Degrees of Democracy
Legislative Institutions and the Opinion-Policy Link
Stuart N. Soroka
Christopher Wlezien
McGill University
Montréal, Québec
Temple University
Philadelphia, PA
E-mail: [email protected]
E-mail: [email protected]
Abstract
A central function of legislatures in representative democracies is the development of policy that is
broadly representative of the public interest. It follows that one critical measure of the success of
legislative institutions is the extent to which they encourage (or discourage) the representation of
public preferences in policymaking. This paper explores this theme. First, we develop a theoretical
model implying differences in representation across countries owing to differences in legislative
institutions. Most importantly, we argue that (1) policymakers are less responsive in parliamentary
systems than in presidential systems, and (2) public responsiveness to policy change is diminished in
federal rather than unitary systems. We then test these propositions, using directly comparable
longitudinal measures of public preferences and government spending in the US, UK, and Canada. In
the process, we draw conclusions about the success of these countries’ legislative institutions, and
distil broader implications for the functioning of modern democracies.
Paper prepared for presentation at the meetings of the International Political Science Association, Fukuoka,
Japan, July 2006. Portions of this research were presented at the Annual Meetings of the Southwest Political
Science Association, Corpus Christi, Texas, March, 2004, the American Political Science Association,
Philadelphia, August, 2003, the Midwest Political Science Association, Chicago, April 2003, the Canadian
Political Science Association, Halifax, June 2003, the Political Studies Association, Edinburgh, April 2002, the
American Political Science Association, Boston, August-September 2002, and the Annual Conference of the
Elections, Public Opinion, and Parties Group, Salford, UK, September 2002. We are thankful to various
discussants and audiences for comments, especially Kevin Arceneaux, Keith Banting, Frank Baumgartner, Chris
Carman, Diana Evans, Peter John, Martin Johnson, Bryan Jones, Mark Kayser, Jose Maria Maravall, Iain
McLean, Jose Ramon Montero, Robert Shapiro, Robert Stein, and James Stimson.
Related project papers are available online at http://www.degreesofdemocracy.mcgill.ca.
Perhaps the primary concern of empirical democratic theory is the relationship between
public preferences and public policy. Scholars have stressed the importance of ‘responsive
rule’—the correspondence between citizens’ preferences and government actions—in
democratic governance (the literature is massive, but see, e.g., Pitkin 1967; Dahl 1971;
Saward 1994; Burstein 1998; Manza and Cook 2002). Yet scholars also have questioned
whether citizens are sufficiently informed and/or reactive enough to play a role governing the
state (Plato’s Republic; Mill 1861; Lippmann 1925; Schumpeter 1950; Converse 1964; Page
and Shapiro 1992; Dahl 1998). This work highlights the importance of the interrelationships
between public opinion and policy, both from opinion to policy and also from policy to
opinion. It implies that those interested in studying democracy should be concerned with
policy representation—whether and how policy follows public preferences—and public
responsiveness—whether preferences react to policy itself.
There already is a large and growing body of work on the correspondence between public
opinion and policy behaviour. 1 And work on ‘thermostatic’ models of the opinion-policy
relationship suggests not just that policymakers respond to the public, but that the public
adjusts its preferences over time in reaction to policy change (Wlezien 1995, 1996a; Erikson
et al. 2002). This work is important. Narrowly conceived, it suggests that policymakers are
attentive to public preferences and that the public is aware of and reactive to policy change, at
least in certain domains. More broadly conceived, it offers empirical evidence on the nature
and quality of representative democracy.
Most empirical work to date focuses only on the US, though there is a growing body of work
examining opinion-policy connections elsewhere. 2 Owing largely to data limitations,
however, there are relatively few direct comparisons of results across countries. Comparative
studies of opinion-policy links are nevertheless fundamental to our understanding of how—or
if, or how well—democracies work. Indeed, differences in representation (and public
responsiveness) across countries may be linked to a variety of institutional factors, such as
the level of federalism, the relative power of the executive and the legislature, and party
competition itself. Highlighting differences in the opinion-policy link and connecting these
with institutional differences may therefore offer crucial information on the comparative
effectiveness of political institutions.
The current paper examines political institutions as mediators of the opinion-policy link.
Using a body of data on public opinion and government spending, we extend research on the
dynamics of spending opinion and policy in the US to Canada and the UK. Specifically, we
explore (1) public responsiveness—the degree to which the public responds
‘thermostatically’ to public expenditures, and (2) policy representation—the extent to which
budgetary decisions reflect public preferences for spending. Our results provide insight into
the comparative dynamics of representation and responsiveness across policy domains, and
across countries. We begin, however, with our theoretical model of opinion and policy, and
1
The literature is vast, but see, e.g., Miller and Stokes 1961; Weissberg 1976; McCrone and Kuklinski 1979;
Monroe 1979; Bartels 1991; Page and Shapiro 1992; Hartley and Russett 1992; Erikson, Wright, and McIver
1993; Goggin and Wlezien 1993; Jacobs 1993; Stimson, MacKuen, and Erikson 1995; Wlezien 1996a; Wood
and Hinton-Anderson 1998; Hill and Hurley 1998; Smith 1999; Sharpe 1999; Erikson, MacKuen, and Stimson
2002; Soroka 2003; Wlezien, 2004. See also Weakliem’s 2002 excellent review and assessment of the
literature.
2
See, e.g., Brooks 1987, 1990; Petry 1999; Franklin and Wlezien 1997; Eichenberg and Stoll 2003; Soroka
2003; Soroka and Wlezien 2004, 2005.
2
an outline of the ways in which opinion-policy congruencies may be enhanced—or
hindered—by political institutions.
The Thermostatic Model of Opinion and Policy
The representation of public opinion presupposes that the public actually notices and
responds to what policymakers do. Without such responsiveness, policymakers would have
little incentive to represent what the public wants in policy—there would be no real benefit
for doing so, and there would be no real cost for not doing so. Moreover, without public
responsiveness to policy, expressed public preferences would contain little meaningful
information. There not only would be a limited basis for holding politicians accountable;
expressed preferences would be of little use even to those politicians motivated to represent
the public for other reasons. We need a responsive public. Effective democracy depends on
it.
A responsive public behaves much like a thermostat (Wlezien 1995). 3 That is, the public
adjusts its preferences for ‘more’ or ‘less’ policy in response to what policymakers do. When
policy increases (decreases), the preference for more policy decreases (increases). For
expository purposes, the public can be viewed as a collection of individuals distributed along
a dimension of preference for policy activity, say, spending on defence. This characterization
is not meant to imply that individuals have specific preferred levels of spending in mind;
rather, it is intended to reflect the fact that some people want more than others. The ‘public
preference’ is then represented by the median along this dimension, which implies a certain
‘ideal’ level of defence spending.
Now, if the level of policy differs from the level the public prefers, the public favours a
corresponding change in policy, basically, either more or less. If the preferred level is greater
than policy itself, the public favours more spending than currently is being undertaken. If
policymakers respond, and provide more (but not too much) for defence, then the new policy
position would more closely correspond to the preferred level of spending. If the public is
indeed responsive to what policymakers do, then the public would not favour as much more
activity on defence. It might still favour more, on balance, but not as substantially as in the
prior period; if policymakers overshoot the public's preferred level of spending, it would
favour less. In effect, following the thermostatic metaphor, a departure from the favoured
policy temperature (which itself can change) produces a signal to adjust policy accordingly
and, once sufficiently adjusted, the signal stops. This conceptualization of public preferences
has deep roots in political science, including Easton's (1965) classic depiction of a political
system and Deutsch's (1963) models of ‘control.’
These expectations can be expressed formally. The public’s preference for ‘more’ policy—
its relative preference, R—represents the difference between the public’s preferred level of
policy (P*) and the level it actually gets (P):
P
Rt = P*t - Pt
(1)
P
Thus, as the preferred level of policy or policy itself changes, the relative preference signal
changes accordingly. The public is expected to respond currently to actual policy change
when put into effect (at t). This is straightforward, at least in theory. It is less
straightforward in practice.
3
This discussion closely follows Wlezien (1995: 981-983).
3
Most importantly, we typically do not directly observe P*. Survey organizations typically do
not ask people how much policy they want. Instead, these organizations ask about relative
preferences, whether we are spending ‘too little,’ whether spending should ‘be increased,’ or
whether we should ‘do more.’ This, presumably, is how people think about most policies.
(Imagine asking people how much health or education spending they want.) The public
preference, however defined, also is necessarily relative. In one sense, this is quite
convenient, as we can actually measure the thermostatic signal the public sends to
policymakers.
P
We can rewrite the model of Rt as follows:
Rt = a + β1Pt + β2Wt + et,
(2)
where a and et represent the intercept and the error term, respectively and W designates the
instruments for the public’s preferred level of policy (P*). Note that levels of relative
preferences are expected to be associated with current levels of policy; if the thermostatic
model applies, the coefficient (β1) relating the two is expected to be less than 0.
Now, if policymakers are responsive to these public preferences, changes in policy (ΔP) will
be associated with lagged (t-1) levels of the public’s relative preference (R), as follows:
ΔPt = ρ + β3Rt-1 + β4Zt-1 + μt,
(3)
where ρ and μt represent the intercept and the error term, respectively, and Z represents the set
of other determinants of policy such as the partisan control of government. The coefficient β3
captures responsiveness, where the effect of preferences on policy is independent of partisan
control and other factors; if the coefficient is greater than 0, policy ‘responds’ to
preferences. 4 Notice that the change in expenditure for fiscal year t is modelled as a function
of net support in year t-1. This specification is not meant to imply that policies do not
respond to current opinion; rather, it is intended to reflect the reality of budgetary decisionmaking, which largely happens over the course of the previous fiscal year (see Wlezien
1996b; Wlezien and Soroka 2003). Thus, this specification captures responsiveness to
opinion when most budgetary decisions actually are made.
These expectations are general ones and we do not expect the model to apply in all policy
domains in all countries. Indeed, public and policy responsiveness is likely to reflect the
political importance (or ‘salience’) of the different domains, if only due to possible electoral
consequences. 5 Following Wlezien (2004), we might expect the pattern of representation to
be symmetrical to patterns of public responsiveness: where the public notices and responds to
policy in a particular domain, policymakers will notice and respond to public preferences
themselves; where the public does not respond to policy, policymakers will not represent
public preferences. This may help us understand certain patterns of policymaker behaviour.
For instance, it might tell us why one government is more responsive than another
government to public preferences. It also may tell us why one government is more
responsive to preferences for policy in some areas but not others. But it tells us little about
the behaviour of the public. Would we expect the public to behave differently in different
4
This does not mean that politicians actually respond to changing public preferences, for it may be that they and
the public both respond to something else. All we can say for sure is that the coefficient (β3) captures policy
responsiveness in a statistical sense, that is, whether and the extent to which public preferences directly
influence policy change, other things being equal.
5
We can explicitly incorporate salience (S) into our model as follows:
ΔPt = ρ + β3 St-1Rt-1 + β4Zt-1 + μt,
where S ranges between 0 and 1. Here the effect of opinion on policy depends on the level of salience. This
follows Franklin and Wlezien (1997); also see Jones (1994) and Soroka (2003).
4
countries? Under what conditions? And taking these differences into account, might we not
expect governments in some countries to pay relatively less attention to public preferences
than do governments in other countries? What is the structuring role of institutions?
Political Institutions
Polities differ in many ways, and some of these differences should have significant
implications for the nature and degree of public and policy responsiveness. Of fundamental
importance are mass media openness and political competition, for instance. Mass media
openness is essential in modern democracies – the role of media in democracy has been
highlighted by both political theorists (e.g., Dahl 1971) and media analysts (e.g., Arterton et
al. 1990). Without an open media, people cannot easily receive information about what
government actors do, and are less able to respond to policy change.
Likewise, governments have less incentive to respond to public opinion in the absence of
some level of political competition; at the very least, the incentive would be less reliable. 6 Of
course, elections are no guarantee of competition (Golder 2004), and even where we have a
degree of competition in modern democracies the level varies non-trivially. But how much
competition do we need? We likely don’t need to have perfect competition – there is reason
to think that a little competition may go a long way. But the absence of competition quite
clearly removes a critical incentive for policy responsiveness.
Political institutions may also be particularly important. Indeed, the literature on comparative
political institutions suggests that institutions will be very powerful mediators of opinionpolicy connections. There are two institutions that seem especially significant: electoral
systems, and the division of powers.
Electoral Systems
Even where we have basic levels of media and political competition, institutional differences
may have important implications for public responsiveness and policy representation.
Electoral systems are the emphasis of most of the existing relevant literature. Most of this
research focuses on differences between majoritarian and proportional visions, using
Powell’s (2000) language, and mostly on how these differences matter for policy
responsiveness. Let us briefly review the literature in the area, focusing on the two leading
exponents. 7
Lijphart (1999) provides one of the clearest statements on the subject. In his recent and
influential book, he assesses the performance of consensual systems (characterized by
proportional representation, multiparty systems, and coalition governments, as well as
federalism, strong constitutions, and executive-legislative separation) versus majoritarian
systems (characterized by simple plurality election rules, a two-party system, and single-party
government, as well as unitary government and an uncodified constitution). Lijphart
compares the relationship between consensual versus majoritarian government and number of
outcome variables, including, for instance, social welfare policy, foreign aid, women’s
parliamentary representation, and turnout. Results lead him to conclude that the consensual
approach is superior to the majoritarian one.
6
Political competition also may be essential for the flow of information to the public as well. That is, in the
course of election campaigns and policy debates, people learn about what politicians actually have done and are
doing. Consider the now voluminous literature on negative campaigning in the US beginning with
Ansolabehere and Iyengar (1995).
7
For a classification of systems around the world, see Golder (2005).
5
Powell (2000) provides further support of Lijphart’s hypothesis, though he focuses
specifically on differences between majoritarian and proportional election rules and their
implications for representation. Relying on expert surveys relating to party positions and
opinion surveys of the public in different countries, he shows that proportional representation
tends to produce greater congruence between the government and the public. Specifically,
the general ideological complexion of policy and the ideological bent of the electorate tend to
be better matched in proportional systems.
According to Powell, this reflects the greater direct participation of constituencies the vision
affords. The mechanism is the bargaining required to form a majority. Whereas majoritarian
rules lead parties to coalesce in advance of a election, thus producing a clear majority on
Election Day, proportional rules disperse power, thus requiring a set of parties to come
together after an election to form a government. In reaching agreement, the resulting
coalition effectively averages across the component party positions, thereby producing
moderation. In majoritarian systems, governments tend to be further off-center. Perhaps
parties in these systems do not reliably move to the ideological center but instead tend to
position off to the left and right. It also may be that parties do compete for the center but that
voters do not consistently select the most centrist party, that is, election outcomes may reflect
other things, such as economic growth. Regardless of the exact underpinnings, the main
result is clear: Proportional elections produce more representative governments than do
majoritarian systems.
Powell’s results pertain to elections and their immediate consequences. But, what about in
the periods between elections? What if public opinion changes after an election? Are
coalition governments more responsive to the change in opinion? Or is it that the singleparty governments in majoritarian systems are equally, if not more responsive?
During the periods between elections, there is reason to think that governments in
majoritarian systems are actually more responsive to opinion change. There are at least two
reasons. First, it presumably is easier for a single party to respond to changes than a multiparty coalition: coordination in the latter is more difficult and costly. Second, majoritarian
governments may have more of an incentive to respond to opinion change. This generalizes
Rogowski and Kayser’s (2002) argument relating to the comparatively higher seats-votes
elasticities in majoritarian systems, namely, that governments in those systems more
responsive to consumer interests than producer interests. Since a shift in electoral sentiment
has bigger consequences on Election Day in majoritarian systems, governments there are
likely to pay especially close attention to the ebb and flow of opinion. Thus, it may be that
both systems work to serve representation, but in different ways, where proportional systems
provide better indirect representation via elections and majoritarian systems better direct
representation in between elections.
Vertical Divisions of Power
As electoral systems may matter, so too may government institutions. In particular, we
expect that the division of powers—both vertical and horizontal—may structure the
relationships between opinion and policy over time. In short, we argue that:
(1) a vertical division of powers, or decentralization, makes it more difficult for the
public to gauge and react to government policy change, and thus dampens public
responsiveness; and,
(2) a horizontal concentration of powers, or parliamentary government, makes
politicians less responsive to changes in public opinion.
6
Let us first consider the vertical division, or federalism, and its consequences for public
responsiveness to policy. We know that thermostatic public responsiveness requires that
people acquire accurate information about what policymakers are doing. This clearly
depends on the supply of information, as we have discussed. It also depends on the clarity of
that information. More precisely, it depends on the extent to which responsibility for policies
is clear, and this is in part a function of how government itself is organized. Federalism, the
vertical division of powers, increases the number of different governments making policy and
thus makes less clear what ‘government’ is doing (see, e.g., Downs, 1999). Put differently,
the government policy signal may be confused—or, rather, there may be different signals
from multiple sources— at least in policy domains for which different governments have
responsibility. This clearly can dampen public responsiveness.
Federalism can of course come in different forms, varying not just across countries but also
across policy domains. In some cases, separate levels of government will have complete
control for different domains – what is in the American context referred to as dual federalism
(textbooks often invoke the image of a layer cake). In this design, different governments
have different responsibilities and there are no interactions between layers. Much like the
unitary system, there is no mistaking the source of policy in each policy area. It’s just that
the source differs. We thus might expect a comparatively high level of responsiveness on the
part of the public (in politically important domains, at least). This does presume that the
availability of information about the behaviour of governments is fairly equal across levels of
government, however, and this may not be true for at least two reasons: (1) the behaviour of
the national government may receive more attention than lower-level governments; (2) the
flow of information about lower-level governments may vary meaningfully geographically,
e.g., across counties, states, and municipalities.
Even so, responsiveness is more likely in systems characterized by dual federalism than it is
in domains where policymaking within domains is divided between governments. Most
federal systems involve different governments sharing responsibilities in policy areas, for
instance, often referred to as cooperative federalism. (The textbook image is of a marble
cake.) There are two ways in which different levels of government can be involved. One
involves direct involvement by multiple levels. In this arrangement, there are different
sources of policy change and implementation, which can create fairly obvious complications
for the public, especially if policy change over time is not parallel across levels of
government. It may not be that the public is less informed about the sum of policy across
levels of government, though this is a possibility. There is very good reason to think that the
public will be less informed about the behaviour of specific levels, however, as this would
require keeping what’s happening at each level perfectly straight. Where there are
overlapping jurisdictions, we expect a dampening of public responsiveness of preferences for
policy at any particular level of government. For instance, we might expect that individuals’
preferences for more policy at the “national” level are less responsive to changes in national
policy. We would expect the same at lower levels of government as well.
Things may be even worse where one government transfers money to other governments.
Here, multiple governments are involved but the actions of one government are not directly
visible, at least in the actual delivery of services. The nature of transfers should affect the
clarity with which the public can attribute responsibility. In some cases, the recipient
government still may have not much policy discretion. For instance, the recipient
government may simply be the conduit for the transfer of public money, as is the case with
grants-in-aid in the US. And even where the recipient is not a mere go-between, there may
be little discretion, as with categorical grants. In these cases, federalism may make relatively
7
little difference to the public, as change in policy at one level largely—though nevertheless
imperfectly—drives changes in policy at another level.
Things are potentially quite different with block grants or revenue sharing. These
mechanisms allow lower-level governments substantial discretion; indeed, revenue sharing
comes with absolutely no strings. This situation is not very common in the US, at least since
1986. While the national government may provide a broad direction for policy change at
lower levels, what happens at these levels may differ quite a lot from place to place. Largely
unconditional block grants remain very common in Canada, however, where provinces have
ostensibly exclusive jurisdiction over many policy domains, but quite limited taxing power.
The Canada Health and Social Transfer (CHST), for instance, was a massive block grant
from the federal government to provinces in order to help cover the cost of healthcare,
education and welfare, and had very few strings attached. In this situation, it is very often not
clear to the public which government is responsible for what policy. Policy in one province
may have gone up a lot, but does this mean we want the national government to do less? Or
do we think that lower-level governments should do less? Or both? Indeed, the truth might
be that we “should” think that the national government needs to do less and the local
government more. Making such an assessment is, however, rather difficult – we simply
cannot tell which government is principally responsible for a given shift in spending.
—Figure 1 about here—
We thus have pretty clear expectations about the effects of federalism on public
responsiveness to policy. These are illustrated in Figure 1. First, responsiveness to policy at
any particular level of government should be generally lower in federal systems than in
unitary systems. Second, in federal systems, responsiveness should be especially low in
domains where multiple governments are involved in policy. As the mix of responsibility
increases, public preferences for policy at any particular level of government will be
correspondingly less meaningful.
This may have implications for policy representation itself. As we already have discussed,
there is good reason to think that we will have clear policy responsiveness only where we
observe clear public responsiveness to policy. Some (Downs, 1999) also have argued that
federalism actually increases net representation of broad policy preferences, as multiple
levels of government provides greater opportunity. This possibility is perfectly consistent
with our conjecture regarding representation at particular levels of government: it may be that
the representation of opinion at each level of government is lower but representation across
levels of government is higher. Unfortunately, with our measure of broad preferences for
‘government’ policy, we can’t explicitly consider the possibility.
Horizontal Divisions of Power
Now let us consider the horizontal division of powers (between the executive and legislature)
and government responsiveness to public opinion. There are really two main forms of
executive-legislative organization: parliamentary and presidential. In the former, which is
much more common among advanced industrial democracies, the executive is traditionally
chosen from within the parliament and it serves only with the confidence of parliament. 8
That is, the government depends on the support of the parliament to stay in power—in effect,
the executive and legislative branches are fused. This concentration of powers is the defining
characteristic of what is often called ‘responsible government’. (When put together with a
majoritarian electoral system, we have ‘responsible party government’.)
8
For a focused discussion of the differences and their implications, see Strom (2004).
8
In the presidential model, conversely, powers are divided into separate executive and
legislative branches, each separately elected and endowed with separate powers. The
separate institutions thus share power, where each can ‘check or balance’ the other, at least
with respect to statute. Neither branch can effectively act without the other, except where the
legislative support is overwhelming. And even then, implementation is in the hands of the
executive.
—Figure 2 about here—
This difference in institutional arrangements may matter for government responsiveness to
public opinion. In short, there is reason to suppose that parliamentary governments are less
reliable in their attendance to public opinion (see Figure 2). Scholars have long noted the
dominance of cabinets over parliaments (see, e.g., the classic statements by Bagehot 1867
and Jennings 1959; also see Laver and Shepsle 1996); these scholars portray a world in which
cabinet governments exercise substantial discretion. Cox (1987) attributes the role of
cabinets it to heightened electoral competition for seats over time. Other research provides a
more institutional motivation (e.g., Tsebelis 2002). First, in contrast with most presidential
systems, the executive in parliamentary systems typically is the proposer. The government
puts its legislation to the commons. Second, the legislature has only a limited check on what
the executive does. It has what Lupia and Strom (1995) call a “doomsday device”: the no
confidence vote. This is a very big check, obviously, but it can be very costly for the
majority party; undertaking such a vote requires that governing party legislators are willing to
face an early election. Indeed, Huber (1996) finds that the threat of a no confidence vote can
be used to the cabinet’s advantage, and allows it to effectively control the legislative process.
Strom (2004) highlights another doomsday device—the dissolution power—which allows the
Prime Minister or cabinet to threaten the parliament with an immediate electoral competition.
Ultimately, according to Strom, parliamentary government deals much better with “adverse
selection” than it does “moral hazard.” Once established, the cabinet is difficult to control on
a recurring basis.
This has fairly clear implications for government responsiveness. When there are
differences between what the cabinet and parliament want, the latter cannot effectively
impose its own contrary will. It cannot even easily veto. Perhaps more importantly, the
legislature cannot consistently undertake ‘error correction,’ that is, adjusting the
government’s position where it may be going too far or not far enough given public
preferences. This is of particular relevance given the independence of individual ministers in
the policymaking process (Laver and Shepsle, 1996). Some may be more conservative than
others, some more liberal. All can make mistakes. The process of amendment and veto is
crucial. In these parliamentary systems, however, this can be somewhat compromised, at
least by comparison with presidential systems. In the latter, the executive cannot effectively
act without the legislature. The legislature proposes, and bills are the result of compromise
within and, usually, across assemblies. The executive can veto legislation but, with sufficient
support, the legislature can impose its will. Although the separation of powers makes
presidential systems much more deliberate in their actions, then, it would appear to make
them more reliably responsive to public opinion over time. We still expect representation in
parliamentary systems, of course. After all, governments in these systems are more easily
held accountable for their actions, as responsibility is far clearer, particularly in a majoritarian
context. In between elections, however, there is little to make cabinets accountable except
for the prospect of a future electoral competition. Though important, the incentive is
imperfect.
9
The horizontal division of powers also may be important to public responsiveness. On the
one hand, past work suggests that executive-legislation separation confuses responsibility for
government action (especially see Franklin 2004). People may thus notice policy change
without knowing who is responsible. On the other hand, some scholars argue that horizontal
separation actually makes it easier to elicit information about what governments do (Persson,
Roland, and Tabellini 1997). According to Persson, et al., separation induces competition
among institutional actors. 9 Strom (2004) echoes much the same point, focusing on
parliamentary systems and the collusion that copartisanship in those systems encourages.
Following this line of argument, we might expect greater public responsiveness in
presidential systems, as information about policy change should be more readily available.
We cannot be sure that the additional information clarifies more than it confuses, however, as
different actors advocate different positions as the process unfolds. We can entertain these
possibilities, but cannot be sure about one expectation or the other.
In sum, there are various reasons to expect patterns of public and policy responsiveness to
differ across countries and policy domains. In particular, we have hypothesized that public
responsiveness will reflect the division of government powers, particularly the vertical
division. Public responsiveness should be greatest in unitary systems and least in highly
federalism countries, especially in those policy domains where multiple governments are
involved. We also have hypothesized that government responsiveness will reflect the
division of powers, especially the horizontal separation, as well as the proportionalmajoritarian quality of the electoral system. Indeed, government responsiveness should be
greatest in presidential systems with proportional representation, an unusual combination. It
should be least in majoritarian parliamentary systems, which themselves are not that
common. The other two, much more typical, cases—majoritarian presidential and
proportional parliamentary—should lie somewhere in between. It is not clear in what order,
however. Given Powell’s work, we might expect that the latter better serves representation.
In effect, proportionality would powerfully mitigate the autonomy of cabinet governments. 10
In theory, we can explicitly test these hypotheses. The problem is that we have data for but a
handful of countries, namely, Canada, the UK, and the US. These countries are not randomly
drawn—to a large extent, they are ‘most similar.’ This may work in our favour. Notice first
that the countries are majoritarian systems. In effect, the case selection controls for this
difference to begin with. This leaves important differences with respect to government
institutions, however. First, the countries differ greatly on the federal dimension (Joumard
and Kongsrud 2003). The UK is a highly unitary system, among the most unitary modern
democracies at least until recent moves toward devolution. The US and Canada meanwhile
are highly federalized systems, especially Canada, which, along with Switzerland, may be
most in opposition to the UK. Second, the countries differ greatly with respect to separation
of powers. The UK and Canada are classic parliamentary systems with unified powers. The
US is the classic presidential system with separation of powers.
Responsiveness and Representation Combined
Thus far we have considered public responsiveness and policy representation independently.
The two are intimately connected, however. In particular, policy representation is dependent
9
Persson et al. argue further that this can improve political accountability.
There is an alternative view, at least a more complex one. For instance, Rogowski and Kayser (2002) argue
that the comparatively higher seats-votes elasticities in majoritarian systems make governments in those systems
more responsive to consumer interests than those of producers.
10
10
on a certain degree of public responsiveness; it follows that the institutional factors that tend
to dampen public responsiveness (federalism) may also indirectly decrease responsiveness.
—Figure 3 about here—
Generally speaking, representation and responsiveness combine to create a kind of ‘system
efficiency’, where efficiency – following Deutsch – captures the speed and accuracy with
which a political system adapts to a shock and returns to a state of equilibrium. That is,
efficiency is a combination of policymakers’ reactions to preferences and the public’s
capacity to recognize this policy shift and adjust relative preferences accordingly. Efficiency
is conditioned by the political institutions discussed above, of course, as well as issue
salience. Figure 3 presents the general structure. Here, policy systems – or policy domains
within systems – are plotted on two horizontal axes, one for the horizontal division of powers
(at the back), and the other for the vertical division of powers (on the left). A given system’s
location on each axis has bearings on our expectations regarding responsiveness and
representation: representation will tend to decline as a system becomes more parliamentary,
and responsiveness will tend to decline as a system becomes more federal. Note that in this
figure efficiency declines more quickly with federalism than with unitary rather than
presidential government. This is purposeful, and meant to reflect the possibility that
incentives for representation will decline alongside public responsiveness, while in contrast
responsiveness need not decline alongside representation.
These are only very general expectations, of course, as we have little basis for being specific
about how steeply, or how asymmetrically, efficiency declines along either axis. In
addition, issue salience will tend to move the entire ‘efficiency plane’ upwards or
downwards. That is, more salient domains will tend to exhibit more responsiveness and
representation overall, though divisions of power will still matter in roughly similar ways.
The same is true for low salience issues, though in this case the efficiency plane moves
downwards. The whole story is not quite this simple, of course, as Figure 3 does not deal
with electoral systems. Nevertheless, most of our general expectations for representation and
responsiveness across political systems are captured in this figure.
Figure 3 (and the hypotheses discussed above) has clear implications for the study of the
performance of policymaking institutions. Most importantly, it makes clear that we need to
consider both sides of the opinion-policy relationship. Preferences should inform policy
change. But policy change, in turn, should inform preferences. Indeed, reactive preferences
provide the critical motivation for policy representation. The ‘performance’ of policymaking
institutions, we suggest, can thus be conceived as a function of the extent to which these
institutions facilitate both policy representation and public responsiveness.
Thinking about the three countries on which our analyses rely, our expectations are most
clear for the public side of the equation, where the dimensionality is essentially singular.
That is, we expect public responsiveness to policy will vary with federalism and thus will be
greatest in the UK, lesser in the US, and lesser still in Canada. 11 Similarly, within countries,
domains that are particularly federal (such as US education spending) will show less public
responsiveness than domains that are less federal (such as US welfare spending). On the
government side, things are slightly more complex, depending on the role federalism plays.
If federalism has limited effects, we expect representation of opinion to reflect the separation
of powers and be greatest in the US and lesser in Canada and the UK. If federalism also has
11
Effects owing to separation of powers might change things dramatically. If separation decreases information,
we would expect differences between countries to increase, especially between the US and UK; if separation
increases information, we would expect relatively little difference between the three countries.
11
effects, we might expect a different pattern, where the differences between the countries are
seriously attenuated. In the extreme, where only federalism matters, we would expect
government responsiveness to perfectly parallel our expectations about public
responsiveness. Although we cannot fully examine the effects of all possible institutional
differences, an analysis of these three countries thus may offer important insights into the
structuring roles of institutions. Let us see what the data reveal.
Data
The dataset used here includes comparable measures of budgetary policy and public
preferences for spending in various policy domains over time, in Canada, the UK and US.
Our work elsewhere has examined opinion-policy connections across a number of different
policy domains, including defense, welfare, health, education, the environment,
transportation, and cities. Here, we focus on just a few domains for which data are available
all three countries: defense, health, and education. In addition, we look at results for welfare
in the US and Canada.
Measures of spending preferences rely on a common question, asked relatively consistently
in all three countries. The question is as follows: “Do you think the government is spending
too much, too little or about the right amount on [healthcare]?” The question thus asks about
general preferences for government spending, not preferences for spending at particular
levels of government per se. 12 Respondents are asked about spending in various categories
besides healthcare – just three others in the UK, but eight more in the US and seven more
(consistently) in Canada. The question is also asked with varying frequencies across
countries: in Canada, Environics asked these questions between 14 and 16 times (depending
on the spending domain) from 1984 to 2002, and some missing years can be filled in using
similar data from Pollara; 13 in the UK, Gallup asked the question 19 times in only 13 years
and not at all since 1995; 14 in the US, the GSS includes these questions almost every year
from 1973 to 1994, and then every other year until 2004. Using these data, we can create
time series covering from 16 to 23 fiscal years between 1976 and 2004, though note that there
are gaps in many of the series, which we address below.
The simplest, most reliable way to reflect public preferences using these data is to create
percentage difference measures, by subtracting the percentage of people who think we are
spending ‘too much’ from the percentage of people who think we are spending ‘too little’ in
each domain. The measures of ‘net support’ (Rt) thus capture the degree to which the public
wants ‘more’ or ‘less’ spending over time – indeed, the measures capture both direction and
magnitude.
Our measures of budgetary policy (Pt) rely on estimates of spending by function in Canada,
the UK and US. Finding directly comparable data is not easy – unlike in the US, for instance,
12
In Canada, people actually are asked about the ‘federal’ government, which is of some relevance and
presumed benefit, given the theoretical discussion above; that is, it would appear possible to assess the explicitly
national-level responsiveness of both the public and government. Ironically, it is difficult to fully disentangle
national and provincial spending in many Canadian domains, especially the highly decentralized ones (see
Soroka and Wlezien 2004).
13
Environics asked questions about most policy domains from 1984 to 2002. Data are missing in 1986, 1992 ,
and 1996; for domains in which Pollara also asked questions, 1996 data are filled in using Pollara results. (For
further details and a comparison of the Environics and Pollara series, see Soroka and Wlezien 2004.)
14
Gallup did not ask the question in 1980, 1981, 1984, 1987 and 1994, but we are able to use measures from
proximate periods for years 1981 and 1987. Data remain missing in the other three years. Note also that when
more than one poll exists in a single fiscal year, results are averaged.
12
Canadian and UK governments do not provide data on appropriations of budget authority at
any level of aggregation. All that is available is data on expenditures (outlays). Although
expenditures surely are important, they are not policy per se. That is, elected politicians have
only limited control over what is spent, which may reflect things that they cannot anticipate
or manage. Using expenditures thus biases analyses against finding opinion representation
(see Wlezien 1996a; Wlezien and Soroka 2003).
To provide comparability across countries, we rely on expenditures in each of the countries.
US outlays are drawn directly from the Historical Tables in the 2003 Budget, and the specific
definitions of the functions used are described in Wlezien (2004). Our measures of functional
spending in the UK rely on a new database that solves the problem of inconsistent functional
definitions that complicates the use of data available in HM Treasury’s Public Expenditure
Statistical Analyses; details on these data are available in Soroka, Wlezien, and McLean
(2006).
Functional spending figures in Canada are available from various Statistics Canada CANSIM
matrices; details on these data are available in Soroka and Wlezien (2004.). Spending data in
Canada present a particular problem: health, education and welfare are in large part funded
through large federal-provincial block transfers that are not allocated to any one domain until
they reach the provinces (and provinces have considerable discretion). The result is that it is
impossible to measure ‘federal’ spending in each of these individual domains. We
accordingly rely on ‘consolidated’ (federal and provincial) spending for individual domains
in Canada. These consolidated data account for expenditures through block transfers after
they have been allocated by provinces, and thus can be linked to a single policy domain. We
also estimate models for domestic domains together, however, and in this case we can use
federal spending independently. To do so, we combine direct federal spending in health,
education and welfare, as well as federal block transfers to provinces for spending across
these three domains. We still cannot separate out the proportions of these transfers that are
spent on each of health, education and welfare; when we look at all three together, however,
this doesn’t matter. This direct comparison of results using consolidated or federal spending
is quite telling, as we shall see.
—Figure 4 about here—
The resulting spending and preferences data are shown in Figure 4. The figure indicates that
spending on the different functions clusters together, at least in levels. A lot more is spent on
some programs than others; not surprisingly, the latter seem to change only very deliberately
over time and the former show more variance. Spending in most domains does tend to trend
upward over time, documenting a well-known pattern of government growth. Note also that
spending on defense tends to move in the opposite direction, at least in the US and UK,
implying a guns-butter trade-off.
Public preferences for spending exhibit similar structure. Preferences in domestic domains
tend to flow together over time in each country. In contrast with what we see for spending,
the preference series are more strictly parallel, that is, they show very similar variance.
Preferences for defense spending also largely mirror domestic preferences in the UK and US.
Thus, spending and public preferences in the different domains tend to move in the same
liberal-conservative direction over time in each country. These results all are potentially
quite telling for our analysis. Indeed, the patterns imply a certain ‘global’ movement of
opinion that may drive politicians’ behavior in various policy domains. The movement is not
entirely global, however. That is, some of the variance of spending and preferences is
domain-specific. This is easy to overlook but of real importance, as our past work has
demonstrated. Now, let us examine how preferences and spending relate to each other over
13
time.
Using these data, we can estimate equations 2 and 3 (above). For the analysis of feedback
(equation 2), the dependent variable is the current level of net support for spending. To
preserve precious degrees of freedom, we impute values when opinion data is missing using a
straight linear interpolation. This actually has relatively little consequence for the general
pattern of results; that is, it mostly increases the reliability of estimated parameters. These
series are regressed on current corresponding levels of spending (in billions of inflationadjusted 2000 national currency units) and the instruments for the public’s preferred level of
spending. 15
Regarding these instruments, we know that public preferences for defense spending in the US
have closely followed perceived variation in national security over time. Specifically, until
1989 preferences reflected attitudes toward the Soviet Union/Russia as indicated by the
‘likes’ and ‘dislikes’ of that country (Wlezien 1995). There is reason to expect a similar
pattern to hold in the UK and Canada. The problem is that we do not have comparable
measures of likes/dislikes of Russia in those countries. However, if the measure from the US
provides a reliable indication of the actual security threat from the Soviet Union during the
period, or broad public perceptions of that threat, then it should work equally well in the UK
and Canada. We thus use the US measure there. The measure of Net Dislike represents the
percentage of Americans who dislike Russia minus the percentage who like the country. The
data are drawn primarily from the General Social Survey (GSS) but are supplemented using
American Institute for Public Opinion (AIPO) data in 1979 and 1981, when the GSS was not
in the field. Changes in net dislike are no longer relevant after 1989, so the series we use
here takes the 1989 value and simply projects it forwards – net dislike varies until 1989, and
is held constant afterwards.
Identifying specific instruments for the different domestic domains is much less
straightforward, and our attempts to do so were not successful. 16 We do include a linear
counter variable, however, to account for any increase in the underlying preferred levels of
spending over time, other things being equal (Wlezien 1995).
For policy representation (equation 3), we build directly on recent ‘political’ models of policy
change (Stimson, MacKuen, and Erikson 1995; Wlezien 1996a, 2004; Smith 1999), which
include measures of public preferences and party control of government. Following equation
3, the dependent variables used in the analyses represent the first differences of spending (in
constant dollars/pounds) for each of the spending categories. Recall that these changes in
spending are expected to be positively related to the levels of net support for spending, which
capture the public’s relative preferences. Politicians are expected to respond currently. In the
budgetary context, this means that change in spending for fiscal year t follows the level of net
support in year t-1, when the bulk of spending decisions for fiscal year t are made.
15
It is important to note that Dickey-Fuller tests indicate that the measures of net support—the dependent
variables— are stationary.
16
We know that preferences in these areas do exhibit common movement, and previous research indicates that it
follows variation in economic security over time. It is not clear exactly what effect the economy has, as the
findings of different studies differ quite a lot: economic security is positively related to spending opinion in
some studies (Durr 1992) and negatively in others (Erikson, et al. 2002). The studies do rely on different
measures – business expectations and unemployment rates, respectively. Still other research on public opinion
relies on the sum of unemployment and inflation rates (Franklin and Wlezien 1997). We have used this ‘misery
index’ in past work but for most of the analyses here economic conditions have no effect. To preserve degrees
of freedom, we do not include these measures in the analyses here, and doing so makes no real difference.
Results including economic variables are available upon request.
14
The party control variable is fairly standard and takes the value ‘0’ under Democrat/ Labour/
Liberal governments and ‘1’ under Republican/ Conservative governments. The measure of
party control thus taps the levels of partisan control, which might appear to be inconsistent
with the (differenced) dependent variables. Given that budgetary policy feeds back in
‘thermostatic’ fashion on public preferences, however, the specification actually is
theoretically implied (Wlezien 1996a; 2004). It also is supported by separate diagnostic
analyses.
Results
Complete results are provided in the Appendix. 17 Our focus here is on the effect of spending
on preferences (public responsiveness), and the effect of preferences on spending (policy
representation). The relevant coefficients – both in their raw and standardized form – are
gathered together in Table 1. Note first that the thermostatic model works remarkably well.
In every case, the coefficient for spending is negative, implying a thermostatic responsiveness
to policy change on behalf of the public. Also in every case, the coefficient for preferences is
positive, suggesting that policy reacts to the preferences of citizens.
—Table 1 about here—
The magnitude and significance of both public responsiveness and representation clearly vary
across domains, and do so in predictable ways. That said, comparing the coefficients across
countries and domains is somewhat difficult. In each case, the preferences coefficient
indicates the change in spending – in 2000 national current units (NCUs) – that is a
consequence of a one-unit shift in net preferences for spending; similarly, the spending
coefficient indicates the change in preferences given a one billion NCU shift in spending.
The amount spent in each domain will affect the magnitude of the coefficients, however.
Defense is a good example. The US spends much more on defense than either the UK or
Canada. As a consequence, the volume of spending varies by a greater amount, and the
feedback coefficient appears to be comparatively small. Put simply, a one billion dollar shift
in defense spending for the US signifies a much smaller overall policy change than does a
similar shift in UK or Canadian defense expenditures. Similarly, the preferences coefficient
for the US in this domain is quite large in comparison to the other countries, since the
spending change resulting from a one-unit shift in preferences should be relatively large
(given the volume of spending in this domain).
Indeed, the magnitude of coefficients will vary not just due to levels of spending (and
different currencies), but also differences in the variances of each spending and preference
series. What we can say about the Table 1 results is that, based on statistical significance,
feedback across all domains is most robust in the UK and representation most evident in the
US. The pattern is only very faint, however. To draw it out more clearly, and to facilitate
more intelligible comparisons of coefficients across both domains and countries, we
transform these coefficients in two different ways, in Tables 2 and 3.
Table 2 contains results for feedback. The first column shows how the estimated effects differ
across domains and countries when we take into account the variances of spending in the
different domains. Specifically, it lists the effect on preferences of a one-standard deviation
shift in spending. The effects are all predictably much larger than the coefficients in Table 1
but the pattern differs, particularly across countries. We can see that the estimates for the UK
17
While we move rather quickly past these estimations here, they are dealt with in much more detail elsewhere.
See Soroka and Wlezien 2004, 2005.
15
still are clearly larger than those for the US and Canada in each domain. The differences are
much less pronounced, however. The estimates in the US and Canada also are much more
similar, at least on average. The US public is much more responsive to a standardized shift in
defense spending; the Canada public is much more responsive to a comparable shift in health
spending. These differences are as we would expect given the evident salience of issue
domains in the two countries (Soroka and Wlezien 2004).
When we look at the domestic domains combined, feedback appears to be larger in Canada
than in the US—9.3 to 6.5. This is true when focusing on consolidated spending in Canada,
at least; when focusing only on federal spending, in the last row of Table 2, the estimated
effect drops below that of the US. This shift is telling. Consolidated spending – reflecting
policymaking by all levels of government – is likely a relatively clear signal, as opposed to
purely federal spending, which captures just one component of the policy citizens will
observe. Feedback is consequently reduced when the analysis relies on federal spending
alone, in line with our expectations about the effects of federalism on public responsiveness.
—Table 2 about here—
By accounting for the quite different variances in spending (see the third column of Table 2),
results in column one provide more understandable information about differences in feedback
across countries. They also hint at differences across domains. In the UK, for instance, the
public seems to be most responsive to spending on defense and a little less so on health and
education. In the US, however, the public seems more responsive to spending on welfare
than defense. In Canada, responsiveness is greatest on health and then welfare and then
education and defense. Most importantly, the results demonstrate clear differences across
countries, and ones that perfectly fit our expectations: The public is much more responsive to
policy change where government control is largely concentrated, as in the UK, and
significantly less so where it is decentralized, as in the US and Canada. The same is true
within countries. There is remarkably low feedback in the US for education, for instance, the
domain that is by far the most ‘federal’. 18
The second column of Table 2 offers a different transformation and a slightly different
pattern across domains. It shows the shift in net support for spending in each domain that
results from a 1% change in mean spending. This transformation offers more information
about differences across domains, as the mean level of spending across domains differs
substantially, especially in the US and Canada. In the US, the estimated response to defense
and welfare spending is about equal, though still much greater than on health and education.
In Canada, the estimate now is greatest for welfare and then health, education and defense.
The patterns across domains are more in fitting with the literature and our own expectations.
Table 3 displays comparable results for representation. The first column shows the effects of
a standard deviation change in preferences on spending. The most striking result here is that
in each domain the effect is much larger for the US, perfectly in fitting with our conjecture.
The differences are most pronounced in the defense domain, where the US government is
estimated to be 14 times as responsive as the British government and fully 50+ times as
responsive as the Canadian government. This borders on the ridiculous. Indeed, it appears
that the differences at least partly reflect differences in the size of the programs—the larger
the program the greater the estimated responsiveness.
We accordingly adjust for program size in the second column, which shows the effect on
spending of a one-point shift in preferences, expressed as a percent of the mean level of
18
According to the US Census of Governments (1987-2002 editions), federal spending on education is just 11%
of total education spending.
16
spending in each domain. Here, the variation in effects is sharply reduced across all domains.
For defense, the mean-adjusted effects , in percents, are a much more modest and reasonable
.24, .17, and .09 in the US, the UK, and Canada, respectively. The pattern is exactly what we
expect given differences in government institutions—between the US on the one hand and
the UK and Canada on the other—and salience—between the US and UK on the one hand the
Canada on the other. Representation is greatest in our one presidential system where the
issue is highly salient, i.e., there is a high level of feedback, and is weakest in the
parliamentary system where the issue is of relatively low salience. In the domestic domains,
government responsiveness in the US is estimated to be about three times that in the UK and
almost four times that in Canada. This also is exactly what we expect.
—Table 3 about here—
Although they distort differences across countries, the standardized estimates in the first
column of Table 3 do appear to tap real differences across domains within countries. (This is
as we should expect given that the variance of preferences mostly differs across countries.)
In the US, the estimates indicate a comparatively high level of defense opinion representation
and then a sizable drop on health and welfare and then another on education. In the UK,
estimated responsiveness to defense opinion also is very high but there is only a slight drop in
responsiveness on health and education. In Canada, responsiveness is most pronounced for
health, and then welfare and then education and lastly defense. These patterns pretty neatly,
if imperfectly, match what we would expect given differences in the level of federalism and
issue salience. That is, there is considerable symmetery in the structure of policy feedback
and opinion representation within countries.
Conclusions
We have suggested that a critical indication of the ‘performance’ of policymaking institutions
is the connection between public preferences and public policy – specifically, the magnitude
of both public responsiveness (feedback) and policy representation. We have further
suggested that feedback and representation will be mediated by institutional design.
Federalism, by weakening the policy ‘signal’ the public receives, will tend to dampen
feedback. Parliamentary government, by both empowering and isolating a governing cabinet,
will tend to decrease policy representation.
These suggestions are powerfully evidenced in the preceding tables. The UK – our one
unitary system – exhibits the greatest degree of feedback, while the US – our one presidential
system – exhibits the greatest degree of policy representation. These differences in feedback
and representation are pervasive and robust across policy domains. They imply real
differences in democratic performance across countries: while the UK benefits from being a
unitary system and the US from being a presidential one, Canada benefits from neither. At
the same time, there are important differences across domains within countries. Institutions
actually may help us understand some of these differences, especially the level of policy
decentralization in the US and Canada. The salience of issues also may help explain the
differences. Whether and the extent to which the public considers issues to be important
appears to matter. This is hardly surprising.
The results are striking, we believe, and telling about how – and how well – policymaking
institutions work. Of course, we still face the classic problem of comparative politics
research: only three cases and many possible independent variables. This is especially true
about the effects of presidential-parliamentary government. For the effects of federalism,
after all, we can also observe consequences of variation across policy domains. It may be
17
telling that public responsiveness in Canada is higher for welfare than for either health or
education, since welfare is by far the most centralized of these domains. 19 Likewise, the
relatively low public responsiveness to education spending in the US may reflect its high
level of decentralization. 20 In the UK, and looking forward, we might expect decreasing
responsiveness with increasing devolution. 21 This all requires additional research.
More generally, though perhaps most importantly, we have found evidence that the
thermostatic model works in each of these three countries. We observe that the public adjusts
its relative preferences for spending in response to spending itself—there is negative
feedback. We also observe that spending itself follows changes in preferences over time—
there is representation. Albeit to different degrees across countries and policy domains,
democracy works.
19
Federal spending on welfare makes up about 82% of consolidated spending over the period investigated here.
Direct federal spending makes up about 10% of consolidated health or education spending, though there are of
course block grants in these domains as well. Even so, both health and education are characterized by a
complicated, and relatively balanced, mix of federal and provincial spending.
20
Based on data from the US Census of Governments (for 1987, 1992, 1997, and 2002), federal spending on
education made up just 11% of combined federal, state and local spending in this domain. In contrast, federal
spending makes up 36% of welfare spending, and 50% of health spending.
21
Though we cannot assess this for the UK, as Gallup sadly stopped asking the relevant questions in 1995.
18
Bibliography
Ansolabehere, Stephen and Shanot Iyengar. 1995. Going Negative: How Campaign Advertising Shrinks and
Polarizes the Electorate. New York: The Free Press, 1995.
Arceneaux, Kevin. 2003. The Federal Face of Democratic Representation: The Effects of Responsibility
Attribution on Cross-Level Voting Behavior and Government Responsiveness in the United States. Ph.D.
Dissertation, Rice University.
Bagehot, Walter. 1867 (1966). The English Constitution. Ithaca, New York: Cornell University Press.
Bartels, Larry M. 1991. "Constituency Opinion and Congressional Policy Making: The Reagan Defense
Buildup." American Political Science Review 85:429-456.
Blais, Andre, Donald Blake, and Stephene Dion. 1996. ‘Do Parties Make a Difference: A Reappraisal.’
American Journal of Political Science 40:514-520.
Brooks, J.E. 1987. ‘The Opinion-Policy Nexus in France - Do Institutions and Ideology Make a Difference?’
Journal of Politics 49:465-80.
———. 1990. ‘The Opinion-Policy Nexus in Germany. Public Opinion Quarterly 54:508-29.
Burstein, Paul. 1998. ‘Bringing the Public Back in.’ Social Forces 77:27-62.
Converse, Philip. 1964. ‘The Nature of Belief Systems in Mass Publics,’ pp. 201-261 in D. E. Apter, ed.,
Ideology and Discontent. New York: Free Press.
Dahl, Robert A. 1971. Polyarchy. New Haven: Yale University Press.
Deutsch, Karl. 1963. Nerves of Government. New York: Free Press.
Downs, William M. 1999. ‘Accountability Payoffs in Federal Systems? Competing Logics and Evidence from
Europe’s Newest Federation.’ Publius: The Journal of Federalism 29: 87-110.
Durr, Robert H. 1993. ‘What Moves Policy Sentiment?’ American Political Science Review 87:158-70.
Easton, David. 1965. A Framework for Political Analysis. Englewood Cliffs, N.J.: Prentice-Hall.
Eichenberg, Richard, and Richard Stoll. 2003. ‘Representing Defense: Democratic Control of the Defense
Budget in the United States and Western Europe.’ Journal of Conflict Resolution 47:399-423.
Erikson, Robert S., Michael B. MacKuen, and James A. Stimson. 2002. The Macro Polity. Cambridge:
Cambridge University Press.
Franklin, Mark N. 2004. Voter Turnout and the Dynamics of Electoral Competition. New York: Cambridge
University Press.
Franklin, Mark and Christopher Wlezien. 1997. ‘The Responsive Public: Issue Salience, Policy Change, and
Preferences for European Unification.’ Journal of Theoretical Politics 9:247-263.
Goggin, Malcolm and Christopher Wlezien. 1993. ‘Abortion Opinion and Policy in the American States.’ In
Understanding the New Politics of Abortion, Malcolm Goggin, ed. Newbury Park, CA: Sage.
Golder, Matt. 2005. “Democratic electoral systems around the world, 1946–2000.” Electoral Studies 24: 103121
Hartley, Thomas and Bruce Russett. 1992. ‘Public Opinion and the Common Defense: Who Governs Military
Spending in the United States?’ American Political Science Review 86:905-915.
Hill, Kim Quaile, and Patricia A. Hurley. 1998. ‘Dyadic Representation Reappraised.’ American Journal of
Political Science 43: 109-137.
Jankowski, Richard and Christopher Wlezien. 1993. ‘Substitutability and the Politics of Macroeconomic
Policy.’ Journal of Politics 55:1060-1080.
Jennings, Sir Ivor. 1959. Cabinet Government 3rd ed. Cambridge: Cambridge University Press.
Jones, Bryan D. 1994. Reconceiving Decision-Making in Democratic Politics: Attention, Choice, and Public
Policy. Chicago: University of Chicago Press.
Joumard, Isabelle, and Per Mathis Kongsrud. 2003. ‘Fiscal Relations Across Government Levels.’ Economics
Department Working Papers (no. 375). Organization for Economic Cooperation and Development.
(http://www.olis.oecd.org/olis/2003doc.nsf/linkto/eco-wkp(2003)29)
Kiewiet, D. Roderick, and Mathew D. McCubbins. 1988. ‘Presidential Influence on Congressional
Appropriations Decisions.’ American Journal of Political Science 32: 713-736.
Laver, Michael and Kenneth A. Shepsle. 1996. Making and Breaking Governments: Cabinets and Legislatures
in Parliamentary Democracies. Cambridge: Cambridge University Press.
Lijphart, Arend. 1999. Patterns of Democracy. New Haven: Yale University Press.
Lippmann, Walter. 1925. The Phantom Public. New York: Harcourt Brace.
Manza, Jeff, and Fay Lomax Cook. 2002. ‘Policy Responsiveness to Public Opinion: The State of the Debate.’
In Jeff Manza, Fay Lomax Cook, and Benjamin I. Page, eds., Navigating Public Opinion: Polls, Policy, and
the Future of American Democracy. New York: Oxford University Press.
McCrone, Donald J. and James H. Kuklinski. 1979. ‘The Delegate Theory of Representation.’ American
Journal of Political Science 23:278-300.
Monroe, Alan. 1998. “Public Opinion and Public Policy 1980-1993.” Public Opinion Quarterly 62:6-28.
19
———. 1979. ‘Consistency between Constituency Preferences and National Policy Decisions.’ American
Politics Quarterly 12:3-19.
Nevitte, N. and R. Gibbins. 1986. ‘Foreign Policy Debates and Sleeping Dogs: All Quiet on the Public Front.’
Canadian Public Policy 12(3): 401-412.
Page, Benjamin I. and Robert Y. Shapiro. 1992. The Rational Public. Chicago: University of Chicago Press.
Persson, Torsten, Gerard Roland, and Guido Tabellini. 1997. ‘Separation of Powers and Political
Accountability.’ Quarterly Journal of Economics 112:1163-1202.
Petry, F. 1999. ‘The Opinion-Policy Relationship in Canada.’ Journal of Politics, 61:540-550.
Pitkin, Hanna Fenichel. 1967. The Concept of Representation. Berkeley: University of California Press.
Powell, G. Bingham. 2000. Elections as Instruments of Democracy: Majoritarian and Proportional Visions.
New Haven: Yale University Press.
Przeworski, Adam, Susan Stokes, and Bernard Manin, eds. 1999. Democracy, Accountability, and
Representation. Cambridge: Cambridge University Press.
Rogowski, Ronald, and Mark Andreas Kayser. 2002. ‘Majoritarian Electoral Systems and Consumer Power:
Price-Level Evidence from the OECD Countries.’ American Journal of Political Science 46:526-539.
Saward, Michael. 1994. ‘Democratic Theory and Indices of Democratization,’ Pp. 6-24 in David Beetham, ed.,
Defining and Measuring Democracy. Thousand Oaks CA: Sage.
Schumpeter, Joseph A. 1950. Capitalism, Socialism and Democracy, 3rd ed. (London: Allen and Unwin).
Sharpe, Elaine. 1999. The Sometime Connection: Public Opinion and Social Policy. Albany, New York: SUNY
Press.
Smith, M.A. 1999. ‘Public Opinion, Elections, and Representation within a Market Economy: Does the
Structural Power of Business Undermine Popular Sovereignty?’ American Journal of Political Science
43:842-863.
Soroka, Stuart N. 2003. ‘Media, Public Opinion and Foreign Policy.’ Harvard International Journal of Press
and Politics 8(1):27-48.
Soroka, Stuart N. and Elvin T. Lim. 2003. ‘Issue Definition and the Opinion-Policy Link: Public Preferences
and Health Care Spending in the US and UK.’ British Journal of Politics and International Relations 6(1).
Soroka, Stuart N. and Christopher Wlezien. 2005. ‘Opinion-Policy Dynamics: Public Preferences and Public
Expenditure in the United Kingdom.’ British Journal of Political Science 35:665-689.
———. 2004. ‘Opinion-Policy Dynamics: Canada in Comparative Perspective.’ Canadian Journal of Political
Science 37:531-559.
Soroka, Stuart N., Christopher Wlezien, and Iain McLean. 2006. ‘Public Expenditure in the UK: How Measures
Matter.’ Journal of the Royal Statistical Society, Series A 169: 255-271.
Stimson, James A., Michael B. MacKuen, and Robert S. Erikson. 1995. ‘Dynamic Representation.’ American
Political Science Review 89:543-565.
Weakliem, David. 2003. ‘Public Opinion Research and Political Sociology.’ Research in Political Sociology 12:
49-80.
Wlezien, Christopher. 2004. ‘Patterns of Representation: Dynamics of Public Preferences and Policy.’ Journal
of Politics 66:1-24.
———. 1996a. ‘Dynamics of Representation: The Case of U.S. Spending on Defense.’ British Journal of
Political Science 26:81-103.
———. 1996b. ‘The President, Congress, and Appropriations, 1951-1985.’ American Politics Quarterly 24:4367.
———. 1995. ‘The Public as Thermostat: Dynamics of Preferences for Spending.’ American Journal of
Political Science 39:981-1000.
Wlezien, Christopher and Stuart Soroka. 2003. ‘Measures and Models of Budgetary Policy.’ Policy Studies
Journal 31(2):273-286.
20
Figure 1. Policy Representation and the Horizontal Division of Powers
Figure 2. Public Responsiveness and the Vertical Division of Powers
21
Figure 3. Feedback and Representation Combined: System Efficiency
22
Figure 4. Preferences and Spending
100
US
US
400
300
50
t
Ne
fe
Pre
ces
ren
Education
Health
Welfare
Defense
200
S
0
(
i ng
nd
pe
00
20
D)
US
100
Education
-50
1975
100
1980
1985
1990
1995
2000
2005
1975
UK
Health
Welfare
0
1980
1985
1990
Defense
1995
2000
2005
UK
45
40
(20
00
UK
P)
50
Education
Health
Education
Health
35
Defense
Sp
end
ing
Defense
Net Preferences
0
30
-50
25
1975
1980
1985
1990
1995
1975
100
1980
1985
1990
1995
100
CA
CA
80
50
Sp
end
ing
(20
00
CD
N)
60
40
Net Preferences
0
20
-50
1985
1990
1995
Education
Health
Welfare
Defense
2000
Education
2005
1985
23
Health
Welfare
0
1990
1995
Defense
2000
2005
Table 1. Feedback and Representation, Raw Effects
a
Representationb
Feedback
Results, in 2000 NCUs
Feedbacka
Representationb
Results, in 2000 USD
Defense
US
-.154***
.752***
-.154
.752
UK
-9.542***
.048**
-6.453
.071
CA
-3.373
.011*
-5.073
.007
Health
US
-.054*
.685***
-.054
.685
UK
-2.323***
.069**
-1.571
.102
CA
-.938**
.101***
-1.411
.067
Education
US
-.171*
.198**
-.171
.198
UK
-3.497***
.057*
-2.365
.084
CA
-.919**
.094**
-1.382
.062
Welfare
US
-.504***
.359**
-.504
.359
CA
-.956***
.124*
-1.438
.082
All Domestic
US
-.063***
1.464***
-.063
1.464
UK
-1.355***
.149**
-.916
.220
CA
-.452***
.343***
-.680
.228
c
-.638***
.261***
-0.960
0.174
CA
a
Based on model of Feedback in Table A1; coefficient displayed is for Spendingt.
b
Based on model of Spending in Table A1t; coefficient displayed is for Preferencest-1.
c
Using federal spending only (see text).
* p < .05; ** p < .01; *** p < .001.
24
Table 2. Comparing Feedback
a
Feedback
Effect of St Dev
Effect of Spending
Spending Change (in
Change: 1% of Mean
NCUs)
Relevant Descriptivesb
St Dev
Spending
Defense
US
-8.304
-.486
53.9
UK
-22.079
-2.766
2.3
CA
-3.808
-.427
1.2
Health
US
-3.017
-.051
55.9
UK
-14.238
-.829
6.1
CA
-10.412
-.586
11.1
Education
US
-1.765
-.072
10.3
UK
-12.458
-1.213
3.6
CA
-4.480
-.550
4.9
Welfare
US
-19.294
-.539
38.3
CA
-7.718
-.829
8.1
All Domestic
US
-6.477
-.154
102.8
UK
-13.042
-.954
9.6
CA
-9.299
-.934
20.6
c
-4.069
-.458
6.4
CA
a
Based on model of Feedback in Table A1; coefficient displayed is for Spendingt.
b
Based on in-sample values only.
c
Using federal spending only (see text).
Mean
spending
315.4
29.0
12.7
94.9
35.7
62.5
42.4
34.7
59.8
106.9
86.7
244.2
70.4
206.6
71.8
Table 3. Comparing Representation
a
Representation
Effect of St Dev
Effect of 1-point
Spending Change (in
Change in Prefs, as a
NCUs)
% of Mean Spending
Relevant Descriptivesb
St Dev
Prefs
Defense
US
14.320
.238
19.042
UK
1.038 (1.535)
.166
21.628
CA
.257 (.171)
.087
23.344
Health
US
5.406
.722
7.892
UK
.664 (.981)
.193
9.618
CA
1.884 (1.252)
.162
18.651
Education
US
1.875
.467
9.471
UK
.566 (.836)
.164
9.922
CA
.901 (.599)
.157
9.585
Welfare
US
4.377
.336
12.191
CA
1.155 (.768)
.143
9.318
All Domestic
US
12.905
.599
8.815
UK
1.403 (2.075)
.212
9.416
CA
4.077 (2.711)
.166
11.888
CA c
3.103 (2.063)
.363
11.888
a
Based on model of Spending in Table A1t; coefficient displayed is for Preferencest-1.
b
Based on in-sample values only.
c
Using federal spending only (see text).
25
Mean
spending
315.4
29.0
12.7
94.9
35.7
62.5
42.4
34.7
59.8
106.9
86.7
244.2
70.4
206.6
71.8
Appendix Tables. Models of Feedback and Representation
Feedback
Spendingt
Preferencest-1
US-Russia
Post-911
UK
US
A1. Defense
Represent
-.154***
(.043)
.579***
(.136)
.021
(.061)
19.382**
(8.241)
Feedback
Represent
-9.542***
(.963)
2
17.136***
(5.064)
1.386***
(.425)
-.909
(.815)
-1.995
(1.909)
-73.227***
(20.003)
-4.226*
(2.104)
.798
.765
29
.643
.599
28
.870
.852
18
.505
.399
18
.803
.768
21
Represent
Feedback
Represent
Feedback
Represent
.069**
(.030)
-.938**
(.425)
.750***
(.165)
2.663**
(1.059)
.101***
(.019)
Counter
-.054*
(.029)
.762***
(.108)
.578***
(.197)
Govt1t-1
Govt2t-1
Constant
-8.091***
(2.802)
-3.114
(2.751)
-.115
(.158)
.200
.106
20
UK
US
Feedback
Preferencest-1
.256
(.236)
.364**
(.147)
.808
(.754)
A2. Health
Spendingt
.011*
(.006)
.324***
(.082)
46.319***
(6.069)
Spendingt-1
R
Adj R2
N
Represent
.048**
(.017)
.326***
(.059)
Govt2t-1
Feedback
-3.373
(2.364)
.752***
(.133)
Govt1t-1
Constant
CA
CA
-2.323***
(.298)
.685***
(.097)
4.132***
(.343)
-.077
(1.382)
.061
(.123)
-2.762
(5.819)
-1.738*
(.884)
-35.126***
(2.957)
1.545*
(.817)
1.996**
(.715)
-21.306**
(8.612)
-.705*
(.365)
2
.889
.728
.954
.284
.903
.681
R
Adj R2
.876
.694
.948
.189
.881
.635
N
29
28
18
18
17
17
Cells contain OLS coefficients with standard errors in parentheses. Govt1 is the party of the President
in the US, and the party of Governments elsewhere. Govt2 is the Republican proportion of Congress.
* p < .05; ** p < .01; *** p < .001.
26
Feedback
Spendingt
Preferencest-1
Counter
Represent
-.171*
(.085)
.708***
(.125)
.394*
(.195)
.198**
(.095)
2
Counter
.057*
(.033)
-.919**
(.362)
.499**
(.192)
1.580***
(.506)
.094**
(.036)
-31.923***
(4.108)
1.095
(.916)
-12.644***
(4.127)
-.900**
(.385)
.937
.929
29
.168
.064
28
.888
.873
18
.171
.061
18
.899
.876
17
.470
.394
17
-1.232
(1.003)
Represent
Feedback
Represent
.359**
(.164)
-.956***
(.191)
.283*
(.141)
1.962***
(.351)
.124*
(.061)
CA
US
-.504***
(.144)
.525***
(.117)
2.622***
(.675)
Govt1t-1
Govt2t-1
Constant
Represent
-5.515*
(2.766)
Feedback
Preferencest-1
Feedback
.437
(1.438)
-.197
(.154)
8.785
(7.244)
A4. Welfare
Spendingt
CA
Represent
3.756***
(.473)
Govt2t-1
R
Adj R2
N
Feedback
-3.497***
(.709)
Govt1t-1
Constant
UK
US
A3. Education
-36.704***
(9.520)
-4.790
(3.611)
-.444
(.265)
23.035*
(13.208)
6.009***
(1.183)
-14.716***
(2.745)
-1.878**
(.657)
2
.765
.178
.910
.679
R
Adj R2
.736
.075
.888
.630
N
29
28
16
16
Cells contain OLS coefficients with standard errors in parentheses.
Govt1 is the party of the President in the US, and the party of
Governments elsewhere.
Govt2 is the Republican proportion of
Congress. * p < .05; ** p < .01; *** p < .001.
27
2.551***
(.737)
A5. All
Domestic
US
Feedback
Spendingt
Preferencest-1
Counter
-.063***
(.019)
.653***
(.102)
1.044***
(.261)
Govt1t-1
Govt2t-1
Constant
R2
Adj R2
N
A6. All
Domestic
Spendingt
Preferencest-1
Counter
-14.620***
(3.698)
.906
.895
29
UK
Represent
1.464***
(.290)
.149**
(.056)
3.909***
(.343)
-6.569
(4.315)
-.894**
(.393)
44.666**
(19.031)
.533
.475
28
Represent
.343***
(.088)
-3.446**
(1.616)
-33.227***
(2.960)
.953
.946
18
3.063*
(1.483)
.328
.238
18
CA (federal)
Feedback
-.638***
(.139)
.585***
(.115)
1.042***
(.285)
10.797***
(2.180)
Govt1t-1
Represent
-1.355***
(.190)
CA (consolidated)
Feedback
-.452***
(.086)
.504***
(.110)
2.803***
(.495)
Feedback
Represent
.260***
(.059)
-6.120***
(1.468)
Govt2t-1
Constant
-21.026***
(3.769)
-3.374***
(1.113)
-7.817***
(2.256)
-1.936***
(.749)
2
.960
.678
.953
.658
R
Adj R2
.951
.629
.941
.606
N
16
16
16
16
Cells contain OLS coefficients with standard errors in parentheses.
Govt1 is the party of the President in the US, and the party of
Governments elsewhere.
Govt2 is the Republican proportion of
Congress. * p < .05; ** p < .01; *** p < .001.
28