Issue Ownership and Representation in the United States:
A theory of legislative response to constituency opinion
Patrick J. Egan
[email protected]
Visiting Scholar (2006-07)
Center for the Study of Democratic Politics
Woodrow Wilson School of Public and International Affairs
Princeton University
Ph.D. Candidate
Charles and Louise Travers Department of Political Science
University of California, Berkeley
this version: October 2006
Abstract
“Issue ownership” (Budge and Farlie 1983, Petrocik 1996)—the extent to which the
public trusts one party to better handle a particular issue than the opposing party—is a persistent,
important phenomenon in American politics. By adding issue ownership to standard models of
position-taking by policy-oriented candidates, I generate the prediction that candidates will take
divergent, asymmetric positions in the policy space. Specifically, the candidate from the issueowning party will leverage her advantage to take a position close to her own ideal point, while the
other candidate’s position will be highly responsive to voters’ preferences. These predictions are
confirmed by examining the relationship between constituency opinion from U.S. Congressional
districts (generated from the 2000 National Annenberg Election Survey) and key roll call votes
cast by Congress from 1997 through 2002. The positions taken by Democrats on Democraticowned issues such as education and the environment are much less sensitive to district opinion
than those taken by Republicans. Similarly, Republican positions on issues like defense and
crime are less responsive to public opinion than those taken by Democrats.
Thanks to James Adams, Larry Bartels, Henry Brady, Bruce Cain, Rosalee Clawson, Rui
de Figueiredo, Patricia Hurley, Megan Mullin, Abe Newman, Ben Page, Robert Powell, Laura
Stoker, Rob Van Houweling and participants in the American Politics Colloquium and the
Positive Political Theory Seminar at UC Berkeley for comments and suggestions regarding this
paper.
I.
Introduction
The notion that elected leaders represent citizens’ interests is at the core of our
understanding of the democratic process. If, with Dahl (1971), we define democracy as
“popular control of government by political equals,” we naturally seek to examine the
nature and extent of that control. Do the decisions of elected officials correspond to the
public’s preferences? Do elected officials change their actions when public preferences
change?
This paper concludes that members of the United States Congress generally
reflect their constituents' preferences in their roll-call votes, but members of
Congress can and do deviate from the public's preferences when the public believes that
the member's political party "owns" an issue. In the United States, Republicans are
generally thought to “own” national security issues such as “terrorism” and Democrats
are thought to own issues such as the “environment” and “social security.” Issue
ownership makes it possible to explain observed variations in the responsiveness of
politicians across issue domains. It also helps us reach a more nuanced understanding of
the distinction drawn by Stokes (1963) between “position” and “valence” issues. In
addition, it requires that we rethink two broad streams of literature: that on formal
models of candidates in electoral competition and that on the relationship between citizen
opinion and government action.
Formal models have been employed to generate predictions about the relationship
between policy platforms adopted by candidates in democratic electoral competition and
voters’ preferences. In general, this work has focused on extensions and refinements of
the simple “median voter” model variously attributed to Hotelling (1929), Black (1948)
and Downs (1957). A second stream of research consists primarily of empirical studies
that seek to measure the relationship between citizen opinion and government action.
This literature on responsiveness and representation has its origins in the pioneering work
of scholars such as Miller and Stokes (1963), Achen (1977, 1978), and Page and Shapiro
(1983).
1
This paper contributes to both streams of literature, and it addresses some of their
shortcomings. On the one hand, analysts of formal models of party competition have
noted that the central prediction of the classic Downsian spatial model—that candidates’
platforms should converge to that of the median voter in two-party electoral
competition—is seen infrequently in politics (see e.g., Osborne 1995, Roemer 2001,
Grofman 2004, Adams, Merrill and Grofman 2005). In addition, recent models that
predict separation based on the possibility of third-candidate entry, uncertainty about the
location of voters, policy-motivated politicians, and other mechanisms do not provide
very clear predictions beyond the mere possibility of separation of candidate platforms.
On the other hand, recent reviews of the empirical responsiveness literature have found it
lacking in theoretical grounding. In particular, scholars have failed to generate strong
theories that explain variation in responsiveness across issue domains (Manza and Cook
2002, Burstein 2003)
I address these critiques by incorporating the notion of issue ownership in a model
of two-party electoral competition. Issue ownership is a phenomenon seen across
representative democracies in which the electorate tends to associate political parties with
long-term strengths and weaknesses on various issues. A political party “owns” an issue
to the extent that the public assesses it can better “handle” the issue than its opponents.
Budge and Farlie (1983) identified that the public’s assessments of these strengths and
weaknesses were important determinants of election outcomes. Petrocik (1996) extended
this finding to the American context. He discovered that the Democratic and Republican
parties make labored efforts to emphasize their owned issues in presidential campaigns,
and that they perform better to the extent that they succeed in doing so. Issue ownership
can be measured empirically by a commonly asked question on public opinion surveys:
“Which party do you trust to a better job handling [issue x]?” In the United States,
Democrats are generally trusted more than Republicans on issues such as the
environment and education; Republicans are trusted more on foreign policy and taxation.
The concept of issue ownership makes possible a more nuanced understanding of
the commonly-cited distinction made by Donald Stokes between “position issues” and
2
“valence issues” in his 1963 critique of spatial models. Position issues, wrote Stokes, are
those that “involve advocacy of government actions from a set of alternatives over which
a distribution of voter preferences is defined” (373). By contrast, valence issues “merely
involve the linking of parties with some condition that is positively or negatively valued
by the electorate” (373). Stokes argued that spatial models of his day, which excluded
valence issues, failed to capture important aspects of party competition.
Recent formal models have addressed this shortcoming. Generally, these models
build in the assumption that some parties or candidates get a “valence benefit” because
they are viewed as more competent, charismatic or have some other advantage (e.g.,
Ansolabehere and Snyder 2000, Groseclose 2001, Schofield 2003). One problem with
many of these models is that simply posit that one candidate has a valence advantage
over another without providing a particular rationale for that advantage and without
providing any sense of the possible limitations of that advantage.1 These models also
neglect a final observation by Stokes regarding position and valence issues: that their
distinction is not always clear. “It is of course true that position-issues lurk behind many
valence-issues.” (373). Many valence debates (such that over President Bush’s handling
of the war in Iraq) are affected by discussions of ordered sets of alternatives (such as
whether to withdraw forces from Iraq or commit more troops).
Issue ownership provides an intuitive explanation for why a candidate or party
would have a valence advantage by linking this advantage with candidates’ ability to
handle specific policies. In this paper, I conceptualize issue ownership as follows: on any
given issue, lawmakers from the “issue-owning” party are better able to deliver the
promises they make during a campaign than politicians from the opposing party. I
incorporate this idea in a model in which candidates from two parties—one of which is
considered by voters to be the “issue-owning” party—compete for a seat in a national
legislature. In the model, the candidate from the issue-owning party can make policy
choices that deliver precise outcomes. The opposing candidate, by contrast, has less
1
An exception to this is one version of Schofield’s model, which assumes that candidates accrue a valence
advantage to the extent that they take positions favored by party activists who in return provide the
candidates with money and support.
3
ability to make policy in the issue domain and thus delivers the outcome she promises in
the campaign with “random noise.” The competency advantage associated with issue
ownership leads some voters to support the issue-owning candidate, even if they prefer
the outcomes promised by the opposing candidate. This in turn allows the issue-owning
candidate to take a position that is unresponsive to the preferences of the typical voter.
By contrast, the opposing candidate compensates for her lack of competency on the issue
by making promises that closely track the preferences of the median voter. To the extent
that an issue-ownership advantage exists, the parties’ positions diverge.
In this paper, I briefly review the formal literature regarding models of two-party
electoral competition and the empirical literature on representation of constituency
opinion across policy domains. I further develop the notion of issue ownership and
provide data indicating how issue ownership works in American politics. I then present a
model that helps us understand how issue ownership affects representation of
constituency opinion by incumbent legislators. Finally, I present analyses of
Congressional roll call votes and public opinion data from the 2000 National Annenberg
Election Survey that comport with the empirical implications of the theory.
II.
Formal models of electoral competition and representation
Formal models of electoral competition are a broad family of analyses that make
predictions about the policy platforms adopted by office-seeking candidates in
representative democracies. The point of departure for this work is generally the model
proposed by Downs (1957), who found that simple assumptions about two-party electoral
competition yielded the powerful prediction that candidate platforms should converge to
the preferences of the voter located at the median of the distribution of all voters’
preferences. The assumptions required to obtain this result are over the policy space (that
it is one-dimensional), the electorate (that each citizen has single-peaked preferences in
the policy space and that all citizens vote for the candidate located closer to them in the
space), and the candidates (that they are motivated solely by winning office). As is well
known, Downs’ work was inspired by the models of firm location developed by Hotelling
(1929) and of voting in committees by Black (1948).
4
The simplicity of Downs’ model and the intuitiveness of his assumptions have led
to the enduring appeal of the Median Voter Theorem (MVT). It is a critical component
of many contemporary models of politics, including legislative activity (e.g. Krehbiel
1998), economic policy (Meltzer and Richard 1981, Persson and Tabellini 2000) and
political transitions (Acemoglu and Robinson 2005). But its central prediction does not
square with the fact that the platforms of political parties differ sharply from one another
in most representative democracies. Much of the formal work on electoral competition
following Downs has therefore focused on determining the changes to the MVT
necessary to achieve the divergence of candidate platforms in equilibrium. Two
important additions to the MVT—that candidates care about policymaking (Wittman
1973), and that candidates do not know the preferences of the median voter perfectly
(Coughlin 1992)—together generate the prediction that candidate platforms will diverge
in ways similar to what is observed in actual campaigns (Roemer 2001). Further
innovations—for example, the possibility of entry by a third party (Palfrey 1984),
allowing for abstention by disaffected voters (Adams and Merrill 2003), and assuming
that a party’s platform is the result of bargaining among factions that make up its
coalition (Roemer 2001)—have also generated predictions of platform divergence.
An additional line of research explains platform divergence with so-called “nonpolicy” factors. Scholars working in this vein begin with the notion that policy platforms
are only one of many criteria voters use to evaluate candidates. A candidate may be
viewed as more experienced, charismatic, or credible than her opponents and may thus
enjoy an electoral advantage. Two recent papers have incorporated this “valence
advantage” in models of electoral competition, and generate the counterintuitive result
that the valence-advantaged candidate takes a relatively moderate position in equilibrium.
A model developed by Ansolabehere and Snyder (2000) features candidates who are
purely office-seeking and have perfect knowledge of voters’ preferences. It yields the
prediction that the valence-advantaged candidate always wins the election and generally
takes a moderate position in the policy space. By contrast, Groseclose (2001) proposes a
model in which candidates are motivated by both policy and officeholding and have
imperfect knowledge of voters’ policy preferences. He finds that, under certain
5
assumptions about the shape of voters’ utility functions, small increases in valence
advantage lead the advantaged candidate to take more moderate positions. The intuition
behind these unexpected results is that the valence-advantaged candidate seeks to
highlight the differences between herself and her disadvantaged opponent. To the extent
that she takes policy positions that are similar to her opponent, she leads voters to choose
primarily on the valence criterion.
By incorporating the notion of issue ownership, the model outlined here takes a
different approach to motivating valence advantage: it ties such an advantage directly to
candidates’ ability to deliver policy outcomes on a particular issue. In doing so, the
model links valence and position together. In this conceptualization, any valence
advantage is issue-specific: a candidate who is considered better able to handle
environmental issues, for example, is not necessarily similarly advantaged on (say)
foreign policy issues. As we will see, the issue-specificity of this advantage—combined
with the assumption that candidates care about policymaking as well as holding office—
yields the prediction that the position of the issue-owning candidate is less responsive to
public opinion than that of the opposing candidate. This is a result that differs from the
predictions of the two papers discussed above.
III. Empirical analyses of responsiveness and representation
Empirical scholars have found varying degrees of responsiveness to public
opinion across time, jurisdictions, and policies. In their pioneering work examining the
relationship between the policy attitudes of survey respondents and their congressional
representatives, Miller and Stokes (1963) found that a significant relationship existed in
the area of civil rights, but not so in social welfare or foreign policy. Subsequent work
pointing out methodological problems in this research has revised these findings, and
demonstrated less variation in responsiveness across issue domains (Achen 1978, Erikson
1978). Page and Shapiro (1983) find that responsiveness is higher on domestic issues
than foreign policy issues, and argue that this is because domestic issues tend to be more
salient with the mass public. In a recent valuable contribution, Wlezien (2004) finds
strong relationships between public preferences and budget appropriations on defense,
6
welfare, and health programs—and significant, but weaker, relationships on education
and environmental protection programs.
Additional research on this topic has largely been conducted on an issue-to-issue
basis, making comparisons between issues very difficult. Perhaps surprisingly given
Page and Shapiro’s result, defense issues have been found exhibit high degrees of
responsiveness. Bartels (1991) uses National Election Studies data and roll-call votes
cast on the defense budget by House members to estimate that constituency demand for
increased defense spending accounted for about 10 percent of the total 1982 fiscal year
appropriation for the Pentagon. Additional studies have found relationships of varying
degrees on environmental policy, taxation and health (for a review, see Burstein 2003).
Another line of research in this field—Lowi (1963), Wilson (1974), and Arnold
(1990)—concludes that the structure of policies and the organized interests supporting
various policy outcomes should affect the representation of public opinion by elected
legislatures. For example, Arnold argues that diffuse, unorganized interests can be
victorious over concentrated, special interests when costs or benefits can be easily
recognized by the public. This hypothesis corresponds somewhat with Carmines and
Stimson’s (1980) notion of “easy” versus “hard” issues—easy issues being those that are
more symbolic than technical, deal with policy ends rather than means, and have long
been on the political agenda. We might expect a stronger relationship between opinion
and policymaking on easy issues.
Perhaps the biggest gap in the research is that we still do not have empirically
tested hypotheses that identify the policy domains on which we should expect more or
less responsiveness by elected officials to public opinion. Should we expect politicians to
care more about public opinion regarding tax policy or the environment? Defense
spending or trade policy? Education or foreign policy? Political scientists have, at best,
educated guesses as to the answers to these questions—and in fact, much of the most
highly regarded research in the field has yielded conflicting findings. In sum, the
research offers support for the idea that government activity bears some relationship with
7
public opinion regarding policy issues. But empirical work that offers us the opportunity
to compare the strength of this relationship across issues is scarce. If government
responds to public opinion to varying degrees on different policies, this phenomenon has
yet to be decisively demonstrated by political scientists (Manza and Cook 2002 and
Burstein 2003).
IV.
Conceptualizing issue ownership
The idea of issue ownership provides a promising place from which to better
explore variation in the character of representation across policy domains. In the
framework described by Petrocik (1996), parties each “own” a set of issues on which the
public views the party as substantially better able to “handle” than the other party. In
Petrocik’s conceptualization, “handling” is “the ability to resolve a problem of concern to
voters. It is a reputation for policy and program interests, produced by a history of
attention, initiative and innovation toward these problems, which leads voters to believe
that one of the parties (and its candidates) is more sincere and committed to doing
something about them.” (826) Petrocik notes that on a wide range of issues, survey
respondents said they “trusted” either the Republicans or the Democrats to “handle” those
issues more than the other party, and sometimes by substantial margins.
Figure 1 provides an example of recent data on issue ownership. It is a bar chart
indicating results from an ABC/ Washington Post poll conducted in 2002. Because
partisan respondents are highly likely to identify their own party as the trusted party on
any given issue, Figure 1 includes only self-identified independents to better illustrate
variation across issues.
8
Figure 1. Trust in Parties' Ability to Handle Policy Issues
(among self-identified Independents only)
source: ABC News/Washington Post Poll, September 2002
war on terrorism
Iraq and Saddam Hussein
30
70
42
58
taxes
47
53
federal budget
48
52
economy and jobs
51
49
financial fraud by corporations
56
44
social security
57
43
prescription drug benefits for elderly
57
43
health care
60
40
education
60
40
70
environment
% trusting Democrats
30
% trusting Republicans
The data displayed in Figure 1 are generally in line with the expectations of
observers of American politics: voters appear to trust Republicans more on foreign policy
issues and taxes, and Democrats on domestic issues. On newly-emerging issues that
don’t fit pre-existing frameworks (such as the corporate financing scandals that had just
begun to break in 2002) there is little differentiation among the parties. An important
note is that the relative ranking of issues owned by each party has remained steady over
time—that is, the Democrats’ and Republicans’ strongest issues in the 1980s and 1990s
(not shown here) are their strongest issues today. The data in Figure 1 are from 2002,
when the political parties were held in approximately equal esteem by the American
public. When one party falls into general disfavor (as is currently the case with the
Republican Party), it can find itself at a disadvantage on most issues. But the relative
ranking of issues over time has been stable since issue-ownership questions were first
9
included on public opinion surveys in the 1980s. For example, an April 2006 poll by the
Winston Group found that while Democrats were more trusted than Republicans on most
issues, the Republicans’ two best issues remained border security and the war in Iraq.2
Given the strong, consistent nature of issue ownership and its evident importance
to election campaigns and outcomes, a natural hypothesis emerges about the influence of
issue ownership on representation: because voters trust parties more on issues they own,
incumbents from the owning party should have more latitude in deciding how to vote on
roll-calls on the issue that come before them in Congress. After a brief discussion of the
nature of issue ownership and issue trust, I proceed to formalize and test this hypothesis.
V.
The nature of issue ownership and issue trust
As described above, “issue ownership” is defined as the extent to which the public
trusts a party to handle a particular issue. (To distinguish this issue-based notion of trust
from the generalized idea of “trust in government,” I will often refer to the trust
accompanying issue ownership as “issue trust.”) Issue ownership has several
components. It is a long-term phenomenon that is established over decades of position
taking and accomplishments by political parties. Changes in the degree of issue trust
assigned to one party or another are rare. Perhaps the most striking example of such a
change is the “issue evolution” that occurred on race in the 1960s as documented by
Carmines and Stimson (1990). Their findings can be reconceptualized in the issueownership framework as a shift in which party “owned” race issues among whites and
blacks. The shift was precipitated by the dramatic change in the positions taken by the
two parties on race. Until just prior to the Civil Rights Act of 1964, Democrats had the
trust of whites on the issue of race and Republicans had the trust of blacks. The reverse
was true by the close of the 1960s.
2
The origins and dynamics of parties’ issue ownership advantages are beyond the scope of this paper. But
we might intuitively locate the basis for issue ownership in the coalitions that contribute to parties’ “history
of attention, initiative and innovation” toward a particular issue domain. These coalitions are undoubtedly
sources of policy expertise for issue-owning parties. They provide new ideas, research, and people who
can staff the parties’ key policymaking positions on the issue domain. These resources allow issue-owning
parties to better achieve the outcomes they want.
10
But such dramatic changes in the ownership of issues are unusual. Democrats
have arguably owned Social Security since the program’s creation by the Roosevelt
administration, and have owned health care since the inception of Medicare and Medicaid
during the Johnson administration. Republicans’ ownership of the issue of taxation
stretches back at least to the “tax revolts” of the 1970s, and the G.O.P. has owned the
crime issue from the law-and-order campaigns of Richard Nixon onward. Issue trust has
remained steady despite actions taken on issues by the owning parties that were clearly
not desired by a majority of the public, including the disastrous Clinton healthcare reform
plan of 1993-94 and the Iran-Contra scandal that clouded the Republicans’ foreign policy
reputation in the late 1980s.
Another characteristic of issue trust is that it is often correlated with the extent to
which voters agree with the party’s position on an issue. In practice, this means that a
majority of the public tends to agree with the positions taken by a party on the issues it
owns. Majorities of Americans not only trust Republicans over Democrats on taxation
and defense; they also tend to share the Republicans’ preferences for lower tax burdens
and a hawkish foreign policy. The reverse is true for the Democratic-owned issues of
education and healthcare. Table 1 displays two additional examples: on the Democraticowned issue of the environment, a strong majority of Americans sides with the party on
whether federal efforts should be expanded in this area. On the Republican-owned issue
of crime, a substantial majority favors the death penalty for some crimes, a position taken
much more frequently by Republican candidates than Democrats.
Table 1. Americans’ Policy Preferences on Issues Owned by the Two Parties
Amount of federal government effort to
protect the environment
Death penalty sentences
for some crimes
more
66.6
favor
74.7
same, less, or none
33.4
oppose
25.3
source: 2000 National Annenberg Election Study.
11
However, issue ownership is a notion that is broader than simply one of shared
policy preferences between voters and parties. It also incorporates voters’ assessment of
the competency of parties to solve particular policy problems. A voter may be a dove on
defense issues and thus agree with Democrats on those policies, but he may ultimately
believe that Republicans are better at defending the country from a terrorist attack.
Another voter may prefer the free-market approach taken to environmental regulation by
Republicans, but thinks the Democrats will actually do a better job on ensuring the
cleanup of toxic waste sites.
Survey data that allow us to assess whether issue ownership goes beyond shared
policy preferences between voters and parties are scarce, because issue-ownership
questions are rarely asked in concert with policy preference questions in the same public
opinion survey. The American National Election Studies did so in 1992, 1994, and 1998
on four different issue domains, and multivariate analysis of these survey data is
presented in Table 2. If issue-ownership is nothing more than policy congruence between
voters and parties, then we would expect that it would be completely explained by this
congruence and party identification. In Table 2, participants’ responses to issueownership questions were regressed on measures of their policy placements of the parties
as well as their party identification. All variables were scaled -.5 to .5 (Democratic to
Republican), with zero indicating a neutral position between the two parties. The
intercept in each regression is thus the estimated issue-ownership advantage accorded to
the Republican Party (or, if negative, the Democratic Party) that cannot be explained
away by party identification or policy preferences. As shown in the table, the intercepts
are of the expected sign and statistically significant on each of the four issues analyzed:
Republicans are advantaged on the issues of foreign affairs and crime, Democrats on
healthcare and the environment.3 On each issue, the advantage is of roughly the same
absolute magnitude (.09 to .18): that is, the advantaged party enjoys a 9-to-18-point issue-
3
The dependent variable takes on three possible values (trust Democrats, trust both parties equally, trust
Republicans) and could arguably be considered an ordinal, rather than interval level, measure. Analyses of
the data using an ordered probit model produced substantively similar but less easily interpretable results
than the OLS analyses presented here.
12
ownership advantage among Independents whose policy preferences are equally distant
from those of both parties.
Table 2. Determinants of Issue Ownership on Four Issues, 1992 - 1998
defense/
foreign
affairs
(1992)
crime
(1994)
health
care
(1994)
environment
(1998)
party to which R feels closer on issue
.29
(.07)
.09
(.02)
.44
(.05)
.56
(.07)
R’s party identification
.55
(.03)
.46
(.03)
.44
(.03)
.10
(.04)
Republican issue-ownership advantage (intercept)
.18
(.01)
.09
(.01)
-.10
(.01)
-.17
(.01)
N
Adjusted R-squared
SEE
1,710
.29
.33
1,480
.25
.31
1,418
.33
.32
886
.12
.30
Source for data: American National Election Studies.
Dependent variable: Party R trusts to handle issue.
Cell entries are OLS regression coefficients and (in parentheses) their standard errors.
All variables scaled -.5 (in direction favoring Democrats) to .5 (in direction favoring Republicans)4
The data presented here make it clear that issue ownership is an important,
consistent factor in how Americans think about the two major political parties which goes
beyond mere policy congruence between voters and parties. It is a concept that is distinct
from shared policy preferences or party affiliation. In the next section, I explore how
incorporating this notion in a model of electoral competition might influence how the
parties respond to constituency opinion.
4
Limited data availability prevented a perfect alignment of policy preference and issue ownership
questions. In the “defense/foreign affairs” model, the NES issue-ownership question asked about foreign
affairs; the policy question asked about defense spending. For the “crime” model, no spatial-location
question was available. I used respondents’ stated positions on President Clinton’s 1994 crime bill instead.
13
VI.
A Note on Representation
A brief discussion is needed regarding the measurement of representation.
Following Achen (1978), I distinguish between two notions of representativeness.
Centrism is the extent to which an elected representative locates close to the typical voter
in her district. By contrast, responsiveness is the extent to which there is an association
between representatives’ locations and district opinion across districts or over time. To
help clarify these concepts, Figure 2 depicts the stylized relationships between district
opinion (on the x-axis) and the roll-call records established by legislators (the y-axis) in
three hypothetical representative systems.
records established by legislators
Figure 2. Centrism and Responsiveness in Three Hypothetical Representative Systems
B
C
A
ideal points of district median voters
Line A in Figure 2 depicts a system in which legislators are both perfectly centrist
and perfectly responsive. Legislators do exactly as wished by their districts’ median
voters, and thus line A lies directly on the graph’s 45-degree line. Line B is a system in
which legislators are perfectly responsive to district opinion, but not particularly centrist:
each legislator takes a position that is further to the right than desired by her district.
Finally, line C is a system that is not particularly centrist and is completely unresponsive.
Very few legislators take positions near the ideal points of their district median voters,
and there is no relationship between opinion and roll-call records across districts.
14
Ideally, we would like to be able to assess a system’s centrism and its
responsiveness with empirical tests. But a limitation of the data has led scholars to focus
on responsiveness: it is rare that opinion and policymaking are measured in the same
units (particularly in the legislative context, where legislative records are derived from
roll-call votes on many different topics). It is therefore impossible to assess where the
45-degree line lies in an actual representative system, and thus very difficult to measure
the extent to which the system is centrist. By contrast, differing units do not prevent us
from assessing the slope of the relationship between opinion and policymaking across
districts: this can be done easily with measures of correlation or with linear regression. I
will therefore focus on my formal model’s implications for responsiveness in the issueownership context, and limit my empirical tests to assessing responsiveness.
VII.
The Model
Two candidates, N (from Party N) and O (from Party O), compete to represent a
legislative district by announcing binding promises to write laws designed to deliver
outcomes xN and xO in a one-dimensional policy space that is in the interval [1,7].5 The
candidates have known ideal points xN* and xO* over outcomes in this space. The
election for the legislative seat is determined by an (odd-numbered) constituency of J
voters, each of whom has an ideal point xj* in the policy space.
Voters and candidates derive quadratic utility from any outcome, x, such that
Ui(x) = -(xi* – x)2 for any actor i. The profile of voter preferences therefore is singlepeaked, and thus any outcome preferred by the voter whose preference is at the median of
the distribution of voters’ preferences will be preferred by a majority of the voters. Call
this median voter V and his ideal point xV*. I assume that xV* is located between the two
candidates’ ideal points and (without loss of generality) that xN* < xV* < xO*.
5
I employ this interval for easy comparison with the scale used by the American National Election Studies
to assess the locations of voters and parties in an issue space. For another example of this practice, see
Adams, Merrill, and Grofman (2005). The model generalizes to any unidimensional issue space.
15
Issue ownership and policymaking ability
Unlike many models of electoral competition, I assume that some candidates are
better than others at delivering the outcomes they promise during the campaign.
Specifically, O has an issue-ownership advantage due to her party’s superior ability to
make policy. So while O is able to deliver her promised outcome xO with perfect fidelity,
N delivers the policy she promises with random noise such that the outcome experienced
by voters if N is elected is x = xN + ε, where ε is a Normally distributed random variable
with mean zero and variance σ2.
Thus the median voter’s expected utility from O’s promised outcome xO, is:
E[UV ( xO )] = E[−( xV * − xO ) 2 ]
2
= − E ( xV * ) + xO 2 − 2 xV * xO 


(1)
2
= − ( xV * ) − xO 2 + 2 xV * xO
But the voter’s expected utility from N’s promised outcome xN is:
E[UV ( xN )] = E[−( xV * − ( xN + ε )) 2 ]
2
= − E ( xV * ) + xN 2 + ε 2 − 2 xV * xN − 2 xV *ε + 2 xN ε 


(2)
2
= − ( xV * ) − xN 2 + 2 xV * xN − σ 2
Thus, even if N and O take the same positions during the campaign, V’s expected
utility from electing N will be less than his expected utility from electing O.
Candidate preferences
The candidates value outcomes as well as holding office. Following Groseclose
(2001), I model the extent to which candidates value office holding over outcomes with
the parameter λ ∈ [0,1]. Where λ equals one, candidates value winning office only, and
where λ equals zero they value outcomes only. O’s expected utility function is therefore:
16
2
E[U O ( xO , xN )] = λ F − (1 − λ )  F ( xO* − xO ) 2 + (1 − F ) ( xO* ) + xN 2 − 2 xO* xN + σ 2   ,

 


(3)
where F is an indicator variable that takes on the value one if O wins the election
and zero if N wins.
Similarly, the expected utility function for N is:
2
E[U N ( xO , xN )] = λ (1 − F ) − (1 − λ )  F ( xN * − xO ) 2 + (1 − F ) ( xN * ) + xN 2 − 2 xN * xN + σ 2  


 
(4)
Play of the game
1.
N and O simultaneously announce their promised outcomes xN and xO. (As is the
standard case in models of electoral competition, these promises are binding and
the winning candidate must write a law to implement the outcome upon which she
campaigns.)
2.
An election is then held in which the median voter V elects one of the two
candidates to office. V makes his choice by comparing the utility he expects to
derive from the outcomes promised by N and O and electing the candidate
yielding the higher expected utility. (I assume that if both candidates offer
positions yielding equal utility, V breaks the tie by choosing O.)
3.
The winning candidate i writes the law xi that was promised during the campaign.
4.
Candidates and voters then realize utility from the resulting outcome x, where:
x = xO
if O wins the election; and
x = xN + ε
if N wins the election.
The game then ends. The solution concept employed is pure-strategy Nash equilibrium.
17
Results
This game results in the candidates locating apart, with O proposing an outcome
xO that is less responsive to the district’s preferences than is N’s proposed outcome.
Specifically, in equilibrium N proposes the outcome xN equal to xV*, while O proposes the
outcome in the interval [xV*, xO*] that makes V just indifferent between the two candidates.
This assures O’s election. This result holds regardless of the value of λ (the degree to
which candidates value officeholding over policy outcomes). Hence, the issue-owning
party can afford to choose an outcome that is between the median voter and its ideal point.
Thus the model predicts that candidates from the issue-owning party are able to be less
responsive to the median voter than those from the other party.
As an illustration, I examine the case where candidates are completely motivated
by policy and thus λ equals zero. I begin by considering the median voter’s expected
utility from xN and xO, as shown in equations (1) and (2). Where xN = xO, the voter obtains
utility from O’s position that is greater by σ2 than his utility derived from N’s position.
Thus O can always win the election by matching N’s platform. But because O cares
about policy, she can do better than this. Because O is from the right-wing party, her
best reply to any proposed outcome xN played by N is to locate ((xN – xV )2 + σ2) to the
right of xV*, which is the rightmost policy she can offer and still win election. Since N is
from the “left” party, she wishes for xO to be as small as possible and thus locates at xV*.
{
{
The profile xV * , min ( xV * + ( xV * − xN ) 2 + σ 2 ) , xO*
}} is thus this game’s unique Nash
equilibrium.
It can be shown that the same equilibrium holds for all values of λ. O’s issueownership advantage allows it to always win the election even when N’s platform is
equivalent to the ideal point of the median voter. O exploits this advantage by
announcing a platform that is to the right of xV. Finally, on issues where no ownership
advantage exists, both candidates are predicted to announce platforms equal to the
preferences of the median voter.
18
Probabilistic voting
The result described above is somewhat unsatisfying in that the model predicts
that the candidate from party O will always win the election. It is likely that there are
other, unobserved factors at work that may determine election results, and from time to
time cause the candidate from party N to win despite her disadvantage. In addition, the
candidates’ locations are implausibly unaffected by the extent to which they value policy
versus officeholding.
We can incorporate these notions by moving to the “probabilistic voting”
framework (Coughlin (1992)), in which candidates are unsure of how the median voter
will act—and specifically (as emphasized by Roemer 2001), how the quantity
E[UV ( xO )] − E[UV ( xN )] will translate into an election outcome. I now assume that the
candidates assess that the probability O wins the election is equal to the probability that
E[UV ( xO )] − E[UV ( xN )] + δ > 0, where δ is a random variable distributed according to
the logistic distribution function (denoted as Λ)6. Thus the probability O wins the
election is Λ evaluated at E[UV ( xO )] − E[UV ( xN )] , or:
Pr(O wins) = Λ ( E[UV ( xO )] − E[UV ( xN )]) =
exp( E[UV ( xO )] − E[UV ( xN )]])
.
1 + exp( E[UV ( xO )] − E[UV ( xN )])
To complete the setup, I substitute this function for the indicator variable F in the
expected utility functions for N and O (equations 3 and 4).
Solving the issue-ownership game with probabilistic voting
The solution concept of pure-strategy Nash equilibrium requires that we look for
values of xN and xO that are best replies to one another—that is, pairs of platforms in
which neither candidate wishes to change her platform given the other candidate’s
6
Adams, Merrill and Grofman (2005) model probabilistic voting in this way.
19
platform. I first calculate the platforms at which the candidates’ expected utility is
maximized by finding
∂E[U i ( xO , xN )]
= 0 for i ∈ {N , O} :
xi
2
∂E[U N ( xO , xN )]
∂ 
=
λ (1 − F ) − (1 − λ )  F ( xN * − xO )2 + (1 − F ) ( xN * ) + xN 2 − 2 xN * xN + σ 2   



  

∂xN
∂xN 
and
∂E[U O ( xO , xN )]
∂ 
 F ( x * − x ) 2 + (1 − F )  x * 2 + x 2 − 2 x * x + σ 2    ,
=
λ
F
−
(1
−
λ
)
( ) N
O
O
O
N
 O
  

∂xO
∂xO 
where Λ =
exp( E[U V ( xO )] − E[U V ( xN )])
.
1 + exp( E[U V ( xO )] − E[U V ( xN )])
Performing these calculations, setting these expressions equal to zero, simplifying
and collecting terms yields the following first order conditions:
λ ( xV * − xO ) = (1 − λ ) ( xV * − xO )( − xN 2 − xO 2 − σ 2 + 2 xO* xN + 2 xO* xO ) − ( xO* − xO )(1 + ed ) 
and
λ ( xV * − xN ) = (1 − λ ) ( xV * − xN )( − xN 2 + xO 2 − σ 2 + 2 xN * xN − 2 xN * xO ) + ( xN * − xN )(1 + e− d ) 
where d = E[U v ( xO )] − E[U v ( xN )] = xN 2 − xO 2 − 2 xN xV * + 2 xO xV * + σ 2 .
The complex nature of these first order conditions makes the model impossible to
solve analytically. This is not unusual in models of electoral competition with policymotivated candidates. Roemer (2001) notes that there are no general proofs of the
existence of these “Wittman-style” equilibria and that it is rare that these equilibria can be
found analytically. I therefore proceed with numerical solutions to this game for a range
of values of σ and λ. For illustrative purposes, I set xN = 2 and xO = 6. Only one purestrategy Nash equilibrium exists for each {λ, σ} pair.
20
Table 3 displays the candidates’ equilibrium platforms that are associated with a
range of values for λ and σ when V’s ideal point, xV*, is located directly in the middle of
the policy space at 4. The table shows how O is able to exploit her issue-ownership
advantage to announce a platform that becomes closer to her ideal point as her advantage
increases. As was the case in the deterministic model, this holds true in the scenario
where candidates care only about policy (where λ equals zero), the near-Downsian
scenario (where λ = .99) and all scenarios in-between. Where no uncertainty exists (i.e.
where σ = zero), no issue-ownership advantage exists for O and the candidates take
positions that are equidistant from the median voter.
Table 3. Platforms of candidates in the issue ownership game
weight
candidates
place on
winning
office
(λ)
size of O’s
issueownership
advantage
(σ)
candidates’ platforms
(on 7-point scale ,
with median voter ideal point = 4)
O, candidate from
issue-owning party
(ideal point = 6)
N, candidate from
opposing party
(ideal point = 2)
probability
that O wins
0
0
1
2
3
4.6
4.7
5.2
5.9
3.4
3.5
3.6
3.6
.50
.68
.93
.996
.5
0
1
2
3
4.5
4.7
5.2
5.9
3.5
3.5
3.6
3.6
.50
.68
.94
.996
.8
0
1
2
3
4.4
4.6
5.1
5.9
3.6
3.7
3.7
3.7
.50
.69
.94
.996
.99
0
1
2
3
3.96
3.97
3.98
3.98
4.04
4.07
4.56
5.65
.50
.73
.98
.998
A Simulation of Legislative Data
Finally, to envision how we might expect this phenomenon to be manifested in
data from an actual legislature, I conducted the following simulation which assumes that
21
legislators of each party all share the same preferred ideal point, but that district medians
(xV*) and candidates’ concern with office-holding (λ) vary independently across districts.
Holding legislators’ ideal points constant at xN *= 2 and xO* = 6, I varied the values of λ
and xV independently at random and calculated the equilibrium positions of the
candidates in each case. I assumed that O had a decided issue-ownership advantage and
set σ = 2.
To simulate an even distribution of median voters across districts, I drew xV* from
a uniform distribution on the interval [2.5, 5.5]. To simulate a distribution of candidates
who are motivated primarily by winning office but are occasionally motivated primarily
by policy, I drew λ from a beta distribution with parameters α = 4 and β = 2. This
distribution is bounded by [0, 1] and heavily skewed to the right with a mean of .67 and a
variance of .032.
The results obtained from 100 iterations of this simulation (i.e. 200 simulated
legislators, half from the issue-owning party and half from the opposing party) are shown
in Figure 3, where xV* is plotted on the x-axis and the candidates’ platforms (xO and xN)
are on the y-axis. To provide a better sense of the distribution of the points, they have
been drawn with a slight random variation (that is, “jittered” (Cleveland, 1993)). Each
simulated position taken by party O candidates has been plotted with a small ‘x’;
positions taken by party N candidates are plotted with dots.
To interpret the results in Figure 3, recall that (without loss of generality), the
model locates O’s ideal point at 6, on the right side of the policy space. Thus as the ideal
point of the median voter becomes located closer to O’s ideal point, O is less responsive
to district opinion in equilibrium. As discussed above and shown in Table 2, this is often
exactly the case: the public’s policy preferences on various issues tend to lie closer to the
ideal point the public associates with the issue-owning party. So while Figure 3 plots the
model’s predicted relationship between district opinion and legislators’ roll-call records
for the [2, 6] interval, in the issue-ownership context we expect most of district medians
to be located on the side of the graph where the ideal point of the issue-owning party is
22
located. This area is indicated with a circle in Figure 3, and I have plotted the linear fit of
all points to the right of xV*= 4 to characterize the relationship we might see in a typical
scatterplot of district opinion against roll-call votes in the issue-ownership context.
Platforms Taken by Candidates in Equilibrium
2
3
4
5
6
Figure 3. Simulated Platforms of Legislative Candidates in the Issue-Ownership Game
issue-owning
party
opposing
party
2
area of policy
space we expect to
analyze in the
issue-ownership
context
3
4
5
Position of District Median Voter
6
The relationship plotted in Figure 3 is unexpected and illuminating. On average,
candidates from both parties fail to take positions that are perfectly centrist (as is seen by
the points’ departures from a hypothetical 45-degree line on the graph). But the positions
of candidates from the opposing party N are much more faithful to voter opinion than
those from the issue-owning party O. And while the responsiveness of both parties’
candidates to district opinion are approximately the same on the left side of the policy
space, they become markedly different on the right side of the space: O’s position
becomes much less responsive to district opinion. Note that the flatness of O’s
23
responsiveness curve in this region is not due to a “ceiling effect”: if this were the case,
we would see a similar “floor effect” in the pattern of N’s responsiveness on the left-hand
side of the graph. The prediction of this model is therefore that legislators from the issueowning party should be less responsive to district opinion than legislators from the
opposing party.
VIII. An empirical test
I now explore the empirical implications of my theoretical model by examining
the relationship between constituency opinion and the positions taken by members of
Congress on 11 different issues between 1997 and 2002. The goal is to see whether
members of Congress are more responsive to their constituencies on issues which their
parties do not own than on issues which their parties do own. Specifically, we are
looking to see whether we find a scatterplot like Figure 3 when we compare the positions
of members who own an issue with those who do not. To do this, we need measures of
constituency opinion for the horizontal axis and measures of outcomes for the vertical
axis. I obtain estimates of constituency opinion on these issues using the unprecedented
power of the 2000 National Annenberg Election Survey (NAES), and derive measures of
legislative position-taking using the ideal point estimation technique developed by
Clinton, Jackman and Rivers (2004) as implemented in the R statistical package by
Martin and Quinn (2006).
Source of opinion data: the National Annenberg Election Survey
One of the barriers that traditionally stood in the way of thorough research on the
responsiveness of legislators to constituency opinion has been the limited availability of
appropriate survey data. The public opinion surveys employed by academic researchers
usually have too few respondents per district to measure constituency opinion with any
degree of reliability. In addition, the clustered sampling design of some surveys can lead
to biased, as well as unreliable district estimates. (For a discussion, see Stoker and
Bowers 2002.) I avoid these problems by using the unprecedented power of the 2000
National Annenberg Election Survey (NAES) rolling cross-section study. The NAES
24
interviewed a total of 58,373 American adults between December 1999 and January 2001.
Among the questions asked of respondents was a battery of questions regarding a range
of policies at the heart of the political debate during the 2000 presidential campaign.
These included prominent issues—such as taxes, military spending, abortion, and gun
control—as well as significant, but less salient issues such as crime, the environment, and
foreign policy. Although the NAES was designed to obtain a sample of the preferences
of the nation as a whole, the random-digit dialing procedure used by the survey makes it
possible to generate statistically unbiased estimates of opinion at the congressional
district level. The survey included respondents from every congressional district in the
United States except those in Alaska and Hawaii, and averaged about 130 respondents
per district. Not every policy question was asked of every respondent, but on many
policy questions the number of respondents per district was 50 or more, yielding
relatively precise estimates of constituency opinion. The size of the NAES sample, its
breadth of policy questions, and its random digit design make it far superior resource than
any academic study to date for generating estimates of constituency opinion at the
congressional district level and studying its effect on Congress.
I use responses from 11 of these questions in this study (see Table 4). Responses
to all questions were rescaled zero-one (liberal to conservative). The NAES
underrepresented younger Americans and racial minorities (for example, only 8.6% of
the NAES sample was African-American and 9.6% was between 18 and 24 years of age;
corresponding figures from the 1990 Census were 12.1 % and 10.8%). I therefore poststratified the NAES sample by district on the basis of age and race using U.S. Census
congressional district-level enumerations of these variables (U.S. Census Bureau, 1995).
The district-level estimates were calculated using these post-stratification weights, which
are shown in Table 5.
25
Table 4. Policy attitude variables from the NAES used in analysis
issue owned by issue domain (direction of scale)
military spending (increase)
127
intervention in civil wars (oppose)
Republicans
crime (punishments too lenient)
57
125
127
BJ03
BB01
education spending (decrease)
57
environment (reduce gov’t effort)
57
school vouchers (support)
127
gay rights (reduce gov’t effort)
neither party
BJ07
BG12
60
taxes (rates too high)
NAES
var id
BJ08
missile shield (support)
Democrats
How much $ on X
How much effort into X
Should govt do X
Seriousness of X as a
problem
question type (mean N per district)
125
gun control (oppose)
BS01
BD02
123
abortion (restrict)
BD09
BL05
BF02
126
BG06
Table 5. Post-stratification weights used to calculate estimates of district-level opinion
(averaged over all districts)
weight
Race
Asian-American
Black
Hispanic
White/other
1.23
1.38
0.73
0.98
18 – 24
25 – 34
35 – 44
45 – 64
65 and over
1.63
1.30
0.94
0.78
1.09
Age
26
On their face, these estimates of district-level opinion appear valid. Appendix I
includes maps depicting the geographic distribution of opinion on abortion,
environmental protection, and gun control. The maps indicate that the distribution of
opinion regarding these issues is as expected.
Source of policymaking data: key roll-call votes in Congress, 1997-2002
The NAES was conducted between December 1999 and January 2001. To
measure policymaking I used votes cast in the House of Representatives from 1997
through 2002, the six years closest to the survey period that occurred before redistricting.
To identify roll-call votes appropriate for analysis, I consulted the annual lists of
significant votes compiled by two highly regarded resources on Congress: Congressional
Quarterly (which lists what it calls “key votes”) (Congressional Quarterly 1997 through
2002) and the National Journal (which compiles a list to calculate annual “vote ratings”
for each legislator) (National Journal 1997 through 2002). Both sources limit their lists
to votes of substantial import or controversy. In order to be considered for inclusion in
my analysis, a roll call vote had to be included on one of these two lists. To choose votes
from this pool for analysis, I determined whether the subject matter of the vote could
reasonably be deemed to be related to the subject matter of one of the issue domains
under study. I did not use votes whose outcomes were lopsided in one direction or
another. A total of 124 House votes on the 11 different issues were included in my
analysis. A list of all of these votes and the policy questions with which they were paired
can be found in Appendix II.
To determine how these votes translated into records established by members of
the House of Representatives on each of the 11 different issues, I employed the Bayesian
simulation ideal-point estimation technique proposed by Clinton, Jackman and Rivers
(2004). This technique generates estimates of the location of each “aye” and “nay” vote
in the roll-call analysis in the issue space and, consequently, the location of each
legislator’s ideal point in the space. This statistical model has many parameters—the
number of legislators plus two times the number of roll-call votes—making it
27
unidentified. Thus leading standard estimation techniques (such as maximum likelihood)
yield inconsistent estimates of these parameters (Clinton, Jackman and Rivers 2004).
However, with the addition of a few generally accepted constraints, the model can be
estimated using an iterative process known as Markov Chain Monte Carlo (MCMC). The
full details of MCMC estimation are complex and beyond this paper’s scope.7 A brief
description of the MCMC ideal-point estimation technique is provided by Clinton,
Jackman and Rivers:
The fundamental difficulty in roll call analysis is that everything other than the votes is
unobservable: The ideal points, bill parameters, and utilities are unknowns. But if it were
possible to impute values to the bill parameters and utilities, then the ideal points could
be estimated by regression. By the same logic, if we were able to impute values for the
ideal points and utilities, the bill parameters could also be estimated by regression. The
MCMC algorithm repeatedly performs these imputations and regressions, starting from
an arbitrary point and alternating between simulation of the ideal points, bill parameters,
and utilities. Under a wide set of conditions (e.g., Tierney 1996) MCMC algorithms are
guaranteed to generate samples from the posterior density of the model parameters,
regardless of where in the parameter space the algorithm is initialized. (358)
The MCMC technique thus allows for a full exploration of the space of
parameters needed to estimate legislators’ ideal points on the issues studied here, despite
the fact that the space has literally hundreds of dimensions. MCMC has additional
advantages over alternate techniques: an explicit model of roll-call voting is incorporated
as part of MCMC estimation, and missed roll-call votes or incomplete legislative terms
do not pose a problem for MCMC.
Using the MCMC technique produced estimates of the ideal points all 529
representatives who served at least some time in the House between 1997 and 2002 on
the 11 different issues. These estimates typically followed a bimodal distribution
7
One fact that may assuage the doubts of those who are unfamiliar with MCMC is that the algorithm’s
estimates of ideal points tend to be very similar to the estimates obtained using a more familiar technique:
factor analysis. In work not presented here, I estimated legislators’ ideal points on each policy domain as
the factor scores derived from the first eigenvector yielded by factor analyzing the roll-call votes in each
policy domain. (This technique is similar to that employed by Heckman and Snyder (1997) in their paper
on ideal point estimation.) These alternative estimates are highly correlated with the ideal points used in
the present analysis and in fact produce results that more strongly support my theory than those reported
here. For additional detail about the use of MCMC techniques, see Jackman (2000) and Clinton, Jackman
and Rivers (2004).
28
bounded by the interval [-2, 2], with zero generally located in the approximate center of
the issue space.
Results: the responsiveness of legislators to district opinion
Because my theory is that issue ownership should lead representatives from the
two parties to vary in their responsiveness, I estimate a model with the member of
Congress’ (MC’s) position as the dependent variable, and with district mean voter
opinion, the member of Congress’ (MC’s) party affiliation, and an interaction between
these two variables for all legislators i on each issue j as independent variables:
MEMBER’S POSITIONij =
β1j(MEAN DISTRICT OPINIONij) + β2j(DEMOCRATIC MCi)
+ β3j(MEAN DISTRICT OPINIONij x DEMOCRATIC MCij) + εij
If the relationship between mean district opinion and the MC position-taking is
the same for both the issue-owning party and the opposing party, then β3j should always
be zero. But if the relationship between opinion and roll-call records differs substantially
between issue-owning parties and non-issue-owning parties (as predicted in Figure 3),
then β3j should take on the same sign for all Democratic-owned issues and the opposite
sign for all Republican-owned issues.
I performed a separate estimation for each of the J = 11 policy domains in the
study using OLS. The results are reported in Table 6. Because scales of neither the
opinion variables nor the ideal point variables are intrinsically meaningful, I report the
standardized (“beta”) coefficients estimated by OLS and thus no intercept is estimated.
These coefficients indicate the change (in standard deviations of the dependent variable)
associated with a one-standard-deviation change in the independent variable.
The issue ownership theory leads us to be keenly interested in the estimated
coefficient on the interaction term between the legislators’ party affiliations and district
opinion (β3j in equation 7). On issues owned by Democrats, we expect Republicans to be
29
(5)
more responsive to district opinion. On issues owned by Republicans, we expect the
opposite. Finally, on issues about which there is no uncertainty about the translation of
policies into outcomes (and thus neither party holds an issue ownership advantage), both
parties should be equally responsive to district opinion. Because all variables are scaled
in the (Democratic/liberal) to (Republican/conservative) direction, this means β3j should
be negative on Democratic-owned issues, positive on Republican-owned issues, and have
no discernable pattern on issues owned by neither party.
Table 6 shows this to be precisely the case. On the three issues in the analysis
owned by Democrats—education, environment, and school vouchers—the estimated
coefficient on the interaction term takes on the expected sign, and in two of the three
cases it is statistically significantly less than zero. On the five issues owned by
Republicans—intervention in civil wars, crime, military spending, the missile defense
shield, and taxes—all coefficients take on the expected (positive) sign and two of the five
are statistically significantly greater than zero. And on the three issues owned by neither
party—abortion, gay rights, and gun control—there is no pattern to the sign of the
coefficients and all fail to reach statistical significance at the .05 level.8
8
Although it is tempting to compare the size of the coefficients across policy domains, a thorny
methodological problem prevents us from doing so. While we would like to conclude that legislators
exhibit relatively higher degrees of responsiveness to public opinion in issue domains with higher values of
β1j, it is also quite possible that variation has nothing to do with variation in responsiveness. Instead
variation may simply be an artifact of opinion items for some policy domains being less valid measures
than others. Note that my estimates of β3j do not suffer from this problem, as we are only comparing their
sign (and not comparing their size) across policy domains.
30
Table 6. Predicting issue positions taken by Congressional representatives,
1997-2002
Issues owned by Democrats
education
spending
district opinion
Democratic MC
opinion x Democratic MC
R-squared
N
Number of roll-call votes
.10**
-1.58***
-.80**
.67
527
3
environment
.19**
-1.40***
-.72*
.54
528
10
school
vouchers
.07
-.72***
-.08
.62
528
3
Issues owned by Republicans
district opinion
Democratic MC
opinion x Democratic MC
R-squared
N
Number of roll-call votes
intervene
in civil
wars
crime
military
spending
missile
shield
taxes
.01
-.67***
.08
.59
528
13
.11*
-.85
.14
.56
527
8
.14***
-1.90***
1.14**
.70
528
11
.06
-1.69***
.88***
.72
528
7
.06
-1.25***
.37
.80
527
28
Issues owned by neither party
district opinion
Democratic MC
opinion x Democratic MC
R-squared
N
Number of roll-call votes
abortion
gay rights
gun control
.24***
-.79***
.15
.60
528
25
.32***
-1.00***
-.39
.62
527
3
.31***
-.02
.59
.57
527
13
Dependent variables in each analysis are legislators’ estimated ideal points on each issue.
Cells contain standardized (“beta”) coefficients. They indicate the change (in standard deviations
of the dependent variable) associated with a one-standard-deviation change in the independent
variable. Negative values of all variables are associated with liberal/Democratic policy
preferences; positive values are associated with conservative/Republican preferences. *p<.05;
**p<.01; ***p<.001 (two-tailed test).
31
We can see graphically how the data square with the predictions of the model by
examining scatterplots of district opinion and legislator records on four different issues
(Figures 4 and 5). On the environment—an issue owned by Democrats—the owning
party is much less responsive to district opinion than the opposing party. The same is
true for military spending, an issue owned by Republicans. A comparison of these graphs
to those in Figures 2 and 3 shows how the actual relationship between opinion and
position-taking in American politics is similar to the patterns of responsiveness predicted
by my model. Finally, note in Figure 5 that on two issues owned by neither party
(abortion and gun control)—and thus on which neither party is accorded a competency
advantage—legislators from both parties are equally responsive to district opinion.
32
Figure 4. District opinion and Congressional roll-call records
on issues owned by the two parties, 1997-2002
Each point represents one member of Congress;
Democrats (Ds) are in black, Republicans (Rs) in gray
Environment
MC voting record
(Owned by Democrats)
D
R
R R
RRRRRRRRRRRRRRRR RRRR RRRRRDR RR R RR RR RR
R RRRRRRRRR RRRRRRRR
RR R
RR R R
R
R
R
R
R
R
R
R
R
RR
R
RRR R R
R R
R
R
RRRR RRRRRRRRRDRR R R RR RR
R RRRRRRD
RR R
D
R
RR
R R RR DR DRRR R R R R R RRRR
RR
R RRR RRDRR
R RR R
R
RR
DRRRR RR RRRR RRR RR RD
R
D
RRR
DRRR R R RR DRRDRRR DRRDDDRRRRRR R R
R DD R RD R RR
RR
RDRD
RD
R
DR DR RRD RD
D
R D D R RDDRD D R DRDD RR
R
D
D
DR D R D
DR D
DD
D
R
D
D
D
D D DRR DDR R
DRDDD D D
D
D
D
DDDDDDRRDRD D
DDRRDDD D DD
D
R
R
D
D DDDDRDDD
D
DD DRDD
D R D
DDDDDDDDDR D
D
DR R DR R DRD DRDDR D
D
D DD D D
D D R RD DR
D
D D
D
D DDDD
DDDD
DD
DD
DDDD D
DDDRDDDDDDDDRDDDDRD
RD
D DDDRDDDDD D
DDDRD
RDDDDDDD DDD
DD
DDD
DRDRD
RDDD
DD
D
D
D
District opinion
Military Spending
(Owned by Republicans)
R R RRRRRRRRR R RR R R R RR RR
RR R
RR RR RRR RRRR R RRR R R R
RRRRRRRRRRRRRRRRRRRRRRRRR
R
RRRR
R RR RRRR R RRRRR
R R RR RR RD R
R R
RD RRR RRRRRRR RRRR
R R RRR
R
R R R
R
R R
RR R R RRR
RRR R R RR RR R RRRRRR RR
R
R R
D R R RDD
R
R D RR R RR R R
R
R RR R R R R
R
R
R
R
R
R
R R R
R
RR R R
D
RR
R
R R
R
R
R R
DRRD
D
R
R
R
R
R
R
DR
R
R
D D RR DDDR
R
R
D
D
R RRR D
D
R
DRDDDR RD D
DD
RRRRDDR
D
D
R D R RD
R D
RR DDD D
D D
DDD
DD RR DDDDDDD D D DD
D
D
D
DD
DR D DR
DDD DDD
D
D
D
R
D
D
D
DD
D DD D R D D
D
D DD DD D
D
D D D
D D DDDD DD DDDDDR DDD DRD
DD
D D RD
D
DDDDD
D D DDDDRDD
D
D
D
D
D
R D
D
D D D DD D DDDDD D
D
D
DDD D D D
D DD D R
D D
D
D
D
D D
D DD DD D D
D
D DD D DDDD
D
D
DDDD D D DD DD
D
D
D
D D
D
DDD DD
D
D
D
D
D D D
D D
D
D
D
D
R
MC voting record
R
D
DD
D
D
D
R
District opinion
33
R
R
D
D
Figure 5. District opinion and Congressional roll-call records
on two issues owned by neither party, 1997-2002
Each point represents one member of Congress;
Democrats (Ds) are in black, Republicans (Rs) in gray
Abortion
(Owned by neither party)
R
D
RD
R RRD R R
D R RRR RDDR RRRRR RRRRRRRRD DR RDRDRR RD
DR DR
RR RRR
RRRRD R
RRRRRD
RRRR RDRD
RRRRR R R RRRRRRRR RD
RDRRRR
RDRRRRRD
D
RRRRRRRD
RR R RRRRRRRRRR RRRDRRRRRRRRR DRRR RRRRRRRR
R
R
R
R
R
R
R
R
R
D
R RR R R R DR R R R RR R R RRRR DRR RR RR R
R
R
R R
R
R
D
R
RR
R
R
R
R
R
R
R
D
DR
R
DR
RR
R
R R
D
D
R D
R
R
R
R
D R
R
R
R R
RR
DD D DRRD
D DD D
DR
RR RDR D
D
R
D
DR D
D
R
R
D
RR R R R RD DDDD
D
D DDD
D
RD D DR D D D DD D DD D D
R DR
R
R
D
D
D
DD
RD
R
D DDDD
D
D
D
D
D
D
D
D
D D D D D DD D
D
DD
D
D DDD
DR D
D
DRDDDRDRDDDD
DDDDD DRDDDD
DD D DDDDDDD D DD DD DDDDDD D DD
DDDD
D DDDD
DDD
DD
DDRDDRDDD
DD D
DD D
D DDDD DDD DDDD D D
D DDDDDDDDDDDD
D
MC voting record
R
D
D
D DDDD
DD
R
R
R
RRRDRR DR
R
R
RR RR R D
D
D
D D
D
D
District opinion
Gun Control
(Owned by neither party)
MC voting record
D
D
R
R RRRRRR
R
R
RR
R
R
RRR RD RR R
R RRRR RR R
R RRR R
R
RR R RR RRRR R RR
R R
R
RR R R R RRRDRRD
RR
R RRRRR R RR
R
R RR
RR
RR RRRRRDRRRR R RD R R RDRR R RR
R
R
R
R
R
R
R
R
R
R R R R RRR RRRRRD
R RRRRR RRR RDRR RDR
D
D
RDDR R
R R RRR D R D R D
RD RRR R D DR R R RD
RR D R RRRR DD
RDRRR DD
DD D
D
RDD D D
RR RDDRRR D
R
DR
R
D
R
D
R
R
D
R
DD
RRRRRR
DD
R RD RRD
D
R
D
R
R D DR
R
R
R
R
R
R
R
R
R
R
R
D
D
D
D
D
D
R
R
R
R
R
R
R
R
R RR R R
R
R
R D
RDDD DR RR
D RR RR DR R RRRRDR D
D
R D D DDR RD
RD RR R RD R D
D
R
D
D
R DD RR DD D
R
R
R
D
DD
R
R
DR
D D D RD D D
D R DD
D
RR
D
DDD
RRDR DD D D
D DDD D D
DRD D DD DDD
D
DDD D
D
D
D
D
D
D
D
D D
DD
DD
DD DDD
D DD
D
D
D
DD D DD D
D
D
DDDDD DDDDD DDDD D DDDDDD D D D
R
D D
DD
D
D
D
R
D
D D
DDDDD DD D DDD D
DDDD DDDD
D
DDD
DD
D
DD
DDD
RR
District opinion
34
R
R
R
R
D
IX.
Conclusion
In this paper, I demonstrate how issue ownership contributes to our understanding
of the nature of issues and to how government responsiveness to public opinion varies
across policy domains. First, I show that issue ownership is a persistent, important aspect
of American politics that merits inclusion in models of electoral competition. I then
develop a model that explicitly ties the notion of issue ownership to the utility citizens
expect to derive in an election featuring candidates with different levels of issue-specific
competency. Finally, I conduct a test that comports with the empirical implications of the
model: elected representatives are less responsive to constituency opinion on issues their
party is trusted to better handle by the public.
These findings suggest that political scientists need to rethink how we have
conceptualized “issues,” as the decisions of voters and candidates are more nuanced than
simply assessing the distance between one another in a policy space. This research
shows how campaign promises—and voters’ evaluations of those promises—appears to
depend on candidates’ policy expertise. Finally, this work helps explain why—time after
time—the two political parties propose outcomes that are more extreme than voters want
in the policy domains on which they have Americans’ trust. Examples of this practice
abound in contemporary politics, from the tax cuts and defense spending enacted by the
Republicans during the presidency of George W. Bush to the over-reaching health care
reform plan proposed during the Clinton administration. To better understand these
dynamics, we need to incorporate the long-term strengths and weakness of the parties on
specific issues in our analyses of American politics.
35
REFERENCES
Acemoglu, Daron and James A. Robinson. 2005. Economic Origins of Dictatorship and
Democracy. New York: Cambridge University Press.
Achen, Christopher H. 1977. “Measuring Representation: Perils of the Correlation Coefficient.”
American Journal of Political Science 21: 805-815.
---------------------------. 1978. “Measuring Representation.” American Journal of Political
Science 22 (August) 475-510.
Adams, James, Samuel Merrill III, and Bernard Grofman. 2005. A Unified Theory of Party
Competition. New York: Cambridge University Press.
Adams, James, and Samuel Merrill III. 2003. “Voter Turnout and Candidate Strategies in
American Elections.” Journal of Politics 65: 161-189.
Ansolabehere, Stephen D. and James Snyder. 2000. “Valence Politics and Equilibrium in Spatial
Models.” Public Choice 103: 327-336.
Arnold, R. Douglas. 1990. The Logic of Congressional Action. New Haven: Yale University
Press.
Bartels, Larry M. 1991. “Constituency Opinion and Congressional Policy Making: The Reagan
Defense Buildup.” American Political Science Review 85 (June) 457-474.
Black, Duncan. 1948. “On the Rationale of Group Decision-making,” Journal of Political
Economy 56: 23-34.
Budge, Ian and Dennis J. Farlie. 1983. Explaining and Predicting Elections: Issue Effects and
Party Strategies in Twenty-Three Democracies. Boston: Allen & Unwin.
Burstein, Paul. 2003. “The Impact of Public Opinion on Public Policy: A Review and an
Agenda.” Political Research Quarterly 56 (March): 29-40.
Carmines, Edward G. and James A. Stimson. 1980. “The Two Faces of Issue Voting,”
American Political Science Review 74 (March) 78-91.
-----------------------------------------------------. 1990. Issue Evolution: Race and the
Transformation of American Politics. Princeton: Princeton University Press.
Cleveland, William S. 1993. Visualizing Data. Summit, N.J.: Hobart Press.
Clinton, Joshua, Simon Jackman, and Douglas Rivers. 2004. “The Statistical Analysis of Roll
Call Data.” American Political Science Review. 98: 355-370.
Coughlin, Peter. 1992. Probabilistic Voting Theory. New York: Cambridge University Press.
CQ Weekly. 1999. “For the Record: House Key Votes in 1999.” December 4. 2944-2947.
--------------. 2000. “House Key Votes in 2000.” December 9. p. 2876.
36
Dahl, Robert A. 1971. Polyarchy: Participation and Opposition. New Haven: Yale University
Press.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: HarperCollins.
Erikson, Robert S. 1978. “Constituency Opinion and Congressional Behavior: A Reexamination
of the Miller-Stokes Representation Data.” American Journal of Political Science 22
(August) 511-535.
Gilligan, Thomas W. and Keith Krehbiel. 1987. “Collective Decisionmaking and Standing
Committees: An Informational Rationale for Restrictive Amendment Procedures.”
Journal of Law, Economics and Organization 3: 287-335.
Grofman, Bernard. 2004. “Downs and Two-Party Convergence.” Annual Review of Political
Science 7: 25-46.
Groseclose, Tim. 2001. “A Model of Candidate Location When One Candidate Has A Valence
Advantage.” American Journal of Political Science 45: 862-886.
Heckman, James J., and James M. Snyder. 1997. “Linear Probability Models of the Demand for
Attributes with an Empirical Application to Estimating the Preferences of Legislators.” RAND
Journal of Economics 28: S142–89.
Hotelling, Howard. 1929. “Stability in Competition,” Economic Journal 39: 41-57.
Jackman, Simon. 2000. “Estimation and Inference via Bayesian Simulation: An Introduction to
Markov Chain Monte Carlo.” American Journal of Political Science 44: 375-404.
Krehbiel, Keith. 1998. Pivotal Politics: A Theory of U.S. Lawmaking. Chicago: University of
Chicago Press.
Lowi, Theodore. 1964. "American Business, Public Policy, Case Studies and Political Theory,"
World Politics 16 (July) 677-715.
Manza, Jeff, and Fay Lomax Cook. 2002. “A Democratic Polity? Three Views of Policy
Responsiveness to Public Opinion in the United States.” American Politics Research 30
(November): 630-667.
Martin, Andrew and Kevin Quinn. 2006. Markov chain Monte Carlo (MCMC) Package.
Version 0.7-2. Available online at
http://mcmcpack.wustl.edu/wiki/index.php/Main_Page .
Meltzer, Allen H. and Scott F. Richard. 1981. “A Rational Theory of the Size of Government.”
Journal of Political Economy 89: 914-927.
Miller, Warren E. and Donald E. Stokes. 1963. “Constituency Influence in Congress.” American
Political Science Review Vol. 57 (March) 45-56.
National Journal. 2000. “Key Roll-Call Votes Used for Ratings.” February 4. available online:
http://nationaljournal.com/members/news/2000/02/0204nj2b.htmh Votes.” February 2.
37
available online:
http://nationaljournal.com/members/news/2001/02/0202nj_house_votes.htm
Osborne, Martin J. 1995. “Spatial Models of Political Competition under Plurality Rule: A
Survey of Some Explanations of the Number of Candidates and the Positions They
Take.” The Canadian Journal of Economics / Revue canadienne d'Economique 28: 261301.
Page, Benjamin I. and Robert Y. Shapiro. 1983. “Effects of Public Opinion on Policy.”
American Political Science Review 77 (March) 175-190.
Palfrey, Thomas R. 1984. “Spatial Equilibrium with Entry.” Review of Economic Studies
51:139-156.
Persson, Torsten and Guido Tabellini. 2000. Political Economics: Explaining Economic Policy.
Cambridge: MIT Press.
Petrocik, John R. 1996. “Issue Ownership in Presidential Elections, with a 1980 Case Study.”
American Journal of Political Science 40: 825-50.
Roemer, John. 2001. Political Competition: Theory and Applications. Cambridge: Harvard
University Press.
Stoker, Laura and Jake Bowers. 2002. “Designing Multi-Level Studies: Sampling Voters and
Electoral Contexts.” Electoral Studies 21: 235-267.
Stokes, Donald E. 1963. “Spatial Models of Party Competition.” American Political Science
Review. 57:368-377.
Tierney, Luke. 1996. “Introduction to General State-Space Markov Chain Theory.” In Markov
Chain Monte Carlo in Practice, ed.W. R. Gilks, S. Richardson, and D. J. Spiegelhalter.
London: Chapman and Hall, 59–74.
U.S. Census Bureau. 1995. Congressional District Data, 105th Congress of the United States.
Available online: http://goldrush.berkeley.edu/GovData/info/congdist.html
Wilson, James Q. [1974] 1995. Political Organizations. (2nd edition). Princeton: Princeton
University Press.
Wittman, Donald. 1973. “Parties as Utility Maximizers.” American Political Science Review 67:
490-498.
Wlezien, Christopher. 2004. “Patterns of Representation: Dynamics of Public Preferences and
Policy,” Journal of Politics 66: 1-24.
38
APPENDIX I. Maps of the distribution of district opinion on three issues
Restrictions on abortion
Support for restricting abortion
0.0000 to 0.2030 (40)
0.2030 to 0.3330 (114)
0.3330 to 0.4640 (208)
0.4640 to 0.5950 (128)
0.5950 to 1.0000 (33)
Environmental protection
Support for environmentalism
0.0000 to 0.7980 (33)
0.7980 to 0.8434 (105)
0.8434 to 0.8890 (213)
0.8890 to 0.9340 (136)
0.9340 to 1.0000 (33)
Gun control
Support for gun control
0.0000 to 0.6980 (31)
0.6980 to 0.7630 (141)
0.7630 to 0.8270 (193)
0.8270 to 0.8930 (110)
0.8930 to 1.0000 (49)
39
APPENDIX II.
Roll call votes included in analysis
policy domain
environment
year
1997
2000
2000
2000
2001
2001
2001
2001
2001
2002
roll call #
108
179
280
305
181
288
311
317
366
315
policy proposal
Limit proposed waivers to Endangered Species Act
Use oil royalties to establish land conservation
Restrict designations of national monuments
Delay implementation of air quality standards
Bar offshore drilling near Florida
Bar delay of new arsenic standards
Increase CAFE standards for vehicles
Bar ANWR drilling
Reallocate farm subsidies to conservation
Prohibit funds for drilling off California coast
gun control
1999
1999
1999
1999
1999
1999
2000
2000
2001
2002
2002
234
235
236
238
240
244
306
324
244
24
292
Require background checks at gun shows w/in 24 hrs
Require certain gun show dealers to run background checks w/in 3 days
Ban gun sales without safety devices
Ban juveniles from semiautomatic assault weapons
Repeal DC gun ban
Require background checks at gun shows
Block adding cities to HUD safe-guns program
Bar gun-safety agreement with Smith and Wesson
Bar funds to change FBI background checks on gun purchases
Exempt gun ads from campaign finance restrictions
Allow pilots to carry guns during flight
gay rights
2000
2001
2001
471
352
354
Broaden coverage of federal hate crimes to include gays
Bar funds for domestic partner benefits of DC employees
Bar funds for DC to enforce anti-discrimination law against Boy Scouts
crime
1999
1999
1999
2000
2000
2001
2002
2002
211
215
233
115
317
242
63
64
Increase juvenile gun penalties
Bar release of prisoners due to overcrowding
Authorize juvenile justice programs
Mandatory minimum sentences for using firearm in crimes
Reduce funding for truth-in-sentencing grants
Bar funds to deport aliens convicted of certain crimes
Require judges to file report when sentencing to life
Life imprisonment for repeat child molesters
abortion
1997
1997
1998
1999
1999
1999
1999
65
362
325
184
261
301
349
Ban late-term abortions
Lighten proposed ban on funds to aid orgs that provide abortions
Override partial-birth abortion veto
Permit abortions at overseas military hospitals
Criminalize interstate transportation of minor for abortion
Remove prohibitions on abortion coverage in federal health plans
Bar federal funds to foreign orgs that perform abortions
40
policy domain
abortion (cont.)
year
1999
2000
2000
2000
2000
2000
2000
2001
2001
2001
2001
2001
2002
2002
2002
2002
2002
2002
roll call #
465
104
203
318
373
396
422
88
89
115
235
357
97
153
342
343
411
412
taxes
1997
1997
1998
1999
1999
1999
1999
2000
2000
2000
2000
2000
2000
2000
2001
2001
2001
2001
2001
2001
2001
2001
2001
2001
2002
2002
2002
148
245
102
90
331
333
485
15
73
119
127
254
450
458
42
45
66
68
75
84
87
104
149
404
103
219
225
policy proposal
Make it a crime to injure or kill a human fetus
Ban late-term/partial-birth abortions
Allow abortions in military hospitals overseas
End ban on abortions for federal prisoners
Prohibit FDA tests of RU486
Remove ban on federal funds to foreign orgs that perform abortions
Allow coverage for abortion in federal health plans
Federal crime to attack pregnant woman
Criminalize killing of fetus
Support restrictions to int'l agencies providing abortions
End ban on abortions for federal prisoners
Permit abortions at overseas military hospitals
Criminalize transport of minor for abortion
Permit abortions in military hospitals
Require late-term/partial-birth abortion bans to consider health of mother
Ban late-term/partial-birth abortions
Federal grantees cannot withhold abortion services
Prohibit discrimination against federal grantees that refuse abortions
Budget resolution cutting taxes, Medicaid and Medicare
Pass Clinton tax cuts
Constitutional amendment requiring 2/3 vote to raise taxes
Constitutional amendment requiring 2/3 vote to raise taxes
Substitute Democratic version of a proposed tax cut
Tax cuts, including estate tax phase out
Provide tax breaks for medical savings accounts
Eliminate marriage penalty tax
Conservative budget resolution: larger tax cuts
Constitutional amendment requiring 2/3 vote to raise taxes
Abolish the tax code
Repeal estate tax
Repeal tax increase on Soc Sec beneficiaries
Override veto of estate tax elimination
Democratic substitute tax cut
Tax cuts
Progressive Caucus budget resolution
GOP Study Group budget resolution
Reduce marriage penalty
Phase out estate and gift taxes
Constitutional amendment requiring 2/3 vote to raise taxes
Conference report on budget resolution
Approve conference report for tax cuts
Approve $100 billion in tax cuts
Make tax cuts permanent
Make estate tax cuts permanent
Constitutional amendment requiring 2/3 vote to raise taxes
41
policy domain
missile shield
school vouchers
intervention in
civil wars
military
spending
education
spending
year
2002
1999
1999
2001
2002
2002
2002
2002
roll call #
229
58
59
230
145
157
214
269
policy proposal
Make marriage penalty tax cut permanent
Impose requirements on SDI
Support deployment of SDI
Praise success of SDI test
Prohibit SDI funding
Bar funds for nuclear SDI
Prohibit debate on ABM treaty
Cut funding for SDI missile silos
1997
1999
2001
569
521
135
Authorize states to use vouchers
Authorize voucher program for students in low-performing schools
Vouchers for students in poor-performing schools
1997
1998
1999
1999
1999
1999
1999
1999
1999
1999
2000
2000
2001
233
58
49
100
101
103
119
183
189
266
89
193
246
Prohibit funds for Bosnia
Withdraw troops from Bosnia
Authorize Kosovo deployment
Prohibit ground forces in Kosovo without Congressional authorization
Remove U.S. forces from Kosovo
Authorize air operations in Kosovo
Bar funds for invasion of Kosovo by U.S. forces
Prohibit permanent Defense Department presence in Haiti
Strike provision to prohibit funding for Kosovo
Recognize achievement in Kosovo
Withhold funds for Kosovo
Require burden-sharing in Kosovo
Bar financial contributions to UN peacekeeping efforts
1997
1998
1999
2000
2000
2000
2001
2002
2002
2002
2002
228
10
118
70
85
196
172
141
142
158
194
Cut funding for B-2 bombers
Override veto of defense construction spending
Reduce proposed supplemental defense spending increase
CBC budget resolution: defense to domestic spending
Increase supplemental defense spending
Terminate Trident II missile
Cut Air Force budget
Prohibit funding for nuclear earth-penetrator weapon
Repeal ban on developing low-yield nuclear weapons under some conditions
Increase defense spending
Supplemental military appropriation
1999
2001
2001
319
143
238
Increase funds for teachers
Reduce increase in education aid to localities
Funding for disabled students
42