You are on page 1of 21

116 European Journal of Political Research 53: 116–135, 2014

doi: 10.1111/1475-6765.12019

Economic vulnerability and economic voting in 14 OECD countries

DIEGO FOSSATI
Department of Government, Cornell University, USA

Abstract. The economic voting literature shows that good economic performance bolsters the electoral
prospects of incumbents. However, disagreement persists as to whether voters in vulnerable economic
conditions are more likely to engage in economic voting. It is argued in this article that a crucial factor in
explaining individual-level variation in economic voting is the degree of exposure to economic risks,
because risk exposure affects the saliency of the economy in voting decisions. In particular, the focus is on
job insecurity and employability as key determinants of economic voting patterns. The article hypothesises
that the extent of economic voting is greater in voters who are more vulnerable to unemployment and less
employable in case of job loss. Support for these hypotheses is found in a test with a dataset that combines
survey data on incumbent support with occupational unemployment rates and other measures of exposure
to economic risks.

Keywords: economic voting; economic vulnerability; occupational unemployment; skills; employability

Introduction

A vast literature in political science explores the nexus between economic factors and
electoral outcomes. Despite the persistence of various areas of contention in the economic
voting literature (Lewis-Beck & Paldam 2000), the idea of the decisiveness of macroeco-
nomic performance in democratic elections has consolidated to the extent that it is now
widely accepted among the non-specialist public and the popular press. In recent years,
however, this body of knowledge has been challenged on two main empirical and theoreti-
cal dimensions (Anderson 2007). On the one hand, several studies have highlighted the
existence of considerable individual constraints to economic voting as individuals differ
greatly in their ability and willingness to acquire information about the economy (Gomez
& Wilson 2006). On the other hand, the institutional environment in which political behav-
iour takes place can affect economic voting decisively, because political institutions may
shape the extent to which voters hold incumbents responsible for economic performance
(Duch & Stevenson 2008).
It is argued in this article that, while responsibility attribution is a crucial link in the
economic voting process, there is another dimension of economic voting that deserves
equal consideration – namely the fact that voters present different degrees of exposure to
economic risks. Due to a host of micro-level factors and institutional variables, the risks
associated with economic shocks and bad economic performance are unevenly distributed
across the electorate. First, individuals differ from one another in their exposure to eco-
nomic risks due to factors such as wealth, income, education, gender, employment type,
professional and industry affiliation, and so on. Second, the exposure to risks from eco-
nomic volatility is influenced by public policy through social programmes and welfare state
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
Published by Blackwell Publishing Ltd
ECONOMIC VULNERABILITY AND ECONOMIC VOTING IN 14 OECD COUNTRIES 117
institutions. Risk exposure matters for economic voting because, as a growing literature
suggests, it is an important determinant of the saliency of economic performance in voting
behaviour (Hellwig 2001; Dorussen & Taylor 2002; Singer 2011a, 2011b). In particular, the
economy is more important for voters who are more directly exposed to its fluctuations as
their incomes are more closely dependent on macroeconomic performance. For this reason,
it is plausible to expect variation in responses to economic performance across subgroups
of the electorate as some groups are more likely to engage in economic voting than others.1
I argue here that such differences depend on the voters’ position in the economy, and more
precisely on their position in the labour market: economic voting is more pronounced in
individuals who face higher levels of job insecurity and in those who are less employable in
case of job loss.
This article contributes to recent efforts in the economic voting literature to identify
individual-level differences in voting behaviour (Duch et al. 2000; Hellwig 2001; Dorussen
& Taylor 2002; Gomez & Wilson 2006; Kayser & Wlezien 2011; Singer 2013) in two main
ways. First, it draws on the theoretical insights of a well-known but separate literature in
political science to investigate the factors that mediate the relationship between macroeco-
nomic performance and incumbent support. Although several studies have ascertained the
importance of ‘skills’ and professional job insecurity as a source of social policy preferences
and welfare state arrangements (Iversen & Soskice 2001; Estevez-Abe et al. 2001; Rehm
2009, 2011a), the significance of such factors for economic voting is still largely unexplored.2
This article integrates these two threads of research by showing that macroeconomic
performance is more decisive among individuals with more specific professional skills and
among those employed in professions more vulnerable to unemployment. Second, this
article offers new empirical evidence suggesting that voters respond to variations in mac-
roeconomic performance according to their position in the labour market. It tests the
hypothesised relationship between risk exposure and economic voting on a dataset that
combines existing survey data with institutional-level variables and various measures of job
insecurity and employability, offering a systematic empirical estimation of the role of risk
exposure as a key intervening variable in the economic voting process. In doing so, it
contributes with new data to the open debate on whether the salience of the economy as a
political issue varies across groups in the electorate.
The remainder of the article proceeds as follows. In the next section, I review the recent
literature on economic voting, with particular emphasis on the study of individual-level
variation in exposure to economic risk. I then outline a model of economic voting, showing
the specific role that the degree of exposure to the economic cycle plays in shaping voting
behaviour. Following this, I present the data used in this study and the strategies I follow for
the empirical test. Finally, I discuss the results of the regression analysis and conclude with
some remarks on the main avenues for further research.

Literature review

Recent work on economic voting has attempted to qualify in different ways the general
assertion that good macroeconomic performance increases incumbent support. As
Anderson (2007) observes, most of such theoretical innovations have focused on factors
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
118 DIEGO FOSSATI

moderating the relationship between economic performance and attitudes toward the
incumbent. Although a compelling frontier of the current academic debate is the integra-
tion of individual and contextual determinants of responsibility attribution (Duch &
Stevenson 2008), the economic voting field has, to a large extent, ignored the possibility that
the saliency of economic issues may vary over time and across the electorate, and that this
could be an important source of individual-level heterogeneity in economic voting pat-
terns.3 A large literature on public opinion and information in electoral campaigns shows
that the saliency of political issues varies across elections as election-specific events, cam-
paign strategies and agenda setting by the media may influence the degree to which voters
consider a certain issue as crucial for their voting choices (McCombs & Shaw 1972; Iyengar
& Kinder 1987; Iyengar & Simon 2000).4 Furthermore, established conceptualisations of the
process of information processing during electoral campaigns suggest that voters attach
different degrees of importance to different issues (Lavine et al. 2000). For this reason, a
central concern for students of economic voting should be the identification of the sources
of the salience of the economy in voting behaviour: what segments of the electorate are
especially likely to consider the economy as the crucial issue in choosing what candidate to
support?
Many studies of economic voting have been reluctant to study this question. Following
seminal work by Stokes (1963), the economy has often been considered as a ‘valence
issue’, as all voters arguably prefer a good economy to a bad one. As Stokes himself
observes, in debates over valence issues, ‘the argument turns on where the credit or blame
ought to be assigned’ or, in electoral competitions, on ‘which party, given possession of
the government, is the more likely to bring it about’ (Stokes 1963: 373). Since this para-
digm has been preponderant in the economic voting literature, most scholars have
focused on the process of responsibility attribution, assuming lack of variation in the
salience of economic performance (see, e.g., Van der Brug et al. 2007; Wlezien 2005;
McAvoy 2006). Yet some classic studies of economic voting have laid the theoretical
foundations for conceptualising the role of individual economic conditions in shaping
economic voting patterns. Hibbs (1977), in particular, famously showed that there are
differences across groups within the electorate in responsiveness to signals of macroeco-
nomic performance such as inflation and unemployment, and that such divergences are
rooted in the different economic interests of various segments of voters. In a similar vein,
Hibbs et al. (1982: 326–327) argued that: ‘It is natural to expect political responses to
macroeconomic performance to vary across groups because the burdens and rewards
conferred by fluctuations in aggregate economic conditions are very unevenly distributed
within the electorate.’
In recent years, only a few articles have addressed the determinants of exposure to
economic risks as moderators of the relationship between macroeconomic performance
and voting behaviour.5 Most notably, the argument that risk exposure is a determinant of
the salience of the economy in voting behaviour has been explored in work by Singer
(2011b, 2011c). Using survey data, Singer finds that voters in vulnerable economic condi-
tions are more likely to report that the economy is a decisive issue in deciding over
candidates.A limited number of other articles examine more explicitly the role of economic
vulnerability in shaping the extent of economic voting.6 Hellwig (2001), in his study of trade
and economic voting, shows that macroeconomic performance affects voting behaviour less
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
ECONOMIC VULNERABILITY AND ECONOMIC VOTING IN 14 OECD COUNTRIES 119
substantially in voters who are to some extent buffered from the fluctuations of the
economy – namely unionised workers, government workers and employees in ‘traditional’
(primary and secondary) economic sectors. Dorussen and Taylor (2002) find that the
economic voting is more prominent among low-skilled workers. Finally, Singer (2013) uses
survey data from Latin American and Eastern Europe to show that macroeconomic per-
formance affects support for the incumbent more prominently in voters who perceive
higher levels of job insecurity and lower levels of employability.7
The findings from these seminal articles call for a closer integration between the eco-
nomic voting literature and the insights of another well-known, although separate body of
work in political science. Scholarly studies of the determinants of social policy preferences
have presented ample evidence that there are deep inequalities in the distribution of
economic risks across the electorate, and that such disparities are a source of variation both
in preferences over redistribution (Iversen & Soskice 2001; Rehm 2009) and in macro-level
welfare state arrangements (Estevez-Abe et al. 2001; Rehm 2011a). In particular, it is
possible to identify two classes of individual-level factors that determine the exposure to
the economic cycle of individual voters. According to one line of analysis, the specificity of
professional skills developed through education and professionalisation practices is posi-
tively related to the degree of risk exposure because workers with non-transferable skills
are more likely to suffer from protracted unemployment in case of job loss (Iversen &
Soskice 2001). From another angle, risk exposure is a function of job insecurity rather than
employability. On the one hand, industry affiliation shapes job security through various
channels (Frieden & Rogowski 1996; Mayda & Rodrik 2005; Iversen & Cusack 2000). On
the other hand, Rehm (2009, 2011a) has focused on occupational unemployment rates as an
indicator of job insecurity, finding that professional status is a better predictor of risk
exposure than industry affiliation.
This article offers two main contributions to this debate. First, it calls for close integra-
tion between work on economic voting and the literature on social policy preferences. It
argues that an individual’s position in the labour market is a pivotal factor not only in
determining their preferences over social insurance, but also in shaping economic voting
patterns. At the macro-level, such disparities in risk exposure result in significant differ-
ences in voting behaviour across occupational groups. Second, this article performs the first
empirical test of this argument using data on occupational unemployment, which is a key
concept in the debate on social policy preferences and the origins of the welfare state.
Furthermore, it offers a range of alternative measures of the idea of ‘risk exposure’ to test
the robustness of its results to different operationalisation strategies.

Theory and model

In a basic version of the economic voting model, formalised by Alesina and Rosenthal
(1995), voters select the most competent manager for the national economy from a pool of
candidates. The argument I propose in this article is informed by three main assumptions of
this model of economic voting that are common to many empirical studies of the link
between macroeconomic performance and voting behaviour. First, Alesina and Rosenthal
posit that policy makers are, at least to some extent, able to influence macroeconomic
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
120 DIEGO FOSSATI

performance. Economic performance is a function not only of exogenous factors, which are
beyond the control of candidates, but also of the competence of elected officials as eco-
nomic mangers. Second, since the level of competence of each candidate is unknown, voters
can only infer the incumbent’s competence by observing current macroeconomic perform-
ance. Third, the state of the economy is an element of the voter’s utility function also
because voters believe that, to some degree, they will benefit directly from a good economy.
An important link is therefore assumed between macroeconomic performance and the
‘pocketbook’ of individual voters.
These core tenets of the economic voting model suggest that there are two main reasons
why voters may care about macroeconomic performance. On the one hand, the prosperity
of the economy may be considered as a good in itself, and economic voting may follow a
‘sociotropic’ pattern in which voters support incumbents only if they have furthered the
material well-being of their country as a whole (Kinder & Kiewiet 1981). On the other
hand, economic performance also matters because the future financial flows that individual
voters receive depend, at least in part, from the health of the national economy. For the
argument advanced here, it is crucial to assume that, at least to some extent, economic
voting takes place due to ‘pocketbook’ concerns about future repercussions of economic
performance on an individual’s personal finances. If voters did not care about how aggre-
gate economic performance affects future developments in their own personal finances,
exposure to economic risk would not be a relevant factor in economic voting. Rather, I
follow Hibbs et al.’s (1982) insight that the rewards and the burdens associated with
national economy performance are unevenly distributed across the electorate, and I argue
that such disparities, in particular as resulting from differences in risk exposure, are the
source of variation in voting behaviour across groups of the electorate.
To understand why this is the case, we need to consider a factor largely understudied in
literature on economic voting – namely the salience of the economy and its variation across
the electorate. In particular, two fundamental assumptions are made in this article about the
relationship between salience of the economy and voting behaviour. First, it assumes
substantial cognitive limits in the way voters receive and process information about specific
political issues. Following foundational work by Herbert Simon on ‘bounded rationality’
(Simon 1982), empirical research in behavioural economics and social psychology has
shown that individuals, acting under various forms of constraints, tend to consider only a
few items from their broader spectrum of preferences when taking decisions.8 In voting
behaviour, this means that most voters do not evaluate candidates based on their perform-
ance in all policy areas, but only on their performance in some areas they consider of
particular interest (Rabinowitz et al. 1982; Krosnick 1988; McGraw et al. 1990; Lavine et al.
1996; Belanger & Meguid 2008). This insight is important for economic voting because,
although all voters may prefer a more competent economic manager to a bad one, some of
them might care less about this attribute of the candidate than about others.
Second, it is assumed in this article that risk exposure is a key factor in shaping the
salience of the economy in voting decisions. The exposure to economic risk, as conceptu-
alised by Anderson and Pontusson (2007), consists of several dimensions, and it is deter-
mined by a host of institutional and micro-level factors. Following Singer (2013), I
conceptualise risk exposure as consisting of two main dimensions: job insecurity and
employability. On the one hand, individuals differ from one another in the degree to which
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
ECONOMIC VULNERABILITY AND ECONOMIC VOTING IN 14 OECD COUNTRIES 121
their income is at risk due to adverse economic shocks. For instance, some voters work in
industries that are very sensitive to economic fluctuations, while some others are employed
in institutions that are less likely to restructure their workforce due to short-term economic
factors. Empirical research has found that some sectors of the electorate – namely union-
ized workers (Bender & Sloane 1999), workers employed in the public sector (Anderson &
Pontusson 2007), workers in sectors with low levels of international economic integration
(Scheve & Slaughter 2002) and workers with long-term employment contracts (Näswall &
De Witte 2003) – enjoy higher levels of job security. On the other hand, individuals differ
in the resources at their disposal for dealing with such adversities induced by poor eco-
nomic performance. In particular, some workers are more employable than others as they
have invested in more sophisticated and more transferable skills that increase their chances
of finding new employment quickly in case of job loss (Iversen & Soskice 2001; Berntson
et al. 2006).

Data

To test empirically the argument outlined above, I use a dataset that combines macro-level
and survey data. In the models I specify, the degree of exposure to the economy at the
individual level is a moderating factor of the relationship between macroeconomic per-
formance and voting intentions for the incumbent.9 A micro-level measure of voting inten-
tions for the incumbent is included in recent International Social Survey Program (ISSP)
surveys, most of which ask a country-specific question about political parties.10 Respondents
are given a list with major political parties in their country and are asked to indicate their
preference. The exact wording of the question, however, varies substantially across coun-
tries, making cross-sectional analysis very problematic. To allow cross-country comparison,
I drop from the sample surveys in which the question asks about past voting behaviour or
general partisan identification.11 Typically, in the included surveys the list of political parties
is preceded by minor variations of the following question:

If we had a general election in [COUNTRY] today/ tomorrow/ next Sunday/ this


week, which of the following political parties would you vote for?

The dependent variable that measures voting intentions for the incumbent is derived from
responses to this question, and it is coded as a dummy variable assuming the value of 1 if the
indicated party is part of the government coalition and of 0 otherwise.12
To gauge the micro-level degree of exposure to the economic cycle, I adopt several
indicators. Some of them are measures of the degree of job insecurity of the respondent.
Following work by Rehm (2009, 2011a), the first operationalisation of economic risk I use is
a measure of unemployment at the occupational level.13 In particular, occupational unem-
ployment is measured through occupational unemployment rates (OURs), calculated as the
share of unemployed workforce in each of the nine ISCO-88 professional categories.14 As
additional measures of job insecurity, I use dummy variables that indicate whether the
respondent is unemployed, a government worker and a union member. Finally, as a measure
of the respondent’s professional skills and employability, I adopt a four-level indicator of his
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
122 DIEGO FOSSATI

or her education that assumes higher values for higher levels of education attainment,
associated with more transferable professional skills and a higher degree of employability.15
I combine the micro-level dataset with country-level information on macroeconomic
performance measured as the annual change of the national gross domestic product (GDP)
in the year the survey was taken.16 As the previous section has illustrated, there are
structural and institutional variables that moderate the extent of economic voting at the
macro-level. In particular, the degree of clarity of responsibility and the availability of a
comprehensive welfare state system are important factors that may pose an inferential
challenge, and should be included as control variables in the model.17 To measure clarity of
responsibility in a specific country-year, I build an index that includes information on
divided government, minority government, coalition government and federalism.18 As a
measure of the provision of social services by the state, I use data on expenditure on social
services, which include a wide range of social insurance programmes designed to shelter
workers from various forms of economic risks associated with participation in the labour
force.19 Finally, I include GDP per capita in all estimations as a measure of socioeconomic
development, and a battery of micro-level sociodemographic covariates in all models.

Estimation

In a basic version of the model, support for the incumbent is a function of both micro- and
macro-level variables, and the effect of macroeconomic variables on incumbent support is
estimated as a constant parameter that does not change according to the degree of risk
exposure of the respondent. This model could be estimated with a simple logistic regression
in which independence among observations is assumed. However, complication may arise
from the structure of the data, which is both clustered and serially correlated. Violation of
the assumption of independence among observations tends to underestimate the standard
errors of the regression coefficients, increasing the risk of producing ‘significant’ results for
spurious relationships (Hox 2002: 5–6). This can be a particularly serious problem when
some of the fixed effects of interests, like in our case, vary at the cluster level (Arceneaux
& Nickerson 2009). For this reason, an empirical test of the argument of this article requires
an estimation technique that accounts for both geographic and temporal clustering,
although traditional longitudinal data analysis tools are unsuitable since this is not panel
data. The three-level hierarchical logit model I outline below, in which voters are clustered
in countries, and countries in years, should offer a more accurate estimation of coefficients
and standard errors and it constitutes a useful robustness test for the results obtained with
simple logistic regression.20
For estimation, consider first a simple random intercept logit-link model in which no
covariates are included, and support for the incumbent in individual i only depends on the
country (j) and the year (k) in which the survey was implemented:

Prob ( Incumbent Supported ijk = 1 β ) = ϕ ijk


log [ϕ ijk (1 − ϕ )] = Yijk (1)
Yijk = β0 jk + ε ijk

© 2013 The Author(s)


European Journal of Political Research © 2013 European Consortium for Political Research
ECONOMIC VULNERABILITY AND ECONOMIC VOTING IN 14 OECD COUNTRIES 123
where b0jk is the random intercept of Eqn 1, or the average probability for a respondent in
country j, year k, of expressing support for the incumbent. The value of this random
parameter is different for each country j and year k, and it contains both level 2 (country)
and level 3 (year) error terms:

β0 jk = β0 + µ oj + γ ok (2)

This specification of the model, accounting for random effects at both country and
year level, allows the application of standard hypothesis testing procedures. The
principal argument in this article, however, is that economic voting is conditional on the
exposure to economic risks. Such conditionality can be modeled with the introduction
of a cross-level multiplicative interaction effect between macroeconomic performance
and level 1 exposure to economic risk, measured here through the various operationali-
sations of job insecurity and employability previously discussed. Eqn 3 outlines the
reduced form of the random intercept model with the interaction effect and its consti-
tutive terms:

Yijk = β0 + β1 jk ( Macroeconomic Performance) + β 2 ijk (Exposure to the economy)


(Macroeconomic performance) + β 3 ijk (Exposure to the economy) + (3)
β 4 jk (Macro-level controls) + β 5 ijk (Micro-level controls) + µ oj + γ ok + ε ijk

Since I hypothesise economic growth to increase support for the incumbent, I expect b1jk
to be positively signed in all estimations. As for the coefficients of the interaction effects
b2ijk, I expect the sign to vary according to the operationalisation used.

Analysis

In this section, I estimate the model outlined in Eqn 3 using various operationalisation
strategies for the concept of exposure to economic risk. In models 1 (simple logit) and 1r
(with random effects) reported in Table 1, the effect of macroeconomic performance on
incumbent support is conditional on the occupational unemployment rate of the profession
of the respondent, a measure of his or her job insecurity. The reported coefficient for GDP
growth, 0.174, can be interpreted as the marginal effect of a unit increase of GDP growth for
a respondent in a profession with an average occupational unemployment rate, which in
this sample is 6.12 per cent. For example, the expected probability of incumbent support for
a respondent in a stagnating economy (0 per cent growth) is about 0.28, but the same
probability increases to about 0.36 in an economy growing at 2 per cent a year. Further-
more, as expected, the coefficient of the interaction term is positive, which indicates that
GDP growth is a more decisive factor for respondents employed in professions at higher
risk of unemployment. Its magnitude is estimated at 0.020 in the logistic regression model
and at 0.017 in the random effects model, and it is different from zero at the 0.01 level of
significance in both specifications. This is a first indicator that economic voting occurs to
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
124 DIEGO FOSSATI

Table 1. Support for the incumbent, macroeconomic performance and job insecurity

1 1r 2r

Interaction effect Interaction effect Baseline

GDP growth 0.174** (0.013) 0.161** (0.057) 0.168** (0.058)


GDP growth*OUR 0.020** (0.0003) 0.017** (0.0003)
Clarity of responsibility 0.171** (0.020) 0.106 (0.083) 0.092 (0.084)
Social policy expenditure 0.046** (0.004) 0.047* (0.019) 0.050** (0.019)
Per capita GDP -0.032** (0.003) -0.034** (0.012) -0.034** (0.012)
Education 0.131** (0.018) 0.095** (0.019) 0.098** (0.019)
Unemployed -0.354** (0.088) -0.359** (0.088) -0.371** (0.088)
Union member 0.073* (0.030) 0.011* (0.032) 0.007 (0.032)
Government worker -0.052 (0.033) -0.042 (0.033) -0.043 (0.033)
OUR -0.026** (0.004) -0.029** (0.005) -0.027** (0.005)
Constant -0.905** (0.015) -0.888** (0.064) -0.889** (0.065)
Random effect s (country) 0.300** (0.043) 0.304** (0.043)
Random effect s (year) 0.000 (0.128) 0.001 (0.198)
Log-likelihood -18,179.187 -17,966.741 -18,013.478
N 30,277 30,277 30,277

Notes: Dependent variable: Support for the incumbent. Standard errors in parentheses. * p < 0.05;
** p < 0.01. All models include controls for gender, age, income and employment status (dummy variables
for whether the respondent is employed, unemployed, retired).

different extents across occupational groups. To investigate the robustness of this finding, I
compare model 1r with a baseline model in which the effect of macroeconomic perform-
ance on incumbent support does not depend on the degree of economic vulnerability of the
respondent (model 2r). The higher value of the log-likelihood in model 1r suggests a
significant improvement of fit with the introduction of a multiplicative term. Furthermore,
a likelihood-ratio test between the two models delivers a value of 33.48 (associated with a
p-value lower than 0.01), which indicates that the interaction term should not be omitted
from the model. Therefore, the estimations reported in Table 1 provide empirical support
for the hypothesis that risk exposure is an important factor in shaping economic voting
patterns.
For the purposes of this article, we are also interested in understanding exactly how the
marginal effect of macroeconomic performance varies across levels of risk exposure.
Figure 1 plots the estimated marginal effect of GDP growth and its 95 per cent confidence
intervals as a function of the occupational unemployment rate of the respondent. Consist-
ently with expectations, the curve shows a strong positive association between the two
factors. For low levels of unemployment, macroeconomic performance has a relatively
limited impact on support for the incumbent. However, as the risk of unemployment
increases, the magnitude of the marginal effect increases accordingly. For example, while
the estimated marginal effect is about 0.11 for a respondent in a low-risk occupation (3 per
cent unemployment), it increases to about 0.23 for an occupation with a higher unemploy-
ment risk (9 per cent unemployment). To understand what this means in terms of expected
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
ECONOMIC VULNERABILITY AND ECONOMIC VOTING IN 14 OECD COUNTRIES 125

Figure 1. Marginal effect of macroeconomic performance on incumbent support at different occupational


unemployment rates.

probability of incumbent support, consider again the hypothetical difference between a


stagnant and a growing economy. For a respondent in a low-risk occupation, a good
economy increases the expected probability of incumbent support of about 0.05. However,
the increase for respondents in high-risk occupations is much higher, as it is estimated at
0.10. This suggests that, as hypothesised, macroeconomic performance affects support for
the incumbent more substantially in respondents employed in higher risk professions.21
Models 3–5 estimate the model in Eqn 3 using alternative measures of job insecurity.
Model 3 interacts macroeconomic growth with a dummy variable that measures whether
the respondent is unemployed. As unemployed respondents are in particularly vulnerable
economic conditions, we should expect the coefficient for the interaction term to be posi-
tive. The estimation reported in Table 2 shows that this is indeed the case, although the
coefficient is not significant at the 0.05 level in either the simple logistic or the random effect
model. The role of employment in the public sector as an indicator of job security is
explored in model 4. The results reported show that the significance of public sector
employment is very sensitive to model specification as it is barely significant in the random
effect model (model 4r) and not significant in the simple regression model. Furthermore,
contrary to expectations, the coefficient is positively signed and of modest magnitude.
Therefore, there does not seem to be a discernible difference in economic voting patterns
between respondents employed in the private sector and those employed in the public
sector. Finally, estimations for models 5 and 5r offer empirical evidence that being a union
member is an important factor in economic voting behaviour. The coefficient of the inter-
action term is negatively signed and significant at the 0.01 level in both models, although the
magnitude of the coefficient varies somewhat across model specification. In the random
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
126

Table 2. Support for the incumbent, macroeconomic performance and job insecurity: Alternative measures

© 2013 The Author(s)


3 3r 4 4r 5 5r

GDP growth 0.182** (0.013) 0.168** (0.058) 0.183** (0.013) 0.169** (0.058) 0.176** (0.013) 0.163** (0.057)
GDP growth*Unemployed 0.105 (0.055) 0.090 (0.057)
GDP growth*Government 0.029 (0.023) 0.048* (0.024)
worker
GDP growth*Union member -0.135** (0.020) -0.118** (0.023)
Clarity of responsibility 0.152** (0.020) 0.092 (0.084) 0.151** (0.020) 0.092 (0.084) 0.147** (0.020) 0.088 (0.084)
Social policy expenditure 0.049** (0.004) 0.050** (0.019) 0.049** (0.004) 0.051** (0.019) 0.049** (0.004) 0.050** (0.019)
Per capita GDP -0.031** (0.003) -0.034** (0.012) -0.031** (0.003) -0.033** (0.012) -0.031** (0.003) -0.034** (0.012)
Education 0.140** (0.018) 0.098** (0.019) 0.140** (0.018) 0.098** (0.019) 0.141** (0.018) 0.098** (0.019)
Unemployed -0.332** (0.090) -0.334** (0.091) -0.381** (0.087) -0.374** (0.088) -0.366** (0.087) -0.369** (0.088)
Union member 0.084** (0.030) 0.007* (0.032) 0.083** (0.030) 0.004 (0.032) 0.060 (0.030) 0.004 (0.032)
DIEGO FOSSATI

Government worker -0.053 (0.033) -0.043 (0.033) -0.051 (0.033) -0.040 (0.033) -0.044 (0.033) -0.036 (0.033)

European Journal of Political Research © 2013 European Consortium for Political Research
OUR -0.023** (0.004) -0.027** (0.005) -0.023** (0.004) -0.027** (0.005) -0.021** (0.004) -0.027** (0.005)
Constant -0.910** (0.015) -0.888** (0.064) -0.912** (0.015) -0.888** (0.065) -0.918** (0.015) -0.897** (0.064)
Random effect s (country) 0.304** (0.043) 0.304** (0.043) 0.302** (0.043)
Random effect s (year) 0.000 (0.193) 0.000 (0.227) 0.000 (0.139)
Log-likelihood -18,208.096 -18,012.205 -18,209.14 -18,011.402 -18,187.995 -18,000.183
N 30,277 30,277 30,277 30,277 30,277 30,277

Notes: Dependent variable: Support for the incumbent. Standard errors in parentheses. * p < 0.05; ** p < 0.01. All models include controls for gender, age,
income and employment status (dummy variables for whether the respondent is employed, unemployed, retired).
ECONOMIC VULNERABILITY AND ECONOMIC VOTING IN 14 OECD COUNTRIES 127
effect model, the coefficient of the interaction term is estimated at –0.118: this yields
an estimated marginal effect of GDP growth of 0.08 for union members and of 0.20,
more than double, for non-members.22 As hypothesised, macroeconomic performance is a
less decisive factor for supporting the incumbent if the respondent is a member of a trade
union.
Table 3 reports estimations of Eqn 3 in which risk exposure is operationalised with
education as a measure of the respondent’s skills, a proxy for his or her employability. As
respondents with lower levels of education should be less employable in case of job loss,
I expect them to be more likely to engage in economic voting.23 Models 6 and 6r interact
macroeconomic performance with a four-level indicator of the respondent’s education,
with higher values denoting a higher education level. As expected, the coefficient is
strongly negative and significant at the 0.01 level in both specifications: the estimated
marginal effect of macroeconomic performance on the degree of support for the incum-
bent decreases in more employable, better educated respondents. Figure 2 illustrates this
finding by showing the point estimates and the relative 95 per cent confidence intervals of
the marginal effect of GDP growth on incumbent support at different levels of educa-
tional attainment. The chart shows that the estimated marginal effect declines sharply with
increases in education, as it drops from about 0.30 for respondents who have only com-
pleted elementary school to 0.07 for college graduates. A similar pattern can be detected
in the results reported in the last two columns of Table 3, obtained by estimating a base-
line model in which economic growth affects incumbent support evenly across the elec-
torate, regardless of the respondent’s economic vulnerability. If we estimate this model
after partitioning the sample into two groups of respondents – namely low-skill and high-
skill workers – we should see a clear difference in the extent and significance of economic
voting between the two groups.24 The results reported show that this expectation is borne
out by the empirics. For the low-skill group (model 7), the coefficient for GDP growth is
strongly positive and significant at the 0.01 level. The marginal effect of GDP growth on
incumbent support is estimated at 0.342 – a substantially greater magnitude than in
models estimated with the full sample. By contrast, the same coefficient drops to 0.016 for
high-skill respondents (model 8), and it is not different from zero at conventional levels of
significance.25
Figure 3 offers a graphic representation of the difference between the two groups,
plotting two distinct curves of predicted probabilities of incumbent support as a function of
GDP growth. As the chart shows, the curve for low-skill respondents (solid curve) is
substantially steeper than the high-skill curve (dashed curve) as incumbent support for
low-skill respondents is more sensitive to macroeconomic fluctuations. For illustration
purposes, consider the difference in predicted probabilities of incumbent support between
a bad economic conjuncture (e.g., – 1 per cent growth) and a very good one (e.g., 3 per cent
growth). For low-skill respondents, the difference is sharp: poor economic performance
decreases support for the incumbent by about 0.21, as it drops from 0.45 to 0.24. For the
high-skill group, however, the difference is only about 0.08 (from 0.39 to 0.31). The chart
also suggests that the divergence is more marked during economic crises than in times of
economic booms – perhaps an indication that the two groups differ more in their willing-
ness to punish incumbents for bad economic management than in their willingness to
reward them for good performance.
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
128

© 2013 The Author(s)


Table 3. Support for the incumbent, macroeconomic performance and employability

7 8

6 6r Low skills High skills

GDP growth 0.172** (0.013) 0.158** (0.057) 0.342** (0.069) 0.016 (0.064)
GDP growth*Education -0.075** (0.011) -0.070** (0.012)
Clarity of responsibility 0.155** (0.020) 0.092 (0.084) 0.073 (0.094) 0.092 (0.093)
Social policy expenditure 0.046** (0.004) 0.048* (0.019) 0.079** (0.021) 0.008 (0.021)
Per capita GDP -0.031** (0.003) -0.034** (0.012) -0.044** (0.015) -0.033* (0.014)
Education 0.147** (0.018) 0.106** (0.019) 0.558** (0.082) -0.007 (0.040)
Unemployed -0.362** (0.088) -0.361** (0.088) -0.294* (0.120) -0.417** (0.138)
Union member 0.079** (0.030) 0.009 (0.032) 0.128* (0.050) -0.080 (0.041)
DIEGO FOSSATI

Government worker -0.057 (0.033) -0.046 (0.033) 0.086 (0.056) -0.111** (0.041)
OUR -0.021** (0.004) -0.026** (0.005) -0.026** (0.006) -0.042** (0.008)

European Journal of Political Research © 2013 European Consortium for Political Research
Constant -0.906** (0.015) -0.887** (0.064) -0.659** (0.112) -0.807** (0.076)
Random effect s (country) 0.303** (0.043) 0.322** (0.056) 0.328** (0.048)
Random effect s (year) 0.000 (0.164) 0.117 (0.097) 0.000 (0.083)
Log-likelihood -18,185.24 -17,995.296 -8,235.256 -9,572.960
N 30,277 30,277 14,786 15,491

Notes: Dependent variable: Support for the incumbent. Standard errors in parentheses. * p < 0.05; ** p < 0.01. All models include controls for gender, age,
income and employment status (dummy variables for whether the respondent is employed, unemployed, retired).
ECONOMIC VULNERABILITY AND ECONOMIC VOTING IN 14 OECD COUNTRIES 129

Figure 2. Marginal effect of macroeconomic performance on incumbent support at different levels of


educational attainment.

Conclusions

Recent developments in the literature on economic voting have been instrumental in


refining our understanding of the link between macroeconomic performance and voting
behaviour. Most of the recently published studies acknowledge the importance of economic
voting while qualifying its extent, showing substantial differences both across individuals
and countries. In particular, many have focused on the idea of responsibility attribution and
on individual-level characteristics that shape the cognitive processes involved in the eco-
nomic voting model. I have observed in this article that there is another, equally important
and understudied dimension in the process of economic voting. I have argued, in particular,
that the study of economic voting should take into account the valuable insights provided
by the scholarship on social policy preferences and welfare state development. Specifically,
I have proposed in this article that voters be categorised in different groups according to
their vulnerability to economic fluctuations, and hypothesised that economic voting occurs
more extensively in voters who are more directly exposed to the economic cycle. The
analysis performed in the previous section has found important empirical regularities
consistent with this hypothesis: voters facing higher degrees of job insecurity and lower
employability in case of job loss are more likely to condition their support for incumbent
governments to good macroeconomic performance.
The findings of the analysis carried out here suggest that the exposure to the economic
cycle matters in determining the extent of economic voting, and the data analysed show that
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
130 DIEGO FOSSATI

Figure 3. Macroeconomic performance and predicted probability of supporting the incumbent for two
groups of voters.

this conclusion is robust across different model specifications and several operationalisa-
tion choices. Studies of economic voting that ignore this important dimension of the
economic voting causal narrative are prone to the risk of drawing an incomplete picture of
voting behaviour. This finding calls for a more organic integration between studies of
economic voting and the literature on welfare state and preferences over redistribution in
postindustrial societies.
The positive findings reported here, however, should be qualified by some limitations of
the research design adopted for this study, and indeed such shortcomings offer potentially
fruitful avenues for further research on the nexus between economic voting and economic
risk. First, the argument outlined in this article should be tested with actual voting data as
the response variable adopted here is a measure of public opinion rather than voting
behaviour. To be sure, the analysis has been restricted to respondents who expressed the
intention to vote, but discrepancies between such intentions and actual voting choices,
including non-voting, are not controlled for in this study. Second, the sample is limited to a
small group of advanced economies. The inclusion of a larger number of clusters in the
sample could address potential methodological shortcomings in the estimations of standard
errors for macro-level variables, and it could allow the exploration of interaction effects
between cluster-level covariates. Furthermore, campaign-specific dynamics and political
communication strategies, while not studied here, are important factors in determining the
saliency of the economy in electoral contests. Finally, this article has not tested the causal
mechanism postulated. There is, as already mentioned, convincing evidence that links the
© 2013 The Author(s)
European Journal of Political Research © 2013 European Consortium for Political Research
ECONOMIC VULNERABILITY AND ECONOMIC VOTING IN 14 OECD COUNTRIES 131
various steps of the process – for instance, in studies showing that economic issues matter
most to those in precarious employment conditions. However, having a unified dataset with
micro-level measures for each causal link would allow a test of a fully specified model, and
enhance greatly the inferential leverage of future empirical research.

Acknowledgements

I am thankful to Chris Anderson, Bryce Corrigan, Rodolfo Espino, Tim Hellwig, Naomi
Hsu, Kevin Morrison, Tom Pepinsky, Matthew Singer, Chris Way, Jessica Weeks and three
anonymous reviewers for providing very useful feedback at various stages of this project. I
am also grateful to Philipp Rehm for sharing his data on occupational unemployment rates.
All errors are my own.

Notes

1. This view contrasts with the conceptualisation of the economy as a ‘valence issue’ that concerns all voters
regardless of their vulnerability to economic risks. See, e.g., Wlezien (2005); Van der Brug et al. (2007).
2. I will discuss the few exceptions to this claim in the literature review section.
3. See Singer (2011c: 622–623) for a more exhaustive treatment of this point.
4. A key idea in many of these studies is that, due to cognitive limitations and resource constraints, voters only pay
attention to a limited number of issues.
5. In this literature review, I focus on micro-level determinants of risk exposure, although one can identify a similar
paucity of work on the nexus between economic voting and institutional factors such as social policy pro-
grammes and welfare state regimes. Two important exceptions are Pacek and Radcliff (1995) and Singer
(2011a).
6. A related strand of the literature has investigated the distribution of economic risks and assets as a determinant
of partisanship. See Nadeau et al. (2011); Rehm (2011b); Lewis-Beck and Nadeau (2011).
7. Some early work on voting behaviour has articulated a similar argument with an analysis of social class as a
determinant of economic voting. In a seminal article, Weatherford (1978: 919) contended that economic voting
during recessions ‘is not a simple across-the-board rejection of incumbents, but a more subtle process in which
economic effects are mediated by class status at a number of stages’. He found, in particular, that the voting
behaviour of working-class voters shows higher responsiveness to economic fluctuations. Yet the vast literature
on ‘class voting’ has neglected these initial findings, focusing instead on a debate on whether social class has
been losing significance as a determinant of partisanship (Evans 2000). To the extent that lower classes feature
higher degrees of risk exposure, this article might relate to class voting research, suggesting a reason why
working-class voters may exhibit more volatile partisan affiliations than voters with high socioeconomic status.
A test of this hypothesis, however, transcends the scope of the analysis presented here.
8. See Krosnick et al. (2010) for a recent review of this literature, with a particular focus on its contributions to
political science.
9. Unfortunately, the data I am using for estimation does not allow empirical validation of the causal mechanism
outlined in the previous section. Ideally, the micro-level data for the empirical test would include indicators
of the respondent’s assessments of the economic climate, the salience they attribute to the economy and other
political issues, their perceptions of economic vulnerability, and so on. However, using ‘objective’ measures
of economic performance and vulnerability rather than subjective evaluations, this article avoids the possible
endogeneity between voting behaviour and attitude formation identified in several studies (e.g., Anderson
et al. 2004; Evans & Andersen 2006).
10. Some readers may feel that European Social Survey (ESS) data would be more suitable for the purposes of this
article, as it would allow controlling for a wider range of variables. However, ISSP data provides a better

© 2013 The Author(s)


European Journal of Political Research © 2013 European Consortium for Political Research
132 DIEGO FOSSATI

measure of support for the incumbent as some of its surveys ask more specifically about voting intentions rather
than about past voting behaviour or party identification.
11. Surveys implemented in the Unites States are examples of such cases. I have also dropped from the sample
surveys for which a key individual-level measure of exposure to economic risks – namely unemployment at the
occupational level – was not available. The final sample includes a total of 31 national surveys from various
ISSP modules implemented in 14 OECD countries. More precisely, these are the countries included, and the
years of fieldwork of their respective surveys: Canada (2004), United Kingdom (2001–2004), Ireland (2002,
2003), Spain (2001, 2003), Portugal (2003, 2004), Austria (2004), The Netherlands (2008), Germany (2002,
2003), Finland (2001–2004), Sweden (2002–2004), Norway (2002–2004), Denmark (2002, 2003), Australia
(2002, 2003) and New Zealand (2001, 2003).
12. Respondents who declined to answer, declared themselves undecided or expressed the intention not to vote for
any of the parties mentioned are coded as 0.
13. I am very grateful to Philipp Rehm for kindly sharing these data. The specific data sources used to calculate
occupational unemployment rates vary by country. As reported in Rehm (2011a), data for Australia, Canada and
New Zealand is drawn from the International Labour Office’s (ILO) Database on Labour Statistics. As for
European countries, the European Union’s Labour Force Survey provides data for most countries, although ILO
data are used for a limited number of country-years.
14. After obtaining OURs with this procedure, it is possible to match them with available survey data. Each ISSP
survey contains ISCO-88 codes that denote the profession of survey respondents (there are various degrees of
precision for this measure, and I use the one-digit code for the purposes of this article). For each respondent,
I thus match the ISCO-88 code with the respective OUR calculated for the given country and year the survey
was taken.
15. The four categories, from lower to higher values, denote respondents who have completed primary school,
middle school, high school and college, respectively.
16. Data from the World Bank’s World Development Indicators dataset. For surveys taken in the first three months
of the year, I use macroeconomic indicators of the previous year.
17. Ideally, the role of macro-level factors such as clarity of responsibility and welfare state provisions should be
explored more thoroughly. However, the limited number of country-years available in the dataset makes
estimation of multiplicative interactions between cluster-level variables problematic.
18. I use data from the Comparative Political Data Set III (Armingeon et al. 2010) to build this index.
19. Expenditure is measured as the total government and private expenditure on social services as share of the gross
domestic product. This indicator includes government expenditure for: ‘Old age, Survivors, Incapacity-related
benefits, Health, Family, Active labor market programs, Unemployment, Housing, and Other social policy
areas.’ Data are from the OECD Social Expenditure Database (SOCX).
20. The larger value of the standard error of point estimates calculated with this method, however, may be an artifact
of the relatively small number of countries included in the dataset. In particular, the limited number of clusters
could be problematic in estimating standard errors of higher level fixed effects, and of interactions involving two
higher level covariates (Hox 2002: 174–177). For this reason, in the next section I report both the simple and
the random effect logistic model.
21. Some information about the empirical variation of OURs in the sample may be helpful to understand the
applicability of the estimations displayed in the figure. Most respondents (about 80 per cent of the sample) are
associated with OURs ranging between 2 and 10 per cent. There is significant variation in OURs across
professions, as professional groups with more specific skills show higher levels of occupational unemployment.
There are also substantial differences across countries, and more limited variation across years. As for the
variability over time of OURs in each profession, most professional groups show a rather stable OUR across
time.
22. Again, the effect of union membership can be expressed in terms of expected probabilities if we think about the
difference between a stagnant and a growing economy. For union members, being in a good economy increases
the expected probability of incumbent support of 0.03; for non-members, the increase is significantly higher
(0.09).
23. Note that this measure of employability is different from operationalisations often found in the ‘skill specificity’
literature discussed earlier. ‘High-skill’ is therefore not a synonym of ‘high skill specificity’, although the two
distinctions are, to some extent, related.

© 2013 The Author(s)


European Journal of Political Research © 2013 European Consortium for Political Research
ECONOMIC VULNERABILITY AND ECONOMIC VOTING IN 14 OECD COUNTRIES 133
24. Respondents with an education level inferior to a high school diploma are considered ‘low-skill’, while others
are grouped into the ‘high-skill’ group.
25. Estimations with simple logistic regression – not reported here – yield very similar results.

References

Alesina, A. & Rosenthal, H. (1995). Partisan politics, divided government and the economy. Cambridge:
Cambridge University Press.
Anderson, C.J. (2007). The end of economic voting? Contingency dilemmas and the limits of democratic
accountability. Annual Review of Political Science 10: 271–296.
Anderson, C.J. & Pontusson, J. (2007). Workers, worries and welfare states: Social protection and job
insecurity in 15 OECD countries. European Journal of Political Research 46(2): 211–235.
Anderson, C.J. et al. (2004). Endogenous economic voting: Evidence from the 1997 British election. Elec-
toral Studies 23(4): 683–708.
Arceneaux, K. & Nickerson, D.W. (2009). Modeling certainty with clustered data: A comparison of
methods. Political Analysis 17(2): 177–190.
Armingeon, K. et al. (2010). Comparative political data set III, 1990–2009. Berne: Institute of Political
Science, University of Berne.
Belanger, E. & Meguid, B.M. (2008). Issue salience, issue ownership and issue-based vote choice. Electoral
Studies 27(3): 477–491.
Bender, K.A. & Sloane, P.J. (1999). Trade union membership, tenure and the level of job insecurity. Applied
Economics 31(1): 123–135.
Berntson, E., Sverke, M. & Marklund, S. (2006). Predicting perceived employability: Human capital or
labour market opportunities? Human Resources Abstracts 42(1): 223–244.
Dorussen, H. & Taylor, M. (2002). Group economic voting:A comparison of the Netherlands and Germany.
In Economic voting. London: Routledge.
Duch, R.M. & Stevenson, R.T. (2008). The economic vote: How political and economic institutions condition
election results. Cambridge: Cambridge University Press.
Duch, R.M., Palmer, H.D. & Anderson, C.J. (2000). Heterogeneity in perceptions of national economic
conditions. American Journal of Political Science 44(4): 635–652.
Estevez-Abe, M., Iversen, T. & Soskice, D.W. (2001). Social protection and the formation of skills: A
reinterpretation of the welfare state. In P.A. Hall & D.W. Soskice (eds), Varieties of capitalism: The
institutional foundations of comparative advantage. Oxford: Oxford University Press.
Evans, G. (2000). The continued significance of class voting. Annual Review of Political Science 3: 401–417.
Evans, G. & Andersen, R. (2006). The political conditioning of economic perceptions. Journal of Politics
68(1): 194–207.
Frieden, J.A. & Rogowski, R. (1996). The impact of the international economy on national policies: An
analytical overview. In R.O. Keohane & H. Milner (eds), Internationalization and domestic politics.
Cambridge: Cambridge University Press.
Gomez, B.T. & Wilson, J.M. (2006). Cognitive heterogeneity and economic voting: A comparative analysis
of four democratic electorates. American Journal of Political Science 50(1): 127–145.
Hellwig, T.T. (2001). Interdependence, government constraints and economic voting. Journal of Politics
63(4): 1141–1162.
Hibbs, D.A. (1977). Political parties and macroeconomic policy. American Political Science Review 71(4):
1467–1487.
Hibbs, D.A., Rivers, R.D. & Vasilatos, N. (1982). The dynamics of political support for American Presidents
among occupational and partisan groups. American Journal of Political Science 26(2): 312–332.
Hox, J.J. (2002). Multilevel analysis: Techniques and applications. Mahwah, NJ: Lawrence Erlbaum Associ-
ates.
Iversen, T. & Cusack, T.R. (2000). The causes of welfare state expansion: Deindustrialization or globaliza-
tion? World Politics 52(3): 313–349.

© 2013 The Author(s)


European Journal of Political Research © 2013 European Consortium for Political Research
134 DIEGO FOSSATI

Iversen, T. & Soskice, D. (2001). An asset theory of social policy preferences. American Political Science
Review 95(4): 875–893.
Iyengar, S. & Kinder, D.R. (1987). News that matters: Television and American opinion. Chicago, IL:
University of Chicago Press.
Iyengar, S. & Simon, A.F. (2000). New perspectives and evidence on political communication and campaign
effects. Annual Review of Psychology 51(1): 149–169.
Kayser, M.A. & Wlezien, C. (2011). Performance pressure: Patterns of partisanship and the economic vote.
European Journal of Political Research 50(3): 365–394.
Kinder, D.R. & Kiewiet, D.R. (1981). Sociotropic politics: The American case. British Journal of Political
Science 11(2): 129–161.
Krosnick, J.A. (1988). The role of attitude importance in social evaluation: A study of policy preferences,
presidential candidate evaluations and voting behavior. Journal of Personality and Social Psychology
55(2): 196–210.
Krosnick, J.A., Visser, P.S. & Harder, J. (2010). The psychological underpinnings of political behavior. In S.T.
Fiske et al. (eds), Handbook of social psychology, 5th edn. Hoboken, NJ: Wiley.
Lavine, H., Borgida, E. & Sullivan, J.L. (2000). On the relationship between attitude involvement and
attitude accessibility: Toward a cognitive-motivational model of political information processing. Politi-
cal Psychology 21(1): 81–106.
Lavine, H. et al. (1996). The relationship of national and personal issue salience to attitude accessibility on
foreign and domestic policy issues. Political Psychology 17(2): 293–316.
Lewis-Beck, M.S. & Nadeau, R. (2011). Economic voting theory: Testing new dimensions. Electoral Studies
30(2): 288–294.
Lewis-Beck, M.S. & Paldam, M. (2000). Economic voting: An introduction. Electoral Studies 19(2): 113–121.
Mayda, A.M. & Rodrik, D. (2005). Why are some people (and countries) more protectionist than others?
European Economic Review 49(6): 1393–1430.
McAvoy, G. (2006). Stability and change: the time varying impact of economic and foreign policy evalua-
tions on presidential approval. Political Research Quarterly 59(1): 71–83.
McCombs, M.E. & Shaw, D.L. (1972). The agenda-setting function of mass media. Public Opinion Quarterly
36(2): 176–187.
McGraw, K.M., Lodge, M. & Stroh, P. (1990). On-line processing in candidate evaluation: The effects of
issue order, issue importance and sophistication. Political Behavior 12(1): 41–58.
Nadeau, R., Foucault, M. & Lewis-Beck, M.S. (2011). Assets and risk: A neglected dimension of economic
voting. French Politics 9(2): 97–119.
Näswall, K. & De Witte, H. (2003). Who feels insecure in Europe? Predicting job insecurity from back-
ground variables. Economic and Industrial Democracy 24(2): 189–215.
Pacek, A.C. & Radcliff, B. (1995). Economic voting and the welfare state: A cross-national analysis. Journal
of Politics 57(1): 44–61.
Rabinowitz, G., Prothro, J.W. & Jacoby, W. (1982). Salience as a factor in the impact of issues on candidate
evaluation. Journal of Politics 44(1): 41–63.
Rehm, P. (2009). Risks and redistribution: An individual-level analysis. Comparative Political Studies 42(7):
855–881.
Rehm, P. (2011a). Social policy by popular demand. World Politics 63(2): 271–299.
Rehm, P. (2011b). Risk inequality and the polarized American electorate. British Journal of Political Science
41(2): 363–387.
Scheve, K. & Slaughter, M.J. (2002). Economic insecurity and the globalization of production. Cambridge,
MA: National Bureau of Economic Research.
Simon, H.A. (1982). Models of bounded rationality. Cambridge, MA: MIT Press.
Singer, M.M. (2011a). Economic voting and welfare programmes: Evidence from the American states.
European Journal of Political Research 50(4): 479–503.
Singer, M.M. (2011b). When do voters actually think ‘It’s the economy’? Evidence from the 2008 presiden-
tial campaign. Electoral Studies 30(4): 621–632.
Singer, M.M. (2011c). Who says ‘It’s the economy’? Cross-national and cross-individual variation in the
salience of economic performance. Comparative Political Studies 44(3): 284–312.

© 2013 The Author(s)


European Journal of Political Research © 2013 European Consortium for Political Research
ECONOMIC VULNERABILITY AND ECONOMIC VOTING IN 14 OECD COUNTRIES 135
Singer, M.M. (2013). What goes around comes around: Perceived vulnerable employment and economic
voting in developing countries. European Journal of Political Research 52(2): 143–163.
Stokes, D.E. (1963). Spatial models of party competition. American Political Science Review 57(2): 368–377.
Van der Brug, W., Van der Eijk, C. & Franklin, M.N. (2007). The economy and the vote: Economic conditions
and elections in 15 countries. Cambridge: Cambridge University Press.
Weatherford, M.S. (1978). Economic conditions and electoral outcomes: Class differences in the political
response to recession. American Journal of Political Science 22(4): 917–938.
Wlezien, C. (2005). On the salience of political issues:The problem with ‘most important problem’. Electoral
Studies 24(4): 555–79.

Address for correspondence: Diego Fossati, Department of Government, Cornell University, 214 White
Hall, Ithaca NY, 14853, USA. E-mail: df275@cornell.edu

© 2013 The Author(s)


European Journal of Political Research © 2013 European Consortium for Political Research
Copyright of European Journal of Political Research is the property of Wiley-Blackwell and
its content may not be copied or emailed to multiple sites or posted to a listserv without the
copyright holder's express written permission. However, users may print, download, or email
articles for individual use.

You might also like