Economic voting and political context: a comparative

Electoral Studies 19 (2000) 151–170
www.elsevier.com/locate/electstud
Economic voting and political context:
a comparative perspective
Christopher J. Anderson
Department of Political Science, Binghamton University (State University of New York), Binghamton,
NY 13902-6000, USA
Abstract
Based on individual-level survey data collected in 13 European democracies, this study
analyzes three alternative ways of modeling how political context affects the relationship
between economic perceptions and vote intention. The three approaches are (1) institutional
clarity of responsibility; (2) governing party target size; and (3) clarity of available alternatives.
The results reveal that political context interacts with economic perceptions to affect voting
behavior. When the institutional context clarifies who is in charge of policymaking, when the
target of credit and blame is large, and when citizens have fewer viable alternative choices,
economic effects are stronger. Taken together, these findings suggest that voters’ ability to
express discontent with economic performance is enhanced when mechanisms of accountability are simple.  2000 Elsevier Science Ltd. All rights reserved.
Keywords: Economic voting; Europe; Accountability; Economic perceptions
In recent years, political scientists have begun to systematically tackle the issue
of how differences in political conditions affect the relationship between the economy
and government support (Anderson, 1995a; Lewis-Beck, 1988; Lewis-Beck and Mitchell, 1993; Pacek and Radcliff, 1995; Paldam, 1991; Powell and Whitten, 1993).
In different ways, this line of research has sought to integrate economic and political
factors by examining how a country’s political context mediates the relationship
between the economy and support. Specifically, scholars have sought to explain differences in economic effects across countries by arguing that the economy affects
government support more strongly the clearer a government’s (or party’s) responsibility for policymaking and economic management or the clearer the available alternatives are in a country.
This paper continues this line of inquiry by focusing on the three main ways
in which political context has been operationalized: formal institutional (clarity of
0261-3794/00/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 6 1 - 3 7 9 4 ( 9 9 ) 0 0 0 4 5 - 1
152
C.J. Anderson / Electoral Studies 19 (2000) 151–170
responsibility), governing party target-size, and clarity of available alternatives.
These three models of politically-mediated economic voting are based on different
notions of how voters assign credit and blame for the state of the economy—i.e.
different logics of how political context and individual-level behavior interact. All
three view voters as purposive agents who seek to assign credit or blame to incumbents. They differ, however, with regard to the specific elements of the political
environment that are hypothesized to affect economic voting.
The next section reviews these approaches and discusses how they differ. Based
on individual-level survey data collected in 13 European democracies, the study subsequently analyzes the interaction of political and economic effects and the validity
of the proposed explanations of political effects. A concluding section discusses the
findings and spells out avenues for future research.
1. How political context affects the structure of credit and blame
Although the literature on vote and popularity functions in western democracies
is varied and extensive (see Nannestad and Paldam, 1994; Norpoth, 1996 for
overviews), it seldom has directly addressed questions of institutional design and
the role political contexts play in the assignment of credit and blame to incumbent
governments. Instead, it usually has examined the relations among economic perceptions or conditions on one hand and government vote or popularity on the other with
the help of single-country studies that implicitly have held many critical political
factors constant. Summarizing the findings of this literature in the early 1990s,
Michael Lewis-Beck (1991, p. 2) concluded that, in general, “when economic conditions are bad, citizens vote against the ruling party”.
The evidence for economic effects has not always been clear-cut nor consistently
significant across countries, however (Paldam, 1991). In fact, the economy works
quite well as a predictor of party and government support in some countries at some
points in time, but fails to do so on a number of occasions under different conditions
(Anderson, 1995a). Why? To answer this question—also termed the “instability
dilemma” by some commentators—researchers originally devoted considerable
energy to improving modeling strategies, measurement and operationalization, as
well as data quality. An assumption implicit in such strategies was that the ways in
which the data originally had been collected, modeled, or estimated occasionally was
flawed, but that there were no obvious shortcomings in the way economic effects
were hypothesized to actually affect political behavior.
Such efforts have made considerable strides toward accounting for a number of
inconsistent and null-findings, but they have not usually provided compelling conceptual solutions to the instability problem. Put differently, they have not assumed that
the inconsistent economic effects actually may have been the “true” effects, and that
part of the challenge may lie in accounting for these differences on theoretical
grounds. In contrast to this approach, several recent studies have started with the
assumption that economic effects will vary across countries and across time in systematic ways, and that political economists need to account for differential economic
C.J. Anderson / Electoral Studies 19 (2000) 151–170
153
effects with the help of theories that incorporate differences in political contexts
across countries.
The idea that political contexts matter and may help explain differences in economic effects is not new. In fact, Eulau and Lewis-Beck (1985, p. 4) reminded us
more than a decade ago that “attribution of responsibility emerges as a key condition
in joining economic grievance to political preferences …. Still, we can count on
certain differences across nations, simply because of institutional variations.” Along
similar lines, Michael Lewis-Beck (1988, in particular pp. 108ff) speculated in Economics and Elections that contextual variation may be crucial for understanding differences in economic effects across countries. Several recent studies have sought to
systematically test the idea that political contexts can be systematically measured
and incorporated into economic voting models. How these studies have gone about
incorporating politics formally in economic voting models is reviewed below.
1.1. Institutional differences and clarity of responsibility
How do voters assign responsibility to incumbents for managing the economy?
In recent years, scholars have started to point out that responsibility for economic
management and policymaking seldom is as transparent as simple models of the
relationship between economy and government support would imply. In fact,
responsibility frequently is shared by competing political actors through mechanisms
such as coalition government or simply obscured because of multiple levels of
decision-making and political control (Anderson, 1995b). As a result, political economists have started to examine in more detail how political contexts help structure
the way citizens assign credit and blame to governments for macroeconomic performance.
Upon closer inspection, it turns out that there are different ways to conceptualize
how political contexts may help to mediate the relationship between economic performance and political preferences. For the purposes of this study, I refer to them
as institutional clarity of responsibility, governing party target size, and clarity of
available alternatives. All three are closely related but differ in the specific ways of
how citizens and political context interact. At the level of individuals, all three
approaches assume that individual voters seek to reward or punish incumbents for
economic performance and are variably helped or hindered by the political context
that exists. That is, they hypothesize that voters’ efforts to assign responsibility for
the country’s economy and express approval or disapproval are systematically filtered
by the political context (see also Bellucci, 1991; Sørensen, 1987; Paldam, 1991).
Powell and Whitten (1993) were among the first and also among the most general
to measure institutional differences across countries in a systematic fashion and
model economic effects as being mediated by how much a country’s formal institutional structure allows citizens to establish who is responsible for economic management. Powell and Whitten’s operationalization and measurement of the clarity of
responsibility concept is quite encompassing. It involves the classification of political
systems into those where responsibility is clear and those where it is not based on
factors such as one-party vs multi-party rule, whether there is bicameral opposition,
154
C.J. Anderson / Electoral Studies 19 (2000) 151–170
decision making powers for opposition parties in parliament, or party cohesion. Their
analysis of objective economic conditions and aggregate election outcomes found
that economic effects were stronger in those countries that had clearer levels of
responsibility (see also Lewis-Beck, 1988, pp. 108–110; Powell, 1989).
1.2. Economic voting and governing party target size
Powell and Whitten’s clarity of responsibility index involves several institutionally-based aspects of political systems that may mediate the economy–vote
relationship. Unfortunately, the generality of the index prevents a concise analysis
of what exactly it is about the political context that mediates the relationship between
economic performance and support for incumbents. Moreover, the index is static to
the extent that it measures political context as varying across countries in predictable
ways without accounting for changes that may occur in the political context over
time. An alternative view of how differences in political context may mediate the
relationship between economy and government support is the governing party target
size approach. While it, too, assumes that individual voters seek to reward or punish
incumbents for economic performance, it differs from Powell and Whitten by suggesting that clarity of responsibility not only varies across countries because of differences in formal decisionmaking procedures that are unlikely to vary much over time,
but that it also is variable within and across countries because of election outcomes
and elite bargaining that periodically reshape the political landscape (Anderson,
1995a; Lewis-Beck and Mitchell, 1993). Specifically, democracies shift the power
to govern and enact policy at more or less regular intervals by way of the electoral
process. Elections shuffle the cards of government and—depending on the political
system—they install new actors, change the composition of governing coalitions, or
confirm political parties and executives in office with an expanded (reduced) or large
(small) mandate. Thus, the extent to which voters assign responsibility to political
actors changes over time and it differs across countries because of election outcomes.
An example may help fix ideas: Imagine two countries with identical institutional
designs and economic performance. Country A has a single-party government that
holds 60% of the seats and 100% of all cabinet posts, Country B a minority coalition
government consisting of four parties, each of which occupies 10 percent of the seats
in parliament and 25% of all cabinet posts. In this hypothetical case of two countries
with the same formal policymaking institutions, we would still expect different economic effects because of the differences in target size. Voters in both countries presumably seek to target parties for the assignment of credit and blame, but they should
have an easier time doing so in Country A than in Country B because the target is
simply larger in the former.1 It would also seem plausible that parties who have a
greater stake in policymaking and decisionmaking should be held more responsible
for the outcomes produced, in this case economic performance.
1
Thus, a correlation between the Powell–Whitten index and target size may exist empirically, but
there does not have to be one on conceptual grounds.
C.J. Anderson / Electoral Studies 19 (2000) 151–170
155
This approach is similar to the Powell and Whitten clarity of responsibility argument in that voters are expected to have an easier or harder time assigning credit
and blame depending on how easily they can identify who is in charge. In this sense,
both the Powell–Whitten and the target size approach can be classified as being
about how clear responsibility is. The target size approach differs, however, in that
voters are not required to be particularly sophisticated about formal institutional rules
or other aspects of the political process. All they need to know is which party has
the most power in the government, and reward and punish this party to a greater
degree. Thus, all that is necessary is for voters to know who is most in charge of
decisionmaking at the national level.2
1.3. Clarity of available alternatives
Both the Powell–Whitten argument and the target size approach assume that those
in power are the targets of citizens’ intentions to assign credit and blame and that
those who can be more easily identified as having responsibility also will be the
more likely targets of citizens’ voting behavior. That is, it is assumed that there are
alternatives available when voters seek to throw incumbents our of office. Note,
however, that this can work only if there are other credible actors who can benefit
from such behavior.
There is considerable variation in the structure of party systems across contemporary democracies, and not all systems provide clearcut or easily predictable alternatives to incumbent governments. Political scientists frequently have measured the
nature of party systems with the help of such indicators as party system fractionalization based on as the effective number of parties in the system (Rae, 1967; Lijphart,
1984; Taagepera and Shugart, 1993). The nature of the party system should matter
for the effects of macroeconomic performance on governing party support because
fragmented party systems—identified by a large effective number of parties—should
make it more difficult for voters to identify a clear alternative to the incumbent
government. Put simply, in such countries there tends to be greater uncertainty about
the likely shape of an alternative future government that will form after the election
has been held. Translated into the research on economic voting, this would mean
that, aside from the clarity of responsibility and the governing party target size, the
effects of economic conditions also should be mediated by the clarity of available
alternatives, or what Lewis-Beck has called “incumbent alternatives for dissent”
(Lewis-Beck, 1986, 1988). Economic effects should be stronger when the number of
2
The target size theory also could be empirically correlated with what is known as the Prime Minister
party effect (cf. Anderson, 1995a), given that in nearly all coalition governments the vote share of the
Prime Minister’s party grows at the expense of the other coalition parties. However, it is worth keeping
in mind that this does not have to be the case, and I would argue that in coalitions where the Prime
Minister’s party is not the largest party, voters should take both the Prime Ministership and governing
party size into account when assigning credit and blame. Unfortunately, there are too few cases that
would permit systematic analyses across these kinds of distinctions (i.e. prime minister party vs governing
party size).
156
C.J. Anderson / Electoral Studies 19 (2000) 151–170
effective parties—i.e. available alternative governments—is smaller. A more clearly
defined set of viable alternatives to the incumbent government should lead citizens
to more readily express content or discontent with the ruling party or parties (see
also McDonald et al., 1999).
An example may again be useful: In a multi-party (11) system with one governing
party that holds 51% of legislative seats and ten parties that share the remaining
49% equally, citizens have a multitude of ways to blame the governing party by
supporting one of the opposition parties. However, voting for any one of these opposition parties is unlikely to throw the incumbent government out of office, and the
significance of a dissenting vote during bad times is decreased. In contrast, voters
in a country with a simple 2 party system, where one party holds 51% and the other
49% of the seats, will find it easy to identify who the alternative government might
be. Note that in both of these hypothetical systems, the institutional structure à la
Powell and Whitten and the governing party target size (in this case 51%) may well
be identical.3 However, what matters are the available alternatives. This available
alternatives approach assumes that a greater effective number of parties leads to an
decreased probability that citizens desert the governing party when macroeconomic
performance is considered unsatisfactory.
1.4. Taking stock
The three approaches outlined here share important assumptions regarding the
relationships among economy, political context, and government support. Specifically, they view voters as purposive agents who seek to assign credit and blame for
economic performance to incumbents. Moreover, they regard the political context as
a constraint on individual voters’ choices and thus seek to model the interactive
relationship of economics and politics on support. Aside from these similarities, these
also differ with regard to precisely what it is about different political systems that
facilitates or complicates the assignment of credit and blame. The institutional clarity
of responsibility approach argues that the structure of formal and informal power
relationships and policymaking authority matters. This approach is based on features
of political systems that are fairly stable, and it assumes an electorate that understands
the intricacies of parliamentary government with a reasonable degree of accuracy.
In contrast, the target size and clarity of available alternatives approaches neither
presume such awareness nor are they based on fairly immutable country-characteristics. Instead, they view voters as willing to reward and punish and doing so to a
greater extent given the simple condition that they know who is in charge of policymaking and who the alternatives are. Finally, the approaches differ in that the
target size and available alternatives approaches are much more specific than the
Powell and Whitten clarity of responsibility argument regarding what it is about a
3
Note, again, that there may be a cross-national correlation between the Powell–Whitten index and
target size measure on one hand and the available alternatives measure on the other. However, such a
correlation is not an automatic consequence of how political context has been conceptualized.
C.J. Anderson / Electoral Studies 19 (2000) 151–170
157
political system voters take into account when seeking to reward or punish governments. Thus, the Powell and Whitten approach is more general than the other two,
but also less able to model how citizens reason at the individual level about various
aspects of the political context because it groups together a variety of institutional
elements that make up that context.
2. Data and measures
To date, these approaches have only been put only to separate tests based on
aggregate data. That is, we have not examined the validity of these models with the
help of individual-level data, nor have we examined their usefulness side-by-side
with identical data. To test the empirical validity of these conceptualizations, I rely
on cross-national individual-level data. Specifically, I make use of the Eurobarometer
42.0, collected between 28 November and 31 December 1994. The Eurobarometer
is a semi-annual set of public opinion surveys conducted in the member states of
the European Union since 1973. Based on multistage probability samples with random selection of respondents, these surveys are based on the population of each
member state, aged 15 and over. The countries included in the study thus were the
(then) twelve member states of the European Union (Belgium, Denmark, France,
Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, and
the UK). This particular survey also was carried out in Norway; which also is
included in this study.
Each country’s survey usually includes about 1000 respondents. There are two
small exceptions to this: Germany, where the Eurobarometer has conducted surveys
with about 2000 respondents (1000 in East and West Germany each) since unification
in 1990, and Luxembourg, the smallest member state, where the sample usually
involves about 500 respondents. I treat the German sample as if it originated from
one population and excluded respondents from Northern Ireland because of the special political situation there.
2.1. Dependent variable
The dependent variable is incumbent vote choice, measured 0,1 with 1 denoting
vote for the incumbent. It is based on the question: “If there were a general election
tomorrow, which party would you vote for?” Based on individuals’ vote choice, two
variants of this dependent variable are analyzed: One categorizes a vote for the governing parties (in the case of single-party government) or one of the governing parties
(in the case of coalition parties) as a vote for the incumbent government (scored 1;
all others 0). The other focuses only on prime minister parties, scoring respondents
who would vote for the prime minister’s party 1, and all others 0.4
4
There are reasons to believe that government support should be disaggregated by prime minister and
coalition parties. To test for possible differences, I performed both kinds of analyses (see also Anderson,
1995b; Wilkin et al., 1997). When different results obtained, this is specified in the text.
158
C.J. Anderson / Electoral Studies 19 (2000) 151–170
2.2. Individual-level independent variables
Economic perceptions are measured with the help of sociotropic and egocentric
measures. The sociotropic question wording is “Compared to 12 months ago, do you
think that the general economic situation in this country now is a lot better, a little
better, stayed the same, a little worse, a lot worse”. The egocentric measure reads:
“Compared to 12 months ago, the financial situation of your household now is a lot
better, a little better, stayed the same, a little worse, a lot worse”.
To be consistent with the rich literature on voting behavior in Europe, I included
several control variables: ideology, religiosity, and social class (for a similar
approach, see Lewis-Beck, 1988). Consistent with cleavage theories of electoral
behavior, these variables are intended to capture long-standing predispositions to
support political parties of different leanings. Ideology is measured by left–right selfplacement: “In political matters people talk of ‘the left’ and ‘the right’. How would
you place your views on this scale?” (left: 1; right: 10) Religiosity is intended to
measure the religious cleavage, differentiating between those who consider themselves religious and those who do not. This is measured by the following question:
“Whether you do or you don’t follow religious practices, would you say that you
are religious, not religious, an agnostic, an atheist?” Finally, social class is selfidentification with a particular social class, measured with the help of the following
question: “If you were asked to choose one of these five names for your social class,
which would you say you belong to? Middle class, lower middle class, working
class, upper class, upper middle class”. Respondents who were unwilling or unable
to answer any of these questions were excluded from the analysis. These variables
were included simply as controls in order to avoid misspecification. Their effects
should vary by country. Because they are not the focus of the present analysis, they
will not be discussed further below.
2.3. Political context variables
Clarity of responsibility is taken from Powell and Whitten’s (1993) study, which
ranked 19 countries on a scale from less clear to more clear. The countries included
in that study that also are included in the Eurobarometer data set are: Belgium,
Britain, Denmark, France, Germany, Greece, Ireland, Italy, Netherlands, and Norway. The original scale theoretically ranged from 0 (more clear) to 5 (less clear);
the actual cases included in this analysis range between 0 and 3.2 (see Table 1).
Governing party target size is measured in two ways: parliamentary strength and
cabinet strength (for prime ministerial parties). Parliamentary strength measures the
percentage of seats held by the prime minister’s party, whereas cabinet strength
measures the percentage of total cabinet positions held by the prime minister’s party.
These percentages were rescaled to vary between 0 (for 0%) and 1 (for 100%). These
data were taken from various issues of the Annual Data Yearbook of the European
Journal of Political Research.
Finally, available alternatives were measured with the help of the Laakso and
C.J. Anderson / Electoral Studies 19 (2000) 151–170
159
Table 1
Descriptive statistics
Variable
Minimum
Maximum
Mean
SD
Retrospective egocentric economic evaluations (5=high)
Retrospective sociotropic economic evaluations (5=high)
Left–right self-placement (0=left; 10=right)
Religiosity (4=high)
Social class—subjective (5=high)
Powell–Whitten index (high=less clear)
Effective number of parliamentary parties (available
alternatives)
Prime minister party target size—cabinet
Prime minister party target size—seats
Government target size—seats
Vote intention governing parties
Vote intention prime minister party
1.00
1.00
1.00
1.00
1.00
0.00
2.17
5.00
5.00
10.00
4.00
5.00
3.20
5.41
2.90
2.88
5.16
3.45
2.41
1.90
3.45
0.84
0.99
1.98
0.89
1.17
1.27
1.02
0.35
0.16
0.41
0.00
0.00
1.00
0.58
0.80
1.00
1.00
0.73
0.40
0.55
0.30
0.22
0.25
0.12
0.01
0.46
0.42
Taagepera (1979) index calculating the effective number or parliamentary parties in
a system. The effective number of parties is computed as follows:5
N⫽
1
冘
(1)
n
p
2
i
i⫽1
where pi is the proportion of parliamentary seats for the ith party (Lijphart, 1984,
p. 120). This formula contains information about the number and relative size of the
parties in the system. It thus helps to differentiate not only between two- and multiparty systems, but it is also a more subtle measure than simply counting the number
of parties that gain representation or receive votes. The proposed measure takes the
relative strength and parliamentary viability of parties into account. Or, as Taagepera
and Shugart (1993, p. 455) have put it:
The advantage of using the effective, rather than the actual, number of parties is
that it establishes a nonarbitrary way to distinguish “significant” parties from less
significant ones. The construction of the index is such that each party weights
itself by being squared. Tiny parties contribute little to the index, while large
parties contribute relatively more.
Again, these data were taken from various issues of the Annual Data Yearbook of
5
Instead of using the more widely known Rae/Taylor index of party system fractionalization, I measure
the effective number of parliamentary parties in a system. Note, however, that the formula for the effective
number of parties carries the same information as Rae’s index of party system fractionalization, only
expressed in a different metric. The reader can obtain Rae’s fractionalization index (F) by substituting
the value for the effective number of parties (N) in the following formula: F=1⫺(1/N).
160
C.J. Anderson / Electoral Studies 19 (2000) 151–170
the European Journal of Political Research. The descriptive statistics for the variables used in the analysis are shown in Table 1.
3. Analysis
I first estimated general multivariate logistic regression models of government and
prime minister party support that did not include any of the contextual variables.
The dependent variable is coded 1 if the respondent supports the incumbent(s) and
0 otherwise. Table 2 shows the results.
As hypothesized, the coefficients for the economic variables turned out to be statistically significant and in the expected direction. Sociotropic effects were stronger
than egocentric ones, with only marginal differences across the two dependent variables (prime minister support/government support). As it turned out, this difference
Table 2
Multivariate economic voting models for Western Europe, 1994 (logistic regressions)
Independent variable
Social class
Religiosity
Left–right self-placement
Sociotropic economic evaluations
(retrospective)
Egocentric economic evaluations
(retrospective)
France
Belgium
Netherlands
Germany
Italy
Luxembourg
Denmark
Ireland
Britain
Greece
Spain
Constant
N
⫺2 log likelihood
Goodness of fit
Cox and Snell pseudo R2
c2 (d.f.)
a
Dependent variable:
Government support
Dependent variable:
Prime minister party support
Parameter
estimate
Parameter
estimate
SE
SE
⫺0.0798***a
0.1847***
0.0967***
0.3361***
0.0193
0.0276
0.0115
0.0254
⫺0.1171***
0.2629***
0.0858***
0.3333***
0.0212
0.0318
0.0125
0.0280
0.0994***
0.0291
0.1283***
0.0316
0.1046
0.0971
0.0927
0.0813
0.0994
0.1272
0.0897
0.0957
0.1035
0.1010
0.1082
0.1559
10,374
12433.092
10248.499
0.076
99.9951 (8)***
⫺0.8017***
⫺0.6279***
⫺0.5675***
⫺0.1155***
⫺1.0948***
⫺0.2860**
0.0522
⫺0.4291***
⫺0.5254***
⫺0.1231***
⫺0.3711***
⫺3.2433***
⫺0.1625
0.2527*
1.0414***
0.0583
⫺0.0297
0.6306***
0.3746***
0.1774*
⫺0.5262***
⫺0.1148
⫺0.3823***
⫺3.0254***
0.1206
0.1104
0.1064
0.0833
0.1227
0.1411
0.0919
0.1012
0.1038
0.1014
0.1085
0.1723
10,347
10981.230
10293.217
0.062
21.0103 (8)**
***P⬍0.001; **P⬍0.05; *P⬍0.1—two-tailed. Portugal is the reference category.
C.J. Anderson / Electoral Studies 19 (2000) 151–170
161
across sociotropic and egocentric effects held throughout the analysis. Sociotropic
effects were consistently more significant, both statistically and substantively, than
egocentric ones. I will therefore focus primarily on the description of sociotropic
effects in the remainder of the paper.
To test the three conceptualizations of how political context and economic effects
interact, I estimated three different models for each contextual factor: Models 1 and
2 estimated the effects of the economic variables on government (or prime minister
party) support for split samples (high/low clarity, large/small target size, many/few
alternatives), whereas a third model estimated the interactive effects on the basis of
the total sample.
3.1. Clarity of responsibility effects
Tables 3 and 4 shows the results of logistic regressions that estimated economic
effects on separate samples of countries that have low and high levels of clarity of
responsibility as defined by Powell and Whitten. Following Powell and Whitten’s
design, I split the sample into those countries scoring 2 or below (high clarity) and
those scoring 2 and above (low clarity). The table also shows the results of logistic
regression models that tested the Powell Whitten index in interaction with the economic variables. If the Powell and Whitten argument holds, we should see stronger
economic effects under conditions of clear responsibility than under conditions of
less clear responsibility.
Turning first to the effects on total government support (Table 3), the results confirmed the hypothesis. Sociotropic economic effects—which always were more significant substantively than egocentric effects—were stronger under conditions of
clear responsibility (0.39 vs 0.27). Egocentric effects were stronger under low
responsibility, but they also were not particularly powerful. Moreover, the high clarity model accounted for more of the variance in the dependent variable.
The interactive model estimated with the help of the total sample (Table 4) confirmed these findings. By itself, sociotropic economic assessments were highly significant (egocentric effects did not achieve statistical significance) and the coefficient
was in the expected direction. Moreover, when economic perceptions were estimated
in interaction with the Powell and Whitten clarity of responsibility index, the results
showed that ambiguous responsibility tempers the economic effects considerably.
Put differently, as countries move from clear to obscure responsibility, the economic
effects had less impact. When government support was examined as a whole, the
findings reported here on the basis of individual level data thus were consistent with
those reported in Powell and Whitten’s earlier study based on aggregate election
results.
As a side note, it is worth pointing out that governments in countries with low
levels of clarity tended to benefit from institutional ambiguity. That is, governments
in countries where it was easier for voters to identify who was in charge of decisionmaking tended to have lower levels of support than governments in countries where
such assignment of responsibility is less clear (Table 4).
To examine whether the effect held when we considered prime minister party
a
0.0378
⫺4.3230
0.0381
⫺2.3272***
0.1776
5521
7184.068
5509.046
0.029
21.7286 (8)**
0.2856
3126
3506.730
3090.986
0.072
6.7884 (8)
0.0515
0.0369
0.0612
0.0215
0.0451
0.0669*
0.0256
0.4152***
0.1142***
0.3888***
0.0424
0.0268
0.0368
0.0167
0.0371
⫺3.0012***
0.2037
5521
6129.758
5522.499
0.037
4.7100 (8)
0.1464***
⫺0.1824***
0.1836***
0.0482**
0.3521***
0.0551
0.0394
0.0713
0.0229
0.0477
⫺4.9918***
0.3272
3126
3164.755
3065.670
0.065
11.5435 (8)
0.1035*
0.0388
0.4892***
0.1344***
0.3050***
Parameter
estimate
SE
0.0236
0.0319
0.0150
0.0325
SE
⫺0.0318
0.1399***a
0.0821***
0.2709***
Parameter
estimate
Parameter
estimate
SE
Parameter
estimate
SE
Low clarity
High Clarity
Low clarity
High Clarity
Dependent variable: Prime minister party support
Dependent variable: Government support
***P⬍0.001; **P⬍0.05; *P⬍0.1—two-tailed.
Social class
Religiosity
Left–right self-placement
Sociotropic economic
evaluations (retrospective)
Egocentric economic
evaluations (retrospective)
Constant
N
⫺2 log likelihood
Goodness of fit
Cox and Snell pseudo R2
c2 (d.f.)
Independent variable
Table 3
Economic voting models by clarity of responsibility (logistic regressions)
162
C.J. Anderson / Electoral Studies 19 (2000) 151–170
C.J. Anderson / Electoral Studies 19 (2000) 151–170
163
Table 4
Economic voting models by clarity of responsibility (logistic regressions)
Independent variable
Social class
Religiosity
Left–right self-placement
Sociotropic economic evaluations
(retrospective)
Egocentric economic evaluations
(retrospective)
Clarity of responsibility
(high=less clear)
Clarity of responsibility×sociotropic
economic evaluations (retrospective)
Clarity of responsibility×egocentric
economic evaluations (retrospective)
Constant
N
⫺2 log likelihood
Goodness of fit
Cox and Snell pseudo R2
c2 (d.f.)
a
Dependent variable:
Government support
Dependent variable:
Prime minister party support
Parameter estimate SE
Parameter estimate SE
⫺0.0246
0.1999***a
0.0975***
0.3899***
0.0199
0.0279
0.0123
0.0495
⫺0.1171***
0.2458***
0.0807***
0.3385***
0.0220
0.0321
0.0134
0.0517
0.0450
0.0556
0.1144*
0.0586
0.2992***
0.0735
0.0272
0.0812
⫺0.0428**
0.0211
⫺0.0031
0.0227
0.0057
0.0241
0.0071
0.0260
⫺3.4366***
0.1958
8647
10690.931
8597.428
0.054
39.5101 (8)***
⫺3.5383***
0.2126
8647
9358.161
8600.590
0.042
9.9358 (8)
***P⬍0.001; **P⬍0.05; *P⬍0.1—two-tailed.
support in isolation, Tables 3 and 4 also shows identical models with the differently
defined dependent variable. The results reveal no institutional effects on prime minister party support. That is, there was no discernible difference in economic effects
between low and high clarity of responsibility countries. The coefficients across the
split sample were virtually identical, and the combined model estimated on the basis
of the complete sample showed no significant effects for the interaction terms.
The logic of the clarity of responsibility argument thus found support only when
we considered total government support—which oftentimes consists of the sum of
support for several political parties. Because coalition government is at the heart of
why it is more difficult for voters to assign credit and blame in some countries,
prime minister parties may thus be sheltered from the clarity of responsibility effects
as voters shift support within the coalition and not necessarily from governing to
opposition parties (Anderson, 1995b). At minimum, these tests revealed that the
Powell and Whitten index has the hypothesized effects when we examined the sum
total of government support, thus providing additional validation for their analysis
on the basis of individual-level data.
164
C.J. Anderson / Electoral Studies 19 (2000) 151–170
3.2. Governing party target size
Target size was measured by the parliamentary and cabinet strength of the prime
minister’s party. Higher levels of strength in both should make that party—and, for
that matter, any party—a bigger target for the assignment of credit and blame. Economic effects should therefore be stronger, the higher the level of responsibility—i.e.
the bigger the target.
Tables 5 and 6 shows the results of split-sample and interacted model estimations,
identical to those conducted for the clarity of responsibility variable. The sample
was split at the mean into the following two groups. Parliamentary strength: 40%
and below/above. This differentiated between the truly sizable prime minister parties
and the moderate to small ones. Cabinet strength: 75% of all cabinet posts. It also
differentiated between dominant prime minister parties and more equal (to small)
ones.6
The results consistently were in the hypothesized direction. Regardless of whether
we examined parliamentary or cabinet strength, bigger targets engendered stronger
economic effects. That is, governing parties with more objective political responsibility suffered greater losses of support among those dissatisfied with economic performance and achieved greater gains among those who evaluated the state of the
nation’s economy positively. Note, however, that this conclusion again was largely
confined to sociotropic effects.
In addition, target size had a significant independent effect (Table 6): Regardless
of economic assessments, bigger targets were more likely to experience losses in
support than smaller ones. Both the split sample analysis as well as the interactive
models displayed significant coefficients in the expected direction, thus validating
the target size hypothesis across types of measures (parliamentary vs cabinet
strength) and types of analysis (total vs split samples).
3.3. Clarity of available alternatives
Finally, I tested the clarity of available alternatives argument measured by the
effective number of parties in the system. Economic effects were expected to be
stronger, when there were fewer viable alternatives to the incumbent government in
the party system. Table 7 shows the results of analyses that are identical to those
presented so far (i.e. split-sample and combined models).
They reveal consistent and statistically significant coefficients across the two different types of analysis. Moreover, the relationships were in the hypothesized direction. Specifically, a smaller set of available viable alternatives was associated with
larger economic effects. That is, when the effective number of parties in the system
was large, economic effects were attenuated. Put differently, when citizens had many,
6
Note that the effects reported below are immune to slight changes in the definition of the cut-off.
Additional analyses splitting the samples at the median revealed no differences. This is also evidenced
by the results obtained on the basis of the interaction models.
a
⫺2.4583***
0.2803
3545
3731.064
3548.512
0.019
2.0209 (8)
0.0537
0.1311**
0.0387
⫺4.1597***
0.1891
6829
7368.787
6744.238
0.066
10.860 (8)
0.1327***
0.0259
0.0371
0.0152
0.0328
0.0345
0.0526
0.0209
0.0466
⫺0.1608***a
0.1325**
0.0092
0.2474***
⫺0.0925***
0.2936***
0.1328***
0.4065***
Parameter
estimate
SE
Parameter
estimate
0.0525
0.0333
0.0571
0.0206
0.0439
SE
⫺4.0625***
0.2827
4149
3923.530
4171.353
0.045
27.2735 (8)***
0.0698
⫺0.0824**
0.2944***
0.1911***
0.1937***
Parameter
estimate
Small target
Large target
Small target
SE
Cabinet posts
Seats in parliament
***P⬍0.001; **P⬍0.05; *P⬍0.1—two-tailed.
Social class
Religiosity
Left–right self-placement
Sociotropic economic
evaluations (retrospective)
Egocentric economic
evaluations (retrospective)
Constant
N
⫺2 log likelihood
Goodness of fit
Cox and Snell pseudo R2
c2 (d.f.)
Independent variable
0.0444
0.0295
0.0406
0.0170
0.0378
SE
⫺4.0599***
0.2113
5329
5879.626
5302.966
0.066
9.0688 (8)
0.1352**
⫺0.0807**
0.2869***
0.0892***
0.4784***
Parameter
estimate
Large Target
Table 5
Economic voting models by governing party target size (parliamentary seats and cabinet posts) (logistic regressions). Dependent variable: Prime minister
party support
C.J. Anderson / Electoral Studies 19 (2000) 151–170
165
166
C.J. Anderson / Electoral Studies 19 (2000) 151–170
Table 6
Economic voting models by governing party target size (parliamentary seats and cabinet posts) (logistic
regressions) dependent variable: Prime minister party support
Independent variable
Seats in parliament
Cabinet posts
Parameter
estimate
Parameter
estimate
SE
Social class
⫺0.1120***a 0.0207
Religiosity
0.2471***
0.0302
Left–right self-placement
0.0906***
0.0123
Sociotropic economic evaluations
0.0053
0.0977
(retrospective)
Egocentric economic evaluations
0.1147
0.1178
(retrospective)
Target size (high=large)
⫺1.3309
0.8381
Target size×sociotropic economic
0.8894***
0.2349
evaluations (retrospective)
Target size×egocentric economic
0.0239
0.2800
evaluations (retrospective)
Constant
⫺3.0875***
0.3853
N
10,374
– 2 log likelihood
11106.611
Goodness of fit
10325.746
0.051
Cox and Snell pseudo R2
4.0326 (8)
c2 (d.f.)
a
SE
⫺0.1042***
0.2565***
0.0914***
⫺0.0173
0.0549
0.0208
0.0305
0.0124
0.0842
0.0999
⫺0.8357**
0.5036***
0.0812
0.3978
0.1076
0.1277
⫺3.0177***
0.3398
8647
11045.552
10302.062
0.056
17.7418 (8)**
***P⬍0.001; **P⬍0.05; *P⬍0.1—two-tailed.
but less sizable, choices to defect from the governing party, economic effects actually
were weaker. The same results were obtained when support for governing parties as
a whole was examined (not shown here).
The differences in effects again was confined to the sociotropic item, with the
size of the coefficient in the sample with a large number of available alternatives
about twice that of the sample with fewer alternatives (0.5 vs 0.25). Moreover, the
interaction model shows that the effective number of parties in a system had a strong
and positive independent effect on governing party support. The more viable alternatives there were in a system, the more governing party support was strengthened.
Thus, consistent with the hypothesis, the results point to what one might call a
stabilization effect or a clarity of alternatives effects. In the limit, it suggests that
too many alternatives are associated with higher levels of support for prime ministers’ parties and a reduction in defection rates from them. Why multiparty systems
appear to shelter governing parties from economic effects is not altogether clear and
cannot be answered with the data at hand. We may speculate that voters in multiparty
systems also have more deeply entrenched partisan attachments, thus making it less
likely that they will defect from any party, including governing parties. In this case,
economic voting would not be about how many alternatives citizens have available
for expressing discontent, but simply about long-standing attachments to particular
a
0.2280
5320
5320.193
5307.475
0.090
16.4703 (8)**
⫺5.1936***
0.0456
0.1224**
0.0438
0.2289
5054
5682.205
5069.149
0.026
11.8611 (8)
⫺1.8987***
0.1133**
0.0298
⫺0.0135
0.3515
10,374
11143.487
10330.860
0.047
8.3481 (8)
0.0945
0.0266
0.3042**
⫺0.1009***
⫺4.6311***
0.1090
0.1721
0.0206
0.0301
0.0123
0.0969
⫺0.1242***
0.2486***
0.0927***
0.7035***
0.0280
0.0434
0.0171
0.0374
⫺0.2370***
0.0729*
0.0109
0.2496***
0.0306
0.0440
0.0179
0.0400
Parameter estimate SE
Parameter estimate SE
0.0024
0.3727***a
0.1829***
0.4927***
Parameter estimate SE
Large
Interaction model
Small
Effective number of parties
***P⬍0.001; **P⬍0.05; *P⬍0.1—two-tailed.
Social class
Religiosity
Left–right self-placement
Sociotropic economic evaluations
(retrospective)
Egocentric economic evaluations
(retrospective)
Number of of parties (high=large)
No. of parties×sociotropic economic
evaluations (retrospective)
No. of parties×egocentric economic
evaluations (retrospective)
Constant
N
⫺2 log likelihood
Goodness of fit
Cox and Snell pseudo R2
c2 (d.f.)
Independent variable
Table 7
Economic voting models by clarity of available alternatives (logistic regressions). Dependent variable: Prime minister party support
C.J. Anderson / Electoral Studies 19 (2000) 151–170
167
168
C.J. Anderson / Electoral Studies 19 (2000) 151–170
parties. In addition, consistent with the other contextual effects shown above, we
may speculate that a simply two-party system facilitates economic voting because
the alternatives are clearer. Such an interpretation would be compatible with both
Powell and Whitten’s clarity of responsibility thesis and the target size explanation
because it implies that voters prefer simple to complicated sets of choices.
4. Discussion
This study started with the assumption that integrating economic and political
factors in economic voting models is likely to yield more comprehensive explanations of how citizens, their political environments, and the economy are related.
Specifically, this paper sought to test three related but alternative ways of modeling
how different aspects of political contexts affect the relationship between economic
perceptions and vote choice. The three approaches tested here on the basis of individual-level cross-national data were the institutional clarity of responsibility, governing
party target size, and clarity of available alternatives approaches.
The results reveal that political context interacts with economic perceptions to
affect voting behavior. When the institutional context clarifies who is in charge of
policymaking and when the target of credit and blame is large, economic effects are
stronger. Conversely, when clarity of responsibility is obscured and when the level
of responsibility is low, governing parties are less affected by how citizens evaluate
the nation’s economy. Note, however, that the results also revealed that many alternatives available to express (dis)content does not mean that citizens are more likely to
defect from governing parties. In fact, when citizens have fewer choices available
to assign blame to incumbents and credit the opposition, economic effects are
stronger. Taken together, these findings suggest that voters’ ability to express discontent with economic performance is enhanced when accountability is simple. Voters’
economic assessments have stronger effects on government support when it is clear
who the target is, when the target is sizable, and when voters have only a limited
number of viable alternatives to throw their support to.
The results also showed that sociotropic economic effects consistently outperformed egocentric ones, and that sociotropic perceptions were affected by contextual
differences whereas egocentric ones were not. Pocketbook voting—while statistically
significant in a number of models—did not have the same substantive or statistical
effects as sociotropic economic voting. The analysis thus adds another piece of evidence in favor of the sociotropic economic voting thesis.
Maybe more importantly, the research reported on here also identifies important
boundary conditions for economic voting research and highlights the need to be
sensitive to, and specific about, the political environment in which economic voting
takes place. Whether and to what extent citizens assign credit and blame for the
country’s economic performance depends on whether they can identify who is in
charge, how much responsibility they have, and what the alternatives are. This finding is important for understanding the nexus between macroeconomy and government support because the political context citizens, policymakers, and parties act in
C.J. Anderson / Electoral Studies 19 (2000) 151–170
169
differs across countries and changes over time. It is seldom obvious how citizens
should react to economic performance given that most democracies exhibit some
institutional ambiguity and political change. Overall, the analysis thus modifies the
simple reward–punishment model of economic voting and suggests that students of
economic voting and government support take systematic account of a country’s
political context. Moreover, the analysis shows that pointing out specifically what
it is about the political context that matters is likely to be a fruitful avenue for
future research.
Acknowledgements
I would like to thank Joann Cruz for help with handling the data set. The data
used in this study are available from the University of Michigan’s Inter-university
Consortium for Political and Social Research (Study No. 6518). Thanks also to the
participants of the Sønderborg conference for stimulating discussions and helpful
comments. In particular I am grateful to Michael Lewis-Beck, Peter Nannestad, and
Martin Paldam for their feedback. Any remaining errors should be considered poor
judgment in the face of good advice.
References
Anderson, C.J., 1995a. Blaming the Government: Citizens and the Economy in Five European Democracies. Sharpe, Armonk, NY.
Anderson, C.J., 1995b. The dynamics of public support for coalition governments. Comparative Political
Studies 28, 350–383.
Bellucci, P., 1991. Italian economic voting: a deviant case or making a case for better theory? In: Norpoth,
H., Lewis-Beck, M.S., Lafay, J.D. (Eds.) Economics and Politics: The Calculus of Support. University
of Michigan Press, Ann Arbor, MI, pp. 63–84.
Eulau, H., Lewis-Beck, M.S., 1985. Economic Conditions and Electoral Outcomes in the United States
and Western Europe. Agathon, New York.
Laakso, M., Taagepera, R., 1979. Effective number of parties: a measure with application to West Europe.
Comparative Political Studies 12, 3–27.
Lewis-Beck, M.S., 1986. Comparative economic voting: Britain, France, Germany, Italy. American Journal of Political Science 30, 315–346.
Lewis-Beck, M.S., 1988. Economics and Elections: The Major Western Democracies. University of Michigan Press, Ann Arbor, MI.
Lewis-Beck, M.S., 1991. Introduction. In: Norpoth, H., Lewis-Beck, M.S., Lafay, J.D. (Eds.) Economics
and Politics: The Calculus of Support. University of Michigan Press, Ann Arbor, MI, pp. 1–8.
Lewis-Beck, M.S., Mitchell, G., 1993. Bringing politics back in: the french lesson for popularity functions.
Paper presented at the Annual Meeting of the Midwest Political Science Association, Chicago, IL.
Lijphart, A., 1984. Democracies. Yale University Press, New Haven, CT.
McDonald, M.D., Budge, I., Hofferbert, R.I., 1999. Governments, parties, elections, and votes: balancing
past performance and future choice. Ms. Binghamton University (State University of New York).
Nannestad, P., Paldam, M., 1994. The vp-function: a survey of the literature on vote and popularity
functions after 25 years. Public Choice 79, 213–245.
Norpoth, A., 1996. The economy. In: LeDuc, L., Niemi, R.G., Norris, P. (Eds.) Comparing Democracies:
Elections and Voting in Global Perspective. Sage, Thousand Oaks, CA.
170
C.J. Anderson / Electoral Studies 19 (2000) 151–170
Pacek, A., Radcliff, B., 1995. Economic voting and the welfare state: a cross-national analysis. Journal
of Politics 57, 44–61.
Paldam, M., 1991. How robust is the vote function? In: Norpoth, H., Lewis-Beck, M.S., Lafay, J.D. (Eds.)
Economics and Politics: The Calculus of Support. University of Michigan Press, Ann Arbor, MI, pp.
9–31.
Powell, G.B., 1989. Constitutional design and citizen electoral control. Journal of Theoretical Politics 1,
107–130.
Powell, G.B., Whitten, G.D., 1993. A cross-national analysis of economic voting: taking account of the
political context. American Journal of Political Science 37, 391–414.
Rae, D., 1967. The Political Consequences of Electoral Laws. Yale University Press, New Haven, CT.
Sørensen, R.J., 1987. Macroeconomic policy and government popularity in Norway, 1963–1986. Scandinavian Political Studies 10, 301–321.
Taagepera, R., Shugart, M.S., 1993. Seats and Votes. Yale University Press, New Haven, CT.
Wilkin, S., Haller, B., Norpoth, H., 1997. From Argentina to Zambia: a world-wide test of economic
voting. Electoral Studies 16, 301–316.