Elections.” American Politics Quarterly 22: 125-153.

Goldenberg, Edie N., Michael W. Traugott, and Frank R. Baumgartner. 1986.
“Preemptive and Reactive Spending in U.S. House Races.” Political Behavior 8:
3-20.

143

Goodliffe, Jay. 2001. “The Effect of War Chests on Challenger Entry in U.S. House
Elections.” American Journal of Political Science 45: 830-844.

Green, Donald Philip and Jonathan S. Krasno. 1988. “Salvation for the Spendthrif’t
Incumbent: Reestimating the Effects of Campaign Spending in House Elections.”
American Journal of Political Science 32: 884-907.

Grier, Kevin. 1989. “Campaign Spending and Senate Elections, 1978-1984.” Public
Choice 63: 201-219.

Groseclose, Timothy and Keith Krehbiel. 1994. “Golden Parachutes, Rubber Checks,
and Strategic Retirements from the 102d House.” American Journal of Political
Science 38: 75-99.

Hall, Richard L. and Robert P. Van Houweling. 1995. “Avarice and Ambition in
Congress: Representatives’ Decisions to Run or Retire from the U.S. House.”
American Political Science Review 89: 121-136.

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

Hersch, Philip L. and Gerald S. McDougall. 1994. “Campaign War Chests as a Barrier
to Entry in Congressional Races.” Economic Inquiry 32: 630-641.

Hibbing, John R. 1982a. “Voluntary Retirement from the U.S. House of
Representatives: Who Quits?” American Journal of Political Science 26: 467-
484.

l-libbing, John R. 1982b. “Voluntary Retirements from the House in the Twentieth
Century.” The Journal of Politics 44: 1020-1034.

Hibbing, John R. 1991. Congressional Careers: Contours of Life in the U.S. House of
Representatives. Chapel Hill: The University of North Carolina Press.

Hibbing, John R. and Sara L. Brandes. 1983. “State Population and the Electoral
Success of U.S. Senators.” American Journal of Political Science 27: 808-819.

Hinckley, Barbara. 1980a. “House Re-elections and Senate Defeats: The Role of the
Challenger.” British Journal of Political Science 10: 441-460.

Hinckley, Barbara. 1980b. “The American Voter in Congressional Elections.”
American Political Science Review 74: 641-650.

144

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

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

Jacobson, Gary C. 1987. “The Marginals Never Vanished: Incmnbency and
Competition in Elections to the U.S. House of Representatives, 1952-1982.
American Journal of Political Science 31: 126-41.

Jacobson, Gary C. 1989. “Strategic Politicians and the Dynamics of U.S. House
Elections, 1946-86.” American Political Science Review 83: 773-793.

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

Jacobson, Gary C. 1993. “Deﬁcit-Cutting Politics and Congressional Elections.”
Political Science Quarterly 108: 375-402.

Jacobson, Gary C. 2000. “Party Polarization in National Politics: The Electoral

Connection.” In Polarized Politics: Congress and the President in a Partisan Era.
Jon R. Bond and Richard Fleisher, Editors. Washington, DC: CQ Press, 9-30.

Jacobson, Gary C. 2001. The Politics of Congressional Elections. 5th edition. New
York, NY: Addison Wesley Longman.

Jacobson, Gary C. and Gregory L. Bovitz. 1998. “The Electoral Politics of Budgets and
Deﬁcits, 1980-1996.” Paper delivered at the American Political Science
Association Annual Meeting.

Jacobson, Gary C. and Michael A. Dimock. 1994. “Checking Out: The Effects of Bank
Overdrafts on the 1992 House Elections.” American Journal of Political Science
38: 601-624.

Jacobson, Gary C. and Samuel Kemell. 1983. Strategy and Choice in Congressional
Elections. 2nd edition. New Haven: Yale University Press.

Johannes, John R., and John C. McAdams. 1981. “The Congressional Incumbency
Effect: Is It Casework, Policy Compatibility, or Something Else?” American
Journal of Political Science 25: 512-542.

145

Kahn, Kim Fridkin and Patrick J. Kenney. 1999. The Spectacle of U.S. Senate
Campaigns. Princeton, NJ: Princeton University Press.

Kazee, Thomas A. 1980. “The Decision to Run for the U.S. Congress: Challenger
Attitudes in the 19708.” Legislative Studies Quarterly 5: 79-100.

Kazee, Thomas A. 1983. ”The Deterrent Effect of Incumbency on Recruiting
Challengers in U.S. House Elections.” Legislative Studies Quarterly 8: 469—480.

Kazee, Thomas A. 1994. Who Runs for Congress? Ambition, Context, and Candidate
Emergence. Washington, DC: CQ Press.

Kennedy, Peter. 1998. A Guide to Econometrics. 4th Edition. Cambridge: The MIT
Press.

Kiewiet, D. Roderick and Langche Zeng. 1993. “An Analysis of Congressional Career
Decisions, 1947-1986.” American Political Science Review 87: 928-941.

King, Gary, Michael Tomz, and Jason Wittenberg. 2000. “Making the Most of
Statistical Analyses: Improving Interpretation and Presentation.” American
Journal of Political Science 44: 347-361.

Kingdon, John W. 1989. Congressmen's Voting Decisions. Ann Arbor: University of
Michigan Press. '

Krasno, Jonathan S. 1994. Challengers, Competition, and Reelection: Comparing
Senate and House Elections. New Haven, CT: Yale University Press.

Krasno, Jonathan S. and Donald Philip Green. 1988. “Preempting Quality Challengers
in House Elections.” Journal of Politics 50: 920-936.

Kuklinski, James H. and Lee Sigelrnan. 1992. “When Objectivity is Not Objective:
Network Television News Coverage of U.S. Senators and the ‘Paradox of
Objectivity’” The Journal of Politics 54: 810-833.

Leblang, David A. 2003. “To Devalue or to Defend? The Political Economy of
Exchange Rate Policy.” International Studies Quarterly, Forthcoming.

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

Levitt, Steven D. and J arnes M. Snyder. 1995. “Political Parties and the Distribution of
Federal Outlays.” American Journal of Political Science 39: 958-980.

146

Lin, Tse-min and Montserrat Guillen. 1999. “The Rising Hazards of Party Incumbency:
A Discrete Renewal Analysis.” Political Analysis 7: 31-57.

Loomis, Burdett A. 1999. The Contemporary Congress. 3rd edition. New York:
Wadsworth Publishing Company.

Lublin, David Ian. 1994. “Quality, Not Quantity: Strategic Politicians in U.S. Senate
Elections, 1952-1990. The Journal of Politics 56: 228-241.

Maisel, L. Sandy and Walter J. Stone. 1997. “Determinants of Candidate Emergence in
U.S. House Elections: An Exploratory Study.” Legislative Studies Quarterly 22:
79-96.

Mann, Thomas E. 1978. Unsafe at Any Margin: Interpreting Congressional Elections.
Washington, DC: American Enterprise Institute.

Mann, Thomas E. and Raymond E. Wolﬁnger. 1980. “Candidates and Parties in
Congressional Elections.” American Political Science Review 74: 617-632.

Matthews, Donald R. and James A. Stimson. 1975. Yeas and Nays: Normal Decision-
Making in the U.S. House of Representatives. New York: Wiley.

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

Mayhew, David R. 1974b. “Congressional Elections: The Case of the Vanishing
Marginals.” Polity 3: 295-317.

McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 1997. Income Redistribution
and the Realignment of American Politics. Washington DC: AEI Press.

McKelvey, Richard D. and Thomas R. Palfrey. 1995. “Quantal Response Equilibria in
Games.” Games and Economic Behavior 10: 6-3 8.

McKelvey, Richard D. and Thomas R. Palfrey. 1996. “A Statistical Theory of
Equilibrium in Games.” Japanese Economic Review 47: 186-209.

McKelvey, Richard D. and Thomas R. Palfrey. 1998. “Quantal Response Equilibria for
Extensive Form Games.” Experimental Economics 1: 9-41.

Milyo, Jeffrey. 1998. The Electoral Eﬂ'ects of Campaign Spending in House Elections.
Los Angeles: Citizens’ Research Foundation.

147

Moore, Michael K. and John R. Hibbing. 1998. “Situational Dissatisfaction in Congress:
Explaining Voluntary Departures.” The Journal of Politics 60: 1088-1107.

Nelson, Garrison. 1993. Committees in the U.S. Congress, 1947-1992. Volume 1.
Washington, DC: Congressional Quarterly Press.

Oppenheimer, Bruce I. 1996. “The Representational Experience: The Effect of State
Population on Senator-Constituency Linkages.” American Journal of Political
Science 40: 1280-1299.

Page, Benjamin I. 1978. Choices and Echoes in Presidential Elections. Chicago:
University of Chicago Press.

Peters, John G. and Susan Welch. 1980. “The Effects of Charges of Corruption on
Voting Behavior in Congressional Elections.” American Political Science Review
74: 697-708.

Poole, Keith T. and Howard Rosenthal. 1997. Congress: A Political-Economic History
of Roll Call Voting. New York: Oxford University Press.

Ragsdale, Lyn and Timothy E. Cook. 1987. “Representatives' Actions and Challengers'
Reactions: Limits to Candidate Connections in the House.” American Journal of
Political Science 31: 45-81.

Reed, William L. 2000. “A Uniﬁed Statistical Model of Conﬂict Onset and Escalation.”
American Journal of Political Science 44: 84-93.

Rivers, Douglas and Morris P. Fiorina. 1989. “Constituency Service, Reputation, and
the Incumbency Advantage.” In Home Style and Washington Work: Studies of
Congressional Politics, Morris P. F iorina and David W. Rohde, editors. Ann
Arbor: University of Michigan Press.

Rohde, David W. 1979. “Risk Bearing and Progressive Ambition: The case of Members
of the United States House of Representatives.” American Journal of Political
Science 23: 1-26.

Rohde, David W. Roll Call Voting Data for the United States House of Representatives,
[953-2000. Compiled by the Political Institutions and Public Choice Program,
Michigan State University, East Lansing, MI, 2003.

Rothenberg, Lawrence S. and Mitchell S. Sanders. 2000. “Severing the Electoral
Connection: Shirking in the Contemporary Congress.” American Journal of
Political Science 44: 316-325.

148

Schlesinger, Joseph A. 1966. Ambition and Politics: Political Careers in the United
States. Chicago: Rand McNally.

Schoenberger, Robert A. 1969. “Campaign Strategy and Party Loyalty: The Electoral
Relevance of Candidate Decision-Making in the 1964 Congressional Elections.”
American Political Science Review 63: 515-520.

Sellers, Patrick J. 1997. “Fiscal Consistency and Federal District Spending in
Congressional Elections.” American Journal of Political Science 41 : 1024-1041.

Serra, George and David Moon. 1994. “Casework, Issue Positions, and Voting in
Congressional Elections: A District Analysis.” Journal of Politics 56: 200-213.

Signorino, Curtis S. 1999. “Strategic Interaction and the Statistical Analysis of
International Conﬂict.” American Political Science Review 93: 279-297.

Si gnorino, Curtis S. 2001. STRAT: A Program for Analyzing Statistical Strategic
Models. Department of Political Science, University of Rochester.
(http://www.rochester.edu/College/PSC/signorino/programs/strat.zip)

 

Signorino, Curtis S. 2002. “Strategy and Selection in International Relations.”
International Interactions 28: 93-115.

Squire, Peverill. 1989a. “Competition and Uncontested Seats in U.S. House Elections.”
Legislative Studies Quarterly 14: 281-295.

Squire, Peverill. 1989b. “Challengers in U.S. Senate Elections.” Legislative Studies
Quarterly 14: 531-547.

Squire, Peverill. 1991. “Preemptive F und-raising and Challenger Proﬁle in Senate
Elections.” The Journal of Politics 53: l 150-1 164.

Stein, Robert M. and Kenneth N. Bickers. 1994. “Congressional Elections and the Pork
Barrel.” Journal of Politics 56: 377-399.

Stokes, Donald S. and Warren E. Miller. 1962. “Party Government and the Saliency of
Congress.” Public Opinion Quarterly 26: 531-546.

Sullivan, John L., L. Earl Shaw, Gregory E. McAvoy, and David G. Barnum. 1993.
“The Dimensions of Cue-Taking in the House of Representatives: Variation by
Issue Area.” Journal of Politics 55: 975-997.

149

Tomz, Michael, Jason Wittenberg, and Gary King. 1999. CLARIF Y: Software for

Interpreting and Presenting Statistical Results. Version 1.21. Cambridge, MA:
Harvard University, June 1. http://gking.harvard.edu/.

Welch, Susan and John R. Hibbing. 1997. “The Effects of Charges of Corruption on
Voting Behavior in Congressional Elections, 1982-1990.” The Journal of Politics
59: 226-239.

Westlye, Mark C. 1991. Senate Elections and Campaign Intensity. Baltimore, MD: The
Johns Hopkins University Press.

Wright, Gerald G, Jr. 1978. “Candidates’ Policy Positions and Voting in U.S.
Congressional Elections.” Legislative Studies Quarterly 3: 445-464.

Wright, Gerald C ., Jr. and Michael B. Berkman. 1986. “Candidates and Policy in United
States Senate Elections.” American Political Science Review 80: 567-588.

Wrighton, J. Mark and Peverill Squire. 1997. “Uncontested Seats and Electoral

Competition for the U.S. House of Representatives Over Time.” The Journal of
Politics 59: 452-468.

150