主页 European Journal of Political Research Candidate-centred electoral systems and change in incumbent vote share: A cross-national and...

Candidate-centred electoral systems and change in incumbent vote share: A cross-national and longitudinal analysis

你有多喜欢这本书?
下载文件的质量如何?
下载该书,以评价其质量
下载文件的质量如何?
卷:
55
语言:
english
日志:
European Journal of Political Research
DOI:
10.1111/1475-6765.12132
Date:
May, 2016
文件:
PDF, 577 KB
Conversion to is in progress
Conversion to is failed
0 comments
 

要发布评论,请登录注册
您可以留下评论,分享你的经验。其他读者也会有兴趣了解您对您所读书籍的看法。不管你喜不喜欢这本书,只要您如实、详细地告诉他们,大家就能找到感兴趣的新书。
European Journal of Political Research 55: 321–339, 2016

321

doi: 10.1111/1475-6765.12132

Candidate-centred electoral systems and change in incumbent vote share: A
cross-national and longitudinal analysis
PETER SÖDERLUND
Social Science Research Institute, Åbo Akademi University, Finland

Abstract. This study examines if prime minister’s parties are punished or rewarded by voters to a
lesser extent in candidate-centred electoral systems compared to party-centred systems. Candidate-centred
systems allow the voters greater choice in determining the fate of individual candidates at the district
level and create incentives for candidates to cultivate a personal vote rather than pursuing a party vote.
Voters in these systems are more likely to focus on individual candidates than on parties, thus fostering
individual accountability at the expense of collective (party) accountability. Cross-sectional time-series data
for 23 OECD countries between 1961 and 2014 were analysed. Two indices of intraparty efficiency (the
Farrell–McAllister Index and the Shugart Index) were used to capture the candidate-centredness of electoral
systems. The analysis of aggregate-level data with almost 300 observations showed that incumbent parties
tend to win or lose fewer votes in candidate-centred electoral systems. This effect has become stronger over
time. Candidate-centredness has a weak moderating impact on the state of the economy on the degree of
public sanctioning of government parties.
Keywords: electoral systems; candidate-centredness; intraparty competition; incumbent vote; accountability

Introduction
In order to advance our understanding of voting behaviour, it is necessary to discover how
political context can affect the way in which government performance is evaluated by voters
(Duch & Stevenson 2008: 358; Hellwig 2011: 150). There is much agreement in the literature
to show that clarity of responsibility is an important issue in terms of how voters hold
governments accountable for their actions. When the clarity ; of responsibility is high, voters
are able to identify who is responsible for policy outcomes and performance and thus punish
(or reward) governing parties accordingly. However, scholars are less in agreement on the
relative importance of formal institutional rules and dynamic government characteristics
concerning how strong the link is between performance evaluations and voting choice. A
multitude of long-, medium- and short-term factors, which are thought to either enhance
or undermine the clarity of government responsibility, have been identified and tested (e.g.,
Duch & Stevenson 2008; Hellwig & Samuels 2008; Hobolt et al. 2013; Nadeau et al. 2002;
Whitten & Palmer 1999).
An institutional factor receiving less attention is the candidate-centredness of electoral
systems. At present, there are no systematic cross-national and longitudinal studies available
that examine the extent that incumbent parties either win or lose votes in candidate-centred
systems. What is already known, however, is that by increasing the focus on individual
candidates rather than political parties, electoral institutions foster individual representation
at the expense of collective representation (Colomer 2011; Shugart 2001) and impact on

C 2016 European Consortium for Political Research

322

PETER SÖDERLUND

accountability mechanisms (Carey 2009; Kitschelt 2000). Adding candidate-centredness
into the equation can help improve current understandings of electoral accountability
generally, and incumbent performance at the ballot box specifically. In this study, I will
develop a theoretical model to predict whether incumbents are punished or rewarded to a
lesser degree in candidate-centred electoral systems compared to party-centred systems. The
major argument advanced here is that electoral systems, fostering personal representation
and individual accountability, decrease the ability or willingness of citizens to monitor
and hold parties in government accountable for national policy outcomes. The degree
of performance voting is thus lower in candidate-centred electoral systems compared to
party-centred electoral systems that encourage responsible party government and collective
accountability. An empirical implication is that, at the aggregate level, there should to be less
election-to-election variability in incumbent party support between elections.
This study empirically explores whether the party of the prime minister is punished or
rewarded by voters to a lesser extent in candidate-centred electoral systems. It does this by
analysing cross-sectional time-series data for 23 Organisation for Economic Cooperation
and Development (OECD) countries between 1961 and 2014. The study goes beyond
the classic distinction between the three main types of electoral systems – majoritarian,
proportional representation and mixed systems – by stressing the intraparty dimension of
competition measured along a multi-point scale. Various electoral system components shape
the incentives for individual candidates at the district level in order to cultivate a personal
vote rather than to pursue a party vote. Candidate-centred systems (as opposed to partycentred ones) allow voters to have a greater choice in determining the fate of individual
candidates at the district level and create incentives for candidates to cultivate a personal
vote rather than to pursue a party vote (Carey & Shugart 1995; Farrell & McAllister 2006;
Shugart 2001).
Three hypotheses are proposed and evaluated. First, the analysis will show that the
shifts in incumbent parties’ electoral support are smaller in candidate-centred electoral
systems. Second, vote shifting has increased in magnitude in party-centred systems relative
to candidate-centred systems over time. Third, the conditional effect on economic voting is
weak, implying that incumbent parties are less punished in candidate-centred systems and
more so in party-centred systems in economically bad times.

Theoretical model
‘Electoral accountability’ refers to the mechanism whereby citizens hold elected officials
accountable for their actions via elections. A central concern is how easy it is for voters
to identify which parties are responsible for which decisions and policy outcomes. In the
clarity of responsibility and economic voting literature, the underlying intuition is that the
link between economic performance and electoral outcomes is conditioned by political and
institutional factors (Powell & Whitten 1993; Whitten & Palmer 1999; cf. Duch & Stevenson
2008: 25–26). In this study, the candidate-centredness of an electoral system is proposed
as a complementary factor affecting clarity of responsibility. The propensity to discern
which parties are collectively responsible for performance should be lower in candidatecentred systems. If electoral institutions provide incentives for individual candidates to
pursue personal votes, voters are more likely to focus on the candidates’ characteristics and

C 2016 European Consortium for Political Research

CANDIDATE-CENTRED ELECTORAL SYSTEMS

323

Figure 1. Theoretical model.

qualities rather than on collective party performance. My theoretical generalisation relates
to a broader category of performance voting involving evaluations of both economic and
non-economic policy outcomes (see Hellwig 2011; Hobolt et al. 2013).
Figure 1 summarises the theoretical model with linkages from electoral institutions to
electoral accountability in terms of punishing or rewarding incumbent parties: (1) Electoral
institutions offer incentives and constraints that shape the strategies candidates use to
cultivate personal reputations rather than to stress party reputation; (2) electoral institutions
and (3) candidates themselves affect the extent to which voters evaluate the qualities of
individual candidates rather than parties; (4) voters may hold either individual candidates
or parties primarily to account as they cast their votes; and (5) the extent to which incumbent
parties are ultimately held accountable in elections for performance varies accordingly.
Candidate-centred electoral systems
I begin by presenting how the type of electoral system contributes to a greater focus on
candidates at the district level. A major concern in the vast amount of literature focusing
on personal voting has been on how electoral rules shape the incentives for individual
candidates at the district level to cultivate a personal vote rather than to pursue a party vote.
A variety of electoral system components – ballot access, ballot structure, ballot type and
district magnitude – help to push electoral systems either in a candidate-centred direction
or in a party-centred direction. ‘Personal vote-seeking’ refers to candidates distinguishing
themselves from other candidates by investing in personal campaigns and increasing their
visibility in their electoral districts. If the electoral fate of candidates is tied to the number
of personal votes they receive, the incentives for candidates to cultivate a personal vote
will be strong. This is particularly the case if co-partisans must compete against each
other for votes and seats in multi-member districts. Intraparty competition is thus seen
as a key mechanism for personal vote-seeking incentives. Whereas a party label will help
candidates to differentiate themselves from candidates of other parties, self-promotion
based on personal qualities will differentiate them from fellow party candidates. In partycentred systems, on the other hand, the candidates are ranked in a fixed order on the party
ballot and the voters are unable to disturb this ranking. The prototype of a party-centred
system is closed-list proportional representation (PR) in which the voters cast party (or list)
votes only (Carey & Shugart 1995; Farrell & McAllister 2006; Shugart 2001).
Elections are truly candidate-centred if there is a supply of district-level candidates
actively developing personal reputations in order to attract personal votes. There must also

C 2016 European Consortium for Political Research

324

PETER SÖDERLUND

be a demand for candidate information among voters casting personal votes on the basis of
candidate evaluations. Hence, if voters must or may cast candidate ballots, they are more
likely to consider candidate traits and qualities when forming voting preferences. These
factors of supply and demand are mutually reinforced, as evident from Figure 1. If candidates
invest in personal campaigns to promote themselves at the district level, voters are likely to
respond by focusing on the candidates’ personal qualities. And vice versa, if voters focus on
candidate qualities, candidates respond by communicating information about themselves to
the electorate (Shugart et al. 2005: 439–441). In the literature on personal voting, the concept
of a ‘personal vote’ refers to ‘that portion of a candidate’s electoral support which originates
in his or her personal qualities, qualifications, activities, and record’. The part of the vote that
is not personal is ‘support based on shared partisan affiliations, fixed voter characteristics
such as class, religion and ethnicity, reactions to national conditions such as the state of
the economy and performance evaluations centred on the head of the governing party’
(Cain et al. 1984: 111).
The next step in the theoretical model deals with how the electoral system cultivates
individual accountability. First, the electoral system affects the balance between party
representation and personal representation (Colomer 2011). At one end of the extremes,
representation is centred on political parties (collective-goods programmes or broad policy
platforms), while at the other end, representation is centred on individual legislators
(personal ties to constituents) (Shugart 2001). In turn, party-dominated representation
encourages collective accountability, while personal representation fosters individual
accountability. ‘Collective accountability’ means that the party and party leadership are
directly accountable to the voters. As teams of legislators act collectively to promote
policy agendas, the electorate may discern which parties are responsible for performance
and reward or punish parties collectively. ‘Individual accountability’ implies that the link
between citizens and individual legislators is more direct, both through the ability of voters
to vote for individual candidates and the incentives of legislators to develop personal ties
to their constituents by providing particularistic goods and services. Voters are thus more
inclined to reward or punish individual legislators in elections (Carey 2009: 16–17; Kitschelt
2000: 859; Weßels 2007: 836).
Electoral institutions which stimulate the personalisation of electoral choice at the
district level arguably should lead to systematic differences in the calculus of voting between
party- and candidate-centred systems. The various criteria voters use to evaluate political
choices should affect the voter’s acquisition of information. If candidates stress their
personal attributes rather than party performance, and voters focus on the characteristics of
individual candidates and legislators, there will be less information about the competence
of incumbent parties in policy making. Alternatively, there will be greater voter ignorance
as to who is responsible for policy outcomes. Disproportionate attention on individual
representatives means that there will be less focus on national events and collective interests.
Voters are thus less able or induced to hold the ruling parties in government accountable.
This is in tune with the clarity of responsibility argument stating that voters have less
ability to make informed choices, and attribute responsibility to any party, if responsibility
is dispersed among various actors (Powell & Whitten 1993; see also Duch & Stevenson
2008: 227–228). Although candidate focus is considered as the primary mechanism in my
theoretical model, it should be mentioned that Powell and Whitten (1993) argue that weak

C 2016 European Consortium for Political Research

CANDIDATE-CENTRED ELECTORAL SYSTEMS

325

party cohesion diffuses the connection between voting choice and government performance.
In this context, ‘weak party cohesion’ refers to candidates and factions being less dependent
on their national parties and relying more on geographic or clientelist platforms for electoral
success. If legislators choose to deviate from the party line in order to act on behalf of
their constituency, the importance of the party as a voting cue will be lower. In a similar
vein, Huber et al. (2005) assume that weak party cohesion impacts on the amount of
information available about party activities which, in turn, affects clarity of responsibility.
First, internal discipline is lower in national parties if the electoral system provides incentives
for representatives to ‘adopt divergent positions and pursue various goals’ (Huber et al.
2005: 370). Second, this lack of party cohesion in candidate-centred systems will lead to less
clarity of responsibility because citizens receive ‘noisy signals about the actions and positions
of the parties’ and thus make it ‘more difficult for voters to make retrospective judgments
about parties’ performance’ (Huber et al. 2005: 370).
Hypotheses
To sum up, the first hypothesis is that voting decisions based on a party’s past performance
are less prevalent in candidate-centred electoral systems. Candidate-centred electoral
systems encourage individual politicians to advance their own personal reputations instead
of promoting party reputation and collective goals. Furthermore, the willingness and ability
of citizens to monitor how well the government parties are doing their jobs will be lower.
Voters are instead more likely to make voting decisions based on the performance of
individual representatives. This implies that the magnitude of the incumbent vote swing –
both negative and positive – between two successive elections should be lower. The formal
hypothesis is as follows:
H1: Shifts in the electoral support of incumbent parties are smaller in candidatecentred electoral systems, all other things being equal.
The second hypothesis states that the tempering impact of candidate-centred electoral
systems on the incumbent vote has become stronger over time. Plausible suggestions for
such an effect include partisan dealignment and the personalisation of politics. With the
weakening of long-term loyalties to parties over recent decades (Dalton et al. 2000), valence
issues and government performance have become more important determinants of voting
choice (Clarke et al. 2004). A parallel process has been the personalisation of politics,
which, broadly, refers to placing greater focus on individual politicians at the expense of
collective actors (Karvonen 2010). Allegedly, this development is particularly evident in
candidate-centred systems. The disposition of voters to hold governments accountable for
their actions may have been similar across countries in the 1960s, but with dealignment and
the personalisation of politics, it can be assumed that the clarity of responsibility has declined
particularly in candidate-centred systems over time. To test this hypothesis, I investigate
whether the candidate-centredness effect on the incumbent vote has significantly changed
in recent decades.
H2: The tempering effect of candidate-centredness on the incumbent vote change has
become more evident over time, all other things being equal.

C 2016 European Consortium for Political Research

326

PETER SÖDERLUND

The third hypothesis asserts that electoral systems condition the effect of economic
performance on incumbent support. The magnitude of the economic vote should be lower
in candidate-centred electoral systems. This aligns with the basic argument in the clarity
of responsibility and economic voting literature: voters are less likely to punish/reward
the government for poor/good economic performance if clarity of responsibility is low
(Anderson 2000; Hellwig & Samuels 2008; Nadeau et al. 2002; Powell & Whitten 1993;
Whitten & Palmer 1999). In economic voting studies, the dependent variable usually ranges
from negative (vote loss) to positive (vote gain) values. Thus the relationship is assumed
to be linear since economic downturn should lead to reduced government support and
economic growth to increased government support. In this study, assessing the economic
effects is more challenging since the dependent variable measures variability in the electionto-election change in the vote share of the prime minister’s party. A higher score indicates
that a party has either gained or lost a greater amount of votes. Therefore variability (or
volatility) is likely to increase in response to both positive and negative changes in economic
performance. Empirically there should be a curvilinear relationship between economic
performance and variability in the vote share, but the relationship is not necessarily perfectly
U-shaped if voter responses to economic performance are asymmetric. In that case, voters
punish incumbents disproportionately harshly for bad economic times, but fail to reward
them for good times (see Dassonneville & Lewis-Beck 2014; Duch & Stevenson 2008: 107–
108). The basic assumption is, however, that incumbent parties in candidate-centred systems
are not punished or awarded as much due to economic performance – a dampening effect
that becomes more evident with greater economic decline or growth. The hypothesis is:
H3: Candidate-centred electoral systems dampen the effect of economic performance
on election-to-election variability in incumbent support.

Empirical design
Data
The data covers approximately 300 parliamentary (lower house) elections in 23 countries
from 1961 to 2014. Included are countries traditionally identified as advanced industrial
democracies: parliamentary democracies in Western Europe as well as Australia, Canada,
Japan and New Zealand. A longer time span allows for a greater number of elections, while
the number of countries is limited since more recent democracies in Central and Eastern
Europe are excluded. Democracies that were established in the mid-1970s (Spain, Portugal,
Greece and Cyprus) are included.
Dependent variable
The extent to which citizens punish or reward parties in government is measured by
modelling changes in the aggregate-level vote shares. In this study, I focus on the change
in the chief executive vote, which is often used in empirical studies (e.g., Anderson 2000;
Chappell & Veiga 2000; Hellwig & Samuels 2008). First, the prime minister’s party is likely
to be held most accountable, having the primary administrative responsibility. Second, I will

C 2016 European Consortium for Political Research

CANDIDATE-CENTRED ELECTORAL SYSTEMS

327

avoid potential problems of aggregation in coalition government cases. Summing the vote
shares for all government parties becomes a problem if the extent of performance voting
is lower for parties with less administrative responsibility (Duch & Stevenson 2008: 57–59).
However, the dependent variable is a non-directional measure: |Vit | is the absolute value of
the election-to-election change in country i at time t, whereby negative values are converted
into positive values.1 Hence, both negative and positive swings in incumbent vote support
are expected to be smaller in candidate-centred systems. Only the incumbent chief executive
parties that were in the cabinet immediately before the elections are included in the sample.
Furthermore, the parties must have been in office for at least 12 months to assure that they
could be held responsible. Incumbent parties that held office for less than a year are purged
from the sample.
Candidate-centredness of electoral systems
The independent variable of central interest is the candidate-centredness of the electoral
system. The aim here is to quantify the extent to which electoral systems provide personal
vote-seeking incentives at the district level. To achieve more reliable results, two indices of
intraparty efficiency are tested: Farrell and McAllister’s (2006) and Shugart’s (2001) indices.
Both build on the work by Carey and Shugart (1995), who provided a rank-ordering of
electoral formulas according to how they foster personal vote-seeking rather than party
reputation-seeking. Intraparty efficiency is specified as a function of three components:
Ballot that captures the extent to which the party leadership exercises control over ballot
access and the final rank of candidates on the ballot; Vote that concerns whether list votes or
nominal (candidate) votes are cast and whether the nominal votes may pool or transfer to
other candidates; and District that combines the type of districts (single-seat districts versus
multi-seat districts) and the type of votes cast (list versus nominal votes).
The Farrell–McAllister index is a refined version of Shugart’s index. The major
differences are whether single non-transferable vote systems (e.g., Japan before 1993) or
single transferable vote systems (e.g., Ireland and Malta) should be considered the most
candidate-centred systems, and whether the alternative vote system (Australia) should be
near the extreme or near the middle of the scale. These differences are apparent from
the coding of the ‘vote’ component (see Table 1A in the Online Appendix). Farrell and
McAllister have systems with nominal and non-transferable votes in the middle category
and systems with nominal votes, which may pool or transfer to other candidates. For Shugart
it is the other way around.
For Farrell and McAllister, the component scores are not actual numerical values,
meaning that they are not summed up to form an interval index. Instead, the component
scores and the index are ordinal in nature. In other words, electoral systems are logically
categorised according to their party- and personal-reputation incentives before being
assigned ordinal values based on the rank positions. I have constructed a similar index which
varies from 1 (most party-centred) to 10 (most candidate-centred) (see Table 1). Shugart,
on the other hand, codes Ballot and Vote from 2 to –2 and District from 1 to –1. Negative
values indicate candidate-centredness, while positive values indicate party-centredness. On
the basis of the values of these three measures, Shugart constructed an index ranging from
1 to –1, where the latter value indicates the highest possible degree of candidate-centredness.

C 2016 European Consortium for Political Research

328

PETER SÖDERLUND

Table 1. Scoring of electoral systems on the intraparty dimension
Electoral system and country
Single transferable vote

Farrell–McAllister Index

Shugart Index

10

7.85

10

7.85

9

5.70

Japan (pre-1993)
Quasi-list proportional representation

8

10.00

Finland1
Open list proportional representation
Cyprus (post-1979)

7

7.15

Ireland
Malta
Alternative vote
Australia
Single non-transferable vote

6

8.55

Denmark
Greece
Italy (pre-1993)

6
6
6

4.00
8.55
8.55

Switzerland
Luxembourg

6
6

8.55
8.55

5

6.45

4
4
4

5.70
5.70
5.70

3
3

4.50
4.50

3
3

5.00
4.50

Austria
Belgium
Iceland (post-2000)

2
2
2

3.00
3.00
3.00

Netherlands
Sweden (post-1996)

2
2

3.00
3.00

1
1

1.00
1.00

Italy (post-2004)
Portugal
Spain

1
1
1

1.00
1.00
1.00

Iceland (pre-2000)2

1
1
1

1.00
1.00
1.00

Single-seat districts, two rounds
France
Single-seat districts, plurality with party control
Canada
New Zealand (pre-1993)
United Kingdom
Mixed-member system with plurality rule
Germany
Italy (1993–2004)
Japan (post-1993)
New Zealand (post-1993)
Ordered list proportional representation

Closed list proportional representation
France (1986)
Greece (1985)

Norway2
Sweden (pre-1996)2

Notes: 1 Preferential voting for a single candidate is mandatory in Finland and therefore classified as a quasi-list
system (see Farrell & Scully 2007: 130; Shugart 2001: 185). 2 Iceland (pre-2000), Norway and Sweden (pre-1996) are
closed list systems due to high thresholds for voters to alter a party’s ordering of candidates.

C 2016 European Consortium for Political Research

CANDIDATE-CENTRED ELECTORAL SYSTEMS

329

For the purposes of this study, however, the index was recoded on a 1–10 scale so that
1 corresponds to maximal party-centredness and 10 to maximal candidate-centredness (see
also Table 1). Despite the differences, the Farrell–McAllister and Shugart indices are highly
correlated (rp = 0.861, p < 0.01, N = 32).
Control variables
A number of control variables were created to account for contextual, economic and
temporal impacts on the incumbent vote. Electoral formula captures the three major
families of electoral systems: majoritarian, proportional representation and mixed systems.
Majoritarian systems (few parties in government) are assumed to reveal high levels of clarity,
while proportional systems (often coalition governments) reveal low levels of clarity (with
mixed systems in between) (De Vries et al. 2011). Type of government combines the number
of government parties and parliamentary support: (1) majority single-party government;
(2) majority coalition government; (3) minority coalition government; (4) and minority
single-party government. The first category is used as a reference category. A majority
single-party government should reveal the greatest clarity of responsibility, while majority
and minority coalitions reveal intermediate degrees of responsibility, and minority singleparty governments reveal the least degree of responsibility (Powell 2000). An additional
institutional effect controls for an Index of bicameralism going from unicameralism (1) to
strong bicameralism (4) (Lijphart 2012). The ability to assign responsibility for economic
outcomes is assumed to be lower if the upper house is not controlled by the government.
The index used here is not a close approximation of the one employed by Powell and Whitten
(1993), and Whitten and Palmer (1999). The data for strength of bicameralism are from
Armingeon et al. (2015). Effective number of parties has been previously included to capture
available alternative governments. A greater number of parties ‘should lead citizens to more
readily express content or discontent with the ruling party or parties’ (Anderson 2000: 156).
Measures of Macroeconomic performance are used to assess the extent to which
incumbent governments are held accountable for economic conditions. Two macroeconomic
indicators central in the extant economic voting literature are used: GDP growth and
unemployment change (World Bank 2015). The squared terms for both variables are added
to capture curvilinear effects on the dependent variable (which is set to the absolute value
of the difference). It will also be possible to detect if the economic vote is asymmetric – that
is, voters allocate greater responsibility to politicians for poor economic performance than
for good economic performance (Dassonneville & Lewis-Beck 2014; Duch & Stevenson
2008: 107–108). The empirical measures for economic performance are based on changes
in economic conditions during the year prior to the election. Since I rely on annual data, as
opposed to quarterly data, each macroeconomic indicator was weighted based on the month
the election was held.2 Trade openness is the ratio of total trade (exports and imports) to
GDP and controls for the impact of globalisation on economic voting between time and
over time (World Bank 2015). The natural logarithm is used to scale down very high values
for some countries. Voters in more open economies, with higher exposure to global trade
influences, are expected to hold incumbents less accountable for economic outcomes. Voters
take exogenous shocks to the economy into account when they form expectations about the
competence of an incumbent (Duch & Stevenson 2008: 180–184; Hellwig & Samuels 2007).3

C 2016 European Consortium for Political Research

330

PETER SÖDERLUND

Time since last election is the number of days between the current and previous elections.
First, the variable controls for the probability that vote switching increases with the passage
of time since personal and contextual conditions are more likely to change over time
(Bischoff 2013). Second, it accounts for the ‘length of time since the electorate last had
an opportunity to hold the governing parties accountable for policy decisions’ (Palmer
& Whitten 2002: 76). A log transformed (ln) variable is used since the effect is expected
to diminish over time (see Bélanger & Gélineau 2010: 87). Years measures the number
of years elapsed between an election and 1960 (divided by ten to make the numbers
more manageable). The variable is used in an interaction term, together with the level of
candidate-centredness. Finally, I control for the Previous vote share of the current prime
minister’s party because incumbent parties with a stronger support in the previous election
are more likely to lose votes in the current election (Palmer & Whitten 2002).
Method
In the data, each case is an incumbent prime minister’s party in country i at time t. The
resulting dataset is a pool of repeated observations across countries. The choice of a
particular statistical method is of great importance since it may influence the reliability of the
empirical findings. However, the analysis of pooled data is plagued by problematic regression
errors such as serial correlation errors, panel heteroskedasticity, contemporaneous
correlation errors, and simultaneous autocorrelation and heteroscedasticity errors (Plümper
et al. 2005: 329).
Linear cross-sectional time-series models will be fitted to explain why the electoral
performance of incumbent governments varies across time and across countries. Although
various methods for analysing such clustered data are available, I will employ panelcorrected standard errors to deal with the existence of panel heteroskedasticity (Beck &
Katz 1995). The Breusch–Pagan test shows the presence of heteroskedasticity (i.e., different
variances in error terms across countries). Including a lagged dependent variable on the
right-hand of the equation turns the model into a dynamic panel model. Simultaneously,
such a procedure will deal with any potential problem of serial correlation in the error term
found in pooled time-series data. A Prais–Winsten estimator is also chosen as a remedy
for serial correlation in the error terms. It would also be possible to include country fixed
effects in all specifications (a dummy variable for each spatial unit) to remove potential bias
related to unobserved heterogeneity across clusters (Beck & Katz 1996). Unfortunately, such
a procedure would absorb cross-sectional variance (Plümper et al. 2005). No fixed effects are
specified in this study’s models since it makes little sense to focus on within-cluster effects;
candidate-centredness is in fact a time-constant variable in many countries.

Results
A total of 294 observations from 23 countries between 1961 and 2014 were analysed.4 The
dependent variable – variability in prime minister parties’ vote shares across elections –
had an average value of 4.19 percentage points (min = 0, max = 19.04, SD = 3.71).5 Three
hypotheses formulated for the effect of candidate-centredness on the incumbent vote share
were tested in a series of regression models and are presented below. The reported R2 values

C 2016 European Consortium for Political Research

331

CANDIDATE-CENTRED ELECTORAL SYSTEMS

Table 2. Explaining variability in prime minister party’s vote share
Interaction models
(Index × Time)

Main effects models
(1)
Farrell–McAllister Index
Shugart Index
Majoritarian system
Mixed system

–0.17** (0.06)
–
1.84* (0.83)

(2)

(3)

–

0.07 (0.12)

–

–

0.12 (0.14)

1.65 (0.85)

1.72* (0.85)

–0.16* 0.07)
1.90* (0.83)

(4)

–1.02 (0.84)

–0.68 (0.81)

–1.39 (0.88)

–0.88 (0.84)

Minority coalition

0.13 (0.82)

–0.11 (0.83)

0.05 (0.83)

–0.31 (0.85)

Minority single-party government

0.53 (0.69)

0.47 (0.72)

0.49 (0.70)

0.44 (0.71)

Majority coalition

0.37 (0.62)

0.25 (0.62)

0.35 (0.62)

0.25 (0.61)

Effective number of parties

0.51* (0.21)

0.55** (0.21)

0.43* (0.21)

0.45* (0.22)

Trade openness (ln)

–0.12 (0.45)

–0.15 (0.45)

–0.26 (0.47)

–0.23 (0.49)

Index of bicameralism

–0.31 (0.23)

–0.36 (0.24)

–0.26 (0.24)

–0.34 (0.24)

GDP growth

–0.69** (0.16)

–0.71** (0.16)

–0.55** (0.16)

–0.58** (0.15)

GDP growth2

0.04* (0.02)

0.04* (0.02)

0.03* (0.02)

0.04* (0.02)

Unemployment change

–0.25 (0.34)

–0.21 (0.34)

–0.11 (0.33)

Unemployment change2

0.14 (0.20)

0.11 (0.20)

0.11 (0.20)

0.08 (0.19)

Time since last election (ln)

0.73 (0.71)

0.97 (0.73)

0.61 (0.71)

0.75 (0.74)

Previous vote

0.10** (0.03)

0.10** (0.03)

0.10** (0.03)

0.09** (0.03)

Years/10

–

–

0.69

**

(0.26)

–0.08 (0.34)

0.78** (0.28)

Farrell–McAllister Index × Years/10

–

–

–0.09 (0.04)

–

Shugart Index × Years/10

–

–

–

–0.10* (0.05)

Constant

–3.93 (5.49)

*

–5.43 (5.61)

–4.40 (5.44)

–5.30 (5.54)

Observations

294

294

294

Number of countries

23

23

23

23

0.16

0.16

0.18

0.18

60.55**

57.28**

79.65**

78.19**

R2
Wald chi2

Notes: Prais-Winsten estimation results with panel-corrected standard errors in parenthesis.
*
p < 0.05.

294

**

p < 0.01;

can be compared with those from cross-sectional models, but not with time-series models.
The total explanatory power of the models is modest (below 20 per cent), but in line with
previous cross-sectional models (e.g., Chappell & Veiga 2000).
The estimates in Table 2 provide support for the hypothesis that election-to-election
variability in prime minister’s party support is smaller in candidate-centred systems.
Model 1 estimates the effect of the Farrell–McAllister Index and model 2 the effect of
the Shugart Index. The reported coefficients can be interpreted as the amount of change
in the outcome variable for each one-unit increase in the index. A negative estimate
indicates that incumbent chief executive parties lost or gained votes in candidate-centred
systems to a lesser extent than in party-centred systems. Both indices are expressed on a
1–10 scale, whereby the most party-centred system scores 1 and the most candidate-centred

C 2016 European Consortium for Political Research

332

PETER SÖDERLUND

Figure 2. Plots of election-to-election variability (prime ministerial party) according to candidatecentredness (95 per cent confidence intervals).

system scores 10. Similar linear effects of the Farrell–McAllister and Shugart indices are
observed, although the indices differ in terms of which systems are deemed the most
candidate-centred. For each one-unit increase in the index, variability decreased by 0.17
(Farrell–McAllister) or 0.16 (Shugart) points. The substantial effects are displayed in
Figure 2, which reports average predictive margins calculated at fixed values for candidatecentredness. Variability decreases by about 1.5 percentage points, going from the most partycentred to the most candidate-centred electoral systems. Additional tests for curvilinearity
(not reported in any table) are negative and thus show that the assumption of linearity is
correct. Thus prime minister’s parties in closed list and ordered list PR systems generally

C 2016 European Consortium for Political Research

CANDIDATE-CENTRED ELECTORAL SYSTEMS

333

experience the greatest shifts in their vote shares, while incumbents in PR systems with
open lists or the single transferable vote are least likely to experience any shifts in vote
shares. It also makes sense to place systems with single-member districts in the middle of the
scale.
The effects of the control variables on electoral volatility warrant closer inspection.
Fluctuations are greater in Majoritarian systems, all else being equal. These systems
tend to produce single-party governments (greater clarity of responsibility) as well as
disproportionate vote-to-seat ratios (greater vote swings). Type of government – whether
minority or majority, single-party or coalition – has no evident effect, however. When
controlling for other factors, the effective number of parties variable, which proxies the
clarity of available alternatives to the incumbent, has, as expected, a positive effect. I should
emphasise that candidate-centredness predicts lower variability in the prime minister’s
party support, even if Type of electoral system, Number of effective parties and Type of
government variables are dropped. Furthermore, there is evidence for the economic voting
thesis since GDP growth is significantly related to incumbent vote swings. The effect is
asymmetric, however, as the squared term is close to zero. Thus voters reward incumbents
less for good events than they punish them for bad events. The positive coefficient on the
Previous vote variable reflects the fact that prime minister parties with stronger support
in the previous election tend to lose more votes. The remaining control variables (Trade
openness, Bicameralism and Time since last election) have expected signs, but they are not
statistically significant.
The second hypothesis stated that the tempering effect of candidate-centredness on
the incumbent vote has become more evident over time. Interaction variables between
the candidate-centredness indices and number of years since 1960 are therefore tested. A
negative interaction term would indicate that the positive effect of candidate-centredness on
the variability in a prime minister’s party support across elections has increased over time. In
Table 2, the interaction terms in both models 3 and 4 are negative and statistically significant
at the 0.10 level. Figure 3 reports predictive margins (with 95 per cent confidence intervals)
when setting the candidate-centredness indices to 1 (party-centred) and 10 (candidatecentred). The slope for party-centred systems is positive, whereby election-to-election
variability has increased over time. The slope is negative when candidate-centredness is set
to its maximum value. Fixing the value for candidate-centredness at lower values naturally
flattens the slope, but the conclusion that variability in incumbent support has increased,
continues to hold. All in all, the second hypothesis is supported.
The third hypothesis states that the electoral fortunes of prime minister’s parties
are less strongly tied to the national economy in candidate-centred electoral systems.
Interaction models are used to see how the strength of the relationship between economic
performance (x) and variability in a prime minister’s party support (y) is changed by
candidate-centredness (z). The two economic variables are introduced in separate models
to decrease the number of interactions and prevent over-specification. The evidence for the
third hypothesis is weak at best. On the one hand, the moderation effects are statistically
insignificant in the linear interaction models (x × z) and curvilinear interaction models
(x2 × z) (see Tables 2A and 3A and Figure 1A in the Online Appendix). On the other hand,
there are indications that the relationship between economic performance and variability
in a prime minister’s party support differs according to the level of candidate-centredness.

C 2016 European Consortium for Political Research

334

PETER SÖDERLUND

Figure 3. Plots of election-to-election variability (prime ministerial party) over time in party- and candidatecentred systems (95 per cent confidence intervals).

For ease of interpretation, the results are presented as predictive margins with 95 per
cent confidence intervals. First, the plots are based on the linear interaction models since
the curvilinear interaction models produce similar results. As voters appear to respond
more strongly to bad economic times (asymmetric economic voting), the relationship is an
upward-sloping curve rather than a U-shaped curve. Second, only the interaction effect plots
for the Farrell–McAllister index are presented for space considerations. Third, the party- and
candidate-centredness scores are set to their extremes.

C 2016 European Consortium for Political Research

CANDIDATE-CENTRED ELECTORAL SYSTEMS

335

Figure 4. Plots of election-to-election variability (prime ministerial party) as a function of economic
performance in party- and candidate-centred systems (95 per cent confidence intervals).

As evident from Figure 4, variability in a prime minister’s party support is low in all
types of electoral systems in economically good times. The slopes of the lines are different
from each other, however, indicating that candidate-centred electoral systems dampen
variability in a prime minister’s party vote shares. The ‘variability gap’ between candidateand party-centred electoral systems widens with lower or negative GDP growth and higher
unemployment. The prime ministers’ parties in candidate-centred electoral systems thus
appear to be less affected by (poor) economic performance. These results should, however,

C 2016 European Consortium for Political Research

336

PETER SÖDERLUND

be interpreted with caution since confidence intervals overlap across most of the range
for each economic performance variable even when the party- and candidate-centredness
scores are set to their extremes.6

Conclusions and implications
This article has argued that the ability or willingness to assign credit or blame to governments
is lower in candidate-centred electoral systems. The empirical analyses demonstrates that
prime minister’s parties in 23 Western parliamentary democracies have lost or gained fewer
vote shares from one election to another in electoral systems that allow voters to cast ballots
for candidates – intraparty preference votes in particular. Fewer significant shifts imply
that chief executive parties are systematically less punished or rewarded. There is also a
discernible trend showing that while vote shifts have remained at similar levels in candidatecentred systems since the 1960s, prime minister’s parties in party-centred systems have
progressively come to experience greater election-to-election variability in the vote shares.
There are also indications that the prime minister’s parties are less affected by economic
downturns in candidate-centred electoral system. Statistically the differences were not large
enough to draw any firm conclusions.
The empirical findings carry important implications for theories of clarity of responsibility. The general argument is that the political context can obscure the clarity of
responsibility by discouraging, or making it difficult, for voters to identify who is responsible
for political and economic outcomes. Many factors have been suggested that contribute
to lower clarity of responsibility, including institutional arrangements that guarantee
opposition participation in policy making (Powell & Whitten 1993) and the cohesiveness
of the incumbent government (Hobolt et al. 2013). Candidate-centredness should be
seen as an additional, or complementary, contextual factor that that diffuses political
responsibility. This is logical considering that personal representation fosters personal
accountability. First, there will be lesser collective responsibility through the agency
of political parties if individual politicians separate themselves from party actions by
cultivating their personal reputations to attract votes in their electoral district. Second, the
linkage between governmental performance and vote weakens if voters are more inclined
to focus on the personal qualities and characteristics of individual candidates rather than
on collective party performance (Carey 2009; Colomer 2011; Kitschelt 2000). The evidence
of an increasing trend is in line with the personalisation of politics thesis that states that
candidate evaluations have come to exert greater influence on voting choice over time (see
Karvonen 2010).
The evidence for the claim that candidate-centred electoral systems condition the effect
of economic performance on incumbent support was weak. Although there were indications
of dampened variability in prime ministers’ party support in candidate-centred systems
in times of economic decline, the statistical support was not robust. This is discouraging
since a critical test of the clarity of responsibility argument is that the strength of the
relationship between economic performance and change in the chief executive’s support
is weaker in certain contexts (Powell & Whitten 1993; Whitten & Palmer 1999). Aggregate
evidence (or lack thereof) for the clarity of responsibility argument has been questioned
(Chappell & Veiga 2000; Duch & Stevenson 2008). A limitation of this study was the use

C 2016 European Consortium for Political Research

CANDIDATE-CENTRED ELECTORAL SYSTEMS

337

of aggregate-level data from a restricted number of countries. To ascertain whether or not
economic voting is conditioned by the context would ideally require a greater number
of country observations. Future studies could also use cross-national individual-level data
from international comparative surveys, making it possible to examine both individual-level
predictors and macrolevel indicators.

Acknowledgments
I would like to thank the three anonymous reviewers for very constructive comments. A
previous version of this article was presented at the ECPR Joint Sessions in Warsaw 2015,
where I received helpful comments from the participants of the workshop ‘Accountability
without Parties?’ chaired by Ruth Dassonneville and André Blais. This work is part of the
Academy Research Fellow project ‘Personalization of Electoral Competition’ financially
supported by the Academy of Finland, 2014–2019 (project 275854).

Supporting Information
Additional Supporting Information may be found in the online version of this article at the
publisher’s web-site:
Table A1. Coding of electoral systems on the intra-party dimension
Table A2. Explaining variability in prime minister’s party vote share by GDP growth and
candidatecentredness
Table A3. Explaining variability in prime minister’s party vote share by unemployment
change and candidate-centredness
Table A4. Directional models: explaining change in prime minister’s party vote share by
GDP growth and candidate-centredness.

Notes
1. Election results and information about governments are from the Parliament and Government
Composition Database (ParlGov) (Döring & Manow 2015). In case of party splits, party mergers and
electoral alliances, the combined vote of the parties in pairs of elections were compared. If this was
not possible, the number of parliamentary party seats was used to proxy vote shares in preceding or
subsequent elections. Incumbent parties that did not contest two consecutive elections were excluded.
2. The formula is ρ = [ρ∗(t−1) (12 − σ(t) )/12] + [ρ∗(t) (σ(t) /12)], where ρ is the annual economic indicator, σ the
month of the election and t the election year (Bélanger & Gélineau 2010: 98, Note 2).
3. Kayser and Peress (2012) have an alternative explanation which proposes that the more open an economy,
the smaller the national deviations from an international benchmark. Hence, voters (with the help of
information in the media) benchmark national economic performance against performance in other
countries. They punish or reward incumbents depending on whether the national economy underperforms or out-performs compared to an international average.
4. Six prime minister parties that either gained or lost more than 20 percentage points were detected and
removed as outliers. They have high residuals (studentised residuals > 3) and high influence (measured
using Cook’s D) in ordinary least squares models. In addition, the following cases were dropped due
to missing GDP growth values at year t–1: Belgium 1961, Cyprus 1976, Germany 1961, Ireland 1961,
Malta 1971 and Norway 1961. The first elections in Spain (in 1977) and Portugal (in 1975 and 1976) after

C 2016 European Consortium for Political Research

338

PETER SÖDERLUND

democratisation were not included. Malta (in 1976) had a very high GDP growth rate (15 per cent) and
was deemed as an outlier.
5. Incumbent prime minister parties lost, on average, 2.30 percentage points (min = –19.04, max = 15.40,
SD = 5.11).
6. Directional models were also tested where the dependent variable was vote swing or change in the
percentage of votes received by the prime minister party (going from negative to positive values). These
estimates generally support the results presented above. First, the coefficients for the interactions between
the candidate-centredness indices and GDP growth are close to zero. Second, the interaction coefficients
are substantially larger when unemployment change is included, showing that prime minister parties
in party-centred electoral systems are punished much more as unemployment goes up. The results are
presented in Tables 4A and 5A in the Online Appendix.

References
Anderson, C.J. (2000). Economic voting and political context: A comparative perspective. Electoral Studies
19(2–3): 151–170.
Armingeon, K. et al. (2015). Comparative political data set, 1960–2013. Bern: Institute of Political Science,
University of Berne. Available online at: www.cpds-data.org/
Beck, N. & Katz, J.N. (1995). What to do (and not to do) with time-series cross-section data in comparative
politics. American Political Science Review 89(3): 634–647.
Beck, N. & Katz, J.N. (1996). Nuisance vs substance: Specifying and estimating time-series cross-section
models. Political Analysis 6(1): 1–36.
Bélanger, È. & Gélineau, F. (2010). Does perceived competence matter? Political parties and economic
voting in Canadian federal elections. Journal of Elections, Public Opinion and Parties 20(1): 83–101.
Bischoff, C.S. (2013). Electorally unstable by supply or demand? An examination of the causes of electoral
volatility in advanced industrial democracies. Public Choice 156(3–4): 537–561.
Cain, B.E., Ferejohn, J.A. & Fiorina, M.P. (1984). The constituency service basis of the personal vote for US
representatives and British Members of Parliament. American Political Science Review 78(1): 110–125.
Carey, J.M. (2009). Legislative voting and accountability. New York: Cambridge University Press.
Carey, J.M. & Shugart, M.S. (1995). Incentives to cultivate a personal vote: A rank ordering of electoral
formulas. Electoral Studies 14(4): 417–439.
Chappell Jr, H.W. & Veiga, L.G. (2000). Economics and elections in Western Europe, 1960–1997. Electoral
Studies 19(2): 183–197.
Clarke, H.D. et al. (2004). Political choice in Britain. Oxford: Oxford University Press.
Colomer, J.M. (2011). Introduction: Personal and party representation. In Personal representation: The
neglected dimension of electoral systems. Colchester: ECPR Press.
Dalton, R.J., McAllister, I. & Wattenberg, M.P. (2000) The consequences of partisan dealignment. In R.J.
Dalton & M.P. Wattenberg (eds), Parties without partisans: Political change in advanced industrial
democracies. Oxford: Oxford University Press.
Dassonneville, R. & Lewis-Beck, M.S. (2014). Macroeconomics, economic crisis and electoral outcomes:
A national European pool. Acta Politica 49(4): 372–394.
De Vries, C.E., Edwards, E.E. & Tillman, E.R. (2011). Clarity of responsibility beyond the pocketbook: How
political institutions condition EU issue voting. Comparative Political Studies 44(3): 339–363.
Döring, H. & Manow, P. (2015). Parliament and government composition database (ParlGov):
An infrastructure for empirical information on parties, elections and governments in modern democracies.
Available online at: www.parlgov.org/
Duch, R.M. & Stevenson, R.T. (2008). The economic vote: How political and economic institutions condition
election results. Cambridge: Cambridge University Press.
Farrell, D.M. & McAllister, I. (2006). Voter satisfaction and electoral systems: Does preferential voting in
candidate-centred systems make a difference? European Journal of Political Research 45(5): 725–749.
Farrell, D.M. & Scully, R. (2007). Representing Europe’s citizens? Electoral institutions and the failure of
parliamentary representation. Oxford: Oxford University Press.

C 2016 European Consortium for Political Research

CANDIDATE-CENTRED ELECTORAL SYSTEMS

339

Hellwig, T. (2011). Context, political information and performance voting. In R.J. Dalton & C.J. Anderson
(eds), Citizens, context and choice: How context shapes citizens’ electoral choices. Oxford: Oxford
University Press.
Hellwig, T. & Samuels, D. (2007). Voting in open economies: The electoral consequences of globalization.
Comparative Political Studies 40(3): 283–306.
Hellwig, T. & Samuels, D. (2008). Electoral accountability and the variety of democratic regimes. British
Journal of Political Science 38(1): 65–90.
Hobolt, S.B., Tilley, J. & Banducci, S. (2013). Clarity of responsibility: How government cohesion conditions
performance voting. European Journal of Political Research 52(2): 164–187.
Huber, J.D., Kernell, G. & Leoni, E.L. (2005). Institutional context, cognitive resources and party attachments
across democracies. Political Analysis 13(4): 365–386.
Karvonen, L. (2010). The personalisation of politics: A study of parliamentary democracies. Colchester: ECPR
Press.
Kayser, M.A. & Peress, M. (2012). Benchmarking across borders: Electoral accountability and the necessity
of comparison. American Political Science Review 106(3): 661–684.
Kitschelt, H. (2000). Linkages between citizens and politicians in democratic polities. Comparative Political
Studies 33(6–7): 845–879.
Lijphart, A. (2012). Patterns of democracy: Government forms and performance in thirty-six countries. New
Haven, CT: Yale University Press.
Nadeau, R., Niemi, R.G. & Yoshinaka, A. (2002). A cross-national analysis of economic voting: Taking
account of the political context across time and nations. Electoral Studies 21(3): 403–423.
Palmer, H.D. & Whitten, G.D. 2002. Economics, politics and the cost of ruling in advanced industrial
democracies: How much does context matter? In H. Dorussen & M. Taylor (eds), Economic voting.
London: Routledge.
Plümper, T., Troeger, V.E. & Manow, P. (2005). Panel data analysis in comparative politics: Linking method
to theory. European Journal of Political Research 44(2): 327–354.
Powell, G.B. (2000). Elections as instruments of democracy: Majoritarian and proportional visions. New
Haven, CT: Yale University Press.
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(2): 391–414.
Shugart, M.S. (2001). Electoral ‘efficiency’ and the move to mixed-member systems. Electoral Studies 20(2):
173–193.
Shugart, M.S., Valdini, M.E. & Suominen, K. (2005). Looking for locals: Voter information demands and
personal vote-earning attributes of legislators under proportional representation. American Journal of
Political Science 49(2): 437–449.
Weßels, B. (2007). Political representation and democracy. In R.J. Dalton & H.-D. Klingemann (eds), The
Oxford handbook of political behavior. New York: Oxford University Press.
Whitten, G.D. & Palmer, H.D. (1999). Cross-sectional analyses of economic voting. Electoral Studies 18(1):
49–67.
World Bank. (2015). World development indicators. Washington, DC: World Bank. Available online at:
www.data.worldbank.org/

Address for correspondence: Peter Söderlund, Social Science Research Institute, Åbo Akademi University,
Samforsk (B4), PO Box 311, Strandgatan 2, 65101 Vasa, Finland. E-mail: peter.soderlund@abo.fi


C 2016 European Consortium for Political Research