Insider or Outsider? Grand Corruption and Electoral Accountability

710258
research-article2017
CPSXXX10.1177/0010414017710258Comparative Political StudiesBauhr and Charron
Article
Comparative Political Studies
2018, Vol. 51(4) 415–446
Insider or Outsider? © The Author(s) 2017
Reprints and permissions:
Grand Corruption and sagepub.com/journalsPermissions.nav
DOI: 10.1177/0010414017710258
https://doi.org/10.1177/0010414017710258
Electoral Accountability journals.sagepub.com/home/cps
Monika Bauhr1 and Nicholas Charron1
Abstract
While democratic accountability is widely expected to reduce corruption,
citizens to a surprisingly large extent opt to forgo their right to protest
and voice complaints, and refrain from using their electoral right to punish
corrupt politicians. This article examines how grand corruption and elite
collusion influence electoral accountability, in particular citizens’ willingness
to punish corrupt incumbents. Using new regional-level data across 21
European countries, we provide clear empirical evidence that the level
of societal grand corruption in which a voter finds herself systematically
affects how she responds to a political corruption scandal. Grand corruption
increases loyalty to corrupt politicians, demobilizes the citizenry, and crafts
a deep divide between insiders, or potential beneficiaries of the system, and
outsiders, left on the sidelines of the distribution of benefits. This explains why
outsiders fail to channel their discontent into effective electoral punishment,
and thereby how corruption undermines democratic accountability.
Keywords
corruption and patronage, elections, public opinion, voting behavior,
accountability, quality of democracy
1Quality
of Government (QoG) Institute, Department of Political Science, University of
Gothenburg, Sweden
Corresponding Author:
Monika Bauhr, Goteborgs Universitet, Sprängkullsgatan 19, Goteborg 405 30, Sweden.
Email: monika.bauhr@gu.se
416 Comparative Political Studies 51(4)
Introduction
While international organizations, policy experts, and nongovernmental
organizations promote democratic accountability as a critical component to
curb corruption, citizens to a surprisingly large extent opt to forgo their right
to protest and voice complaints, and refrain from using their electoral right to
punish corrupt politicians. The election and reelection of corrupt politicians
pose a major challenge to democratic theorists as it allows democracies to
underprovide public goods, such as health care, economic development, or
environmental protection, which in turn threatens to undermine the legiti-
macy of the democratic system in the eyes of the electorate (Norris, 2011).
This study examines why politicians involved in corruption continue to be
reelected. In doing so, we distinguish between two types of corruption—one,
political corruption scandals among elected officials, and, two, what we refer
to as high-level, “grand corruption” (GC). We explore how the latter influ-
ences voter behavior in response to the former type of corruption, and suggest
that pervasive GC corrodes electoral accountability and explains why institu-
tions rife with venality remain unchecked. In other words, the extent to which
a citizen punishes corrupt politicians is highly contingent upon the context in
which she finds herself. A polity dominated by egregious GC among elites is
organized around informal extended networks of acquaintances, and the per-
sonalized interchange of concrete rewards and punishment, rather than
around abstract impartial principles involving people who may not even be
part of your network (Hale, 2015, p. 20; Karklins, 2005; Ledeneva, 2013).
Such a systemic level of corruption permeates polities and “dictates every
political and economic pursuit” (Hale, 2015), and thereby shakes the very
foundations of liberal democracy, as it builds on “personal bonds where
detached judgment should be” (Piliavsky, 2014, p. 2) and particularism
(Mungiu-Pippidi, 2006) in lieu of impartiality (Rothstein & Teorell, 2008).
We broadly define GC as collusion among the highest levels of govern-
ment that involves major public sector projects, procurement, and large
financial benefits among high-level public and private elites (Rose-Ackerman,
1999, p. 27), transactions which are rarely directly observed by the general
public. We suggest that the level of GC varies considerably in scale and
severity between countries, and, at times, within countries from region to
region and risks corroding electoral accountability through three parallel but
interrelated electoral effects. In environments where GC is systemic, patron-
age-like exchanges are common and enhance loyalty to corrupt politicians, as
defectors may be deprived of spoils and face the moral costs involved of
breaching loyalty obligations. A polity with high or even systemic levels of
GC also produces resignation and voter abstention, as the electorate is more
Bauhr and Charron 417
likely to experience alienation and cynicism about the potential of the politi-
cal system to contain corruption. In addition, polities dominated by egregious
GC craft deep divide between citizens, particularly between the insiders of
collusive networks of exchange and spoils, or the real or potential beneficia-
ries of GC, and outsiders, those excluded from such collusion. In particular,
insiders remain loyal to corrupt politicians, irrespective of the politicians
being involved in a corruption scandal, whereas outsiders are more likely to
abstain from voting altogether. This contributes to explaining the failure of
outsiders to canalize their grievances into effective electoral punishment in
places where such behavior is most needed—in democracies with systemic
corruption.
Our analysis contributes to the growing literature on how corruption
affects voting behavior in democracies. We make a theoretical contribution
by developing a comprehensive multilevel theory, explaining why and how
GC corrodes electoral accountability and maintains corrupt politicians in
power. This contributes toward explaining why both virtuous and vicious
cycles of corruption are maintained and perpetuated. We model individual
characteristics of voters and macrolevel factors to interact to explain electoral
responses to a political corruption scandal. At the individual level, our
insider–outsider framework distinguishes between respondents, and at the
macrolevel we focus on variation in the level of grand (elite-level) corruption
across polities. In addition, as opposed to only focusing on isolated responses
to corruption, such as whether a voter would vote for a corrupt party or
whether they would abstain from voting altogether, we analyze three types of
electoral responses to corruption using a development of Hirschman’s (1970)
framework of Exit, Voice, and Loyalty. How citizens express discontent is of
central importance to the functioning of democracy, and mechanisms of
accountability operate very differently depending on the type of response that
citizens perceive to be most viable.
We also make an empirical contribution. We test our theory with a large
original survey that asks respondents directly how they would respond elec-
torally if their preferred party was involved in a corruption scandal, and com-
bine these survey data with a newly developed, hard measure of GC among
185 European regions in 21 countries. Our findings provide clear empirical
evidence that pervasive GC creates an environment where corrupt incum-
bents are more likely to be reelected, undermining the effectiveness of elec-
tions in curbing political corruption. In such contexts, voters are less inclined
to hold a corrupt incumbent accountable by switching to another party.
Furthermore, insiders are more likely to be loyal to their preferred party,
whereas outsiders become more inclined to abstain from voting. Thus, we
provide systematic evidence for a growing sentiment in the literature that has
mostly been anecdotal—that elections can perpetuate (or even increase) cor-
ruption in places where elite collusion is pervasive (Bauhr & Grimes, 2014;
Mungiu-Pippidi, 2006; Persson, Rothstein, & Teorell, 2013). Conversely, we
find that, in environments where GC is less prevalent, accountability in the
form of voting for an alternative party becomes much more likely, and this is
particularly so among insiders. In other words, we find that insiders perpetu-
ate the status quo and defend the prevailing system in low as well as high
corrupt contexts: In “cleaner” societies, insiders defend the low corrupt equi-
librium, in contexts where GC is pervasive insiders loyally maintain corrupt
politicians in power.
The article proceeds as follows: First, we provide a brief overview of the
literature on electoral responses to corruption, and argue for the importance of
better integrating knowledge about how the functioning and nature of GC
shape loyalty structures and cynicism, and how these effects vary systemati-
cally across different groups of citizens. We then develop a theory about how
GC influences citizens’ responses to corruption scandals, focusing particularly
on loyalty, demobilization, and the divisiveness of GC. Third, we discuss our
data and measurements: We test our hypotheses using novel survey data for
more than 85,000 respondents in 24 European countries along with new data
on high-level administrative corruption in 185 regions. The data uses approxi-
mately 1.45 million awarded public procurement contracts and measure in
particular restrictions on open competition for government contracts to benefit
well-connected companies (Fazekas & Kocsis, 2015). The fourth section pres-
ents our results, and the fifth section concludes the article.
Corruption and Democratic Accountability

Cross-national analysis has so far failed to produce conclusive evidence on
whether democracies are less corrupt than nondemocracies (Sung, 2004;
Treisman, 2007; Fukuyama, 2015; Keefer, 2007; Montinola and Jackman,
2002). The broad differences in the extent to which democracies contain corrup-
tion suggest that fundamental mechanisms of democratic accountability work
very differently in different contexts, and that voters far from always punish
corrupt politicians (Chang, Golden, & Hill, 2010; Manzetti & Wilson, 2007).
While the empirical evidence for whether the electorate punishes corrupt
politicians remains mixed (see, for example, Bågenholm, 2013; Basinger,
2013; Crisp, Olivella, Potter, & Mishler, 2014), several case studies suggest
that voters’ punishment of corrupt politicians remains limited. For instance,
evidence from the United States (Welch & Hibbing, 1997), Brazil (Ferraz &
Finan, 2008), Italy (Chang et al., 2010), the United Kingdom (Eggers &
Fisher, 2011), and Mexico (Chong, De La O, Karlan, & Wantchekon, 2015)
suggests that candidates are punished to only a very limited extent for being
implicated in corruption scandals.1
An emerging literature seeks explanations for the shortcomings of elec-
toral and societal accountability and for why it fails to contain corruption.
This body of research typically assumes that political corruption leads to
some level of indignation, but that this indignation is for various reasons not
canalized into electoral punishment (see, for instance, de Sousa & Moriconi,
2013). At least three explanations seem to be particularly prominent in this
literature. Explanations related to trade-off suggest that the reason why cor-
rupt politicians survive in office is that voters simply do not prioritize having
a noncorrupt political representative. Other factors, such as the politician’s
ability to attract investment, build successful coalitions, generate economic
growth, or make otherwise popular decisions, are often valued more highly
by most voters (see, that is, Konstantinidis & Xezonakis, 2013; Zechmeister
& Zizumbo-Colunga, 2013). Others find limited support for the existence of
trade-offs, and instead suggest that the reason why corrupt officials can main-
tain political support is that voters lack information about wrongdoings
(Winters & Weitz-Shapiro, 2013, see also Chang et al., 2010; Costas-Pérez,
Solé-Ollé, & Sorribas-Navarro, 2012). Furthermore, various institutional
explanations, such as press freedoms, electoral and party systems, the delega-
tion of power and the level of independence of the judiciary (Charron &
Bågenholm, 2016; Ferraz & Finan, 2008; Gerring & Thacker, 2004; Kunicová
& Rose-Ackerman, 2005), may contribute toward explaining reelection, not
least because they can influence the effectiveness of protests and citizen
mobilization. In short, the reason why citizens do not voice complaints is that
they lack information, lack institutional opportunities to do so, or simply do
not prioritize anticorruption efforts.
While all these explanations are plausible, we seek to take a step back in
this article and ask a more fundamental question: What determines citizens’
institutional opportunities to punish corrupt politicians, information about
corruption, and the extent to which they prioritize politicians running on anti-
corruption mandates? We suggest that the importance of these explanations
varies by context, and is in turn partly determined by the level of GC, and in
particular the extent to which such corruption pervades a given society. While
previous studies have distinguished between the effects of experiences of
corruption/bribery and perceptions of societal-level corruption on voting
behavior (Klašnja, Tucker, & Deegan-Krause, 2016) or compared the effects
of different forms of corruption on voting behavior and civic engagement
(Bauhr, 2016; Fernández-Vázquez, Barberá, & Rivero, 2016), our study is
unique in that we model the level of societal corruption as a macrolevel vari-
able using an objective measure, as opposed to the perceptions of individual
respondents. We believe that simple indicators of individual-level percep-

tions of societal levels of GC may be less likely to capture the deep-rooted
societal effects of GC. Furthermore, many studies focus on electoral responses
to corruption in single countries (with data from either experimental or
“actual” scandals), which limits tests of macrolevel contextual factors.2
In other words, this study seeks to bring in insights from the large literature
on GC to explain the reelection of corrupt politicians in Europe (see, for exam-
ple, Stokes, Dunning, Nazareno, & Brusco, 2013; Nichter and Peress, 2016).
GC and other elite types of collusions (see Jain, 2001) often have deep-rooted
effects on the functioning of society, which may have both direct and indirect
effects on the tolerance for political corruption. Electoral responses to corrup-
tion are not necessarily influenced by the nature of the corruption scandal and
how much citizens morally oppose such activities, or by whether politicians
are involved in direct vote buying. We believe that electoral responses to cor-
ruption are also influenced by the norms, benefits, and expectations that high
levels of GC corruption convey to citizens.
Furthermore, the mechanism that explains the willingness to punish a cor-
rupt politician is not necessarily the same as the explanations for why citizens
stay loyal to a corrupt politician, or abstain from voting altogether. Using the
logic of collective action theory (Karklins, 2005; Ostrom, 1998; Persson
et al., 2013), some studies suggest that exposure to corruption leads to resig-
nation and demobilization (Bauhr & Grimes, 2014), voter abstention
(Sundström & Stockemer, 2015), or cynicism and alienation (Pavão, 2015).
These studies suggest that resignation is often a fundamentally different type
of response than the indignation or voice that may constitute a more demo-
cratically desirable response to corruption.
In an attempt to understand different forms of electoral responses, we take
Hirschman’s (1970) conceptualization of Exit, Voice, and Loyalty as a point
of departure. Put simply, citizens are faced with three choices in response to
corruption: Voice, Exit, or Loyalty. We define loyalty as citizens’ staying
loyal to the party despite it facing a corruption scandal, and suggest that citi-
zens can express discontent in two very different ways: Citizens can either
use voice, defined as voting for an alternative party, or exit, defined as abstain-
ing from voting altogether. How citizens express discontent is of central
importance to the functioning of democracy. If citizens use voice, that is,
change political parties, the mechanisms of electoral accountability work to
punish political leaders who abuse their power. It gives the political system a
chance to recover from its mistakes and change directions. Exit, such as not
voting at all, however, may be induced by disillusionment with the ability of
the political system to deal with the problems at hand. It is often associated
with resignation, alienation, and disengagement.
How Grand Corruption Corrodes Electoral

Accountability: Loyalty, Demobilization, and the
Insider–Outsider Divide
Our theory suggests that the nature of GC influences the functioning of
democracy and the reelection of corrupt politicians. GC influences expecta-
tions and voting behavior through partly different channels than experiences
of petty corruption. Similar to Manzetti and Wilson (2007), we are concerned
with the institutional context in which a voter finds herself. Our contextual
focus is on the extent to which a polity is permeated by GC, and thereby col-
lusive corrupt exchanges at high levels of society, and the extent to which
elites in the public and private sectors trade favors for their own benefit,
mainly at the public’s expense. In “cleaner” societies, norms of impartiality
and merit are more likely to dictate high-level exchanges in sectors such as
public investment and procurement (Charron, Dahlström, Lapuente, &
Fazekas, 2017). In particular, we suggest that when GC becomes systemic
citizens expect that the entire elite of one’s polity is involved in collective
collusion at the public’s expense, which corrodes electoral accountability and
the electoral punishment of corrupt politicians as it increases loyalty to cor-
rupt politicians, depresses voter turnout, and divides societies between insid-
ers and outsiders. Several explanations are discussed below.
Grand Corruption and Loyalty

Perhaps the most important key to understanding the deep-rooted effects of
corruption is to understand how it induces loyalty among certain groups of
citizens, and mobilizes citizens to support and maintain corrupt polities.
GC tends to promote patronage-like relationships driven by norms of reci-
procity and a normative sense of obligation toward the network (Auyero,
2001), and perhaps even more decisively by a fear that real or anticipated
flows of benefits will be cut off (Stokes et al., 2013). The broader literature
on voting behavior suggests that an electorate bound by strong loyalty ties
to the incumbent will be more likely to disregard improper or even immoral
behavior, including corruption (De Mesquita, Morrow, Siverson, & Smith,
2002). In high GC contexts, we can expect such loyalty effects to be par-
ticularly strong and forgiving as politicians in these contexts often attain
power with the aid of corruption and paternalistic networks (Golden, 2003;
Wantchekon, 2003).
Recent studies show that elites have used patronage jobs in the public
administration to help U.S. parties to retain power, and that the “degree of
control over patronage jobs affected a political party’s probability of winning
future elections” (Folke, Hirano, & Snyder, 2011, p. 568; Golden, 2003). In
settings with high levels of GC, less formal rules (as opposed to impartial
merit) tend to dictate employment, and winners in the marketplace conse-
quently tend to be privileged insiders (Rose-Ackerman, 1999). GC increases
the importance of informal exchanges and produces strong loyalty ties among
both real and potential “clients” in a GC system, which reduces the extent to
which citizens elect to switch parties. These loyalty codes extend far beyond
the presence of direct vote buying, and may be present even in contexts where
elections are considered “free and fair,” violence and fraud are relatively
uncommon and voters are generally neither directly incentivized nor coerced
to vote for a corrupt incumbent.
Political change in contexts dominated by egregious GC often comes at a
great risk as it threatens potential and future job, career and even subsistence
opportunities for many citizens and their extended families. Thus, high levels
of GC often increase the stakes of defection and voice. Long-term exposure
to business, public career, and recruitment structures built on GC, where pub-
lic contracts and positions are distributed on the basis of personal contacts
rather than merit, fundamentally shapes loyalty structures in society as well
as perceptions of “How things get done” (Hale, 2015; Ledeneva, 2013). This
forms our first hypothesis.
Hypothesis 1: In response to political corruption, a willingness to vote for

an alternative (voice) is less likely the higher the level of GC, ceteris
paribus.
Grand Corruption and Demobilization

If politicians involved in corruption scandals are ultimately voted out of
office, then this depends not only on the extent to which citizens switch par-
ties but also on the extent to which citizens decide to abandon the political
system and abstain from voting altogether. If citizens who are unfairly treated
by a system of corruption and favoritism, as suggested by Pollock (1937) and
others, respond by widespread party switching, then GC is not necessarily
beneficial for the reelection of corrupt politicians. If, however, protests take
other forms, in particular if they lead to voter abstention, citizen discontent
may not be canalized into effective electoral punishment and thereby contrib-
ute to maintaining corrupt politicians in power.3
While mobilization and turnout have traditionally not been granted much
systematic attention in studies on corruption, GC could perhaps in some
instances facilitate the favoring of particular constituents and even vote buy-
ing, and mobilize otherwise unorganized voters by offering them direct and
tangible rewards. This “turnout buying” (Nichter, 2008) or “participation

buying” (Schaffer & Schedler, 2007, p. 25), driven partly by the difficulties
involved in monitoring vote choices when ballots are cast in secrecy (Remmer,
2010), may contribute to maintaining corrupt politicians in power. Studies
that use aggregate perception-based measures of corruption seem, however,
to arrive at a conflicting conclusion. These studies suggest that exposure to
egregious corruption may demobilize the citizenry (Bauhr & Grimes, 2014;
Chong et al., 2015), and that perceptions of corruption lead to voter absten-
tion (Birch, 2010; Davis, Camp, & Coleman, 2004; Stockemer, LaMontagne,
& Scruggs, 2013; Sundström & Stockemer, 2015). One possible explanation
for the demobilizing effect of corruption is derived from the logic of collec-
tive action theory and, in particular, that a citizen’s mobilization is highly
contingent upon evidence that others will do the same (Bauhr & Grimes,
2014; Karklins, 2005; Ostrom, 1998; Persson et al., 2013). In contexts in
which corruption reaches systemic levels, it may be difficult to find individu-
als willing to face the risk and costs of mobilizing against corruption when
the chances of success are slim. Corruption may therefore lead to resignation
and alienation (exit) rather than protest (voice).
In other words, the stronger the dominant network becomes, the more
difficult it is for those who want to protest against it to succeed in their
endeavor. Such contexts may both reduce the supply of non or low corrupt
candidates who run for office, and undermine the potential of competing
patronage-based networks to mobilize support, as such mobilization tends to
work better for incumbents, or dominant parties, than for challengers
(Remmer, 2010; Wantchekon, 2003). Not voting at all may be perceived as
a more viable way to breach loyalty structures than “voice,” as the risk of
doing so can be very tangible in societies dominated by a strong corrupt
network. Citizens who are discontent with the state of affairs therefore opt to
abstain from voting altogether rather than risk the consequences of express-
ing loyalty to competitors.
In a polity dominated by GC, citizens may therefore be more likely to
develop cynicism and resignation about the potential of political systems to
contain corruption, or of challenging networks to contain corruption, and
such citizens may therefore opt to abstain from voting altogether. GC may
therefore channel discontent into various forms of exit options rather than
citizens expressing discontent in ways that may be more democratically
desirable (i.e., changing political parties if your own is involved in a corrup-
tion scandal), and thus depress voter turnout.4
Hypothesis 2: In response to political corruption, voter abstention (exit)

is more likely the higher the level of GC, ceteris paribus.
Grand Corruption and the Insider/Outsider Divide

H1 and H2 suggest that GC corrodes electoral accountability and maintains
corrupt politicians in power as it increases voter abstention (exit) and
decreases the likelihood of abandoning a party involved in corruption (i.e.,
reduces voice). What may be driving these effects? Perhaps the most power-
ful explanation for the detrimental effects of GC is that it often produces a
deep divide in societies between insiders, that is, real and potential beneficia-
ries of the corrupt network, and those left on the sidelines of the distribution
of spoils and benefits, or outsiders. This divisive effect of systems dominated
by GC is not only often deeply polarizing but may also be directly detrimen-
tal to the punishment of corrupt politicians. Some groups of citizens obtain
real or anticipated benefits from the prevailing system, and thereby develop
solid loyalty ties to corrupt politicians. Others left on the sideline may feel
alienated by the political system, and therefore be more likely to abstain from
voting altogether. While the responses of “victims,” including the extent to
which discontent is channeled into extraelectoral forms of protest, are more
often explored, less attention is directed to “. . . how the winners from this
process—the recipients of spoils jobs and their families—become more or
less attached to their patrons as a result” (Gingerich, 2014). A recent study on
how patronage divides societies in Africa furthermore finds that patronage
insiders “simultaneously perceive corruption as ubiquitous and are more tol-
erant of malfeasance,” and such insiders are also less likely to punish corrupt
politicians (Chang & Kerr, 2017).
We suggest that those who are, or anticipate being, future beneficiaries of
the system, develop strong loyalty ties. These loyalty ties may become
increasingly deeply ingrained as the strength of the system increases.
Furthermore, we expect that insiders play a radically different role in societ-
ies depending on the level of GC. While insiders perpetuate corruption in
contexts where it is pervasive, we expect that in contexts with limited GC
insiders instead play the role of vigilant defenders against corrupt politicians.
In other words, we expect insiders to perpetuate the prevailing system,
regardless of what that current system may entail. If the dominant system is
plagued by GC, insiders will support it by staying loyal to corrupt politicians.
If GC is the exception rather than the norm, insiders will instead be more
likely than outsiders to rise up to defy corrupt politicians.
Hypothesis 3: Insiders are more (less) loyal to corrupt politicians in high

(low) GC contexts, ceteris paribus.
Hypothesis 4: Outsiders are more likely to abstain from voting than insid-
ers, ceteris paribus.
Our theory thus posits that GC both decreases voice and increases exit, but
that these effects are strongly heterogeneous between different segments of
the population and depend on the extent to which GC corruption permeates
the system. GC divides societies between insiders and outsiders and, as the
prevalence of GC increases, insiders become increasingly loyal to corrupt
incumbents and less likely to punish corrupt politicians.
Research Design, Sample, and Data

The dependent variable and research designs in general of corruption voting
studies vary substantially. Some cross-country comparative studies building
on multilevel survey data use an indirect proxy for voting, such as govern-
ment support or satisfaction (Manzetti & Wilson, 2007), or whether the
respondent stated to have voted in a previous election for the incumbent
(Ecker, Glinitzer, & Meyer, 2016). Others employ data looking at actual
scandals and electoral turnout (Basinger, 2013; Chang et al., 2010; Peters &
Welsch, 1980; Praino, Stockemer, & Moscardelli, 2013; Stockemer et al.,
2013; Sundström & Stockemer, 2015), or electoral party volatility (Crisp
et al., 2014). Yet, the motivation as to why voters do or do not show up to cast
their ballot for the incumbent, vote for an alternative or stay home is left
unaccounted for due to the aggregation of the data to the macrolevel.
We elect to test the hypotheses empirically employing an observational,
multilevel design, using a large survey of 85,000 individuals in 24 European
countries.5 Europe offers an excellent test case, in that corruption varies sig-
nificantly from country to country (Charron, Dijkstra, & Lapuente, 2014) and
all countries have—to varying degrees—multiparty, democratic systems.
This spatial, comparative design allows us to best test our hypotheses, which
capture spatial variation both within and across countries.
The survey utilized was started in February 2013. The study, which was
conducted in the local language of each country, randomly selected between
1,200 and 10,500 respondents per country (varying by population size of the
country), providing a high degree of reliability due to the large sample size.
In all, 34 questions were posed to the respondents, and several are of particu-
lar interest here in capturing the key concepts.
As per our dependent variable, with no perfect way in which to capture the
concepts of voice, exit, and loyalty, we measure them as closely as possible
using the following questions, which express how an individual intends to
respond to a corruption scandal in his or her preferred party in an upcoming
election.
Question 1: What political party would you vote for if the national parlia-
mentary election were today?
Figure 1. Proportion of “exit,” “voice,” and “loyalty” responses by country.
Question 2: Now imagine that party was involved in a corruption scandal;

which of the following would be most likely? (a) still vote for the pre-
ferred party, (b) vote for an alternative party not involved in the corruption
scandal, (c) not vote at all.
In all, 20.7% of all respondents answered “1,” whereas 33.7% and 38.6%
responded “2” and “3,” respectively. About 6.9% responded “Don’t know.”
The third response most closely represents our idea of exit, as the respondent
is clearly choosing to abstain from the democratic process. Thus, we create a
dummy variable that equals “1” if an individual selected this response and
“0” if otherwise. The second response, “vote for an alternative party not
involved in the corruption scandal,” most closely captures our idea of
accountability, or voice, as the respondent intends to vote, but elects to punish
the corrupt party (despite it being their first preference) by voting for an alter-
native. A dummy variable is thus created with this response as well, where an
individual receives a “1” if he or she gave the second answer and “0” if oth-
erwise. Loyalty is coded “1” if the respondent answered “still vote for the
preferred party” and “0” if otherwise. Figure 1 gives the proportion of each
response by country.
There are of course shortcomings with broad, hypothetical questions such
as these. First, the magnitude of the scandal is not mentioned and, even if
this was the case, individual interpretations of this would be expected to

vary. While we may lose some internal validity with the lack of specificity,
we argue that it is beneficial on the whole. For example, if we had imposed
a certain type of scandal, it could very well be interpreted differently by
respondents across countries. In our case, we permit respondents themselves
to evaluate how relevant a corruption scandal is when using their vote,
resulting in a more general test of our hypotheses.6 Second, the question
pertains to the party and not an individual candidate. In countries that have
single-member districts, such as the United Kingdom, the salience of the
candidate might trump that of the party, which could under/overestimate our
findings in such cases. Third, the question is a hypothetical one, and thus the
respondents might behave differently in real life compared with how they
answer a survey question, resulting in “hypothetical bias.” This is of course
a problem for many other studies in the field of corruption voting, in particu-
lar in experimental studies where many research designs have been based on
hypothetical scenarios. However, we would argue that this is less of a prob-
lem than one might believe at first glance, as several recent empirical studies
have tested the external validity of hypothetical survey questions in relation
to “actual” electoral outcomes and found results to be remarkably similar
(Hainmuller, Hangartner, & Yamamoto, 2015; Johnston, 2006; Vossler &
Kerkvliet, 2003). Another weakness is that one of our “treatments” (the cor-
ruption scandal) is built into the question, and is thus without a quasi-exper-
imental design, where we are comparing the voting pattern with a “control”
group.
Finally, given the content of the question involved, there may be a degree
of “social desirability bias” at work. However, we do observe that almost
21% claim they would continue to support their preferred party despite a cor-
ruption scandal, and thus, underreporting would not necessarily cause such
bias if randomly distributed across individuals. However, this type of bias
could of course produce misleading results if correlated with one of the indi-
vidual-level variables, or if it was systemic in certain countries. Unfortunately,
we cannot directly test the presence of hypothetical bias or social desirability
bias, and we therefore highlight this as a caveat in interpreting the results.
While we are not able to fully address all of these potential problems, our
data do have several advantages in relation to previous studies, in particular
our ability to employ a multilevel design, combining a large number of indi-
viduals in 21 countries, and allowing us to directly test cross-level interac-
tions. Moreover, our study employs a more direct measure of the dependent
variable with the “treatment” (corruption scandal) specified in the question,
as opposed to an indirect measure such as government satisfaction, percep-
tions, or tolerance of political corruption.
Our key independent variables are as follows: First, while there is no clear
consensus in the field as to how this concept should be measured, to avoid
possible endogeneity with a perceptions measure, we capture GC with an
objective measure with the aid of data from Fazekas and Kocsis (2015). This
measure uses public procurement data to assess the risk of high-level corrup-
tion and favoritism, which is the best measure available given that our con-
cept of high-level corruption among elites is of concern here. The data, for
EU28 between 2009 and 2013, contain information on individual public pro-
curement tenders that has to do with regulated administrative procedures in
which public bodies purchase goods and services, including contract value,
the number of bidders, the deadline for submitting bids, and the assessment
criteria used. The measures tap into deliberate restrictions on open competi-
tion for government contracts to benefit well-connected companies, which
we argue is the best available comparative metric of our concept of GC. To
avoid distortion from less competitive markets such as defense, the indicators
are defined only when at least 10 contracts were awarded in the period 2009-
2013, indicating a sufficient demand for bids from at least two operating
companies. In all, the dataset is based on approximately 1.45 million awarded
contracts. The simplest indication of corruption risk is when only one bid was
submitted on a tender in an otherwise competitive market. Hence, the per-
centage of single-bidder contracts awarded in all the awarded contracts is the
most straightforward measure we use to capture our concept and, we argue,
this indicates a climate very favorable for the type of GC corruption we envi-
sion in our theory. As this—like most other measures of corruption—is not a
direct measure of corruption, one clear potential pitfall with this measure is
that it is not actually capturing GC, but something else, such as lack of com-
petitive markets. We find however that correlations between this measure and
alternative measures based on citizen or expert corruption perceptions and
experiences are quite significant at both the country and regional levels for
this sample, which gives us some assurance of validity.7 The data are also
available at the regional level for NUTS 1 and NUTS 2 levels. Country-level
averages are reported in Online Appendix Table A2.
Second, there is also a lack of consensus on how one should operationalize
an insider. In political economy, particularly labor market studies, an insider
is someone with full-time (or willing part-time) employment, whereas out-
siders are those who are outside of the labor market, those who are unem-
ployed (Rueda, 2005). However, this distinction has been used much less so
in studies of voting behavior (Lindvall & Rueda, 2014). Most studies in polit-
ical science and development economics about how corruption affects elec-
toral behavior focus on the related concept of clients (a type of insider),
whereby voters are considered clients if they claim to directly rely on a
powerful patron for help when in need (Stokes et al., 2013). In a recent study
on corruption tolerance in African countries, Chang and Kerr (2016) distin-
guish two types of insiders—“network insiders” (similar to the aforemen-
tioned concept of client) and “identity insiders,” citizens who either share
partisan or ethnic ties with the sitting government. Most of these past studies
however focus on Latin American or African countries, whereas studies of
Europe using the insider–outsider framework focus on one’s position in the
labor market.
Our conception of an insider is similar to Kitschelt and Wilkinson’s (2007,
p. 10) idea of “contingent direct exchange,” in that we attempt to capture
whether a respondent potentially has “something to lose” politically and eco-
nomically by a political change resulting from an election. We therefore
attempt to merge the operationalization of the concept insider from both past
corruption and labor market studies, and thus include two components of
insiderness—economic and political. On the economic side, we follow the
operationalization of most political economy scholars of the labor market
(Rueda, 2005)—insiders are those with full-time employment. In several
robustness checks, we go one step further and include only full-time public
sector workers as a new administration can make political appointments, par-
ticularly within the public sector, making this type of employment potentially
especially relevant in this context. In addition, we add a political element.
Using a similar operationalization for the political element to Chang and Kerr
(2016) whereby we code an individual as an insider if he or she is a partisan
supporter of a sitting party in government.8
Each respondent is provided with a country-specific list of relevant politi-
cal parties (party lists included all sitting parliamentary parties and poten-
tially relevant parties outside parliament, based on country-expert
assessments). If a respondent answered that they would support a party in the
sitting government (at the time of the survey), then they would be coded “1”
and “0” if they said they would support any other party,9 and thus the compo-
nent of political insiderness is nonideological and has only to do with the
current power of one’s party. The binary measures are combined to equal “1”
for respondents with full-time employment who support the sitting party in
power and “0” if otherwise.
As stated in the hypothesis, we expect that insiders will be more inclined
to express loyalty as an electoral response as the level of contextual GC
increases, meaning that their personal gain through political ties would out-
weigh their desire for a cleaner political party. Conversely, the extent to
which insiders are also expected to opt for voice decreases in contexts with
higher levels of GC. We also hypothesize that GC affects insiders differently
than outsiders in terms of electoral outcomes. Therefore, we test the cross-
level interaction term—“insider×GC.”
Several control variables are included at both the individual and country
levels based primarily on empirical research of election studies and voter
turnout (Geys, 2006; Smets & van Ham, 2013). First, we account for gender,
income, age, education, population of one’s residence, one’s self position on
the left–right scale, and general satisfaction with the economy. In addition,
we control for whether the respondent has paid a bribe in the last 12 months
for any public service. Thus, bribe corruption is measured at the microlevel,
whereas GC corruption is measured at the macrolevel (regional and country
level). The second-level fixed effects are population density (logged), ethnic
heterogeneity, recent past levels of voter turnout (average of the past two
election cycles), and the number of parties in the system, which Charron and
Bågenholm (2016) find to be relevant to corruption voting.
Due to our dependent variables being binary, we test the effects of corrup-
tion on voice and exit with logit estimation. Given that GC varies systemati-
cally within and across countries, we test for the effects with both logit
controlling for the country-level fixed effects described above as well as with
hierarchical logit, allowing for random country intercepts. Structuring the
data hierarchically has the added advantage of allowing us to test the cross-
level hypothesis more directly—that is, the insiders are more prone to loyalty
as the context of GC becomes greater.
Estimation
We begin by employing a hierarchical design, whereby the cross-level inter-
action effects are tested directly for the three outcomes separately. We find
that country-level differences in the dependent variable are present, and thus
country-level effects must be taken into account to avoid bias estimates.10
Simply accounting for regional and/or country dummy variables for individ-
ual-level variation in an ordinary-least-squares (OLS) model can lead to
problems because the error terms of the lowest level unit (individuals) within
the same group will still be correlated. Thus, we elect to explain individual
levels of voting in a hierarchical logit model with two levels, “i” and “j,” to
represent individual and country levels, respectively. The basic model used
here is as follows:
voting ij = α j + β2 Insiderij + β2 GC j +
β3 Insider × GCij + ϕnij +
ωn j + µ 0 j + e0ij ,
where voting ij is the log of the odds as a function of a set of individual-

level and country-level fixed effects to estimate our model parameters, β, ϕn ,
and ωn represent individual- and country (or regional)-level fixed effects con-
trols, respectively, µ0j shows the random-level intercepts for the macrolevel,
and e0ij is the error term.11
As the number of second-level observations (countries) is only 21, the
sample thus somewhat violates the so-called “30/30 rule” (Maas & Hox,
2005); we elected to remedy this in two ways: First, we estimate several of
the main models reported as individuals nested in regions (NUTS 1 and
NUTS 2 depending on the country).12 This has several advantages, as we see
it. One, the second level of the estimation becomes 185 (as opposed to 21),
giving us more than a sufficient sample size for hierarchical modeling, greater
degrees of freedom, and greater accuracy in the results as compared with the
limited country-level estimation on the second level (Maas & Hox, 2005).
Two, although the country context is salient, there is sufficient spatial varia-
tion in corruption, loyalty, and voting outcomes to warrant accounting for
regional differences,13 thus giving us a more precise picture of the landscape,
in particular in countries such as Italy, Spain, Belgium, and Germany that
have considerable regional variation. However, we do run models where the
country is used at the second level as well as standard logit with country fixed
effects. We also check for the effects of outliers in each model, and rerun the
analyses removing one country at a time in subsequent models (found in the
online appendix).
Results
Table 1 elucidates the results for each of the three outcomes. For the sake of
space, we report only the effects of GC and essential model statistics. For
each outcome, a baseline model was run (no controls—Models 1, 3, and 5)
along with a second model with full controls (Models 2, 4, and 6). We report
odds ratios to make the results more interpretable. We see that while GC does
not appear to have a significant effect on the probability of choosing loyalty,
it has an independent and significant effect in the full models for voice and
exit. For example, all things being equal, a 1-unit increase in the GC measure
is expected to increase the odds of exit by 13% and decrease the likelihood of
voice by 8%.
Overall, these results are more or less in line with our first two hypothe-
ses—that all things being equal, GC undermines electoral accountability by
decreasing the likelihood of voice and increasing the likelihood of exit,
although the effects for the latter are more robust to the inclusion of controls.
For example, as can be observed in Figure 2, for roughly the “cleanest” 20%
of the regions in the sample, the model predicts that the most likely electoral
432
Table 1. Electoral Responses to Corruption: Hierarchical Estimation Test of H1 and H2.
Loyalty Voice Exit
Fixed effects 1 2 3 4 5 6
Level 1 (individual)
Intercept 0.21 (.000) 0.04 (.000) 0.36 (.000) 0.07 (.000) 0.97 (.76) 3.28 (.000)
Level 2 (region)
Grand corruption 0.94 (.08) 1.01 (.65) 0.85 (.000) 0.92 (.04) 1.25 (.000) 1.13 (.000)
Random effects
Random intercept (SE) 0.45 (0.03) 0.37 (0.02) 0.46 (0.03) 0.42 (0.02) 0.51 (0.03) 0.38 (0.02)
Wald test Pr (χ2) .000 .000 .000 .000 .000 .000
ICC .06 .04 .07 .05 .07 .04
Log likelihood (It. 0) −36,635.6 −33,650.4 −46,215.3 −44,148.2 −47,862.4 −44,859.9
Log likelihood (final) −36,566.9 −33,538.7 −46,147.8 −44,020.7 −47,807.9 −44,729.1
LR test .000 .000 .000 .000 .000 .000
Ind. obs (regions) 74,338 (185) 74,338 (185) 74,338 (185) 71,477 (181) 74,338 (185) 71,477 (181)
Odds ratios reported with p values in parentheses from hierarchical logit estimation. Models 1, 3, and 5 show the baseline interaction terms
without controls, whereas 2, 4, and 6 include full controls and Level 1 and Level 2. Level 1 controls are gender, age, income, education, population
of residence, satisfaction with current economy, unemployment, left–right self-placement, support for government party (insider), and corruption
experience (bribe). Level 2 controls include ethnic heterogeneity, population density (logged), past turnout levels, and the effective number of
parties. ICC stands for “interclass correlation,” and ranges from 0 to 1. ICC = intraclass correlation; LR = likelihood ratio.
Figure 2. Visual summary of the effects of GC on corruption voting in voter

response to political corruption scandal.
Results from Models 2, 4, and 6 in Table 1 (full models). Predicted probability of each
outcome listed on the y axis. Estimate effect (line and dots) with a 95% confidence interval
(dashed lines). GC = grand corruption.
response is voice, whereas for more corrupt regions exit becomes the option
of the plurality of respondents, demonstrating evidence for Hypotheses 1
and 2.
In Table 2, we test Hypotheses 3 and 4 by including a cross-level interac-
tion between GC and whether an individual respondent is an insider or not.
Our anticipation is that being an insider will have an impact on how someone
views a corruption scandal in his or her party, and that this decision is also
conditional on the level of GC in the region where they reside. In Models 1,
3, and 5, we produce baseline estimates with only the main variable and the
interaction, whereas in 2, 4, and 6 all controls are included. Due to the insider
variable being binary, “1” for insiders and “0” for outsiders, the GC coeffi-
cient can be interpreted as the effect of GC among outsiders. Again, odds
ratios are reported.
With regard to loyalty, we find a significant interaction effect, implying a
divergence of behavior between insiders and outsiders, as GC increases. At
almost all levels of GC, insiders are more inclined to show loyalty than
434
Table 2. Electoral Responses to Corruption: Tests of H3.
Loyalty Voice Exit
Fixed effects 1 2 3 4 5 6
Level 1 (individual)
Insider 3.60 (.000) 3.16 (.000) 0.90 (.03) 0.81 (.000) 0.54 (.000) 0.65 (.000)
Intercept 0.15 (.000) 0.04 (.01) 0.37 (.000) 0.07 (.000) 1.09 (.31) 3.55 (.000)
Level 2 (region)
Grand corruption (CG) 0.89 (.003) 0.92 (.02) 0.87 (.000) 0.94 (.11) 1.25 (.000) 1.13 (.000)
Insider × GC 1.23 (.000) 1.23 (.000) 0.94 (.002) 0.93 (.000) 0.94 (.004) 0.96 (.04)
Random effects
Random intercept (SE) 0.42 (0.02) 0.38 (0.02) 0.46 (0.03) 0.42 (0.02) 0.49 (0.03) 0.38 (0.02)
Wald model test Pr (χ2) .000 .000 .000 .000 .000 .000
ICC .05 .04 .06 .05 .07 .04
Log likelihood −35,639.2 −33,532.8 −46,671.1 −44,013.5 −47,486.9 −44,735.4
LR test .000 .000 .000 .000 .000 .000
Ind. obs (regions) 74,338 (185) 71,477 (181) 74,338 (185) 71,477 (181) 74,338 (185) 71,477 (181)
Odds ratios reported with p values in parentheses from hierarchical logit estimation. Models 1, 3, and 5 show the baseline interaction terms
without controls, whereas 2, 4, and 6 include full controls and Level 1 and Level 2. Level 1 controls are gender, age, income, education, population
of residence, satisfaction with current economy, unemployment, left–right self-placement, corruption experience (bribe). Level 2 controls include
ethnic heterogeneity, population density (logged), past turnout levels, and the effective number of parties. ICC stands for “interclass correlation”
and ranges from 0 to 1. ICC = intraclass correlation; LR = likelihood ratio.
Figure 3. Marginal interaction effects.
outsiders. However, as the level of GC increases, the likelihood of loyalty

actually increases rather strongly among insiders, suggesting strong empirical
support for Hypothesis 3. We notice however the reverse trend among outsid-
ers—when comparing two otherwise similar outsiders—one living in a con-
text where GC is pervasive and the other in a context of low GC, the outsider
living in the context of pervasive GC is expected to be slightly less likely to
express loyalty. In turn, we see that outsiders in high GC contexts are most
likely to choose exit over voting. Figure 3 highlights the interaction visually.
With respect to voice, there is also a significant interaction effect between
insider and GC. We find that insiders seem to be driving the negative effects
of GC on voice substantially. At the lowest levels of observed GC, insiders
are actually more likely than outsiders to vote for an alternative party (voice).
However, GC sharply decreases the odds of insiders choosing voice, yet
among outsiders there is no independent effect of GC on voice (p = .11) as
shown in Model 4. We interpret this to mean that as the polity becomes
increasingly permeated by GC, the relationship flips, and insiders become
less likely than outsiders to express voice. In other words, insiders play radi-
cally different roles in society depending on the extent to which the system is
dominated by GC. While insiders will perpetuate corruption in deeply
Table 3. Summary Results of Control Variables by Outcome.
Loyalty Voice Exit

Female −** + +**
Education +** +** −**
Age +(60+)** −(60+)** −**
Income +** +* −**
Population +** −* −**
LR ideology (cen) −** +** +**
Econ. sat. +** + −**
Ethnic heterogeneity −** + −
Population density (log) + + −
Past turnout +** +** −
ENPs + + −**
ENP = effective number of parties; LR = likelihood ratio.

*p < .05. **p < .01.
corrupt contexts, they will choose to hold those who engage in it accountable
in contexts where corruption is already comparatively low.
With regard to the third outcome, exit, we find that both being an insider
and the level of GC have strong independent effects. The odds of exit among
outsiders increase by 13% for every unit increase of GC, and there is even a
slight interaction effect between these two factors (the odds of outsiders exit-
ing compared with insiders increase slightly as a function of greater GC). Yet
in all cases, outsiders are considerably more likely to exit than insiders, as
seen in Figure 3, which corroborates Hypothesis 4.
Table 3 highlights the findings of the effects of the control variables. The
results corroborate much of the existing literature on turnout and party loy-
alty. On average, given a corruption scandal in one’s preferred political party,
females are less prone to loyalty and more prone to voice and exit. People
with higher socioeconomic status (education and income) show a greater pro-
pensity for voting—either expressing loyalty or voice—than for abstention.
Participants older than 60 are more loyal and less likely to switch parties,
while the effects of age on exit are strongly negative, with younger voters
being the most likely to abstain. Residents of larger urban areas show more
loyalty than rural residents, while the latter are more prone to voice and exit.
On a left–right spectrum (5-point scale), we observed clear “u-shaped” (and
“inverted u-shaped” patterns), with center voters least likely to show loyalty
and most likely to either voice or exit relative to voters on the ideological
fringes. The effects of perceptions on the economy work according to our
expectations—less satisfaction leads to exit, and more satisfaction leads to

loyalty, while economic satisfaction is negligible in explaining voice. Of the
macrolevel variables, we find that greater levels of past turnout are associated
with a greater propensity to loyalty and voice, with a greater effective number
of parties (ENPs) being associated with less voter abstention. Finally, on
average, greater levels of ethnic diversity are also associated with slightly
higher levels of loyalty.
Checking for Robustness

We began by checking to see if any of the results reported in Tables 1 or 2 are
driven by observations from a single country. Thus, we reran all models with
a country-wise jackknife to test whether the results are robust to excluding
each country one at a time. We find the results in all cases to be strongly
robust (see Table A5 in the online appendix). Second, we recoded the variable
insider, taking only those who are partisan supporters and employed in the
public sector (and therefore possibly most susceptible to political reappoint-
ments given a change in government). We find no distinguishable changes in
the results (Table A6). Some recent studies find personal experiences with
petty corruption to be more relevant in explaining voting patterns than mac-
rolevel corruption (Chang & Kerr, 2016; Klašnja et al., 2016), and thus the
findings of our model, which focuses on macrolevel variation in corruption,
might be simply proxying for the effect of respondents’ personal experience
with petty corruption (as petty corruption plausibly occurs more often in
places with higher levels of GC). We tested for this by running several mod-
els, including a triple interaction between insider and personal experience
with corruption (petty bribery) and GC. We find that the main effects are not
affected by whether respondents have personal experience with bribery14—
indicating that, at least in the European context, it is not personal experience
with corruption that drives electoral behavior among either insiders or outsid-
ers. We reran the models in the two main tables with standard logit estima-
tion, controlling for country-level fixed effects (Table A7); the main
interaction effects proved to be relevant on average, even among regions
within countries. Next, we combined the three outcome variables into a sin-
gle categorical variable and used multinomial logit estimation (Table A8).
Finally, it is plausible that the choices presented are not completely indepen-
dent but the result of a sequence of decisions.15 For example, when faced with
a corruption scandal in their party, it is possible that individuals first choose
between being loyal or not; if they choose not to be loyal, they then elect
between choice and exit. Or possibly, the first choice is between not voting
and voting (exit or other), and then second, between loyalty and voice. We
estimate these two sequences in Table A9 via sequential logit models. The
main findings are remarkably consistent and robust to these alternatives.
Conclusion
While democratic accountability is perhaps not only the most desirable but
also the most indispensable means to reduce political corruption, democra-
cies clearly vary in their ability to contain corruption. This study suggests that
GC corrodes electoral accountability and explains why institutions rife with
venality remain unchecked. In particular, we suggest that egregious GC is
severely detrimental to the electoral punishment of corrupt politicians as it
canalizes grievances into resignation, alienation, and exit rather than indigna-
tion and voice, enhances loyalty to corrupt politicians, and crafts deep soci-
etal divides between insiders, or beneficiaries of the dominant system, and
outsiders, those left on the sidelines of the spoils of the dominant network.
All these effects—loyalty, demobilization, and divisiveness—we suggest
maintain corrupt politicians in power.
Using a large-N multilevel comparative design that allows us to directly
model cross-level interactions between individual and societal factors, a new
objective measure of GC using approximately 1,45 million procurement con-
tracts and surveys at the regional level for 185 European regions in 21 coun-
tries, we provide several novel and noteworthy findings: First, we provide
clear empirical evidence that GC and elite collusion create an environment
where corrupt incumbents are more likely to be reelected, even if politicians’
involvement in corruption scandals is widely known. In contexts where GC
is high, citizens respond to corruption scandals either by simply staying loyal
to the corrupt politician (loyalty) or by abstaining from voting altogether
(exit). Conversely, in contexts where elite-level GC is low, citizens are more
likely to switch parties in response to a corruption scandal (voice).
Interestingly, we find no evidence for a systematic effect of personal experi-
ence with petty corruption on either voice or loyalty, and only a relatively
small effect on exit, suggesting that the contextual effects of elite-level cor-
ruption are more powerful in affecting voting behavior than personal experi-
ences in a comparative European context.16
Second, we develop an insider–outsider framework and observe important
differences not only in the way in which insiders and outsiders respond to a
corruption scandal but also that these responses are highly contingent upon
the context in which voters find themselves and in particular on the level of
GC in their polity. Insiders in polities where GC reaches systemic levels typi-
cally stay loyal to corrupt politicians, whereas outsiders in such contexts are
instead more likely to refrain from voting and exit the electoral system
altogether in response to a corruption scandal. In other words, outsiders, or

those excluded from the spoils of the dominant network and the benefits of
favoritism, fail to canalize their grievances into effective electoral punish-
ment. Interestingly, insiders in contexts where GC is low are, instead, signifi-
cantly more likely to punish corrupt politicians than outsiders.
The level of corruption in a given political system is at times framed as a
type of systemic equilibrium, varying from low to high (e.g., see Johnston,
1998). Our study helps us understand how the vicious (or virtuous) cycle of
corruption operates in democratic societies, that is, why corrupt equilibria are
maintained and perpetuated. We show that insiders play a different role than
outsiders, particularly that insiders are more likely to remain loyal supporters
of the dominant system and status quo, regardless of what that dominant sys-
tem may be. In polities permeated by corruption, insiders will use their elec-
toral power to reelect corrupt politicians and thereby perpetuate corruption. In
contexts where corruption is low, insiders will use their electoral power to
defend the low corruption system. The vicious cycle of corruption is also per-
petuated by outsiders. We show that a near majority of outsiders chooses to
abstain from voting altogether, thereby contributing to maintaining corrupt
politicians in power. We can speculate that these voters view the system as
“rigged” and, without a proper alternative to their preferred party, are unable
to collectively compete with and outweigh the increased loyalty of insiders. In
the European context, these results mirror and explain the steady decline over
the past two decades in voter turnout trends in the areas where elite corruption
is relatively pervasive, such as Romania, Bulgaria, Slovakia, and Greece.
Furthermore, our design has given us new insights into how “partisanship
biases” may shape electoral responses to corruption. While previous studies
suggest that voters have a higher tolerance of political corruption when com-
mitted by their own party (Anduiza, Gallego, & Muñoz, 2013), we suggest
that this effect may have a larger electoral impact in contexts plagued by
egregious GC. In low corruption settings, respondents are quite willing to
abandon partisan loyalties in favor of alternative parties when their preferred
party has committed malfeasance. Furthermore, in high corruption settings
we find that outsiders are quite willing to abandon partisan ties in the face of
a corruption scandal, yet their response is to abstain from voting altogether.
We interpret both these findings to mean that when viable alternatives exist
a preference to maintain clean governance trumps partisan loyalties; while
in settings with higher corruption—at least for outsiders—partisanship is
trumped by resignation.
Our findings thereby add several insights into the growing literature on
how corruption affects electoral accountability. In many ways, this study
shows the benefit of better integrating knowledge on the societal effects of
corruption, its heterogeneous effects among different segments of the popula-

tion, and the varieties of electoral responses to corruption (exit, voice, and
loyalty) to understand how and why the electorate sometimes fails to punish
corrupt politicians. While interesting and insightful current research explores
these aspects individually (see, that is, Chang & Kerr, 2016; Chong et al.,
2015; Karklins, 2005), this is to our knowledge the first study that integrates
these insights into a single framework and tests it using a large-N compara-
tive design. We believe that moving toward a more integrated approach and
exploring how these factors interact, reinforce, and undermine each other
contribute to a more dynamic understanding of how corruption influences
democratic accountability.
The study does not provide a simple panacea for how to reduce political
corruption. It does indicate, however, that serious reforms that include
increased transparency, competition, and the implementation of laws prohibit-
ing preferential treatment of firms in public bids, or otherwise undermining
opportunities for administrative corruption, can contribute not only to hori-
zontal accountability, that is, elites’ opportunities to monitor each other, as
suggested by previous studies, but also to increased vertical accountability by
reducing the probability of corrupt politicians to become elected and reelected.
Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research,
authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research,
authorship, and/or publication of this article: This project has received funding from
the European Research Council (ERC) under the European Union’s Horizon 2020
research and innovation programme (Grant Agreement No. 339571) called “PERDEM,”
the Swedish foundation for humanities and social sciences and the Seventh Framework
Programme for Research and Development of the European Union (Project Number
290529) called “ANTICORRP” (http://anticorrp.eu/).
Notes
1. See Fernández-Vázquez, Barberá, and Rivero (2016) for an excellent overview
of the findings of these studies.
2. See Fernández-Vázquez, Barberá, and Rivero (2016); Funk (1996); Welch and
Hibbing (1997).
3. However, clientelist parties may also buy voter abstention as a strategy to win
elections (Cox & Kousser, 1981), indicating that the mobilization effect of party
machines may not be easily predictable.
4. While voter abstention is not necessarily associated with resignation from

all attempts to influence political systems, it does contribute to maintaining
corrupt politicians in power (see also Gingerich, 2014; Cornell & Grimes,
2015).
5. More information on methodology, administration, and response rates of the sur-
vey can be found in the online appendix.
6. An alternative would possibly be to incorporate some type of “anchored vignette”
(King, Murray, Salomon, & Tandon, 2004); however, as our response categories
are nominal rather than ordinal there is less utility in any adjustments made from
anchoring such a question.
7. For example, Pearson’s correlation between our measure and the corruption pillar
of the European Quality of Government Index (Charron, Dijkstra, & Lapuente,
2015) is .67 (n = 185), while the Organisation for Economic Co-Operation and
Development (OECD; 2016) provides regional-level data in Europe to a citizen-
level survey question about the “% of respondents who believe the government
is corrupt,” with which our measure correlates at .56 (n = 145). At the country
level, our objective measure correlates with the World Governance Indicator’s
“Control of Corruption” measure at .86 (n = 21).
8. Unfortunately, no question in our survey allows us to operationalize a “network
insider.”
9. Responses were matched with government data at http://www.parties-and-
elections.eu/
10. In addition, we estimate the country effects of the dependent variable with con-
fidence intervals around the estimate in Figure 1 in the online appendix.
11. We tested whether we can reject the null hypothesis that the random slope for
loyalty is equal to “0,” which was rejected. Thus, this serves as the basis for test-
ing our cross-level interaction term estimated by β3 .
12. NUTS stands for “Nomenclature of territorial units for statistics” (for more
information, see http://ec.europa.eu/eurostat/web/nuts/overview). As the survey
from which we draw the individual level was primarily interested in regional-
level variation, there is a sufficient sample size (400-450 respondents) per
region. Countries’ NUTS regions and number of regions per country are found
in the online appendix.
13. The interclass correlation from an empty (constant only) model of the corruption
variable, for example, shows that roughly 27% of the total variation is explained
by subnational differences (p < .0000).
14. However, with respect to exit, we find that personal experience with bribery
among insiders slightly decreases the probability of exit compared with insiders
who have no personal experience with corruption.
15. We would like to thank an anonymous reviewer at Comparative Political Studies
(CPS) for this suggestion.
16. This does not, however, mean that “pocketbook” corruption voting cannot be
present in certain contexts, as suggested by Klašnja, Tucker, and Deegan-Krause
(2016).
References
Anduiza, E., Gallego, A., & Muñoz, J. (2013). Turning a blind eye: Experimental
evidence of partisan bias in attitudes toward corruption. Comparative Political
Studies, 46, 1664-1692. doi:10.1177/0010414013489081
Auyero, J. (2001). Poor people’s politics: Peronist survival networks and the legacy
of Evita. Durham: Duke University Press.
Bågenholm, A. (2013). Throwing the rascals out? The electoral effects of corruption
allegations and corruption scandals in Europe 1981–2011. Crime, Law and Social
Change, 60, 595-609.
Basinger, S. J. (2013). Scandals and Congressional elections in the post-Watergate
era. Political Research Quarterly, 66, 385-398.
Bauhr, M. (2016). Need or greed? Conditions for collective action against corruption.
Governance. Advance online publication. doi:10.1111/gove.12232
Bauhr, M., & Grimes, M. (2014). Indignation or resignation: The implications of transpar-
ency for societal accountability. Governance, 27, 291-320. doi:10.1111/gove.12033
Birch, S. (2010). Perceptions of electoral fairness and voter turnout. Comparative
Political Studies, 43, 1601-1622.
Chang, E. C. C., Golden, M. A., & Hill, S. J. (2010). Legislative malfeasance and
political accountability. World Politics, 62, 177-220. doi:10.1017/S004388
7110000031
Chang, E. C., & Kerr, N. N. (2017). An insider–outsider theory of popular tolerance
for corrupt politicians. Governance, 30(1), 67-84.
Charron, N., & Bågenholm, A. (2016). Ideology, party systems and corruption
voting in European democracies. Electoral Studies, 41, 35-49. doi:10.1016/j.
electstud.2015.11.022
Charron, N., Dahlström, C., Lapuente, V., & Fazekas, M. (2017). Careers, connec-
tions, and corruption risks: Investigating the impact of bureaucratic meritocracy
on public procurement processes. The Journal of Politics, 79(1), 89-104.
Charron, N., Dijkstra, L., & Lapuente, V. (2014). Regional governance matters:
Quality of government within European Union member states. Regional Studies,
48, 68-90. doi:10.1080/00343404.2013.770141
Charron, N., Dijkstra, L., & Lapuente, V. (2015). Mapping the regional divide in
Europe: A measure for assessing quality of government in 206 European regions.
Social Indicators Research, 122, 315-346.
Chong, A., De La O, A. L., Karlan, D., & Wantchekon, L. (2015). Does corruption
information inspire the fight or quash the hope? A field experiment in Mexico on
voter turnout, choice, and party identification. The Journal of Politics, 77, 55-71.
doi:10.1086/678766
Cornell, A., & Grimes, M. (2015). Institutions as Incentives for Civic Action:
Bureaucratic Structures, Civil Society, and Disruptive Protests. The Journal of
Politics, 77, 664-678.
Costas-Pérez, E., Solé-Ollé, A., & Sorribas-Navarro, P. (2012). Corruption scandals,
voter information, and accountability. European Journal of Political Economy,
28, 469-484. doi:10.1016/j.ejpoleco.2012.05.007
Cox, G. W., & Kousser, J. M. (1981). Turnout and rural corruption: New York as a test
case. American Journal of Political Science, 25, 646-663. doi:10.2307/2110757
Crisp, B. F., Olivella, S., Potter, J. D., & Mishler, W. (2014). Elections as instruments
for punishing bad representatives and selecting good ones. Electoral Studies, 34,
1-15. doi:10.1016/j.electstud.2013.08.017
Davis, C. L., Camp, R. A., & Coleman, K. M. (2004). The influence of party systems
on citizens’ perceptions of corruption and electoral response in Latin America.
Comparative Political Studies, 37, 677-703. doi:10.1177/0010414004265879
De Mesquita, B. B., Morrow, J. D., Siverson, R. M., & Smith, A. (2002). Political
institutions, policy choice and the survival of leaders. British Journal of Political
Science, 32, 559-590.
Ecker, A., Glinitzer, K., & Meyer, T. M. (2016). Corruption performance voting and
the electoral context. European Political Science Review. Retrieved from /core/
journals/european-political-science-review/article/corruption-performance-vot-
ing-and-the-electoral-context/655C33877217827725D626BAB83CEAD4
Eggers, A. C., & Fisher, A. C. (2011). Electoral accountability and the UK parlia-
mentary expenses scandal: Did voters punish corrupt MPs? (SSRN Scholarly
Paper No. ID 1931868). Rochester, NY: Social Science Research Network.
Retrieved from https://papers.ssrn.com/abstract=1931868
Fazekas, M., & Kocsis, G. (2015). Uncovering high-level corruption: Cross-national
corruption proxies using government contracting data (SSRN Scholarly Paper
No. ID 2711932). Rochester, NY: Social Science Research Network. Retrieved
from https://papers.ssrn.com/abstract=2711932
Fernández-Vázquez, P., Barberá, P., & Rivero, G. (2016). Rooting out corruption or
rooting for corruption? The heterogeneous electoral consequences of scandals.
Political Science Research and Methods, 4, 379-397. doi:10.1017/psrm.2015.8
Ferraz, C., & Finan, F. (2008). Exposing corrupt politicians: The effects of Brazil’s
publicly released audits on electoral outcomes. The Quarterly Journal of
Economics, 123, 703-745. doi:10.1162/qjec.2008.123.2.703
Fukuyama, F. (2015). Why is democracy performing so poorly? Journal of
Democracy, 26(1), 11-20. Retrieved from https://doi.org/10.1353/jod.2015.0017
Folke, O., Hirano, S., & Snyder, J. M. J. (2011). Patronage and elections in U.S. States.
American Political Science Review, 105, 567-585. doi:10.1017/S0003055411000256
Funk, C. L. (1996). The impact of scandal on candidate evaluations: An experimen-
tal test of the role of candidate traits. Political Behavior, 18, 1-24. doi:10.1007/
BF01498658
Gerring, J., & Thacker, S. C. (2004). Political institutions and corruption: The role of
unitarism and parliamentarism. British Journal of Political Science, 34, 295-330.
doi:10.1017/S0007123404000067
Geys, B. (2006). Explaining voter turnout: A review of aggregate-level research.
Electoral Studies, 25, 637-663.
Gingerich, D. W. (2014). Yesterday’s heroes, today’s villains: Ideology, corrup-
tion, and democratic performance. Journal of Theoretical Politics, 26, 249-282.
doi:10.1177/0951629813495120
Golden, M. (2003). Electoral connections: The effects of the personal vote on politi-
cal patronage, bureaucracy and legislation in postwar Italy. British Journal of
Political Science, 33, 189-212. doi:10.1017/S0007123403000085
Hainmueller, J., Hangartner, D., & Yamamoto, T. (2015). Validating vignette and
conjoint survey experiments against real-world behavior. Proceedings of the
National Academy of Sciences, 112, 2395-2400.
Hale, H. E. (2015). Patronal politics: Eurasian regime dynamics in a comparative
perspective. New York, NY: Cambridge University Press.
Hirschman, A. O. (1970). Exit, voice, and loyalty: Responses to decline in firms,
organizations, and states (Vol. 25). Cambridge, MA: Harvard University Press.
Jain, A. K. (2001). Corruption: A review. Journal of Economic Surveys, 15, 71-121.
Johnston, M. (1998). What can be done about entrenched corruption? In B. Pleskovic
& J. Stiglitz (Eds.), Annual World Bank conference on development economics,
1997 (pp. 69-90). Washington, DC: World Bank.
Johnston, R. J. (2006). Is hypothetical bias universal? Validating contingent valu-
ation responses using a binding public referendum. Journal of Environmental
Economics and Management, 52, 469-481. doi:10.1016/j.jeem.2005.12.003
Karklins, R. (2005). The system made me do it: Corruption in post-communist societ-
ies. Armonk, NY: M.E. Sharpe.
Keefer, P. (2007). Clientelism, credibility, and the policy choices of young democra-
cies. American Journal of Political Science, 51(4), 804-821.
King, G., Murray, C. J., Salomon, J. A., & Tandon, A. (2004). Enhancing the validity
and cross-cultural comparability of measurement in survey research. American
Political Science Review, 98, 191-207.
Kitschelt, H., & Wilkinson, S. I. (2007). Patrons clients policies: Patterns of demo-
cratic accountability and political competition. Cambridge, UK: Cambridge
University Press.
Klašnja, M., Tucker, J. A., & Deegan-Krause, K. (2016). Pocketbook vs. Sociotropic
corruption voting. British Journal of Political Science, 46, 67-94. doi:10.1017/
S0007123414000088
Konstantinidis, I., & Xezonakis, G. (2013). Sources of tolerance towards corrupted
politicians in Greece: The role of trade offs and individual benefits. Crime, Law
and Social Change, 60, 549-563. doi:10.1007/s10611-013-9478-2
Kunicová, J., & Rose-Ackerman, S. (2005). Electoral rules and constitutional struc-
tures as constraints on corruption. British Journal of Political Science, 35,
573-606. doi:10.1017/S0007123405000311
Ledeneva, A. V. (2013). Can Russia modernise? Sistema, power networks and infor-
mal governance. Cambridge, NY: Cambridge University Press.
Lindvall, J., & Rueda, D. (2014). The insider–outsider dilemma. British Journal of
Political Science, 44, 460-475.
Maas, C. J. M., & Hox, J. J. (2005). Sufficient sample sizes for multilevel modeling.
Methodology: European Journal of Research Methods for the Behavioral and
Social Sciences, 1(3), 86-92. Retrieved from http://dx.doi.org.ezproxy.ub.gu.se/
Manzetti, L., & Wilson, C. J. (2007). Why do corrupt governments maintain pub-
lic support? Comparative Political Studies, 40, 949-970. doi:10.1177/001
0414005285759
Mungiu-Pippidi, A. (2006). Corruption: Diagnosis and treatment. Journal of

Democracy, 17(3), 86-99. doi:10.1353/jod.2006.0050
Montinola, G. R., & Jackman, R. W. (2002). Sources of corruption: A cross-
country study. British Journal of Political Science, 32(1), 147–170. Retrieved
from https://doi.org/10.1017/S0007123402000066
Nichter, S. (2008). Vote buying or turnout buying? Machine politics and the secret
ballot. American Political Science Review, 102, 19-31.
Nichter, S., & Peress, M. (2016). Request fulfilling: when citizens demand clien-
telist benefits. Comparative Political Studies, 10414016666838. Retrieved from
https://doi.org/10.1177/0010414016666838
Norris, P. (2011). Democratic deficit: Critical citizens revisited. New York, NY:
Cambridge University Press.
Organisation for Economic Co-Operation and Development. (2016). OECD regions
at a glance 2016. doi:10.1787/reg_glance-2016-en
Ostrom, E. (1998). A behavioral approach to the rational choice theory of collec-
tive action: Presidential address, American Political Science Association, 1997.
American Political Science Review, 92, 1-22. doi:10.2307/2585925
Pavão, N. (2015). The failures of electoral accountability for corruption: Brazil and
beyond (Doctoral dissertation). University of Notre Dame, IN.
Persson, A., Rothstein, B., & Teorell, J. (2013). Why anticorruption reforms fail—
Systemic corruption as a collective action problem. Governance, 26, 449-471.
doi:10.1111/j.1468-0491.2012.01604.x
Peters, J. G., & Welch, S. (1980). The effects of charges of corruption on voting behavior
in congressional elections. American Political Science Review, 74, 697-708.
Piliavsky, A. (2014). Patronage as politics in South Asia. New Delhi, India:
Cambridge University Press.
Pollock, J. K. (1937). The cost of the patronage system. Annals of the American
Academy of Political and Social Science, 189, 29-34.
Praino, R., Stockemer, D., & Moscardelli, V. G. (2013). The lingering effect of scan-
dals in Congressional elections: Incumbents, challengers, and voters. Social
Science Quarterly, 94, 1045-1061. doi:10.1111/ssqu.12046
Remmer, K. L. (2010). Political scale and electoral turnout: Evidence from
the less industrialized world. Comparative Political Studies, 43, 275-303.
doi:10.1177/0010414009352638
Rose-Ackerman, S. (1999). Corruption and government: Causes, consequences, and
reform. Cambridge, UK: Cambridge University Press.
Rothstein, B., & Teorell, J. (2008). What is quality of government? A theory of
impartial government institutions. Governance, 21, 165-190. doi:10.1111/j.1468-
0491.2008.00391.x
Rueda, D. (2005). Insider-outsider politics in industrialized democracies: The chal-
lenge to social democratic parties. American Political Science Review, 99, 61-74.
Schaffer, F. C., & Schedler, A. (2007). What is vote buying. In Elections for sale: The
causes and consequences of vote buying (pp. 17-30). Boulder, CO: Lynne Rienner.
Smets, K., & van Ham, C. (2013). The embarrassment of riches? A meta-analysis
of individual-level research on voter turnout. Electoral Studies, 32, 344-359.
doi:10.1016/j.electstud.2012.12.006
Sousa, L., & de Moriconi, M. (2013). Why voters do not throw the rascals out? A
conceptual framework for analysing electoral punishment of corruption. Crime,
Law and Social Change, 60, 471-502. doi:10.1007/s10611-013-9483-5
Stockemer, D., LaMontagne, B., & Scruggs, L. (2013). Bribes and ballots: The impact
of corruption on voter turnout in democracies. International Political Science
Review, 34, 74-90. doi:10.1177/0192512111419824
Stokes, S. C., Dunning, T., Nazareno, M., & Brusco, V. (2013). Brokers, voters,
and clientelism: The puzzle of distributive politics. Cambridge, UK: Cambridge
University Press.
Sundström, A., & Stockemer, D. (2015). Regional variation in voter turnout in
Europe: The impact of corruption perceptions. Electoral Studies, 40, 158-169.
doi:10.1016/j.electstud.2015.08.006
Sung, H.-E. (2004). Democracy and political corruption: A cross-national
comparison. Crime, Law and Social Change, 41, 179-193. doi:10.1023/
B:CRIS.0000016225.75792.02
Treisman, D. (2007). What have we learned about the causes of corruption from ten
years of cross-national empirical research? Annual Review of Political Science,
10, 211-244. doi:10.1146/annurev.polisci.10.081205.095418
Vossler, C. A., & Kerkvliet, J. (2003). A criterion validity test of the contingent valua-
tion method: Comparing hypothetical and actual voting behavior for a public ref-
erendum. Journal of Environmental Economics and Management, 45, 631-649.
Wantchekon, L. (2003). Clientelism and voting behavior: Evidence from a field
experiment in Benin. World Politics, 55, 399-422.
Welch, S., & Hibbing, J. R. (1997). The effects of charges of corruption on voting
behavior in Congressional elections, 1982-1990. The Journal of Politics, 59, 226-
239. doi:10.2307/2998224
Winters, M. S., & Weitz-Shapiro, R. (2013). Lacking information or condoning cor-
ruption: When do voters support corrupt politicians? Comparative Politics, 45,
418-436. doi:10.5129/001041513X13815259182857
Zechmeister, E. J., & Zizumbo-Colunga, D. (2013). The varying political toll of con-
cerns about corruption in good versus bad economic times. Comparative Political
Studies, 46, 1190-1218. doi:10.1177/0010414012472468
Author Biographies
Monika Bauhr is an Associate Professor at the Quality of Government Institute at the
department of political science, University of Gothenburg. She is the Principal inves-
tigator of the ANTICORRP (Anticorruption Policies Revisited: Global Trends and
European Responses to the Challenge of Corruption), a large-scale multidisciplinary
research program, funded by the European Commission. Her research is mainly con-
cerned with the link between democracy and corruption and the impact of corruption
on international redistribution and foreign aid.
Nicholas Charron is an Associate Professor in political science at Gothenburg
University in Sweden. His primary research interests are in political institutions, cor-
ruption, regional studies, and the European Union.

Insider or Outsider? Grand Corruption and Electoral Accountability

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Insider or Outsider? Grand Corruption and Electoral Accountability

Diunggah oleh

Hak Cipta:

Format Tersedia

710258

Monika Bauhr1 and Nicholas Charron1

Corruption and Democratic Accountability

respondents. We believe that simple indicators of individual-level percep-

How Grand Corruption Corrodes Electoral

Grand Corruption and Loyalty

Hypothesis 1: In response to political corruption, a willingness to vote for

Grand Corruption and Demobilization

tangible rewards. This “turnout buying” (Nichter, 2008) or “participation

Hypothesis 2: In response to political corruption, voter abstention (exit)

Grand Corruption and the Insider/Outsider Divide

Hypothesis 3: Insiders are more (less) loyal to corrupt politicians in high

Research Design, Sample, and Data

Figure 1. Proportion of “exit,” “voice,” and “loyalty” responses by country.

Question 2: Now imagine that party was involved in a corruption scandal;

this was the case, individual interpretations of this would be expected to

where voting ij is the log of the odds as a function of a set of individual-

Loyalty Voice Exit

Figure 2. Visual summary of the effects of GC on corruption voting in voter

Loyalty Voice Exit

Figure 3. Marginal interaction effects.

outsiders. However, as the level of GC increases, the likelihood of loyalty

Table 3. Summary Results of Control Variables by Outcome.

Loyalty Voice Exit

ENP = effective number of parties; LR = likelihood ratio.

expectations—less satisfaction leads to exit, and more satisfaction leads to

Checking for Robustness

altogether in response to a corruption scandal. In other words, outsiders, or

corruption, its heterogeneous effects among different segments of the popula-

Declaration of Conflicting Interests

4. While voter abstention is not necessarily associated with resignation from

Mungiu-Pippidi, A. (2006). Corruption: Diagnosis and treatment. Journal of

Anda mungkin juga menyukai