Академический Документы
Профессиональный Документы
Культура Документы
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/279296293
CITATION READS
1 31
2 authors:
All content following this page was uploaded by Isabela Mares on 29 June 2015.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
1
Isabela Mares
Columbia University
im2195@columbia.edu
Aurelian Muntean
SNSPA Bucharest
and Lucian Blaga University of Sibiu
muntean@politice.ro
Paper prepared for presentation at the Annual Meeting of the European Political Science Association,
Vienna, June 25-27, 2015. We are grateful to Daniel Corstange, Donald Green, Kimuli Kasara, George
Matu and Lauren Young for invaluable help and advice in preparing this study. All outstanding errors
remain our own.
2
Abstract
wide variation in the type of brokers deployed by politicians. This paper formulates and tests a number of
hypotheses that seek to explain the variation across localities in the types of brokers mobilized by
politicians. We examine the relationship between partisan favors and political conditions in a locality and
ability of candidates to mobilize the resources of three types of brokers: employees of the local
administration, ethnic brokers and partisan brokers. To test these propositions, we conducted a survey that
includes a battery of list experiments measuring different clientelistic strategies in 85 Romanian rural
settings. In contexts, where candidates compete on the basis of such clientelistic mobilization, candidates
rely on a variety of brokers to mobilize voters at times of elections and also to influence their political
choice at the ballot box (Stokes et al. 2013). These brokers seek to influence the choices of voters through a
variety of strategies, including the offer of gifts or money, administrative favors or through coercion or
harassment.
In a recent seminal contribution, Susan Stokes and collaborators have brought to the forefront the
importance of political brokers as intermediaries between candidates and voters (Stokes et.al. 2013). Yet
while the recent literature has recognized the importance of brokers, a number of additional questions
pertaining to the operation of such broker-mediated clientelistic relationships remain still insufficiently
understood. Most studies focus primarily on partisan brokers. However, in a variety of contexts, candidates
enlist support from other agents. One important type of brokers are employees of the state – such as
officials of local welfare administration like mayor or councilors, employees of the local city hall like tax
collectors, social workers or policemen. Other types of brokers, like former officials, party activists, NGO
members, clergymen, teachers, postmen, money lenders, informal community / ethnic leaders, or even
candidate’s family members, do not necessarily have a direct employment relation with the local state
institutions, but they are part of the electoral mechanism of brokerage used by candidates and parties. These
brokers can mobilize a sizeable group of voters on the day of elections and rely on a combination of
inducements and pressure to influence the choice made by their co-ethnics in the ballot box.
If candidates competing at elections can mobilize a variety of different types of brokers, what
factors explain the variation in the type of brokers deployed in different communities? When and under
what conditions are politicians more likely to rely on state employees as political brokers? Under what
conditions do candidates turn to ethnic intermediaries? What is the relative importance of locality level
political and economic conditions and partisan factors in explaining the variation in the incidence of
brokers across localities? Turning from analysis of variation across localities to an analysis of variation
across individuals, we will also examine a range of questions about individual level characteristics of voters
In addressing these questions, our paper seeks to contribute to two distinct literatures. First, we
seek to contribute to the rapidly growing literature on electoral clientelism. Our study attempts to explain
the systematic variation across localities in candidates’ reliance on different types of brokers. We also
examine the variation in the strategies pursued by these brokers and seek to identify the particular voters
Secondly, our study also has implications for the literature examining the development of party-
like organizations in contexts with weakly developed political parties. Our study, which seeks to
characterize the systematic variation across localities in the use of different clientelistic strategies, sheds
light on the formation of institutions that are functional substitutes to party organizations. As, such, our
study seeks to explain the variation in political institutions that are functional substitutes to party
organizations across different localities. This micro-level approach to the questions on party organization
complements existing studies on political parties that explain variation in party organizations as a result of
The empirical data used in our paper draws on two surveys conducted in rural Romanian
communities in the aftermath of the 2014 presidential election. To examine a range of illicit electoral
strategies on the basis of which candidates appealed to voters, our surveys made use of item-count
techniques or list experiments, which allow us to elicit truthful responses to highly sensitive political
questions (Kuklinski, Cobb and Gilens 1997, Blair and Imai 2012). We use a variety of such ‘lists’ to
assess both the variety of political brokers deployed by candidates and, in some cases, the strategies
The remaining paper will be organized as follows. We begin by locating our study within the
rapidly growing literature on electoral clientelism. Next, we present a number of theoretical hypotheses
about the variation in the type of brokers deployed by politicians in different localities. The following
section discusses the design of our study and of the various lists included in our surveys. Section 3
discusses the results about the variation in the incidence of non-programmatic strategies across localities,
while the next section discusses the variation in the type of voters targeted by different strategies. We
The study of electoral clientelism is a vibrant and recently growing literature in comparative
politics (Kitschelt and Wilkinson 2007, Stokes et al. 2013). In its first iterations, the literature on
(Kitschelt 2000, Kitschelt and Wilkinson 2007). In the former, candidates appealed to voters on the basis of
policy promises, in the latter, they appealed on the basis of offers of gifts, money or favors. In this early
literature that examined tradeoffs between clientelistic and programmatic promises, political clientelism
remained a highly aggregate category. An important recent development in the literature of electoral
clientelism has attempted to disaggregate among the various clientelistic strategies (Luna 2014; Gans-
Morse, Mazzuca and Nichter 2014; Mares and Petrova 2014; Mares and Young 2014). This paper seeks to
contribute to the growing literature that attempts to disaggregate clientelistic practices. We disaggregate
clientelism by examining the different types of brokers deployed by politicians and by understanding the
political factors that account for the mobilization of different types of brokers.
To understand the menu of possible brokers, we will start by considering the types of brokers that
have been documented in previous studies. The clientelistic exchange that has been investigated most
extensively has been vote-buying, understood as the offer of monetary benefits to voters by partisan
brokers (Brusco et al. 2004, Stokes 2005, Stokes et al. 2013). Stokes defines partisan brokers as “locally
embedded agents of the machine who command the knowledge of voter preferences and partisan
inclinations.” (Stokes et al. 2013: 100) Finan and Schechter describe these brokers as professional in
politics and the backbone of the election campaign who “[…] know their fellow villagers well […]” (Finan
and Schechter 2012: 867). One of the respondents in their study provided the following definition of
partisan brokers. “The operative does his work, buying the conscience of the voter with money, with
alcohol, buying his ID card, a little medicine, sugar, bread, tea and in this way he goes buying and winning
A second type of brokers that can be mobilized by candidates are state employees. While earlier
studies of electoral clientelism have neglected the importance of these brokers, state employees occupy a
central position in more recent research on clientelism (Oliveros 2013, Mares 2015). Examples of state
employees who can be deployed as brokers include policemen, city infrastructure employees, agricultural,
veterinary and forest inspectors, tax collectors or employees of the local social policy administration. These
brokers can attempt to influence the choices of voters through a combination of positive and negative
inducements. Positive inducements can refer to offers of administrative favors delivered in exchange for
political support. Such offers may include assistance with administrative matters, such as certificates of
land-ownership or various licenses for small- and medium-sized enterprises, or provision of privileged
access to social policy benefits in expectation for a vote. By contrast, negative inducements refer to the use
of threats by state employees to withhold access to welfare benefits for voters that made the incorrect
political choice.
Finally, a number of recent studies have discussed candidates’ reliance on local leaders as brokers
(Baldwin 2012, Koter 2013). Such leaders may include traditional chiefs, religious dignitaries, local
leaders, such as leaders of ethnic communities. These brokers control access to goods and services that are
valued by voters and can use this control to threaten to influence voters’ choice. As Koter has argued, the
relationship between local leaders and their followers is complex in that it can be based both on reciprocity
and on some degree of exploitation. Voters can trust and rely on their leaders but also feel trapped in their
subordinate position (Koter 2013: 193). Local ethnic leaders often possess unrivaled information not just
about the political preferences of the voters in the respective community, but also about their respective
personal histories of interaction with state authorities, their economic needs and so on. Ethnic leaders can
use selective positive and negative inducements to influence the political choice of voters or to coerce
In cumulation, existing studies present a very rich characterization of possible brokers that can be
mobilized at elections and of possible clientelistic strategies. However, with few notable exceptions, very
few studies examine how candidates select among the variety of possible brokers, or how they configure
the specific mix of brokers and strategies. The empirical strategy in most studies is to examine one type of
broker deployed by candidates. The most widely studied clientelistic exchange has broker-mediated vote-
7
buying. Studies of the deployment of state employees as brokers (such as Oliveros) have turned to state
employees, but ignored other possible clientelistic strategies that are mediated by partisan brokers. This
empirical strategy focusing on one broker at-a-time has important empirical and theoretical shortcomings.
One significant empirical shortcoming is that such approaches produce biased estimates of the magnitude
of clientelistic phenomenon. Theoretically, these approaches leave important questions still unanswered.
These questions concern the choices among different types of brokers and the political and economic
factors the variation across localities in the political brokers deployed at elections.
We take up these questions in our paper. In contrast to these studies, the starting premise of our
analysis is that candidates can deploy a variety of different brokers during elections. We start by developing
a number of hypotheses about the political conditions and partisan factors that might affect the choice of
different brokers. We discuss the relationship between political conditions in different localities and the
supply of possible brokers and their willingness to offer political support to candidates. The following
In most settings characterized by weakly developed party organizations, candidates competing for
national office cannot rely on support from local party offices during campaigns. Party organizations
employing full-time employees may exist in some large urban communities but they are rare or absent in
smaller urban and rural settlements. In these localities, candidates confront in each election the challenge to
establish organizations that can substitute for a local party branch and carry out some of its functions.
Candidates construct party organizations by deploying and combining the resources that are available in
each locality.
Mayors occupy a central position among the potential actors who can provide electoral support to
candidates competing in national elections. They preside over an extensive network of state employees who
can be mobilized at elections. This includes welfare officials, tax collectors, law and order enforcers and
the like. Mayors also have access to significant local-level political information about the political
8
attachments of various citizens. Due to their informational and logistical resources, mayors often become
the central node in the proto-partisan organization set up by candidates that compete for national elections
in each locality. Mayors differ, however, in the level of resources they can supply to candidates at elections
and in the local political constraints they encounter in attempting to mobilize these resources. Let us
The length of political incumbency is an important factor affecting the political resources that can
be mobilized by mayors. As a vast literature on political bureaucracies has established, longer political
incumbency allows mayors to appoint a higher number of loyal partisan activists in the local
administration. These members of the local administration can become an important political resource
during campaigns. Local government officials can be mobilized as brokers during elections. The interaction
by employees of local administration with citizens can become an opportunity to sway political support on
behalf of a candidate. Two such strategies are available to members of the local bureaucracy. On the one
hand, they can promise future access to certain policies or programs in exchange of voters’ support of
particular candidates. We refer to this strategy as the provision of favors. Alternatively, local state
employees can threaten to cut future access to a policy program they are administering to those voters that
support the ‘incorrect candidates’. We refer to this second strategy as welfare threat or welfare coercion.
While the length of incumbency may affect the supply of state employees that can be mobilized on
behalf of a candidate, other political conditions in the locality may constrain the ability of mayors to deploy
the employees of the local administration during elections. One such variable is the political composition of
the local city council. We conjecture that mayors that encounter a city council with a similar partisan
majority face low levels of constraint in deploying members of the local administration at elections.
(Hypothesis 1) However, we expect that the mayor will be constrained from deploying the resources of the
state for political purposes if the political majority in the local city council is of a different partisanship. In
these cases, we expect to find lower levels of deployment of state employees at elections. (Hypothesis 2)
While the previous two hypotheses conjecture that variation in political conditions across localities
affects the incidence in the mobilization of state employees as brokers, the partisanship of the mayor might
also affect his ability to engage in clientelistic mobilization. We hypothesize that parties that benefit from
national incumbency might have an advantage over opposition parties in directing resources towards
9
mayors that provide additional political support at elections. (Hypothesis 3) Incumbent parties are likely to
transfer material resources (such as budgetary transfers from national and county levels) to the co-partisan
mayors. They can also enact changes in legislation that promise rewards for mayors for particular electoral
results. One example of such legislation, encountered in Romania in September 2014, the site of our study,
is a decree that encourages the ‘migration’ of mayors to the incumbent party prior to the election, which we
will detail below. We expect to find to find a higher incidence of clientelistic mobilization among mayors
Ethnic intermediaries
State employees – such as policemen, administrators of public welfare or tax collectors – are not
the only political brokers that can be deployed by candidates during election. Local leaders of ethnic
minorities can also provide important political services during elections. Such ethnic brokers may use a mix
of coercion and inducements to persuade co-ethnics to support a particular candidate. On voting day, ethnic
leaders may bring out voters to the voting place, engaging thus in turnout buying activities.
The contact between national political campaigns and local ethnic leaders is, usually, mediated by
other political actors. Most often these include the partisan activists, such as mayors or politically appointed
members of the local administration. Local mayors can use a variety of strategies to incentivize leaders of
local ethnic communities to commit time and effort on behalf of a political candidate. On the one hand
mayors can offer policy advantages to leaders of the ethnic minority in exchange for their political support.
Such positive inducements include access to commonly-held land or access to subsidized housing for
members of the small ethnic groups. Other inducements may involve the forbearance of past irregularities
committed by the leader or members of the ethnic group or outright blackmail invoking future prosecution
of irregularities. Furthermore, this collaboration between ethnic leaders and local politicians can be an asset
Local political conditions affect the ability of national candidates to rely on political services
provided by ethnic political brokers and also the willingness of these brokers to supply electoral support.
We will examine three political hypotheses about the factors that affect the reliance of candidates on ethnic
10
leaders. First, we conjecture that long-term incumbents will find it easier to mobilize local ethnic leaders as
brokers. (Hypothesis 4) Longer political incumbency gives these mayors access to more resources that can
be offered to these brokers in exchange for political support. At the same time, a longer incumbency gives
mayors more information about past possible irregularities and a higher ability to blackmail ethnic leaders
Secondly, we conjecture that the political fragmentation in the locality by the presence of partisan
majority that is not aligned with the incumbent mayor will constraint the capacity of local mayors to forge
these electoral alliances with local brokers. (Hypothesis 5a) We conjecture that in localities with a higher
number of local veto players, the ability of mayors to offer a variety of political and legal advantages to
leaders of ethnic groups in exchange for their support at elections is more limited. (Hypothesis 5b)
Finally, we hypothesize that the partisanship of the mayor may affect ethnic based clientelistic
mobilization. Mayors that are aligned with the national incumbent party may access higher levels of
political resources, which they in turn can offer to ethnic leaders for in exchange for the clientelistic
mobilization of the ethnic minority. Expecting these favors, leaders of ethnic groups are also more likely to
exert higher levels of effort on behalf of these mayors. It follows that we should see a higher incidence of
ethnic clientelistic mobilization in localities with mayors aligned with the incumbent party. (Hypothesis 6)
Partisan brokers
In addition to state employees and ethnic leaders, candidates can also enlist political support from
partisan brokers. As discussed above, the existing literature on electoral clientelism has privileged this type
of broker. Broker-mediated clientelism has been documented in a variety of settings, ranging from
Argentina, Colombia and Mexico in Latin America, to Benin and Nigeria in Africa. Despite the significant
emphasis of the literature on broker-mediated clientelistic strategies, there is remarkably little systematic
research that examines how these strategies vary across localities. Most studies have taken an individual
level perspective, examining the voters that are targeted by these strategies.
Does the presence of partisan brokers vary in systematic ways across localities and what factors
explain this variation? Let us begin by considering the relationship between mayor incumbency and the
11
deployment of partisan brokers. Our theoretical predictions are ambiguous. On the one hand, if we consider
available resources and information in the locality, long-term incumbents have an advantage relative to
mayors with shorter incumbency in identifying and recruiting local level brokers. These considerations lead
us to expect to find a higher incidence of partisan brokers in localities with long-term incumbency. By
contrast, other studies have taken into consideration not just the abundance of resource, but the strategic
considerations of mayors, given the mix of available brokers and the relative abundance of campaign
resources, financed by the individual candidates (Mares and Petrova 2014). These studies have conjectured
that in contexts of relative abundance of state employees as brokers, candidates will refrain from the use of
vote-buying as a clientelistic strategy, because the latter strategy is costlier. The corollary of this conjecture
is that one should see a higher relative incidence of partisan brokers in localities that have experienced
Political fragmentation in a locality, in other words, a situation where the city council has a
different partisan majority than the mayor, is likely to constrain the deployment of partisan brokers. In
localities with high levels of political fragmentation, candidates are less willing to deploy partisan brokers
fearing the possible denunciation by local employees affiliated with the other campaign. Thus, we expect to
Table 1
Predictions about variation in the deployment of brokers across localities
Table 1 summarizes our theoretical predictions about the variation in the deployment of different
types of brokers across localities. Before discussing the design of our study that tests these hypotheses, we
present some brief political background on Romania, the site of our study.
To test the above hypotheses, our study examines the variation in the type of political brokers and
resulting clientelistic strategies deployed during the 2014 presidential election in Romania. Romania
presents us with a good opportunity to study the incidence and variation in clientelistic practices. Both
studies monitoring electoral irregularities and academic research on partisan politics concur that electoral
between parties and citizens, a dataset based on expert-coding of different parties in over 100 countries,
Romania is ranked as one of the East European countries that exhibits some of the highest levels of
electoral clientelism among post-communist countries (Kitschelt, Wang and Kolve 2012).
governmental associations or election observers identified the presence of a wide variety of brokers in
Romanian elections. Election monitors revealed a variety of brokers used by candidates and their parties,
members of the local electoral office, or even the incumbent mayor that offered an elector money in
exchange of the vote, former local councilors, or candidates’ family members (Pro Democratia Report
2008, 55-70; (Pro Democratia Report 2012, 45-46). Candidates used brokers to offer gifts and favors to
voters and as agents of electoral intimidation. In addition, brokers paid an important role in breaching voter
secrecy. These reports presented various ways that the secrecy of the ballot could be breached by
politicians and their brokers, especially the party representatives in the polling stations. The Pro
Democratia Association (APD) report of 2012 national elections documented cases of the use of the mobile
ballot box as a mechanism for party representatives to tell voters, especially older ones, to vote for a
13
specific candidate or party1. Similar strategies of breaching voting secrecy as documented by the same
report in the case of voters that requested the help of an outside the station person, which thus could enter
the polling booth with the voter using reasons that voters were unable to read the ballot 2. A different
strategy for breaching the voting secrecy is the so called “suveica votului” (vote shuttle). This strategy
mixes vote buying with breaches in voting secrecy breaching. More specifically the party broker obtains a
blank ballot and a voting stamp, either by stealing them or by counterfeiting 3 . Party brokers give the
stamped ballot to a client voter to place them in the ballot box. The client has to return the unstamped ballot
he/she received at the polling place to brokers. This ballot is then used for the next voter, and so on. In
exchange for taking part in this exchange, voters were reported to have received a payment. Finally another
strategy for breaching voting secrecy involves the use of a smartphones. The client voter takes a snap of the
stamped ballot in order to show it as a proof to the broker. Candidates’ use of gifts and monetary offers to
sway the decision of voters was identified as a pervasive electoral strategy, which was especially
pronounced during the 2008 and 2012 local elections (Interview Focsani March 2013, Interview Bucharest
March 2013). Finally, a recent study has documented the use of workplace intimidation by employers in
urban settings and the existence of significant variation across localities in the incidence of this coercive
These reports also note however on important temporal patterns in the incidence of electoral
clientelism in Romania. The 2012 referendum to impeach Romania’s president Traian Basescu marked a
high point for a variety of clientelistic practices, even for Romanian standards. At this time, the electoral
coalition of political parties that sought to impeach Romania’s president – which included the left wing
Social Democratic Party and the right wing National Liberal Party – used a variety of illicit strategies to
mobilize supporters and also bring turnout above the 50 percent threshold, out of the total number of
registered electors, that was necessary to impeach the president in case a positive vote. Such illicit electoral
strategies included busing, intimidation, turnout buying etc. Yet the blatant and unabashed use of a variety
1 This is a special case of voting regulated by electoral law, according to which a voter that is clinically unable to travel
to the voting station could ask for a special ballot box to be delivered at home by polling station officials in order for
him/her to cast the vote. This special box is sealed and its ballots are counted separately at the end of the elections day,
against the votes casted in the ballot box placed at the polling station
2 Pro Democrația Association 2012: 32.
3 Both the ballot and the stamp bear no security mechanism neither in the paper nor in the ink used for stamping the
ballot
14
of strategies that trespassed the provisions of the electoral law backfired. In the aftermath of the
referendum, which failed to impeach the incumbent president, anti-corruption public prosecutors began to
investigate electoral irregularities. Such investigations resulted in the prosecution of important political
Investigations of other important political leaders – such as Liviu Dragnea, the secretary general of
the Social Democratic party – were ongoing during the 2014 presidential election. Dragnea’s investigation
was concluded only in May 2015 and resulted in an indictment of prison for one year, with probationary
suspension, for the electoral irregularities conducted during the referendum (Romania Libera 15 Mai 2015).
His appeal will be judged by the High Court of Cassation and Justice of Romania. It is unambiguous, that
the shadow of investigations of irregularities committed during the 2012 referendum had a strong deterring
effect on the strategies used by Romanian politicians during the 2014 elections (Interview Bucharest March
2013).
Like previous presidential elections, the 2014 Romanian presidential election was extremely
contested. The two candidates that faced each other in the runoff were Victor Ponta – the incumbent prime
minister and leader of the Social Democratic Party – and Klaus Iohannis, an ethnically German candidate
and mayor of Sibiu, who represented a coalition of political parties on the right. The election was won by
In this election, the Social Democratic candidate made use of his power as incumbent prime
minister to affect electoral outcomes. One such decision imposed severe restrictions on the ability of
overseas Romanians to cast their ballot. This decision led to long waiting and created a mass wave of
protest that was widely publicized on social networks (Gherghina 2015; Mediafax 2014a, 2014b). To
increase his political advantage, the incumbent prime minister issued on August 2014 an emergency decree
which gave mayors, and representatives in city and county councils the ability to switch political parties
prior to the election (Mediafax 2014c, 2014d). The exception allowed a one-time only defection. In
addition, the decree opened a migration window valid for only 45 days, during which the local
representatives could decide whether to switch party or not. Such decree resulted in wide-spread party
switching of a significant number of mayors. An estimated total of 552 mayors (out of 3,180), 4607 city
councilors (out of 40,022) and 184 county councilors (out of 1,338) migrated mainly to PSD, PLR and
15
UNPR (the three governmental parties). There was not a single county where local representatives would
not migrate based on this ruling. (ExpertForum 2015). The decision was clearly an invitation granted to
The incidence of clientelistic strategies such as offers of goods and money, the offer of privileged
access to government policy in exchange of political support or coercive strategies, such as threats to
remove access to a particular policy benefits, is particularly difficult to measure using traditional survey
techniques. A variety of considerations, such as fear of retribution or social desirability bias, may prevent
respondents from providing accurate answers. To overcome these problems, many recent studies of
political clientelism have attempted to measure the incidence of clientelistic practices using list experiments
(Corstange 2008; Imai 2011; Glynn 2013; Gonzalez-Ocantos et al. 2012; Corstange 2011; Oliveros 2013;
Weitz-Shapiro and Winters 2013). Our study joins this large and vibrant literature.
List experiments differ from traditional surveys in two important respects. First, respondents are
randomly divided in a treatment and control group and presented with different versions of the survey.
Secondly, respondents are not asked to address whether a particular behavior is true, but are presented with
various lists and are asked how many items on each respective list apply to them. While respondents in the
control group are given lists that include non-sensitive items only, respondents in the treatment group are
presented with lists that include the same list of non-sensitive items presented to the control group but also
an additional sensitive item. If randomization is successful, one can back out the prevalence of the sensitive
item in the population, by computing the difference in the mean number of items reported by respondents
While the ‘principle’ of list experiments is deceptively simple, its implementation necessitates
extremely careful design. First, significant pre-testing of both the sensitive items as well as of the control
items included in the lists are necessary. Such pretesting both ensures that the wording of the questions
capture the political experiences of citizens in the communities where the studies are conducted and helps
16
improve the precisions of the estimates of the sensitive behavior. To design the questions included in our
lists, we began by conducting interviews with election observers, journalists and non-governmental
organizations. We also conducted focus groups that allowed us to identify the types of strategies used by
politicians and the brokers used at elections. We used the information obtained in the focus groups to word
We carried out two small surveys (of about 120 respondents each) to pretest both individual items
considered for inclusion in our list and several possible lists of control items. We implemented both of
these pretests in rural communities in the immediate aftermath of the presidential election. The pretests had
two different goals. First, we attempted to test among different wording for particular sensitive items.
Secondly, we attempted to improve the design of the control lists to avoid “ceiling” and “floor” effects and
to increase the confidence of our respondents in reporting truthfully the sensitive behavior.
Please think about the last presidential elections held in November 2014. I am now going to read a
few statements about the recent presidential election. For each statement, I would like to know
how many of these events happened here, in your community. You don’t have to tell me which of
When I arrived at the voting place, I saw many observers from Germany that came to see
When I arrived at the voting place, the voting place was open.
Our lists included treatments that attempted to measure the incidence of three types of brokers:
state employees, ethnic brokers and partisan agents. In the case of state-employees, our lists also attempted
to differentiate among the different strategies deployed by these brokers. In this case, we distinguished
among positive strategies (inducements) and coercive electoral strategies (welfare threats). Table 2
Table 2
List of sensitive items included in or survey
located in two counties in Southern Romania (Teleorman and Buzau). The surveys were administered by a
team of graduate students in political science and sociology from SNSPA, Bucharest, with extensive field
experience and training as enumerators. Table 1 in the Appendix presents the list of localities included in
our survey. Within these two counties, we have selected localities in order to maximize the variation along
all political variables that are conjectured to affect the mix in the type of brokers, such as the length of
incumbency, the partisanship of the mayor, the political fragmentation in the locality and the migration
status of the mayor. The localities also differ in the share of their Roma population, an ethnic minority that
Within each locality, we used quota sampling to select the respondents included in our survey. In
each locality, the distribution of respondents is similar to the distribution in the 2011 Romanian census on
covariates such as gender and age. In each locality, enumerators chose a central starting location and then
selected randomly a direction along the walk for the survey. Only one questionnaire was applied per
household and within the household, the enumerator selected the respondent with the birthday closes to the
date of the survey to participate in the survey. Enumerators avoided consecutive households, the main road
in each locality and were given special instructions to interview only respondents in their own home, but
The implementation of our list experiment was successful. In Table 3, we present summary
statistics about the distribution of respondents with different characteristics to the control and treatment
18
version of our list experiment survey questionnaire. We do not find any statistically significant difference
in the observable characteristics of respondents assigned to the control and treatment groups, respectively.
Table 3
Tests of balance results
Note: This table displays the proportion of respondents with different demographic group for different
versions of the questionnaire and the difference between them. Due to rounding, percentages may not add
up to 100%.
Results
How pervasive are the different clientelistic strategies in rural Romania? We begin by presenting
descriptive results on the mean number of list-items for respondents that received the ‘control’ and
‘treatment’ version of the questionnaire. By subtracting the mean number of items of the control group
from the mean number of items in the treatment groups, we obtain the list experiment estimate of the
magnitude of different electoral irregularities. Consider the example of our list that seeks to measure the
incidence of political favors provided by state brokers. The mean response of the respondents in our control
group is 0.98 and the mean response of respondents in our treatment group is 1.05. The difference in these
means is 7% and the t-test of this difference yield a p-value of below 0.001. We infer that the incidence of
Table 4
Estimated incidence of clientelistic strategies
Some discussion about the results in Table 4 are now in order. Our list experiments found
evidence of state brokers and ethnic brokers during the most recent presidential election in Romania. With
respect to state-brokers, we find both evidence of ‘inducements’ – offer of administrative favors and
privileged access to policy benefits in exchange for political support – and also coercive strategies. These
findings are in agreement with results of earlier studies of political clientelism in the region, which suggest
that the dominant clientelistic strategies in East European countries are strategies that deploy the resources
of the state during elections (Mares, Muntean and Petrova 2014, Mares and Young 2014). With respect to
vote buying, we do not find robust evidence of the incidence of this strategy in the entire sample. A brief
analysis of the results on vote-buying suggests that this strategy was used in localities with incumbent
mayors. In localities with long-term mayors, the mean value of the responses to the control version of the
survey is 0.97, while the mean value of the responses to the survey that includes the sensitive question is
1.00 and this difference is significant at conventional levels (p < 0.02). As a result, we will also present the
Do we find variation in the incidence of these clientelistic strategies across localities? What factors
explain this variation? We now turn to an analysis of these questions. In our empirical analysis of each of
the clientelistic strategies, we examine variation in the incidence of clientelistic strategies across localities
and across voters with different characteristics. For each of the four lists, we estimate the following model:
20
where 𝑍𝑗 = N(0, 𝜎𝑗 )
The dependent variable 𝑌𝑖 represents the list response to different lists, subscripted i for different
individuals. 𝑇𝑅𝐸𝐴𝑇𝑀𝐸𝑁𝑇𝑖 is a variable that takes the value 1 for those individuals that have received the
version of the questionnaire that includes the treatment item for workplace intimidation. 𝑍𝑗 is a battery of
locality level variables (that will be discussed below). 𝑊𝑖 is a battery of individual-level characteristics and
Our hypotheses formulated above predicted that both locality-level variables and partisan factors
affect the variation in the incidence of various non-programmatic strategies. Table 5 presents the main
locality-level variables included in our models. With respect to locality-level variables, we include a
variable INCUMBENT that takes the value 1 if the mayor has been in office for more than one term, and 0
otherwise. To control for the fragmentation of local political authority and include the variable OPPOSED
CITY COUNCIL that takes the value 1 if the city council has a different political partisanship than the
mayor.
We include several measures that control for the partisanship of the mayor, which allow us to test
for differences in the types of clientelistic strategies across parties. We include variables that code the
partisanship of mayors in Romania’s largest political parties, the Social Democratic Party (PSD MAYOR),
the National Liberal Party (PNL MAYOR) and the Democratic Party (PD MAYOR). One difficulty we
encountered in coding mayor partisanship was that during Romania’s last local election, the Social
Democratic Party and the National Liberal Party formed an electoral alliance in some but not all localities,
the SOCIAL LIBERAL UNION. This alliance was no longer in place during the 2014 presidential election.
The additional variable SOCIAL LIBERAL UNION MAYOR takes the value 1 for mayors that were
elected under this alliance. A final partisanship variable, MIGRANT MAYOR takes the value 1 for mayors
that changed partisanship immediately prior to the 2014 election, following the governmental decree issued
As discussed in Table 5, all the models reported in Tables 6-8 include a battery of individual level
controls. We include controls for the gender, age and income of the respondents and whether the
respondent is a member of Roma minority. We also include a variable that measures the strength of the
partisanship of the respondent, which takes the value 1 if the respondent “feels close to a particular party”
and 0 otherwise.
Table 5
Locality- and individual-level variables predicting incidence of clientelistic strategies
Table 6 reports the results about the variation across localities in the use of brokers who are
employees of the state. The lists included in our survey included two sensitive items that attempted to
measure two distinct electoral strategies deployed by employees of the state, which involved ‘welfare
pressure’, the threat to cut voters from welfare benefits and the provision of administrative favors in
exchange for political support. The results about the variation across localities in the use of welfare
pressure support our main hypothesis. We find a higher incidence of welfare pressure in localities with
long-term incumbents. We also find that institutional constrains at the level of the locality do not limit
mayors from deploying this coercive strategy. Model 3, which includes the full battery of partisan variables
demonstrate the existence of differences across parties in the use of coercive strategies. We find that
mayors affiliated with the National Liberal Party (PNL) were less likely to use coercive strategies. While
incumbency and partisanship remain significant predictors in the variation in the incidence of welfare
coercion, they do not explain variation across localities in the incidence of the provision of welfare favors.
None of the economic or political variables included in our model explain the variation across localities in
Table 6
Variation across localities in the incidence of state brokers
(0.064) (0.032)
PNL MAYOR X TREATMENT -0.17** 0.043
(0.084) (0.045)
PDL MAYOR 0.053 -0.027
(0.087) (0.044)
PDL MAYOR X TREATMENT -0.094 -0.025
(0.11) (0.062)
USL MAYOR -0.040 -0.031
(0.063) (0.031)
USL MAYOR X TREATMENT -0.13 0.044
(0.087) (0.047)
POPULATION (ln) 0.070 0.085 0.071 0.013 0.013 0.0064
(0.052) (0.054) (0.055) (0.026) (0.027) (0.027)
POPULATION (ln) X TREATMENT -0.080 -0.090 -0.081 0.011 0.016 0.012
(0.067) (0.070) (0.071) (0.036) (0.038) (0.039)
UNEMPLOYMENT 0.0038 0.0047 0.0042 -0.00061 -0.00062 -0.00044
(0.0035) (0.0036) (0.0036) (0.0018) (0.0018) (0.0018)
UNEMPLOYMENT X
TREATMENT -0.0040 -0.0047 -0.0034 -0.000015 0.00028 -0.00033
(0.0045) (0.0047) (0.0047) (0.0024) (0.0025) (0.0025)
ROMASHARE 0.014 -0.079 -0.090 0.18 0.18 0.16
(0.32) (0.33) (0.33) (0.15) (0.16) (0.16)
ROMASHARE X TREATMENT 0.38 0.45 0.44 0.042 0.012 0.019
(0.40) (0.42) (0.42) (0.22) (0.23) (0.23)
Individual level variables included but not reported
CONSTANT 0.74* 0.63 0.68 0.83*** 0.83*** 0.90***
(0.45) (0.46) (0.47) (0.22) (0.23) (0.24)
In Table 6, we examine variation across localities in the reliance on ethnic leaders as brokers. (In
the examination of this strategy, we restrict our sample to Roma respondents). Consistent with our
theoretical expectation, we find that the deployment of ethnic leaders as brokers is higher in localities with
long-term incumbents as compared to localities with significant political turnover of mayors. As discussed
above, the length of their term in office gives incumbents a higher ability to forge political ties with leaders
of ethnic minority groups. This glue linking the interests of mayors and leaders of ethnic groups can be
based on favorable policy decisions, granted to the leaders of the ethnic group, but also forbearance,
decisions of mayors not to prosecute leaders or members of the ethnic minority for violations of existing
regulations.
authority between the mayor and the city council. We examine whether city councils with a different
partisan majority than the mayor can constrain the reliance on ethnic leaders as brokers. We find that such
political fragmentation does not act as a political constraint. Finally, model 3 adds a variety of controls for
the partisanship of the mayor. We do not find any systematic differences among mayors of different
Table 7
The variation across localities in the reliance of ethnic brokers
The final results reported in Table 8, examine the variation in the incidence of vote-buying across
localities. The main finding is this electoral strategy was more likely to be deployed in localities with long-
term incumbents. None of the other locality-level variables, such as a city council with a different partisan
majority, the level of unemployment or the share of minority Roma voters can explain the variation in the
use of this strategy. We do not find a systematic differences in the use of vote-buying among localities with
These findings run counter to our initial conjecture that vote-buying and state coercion represent
substitute clientelistic strategies (Mares and Petrova 2014). We find that vote-buying and state coercion are
complementary strategies that are both deployed in localities with long-term incumbents. In a recent paper,
van Ham and Lindberg report similar results about the complementarity between government intimidation
and vote-buying using cross-national data of a sample of African countries (van Ham and Lindberg 2015).
One possible explanation of these findings is that incumbent mayors have an advantage in identifying and
mobilizing local brokers who can appeal to voters distributing money, goods or small favors. Higher
incumbency, we find, translates into a higher ability to establish any network of brokers, irrespective of the
Table 8
The variation across localities in the use of vote-buying
Individual-level variation
strategies. Who are the voters targeted by different clientelistic strategies? Do politicians and brokers
operating on their behalf target voters with similar characteristics using different non-programmatic
strategies or do brokers deploy different non-programmatic strategies towards different groups of the
electorate?
To examine these questions, Table 9 reports the results of the individual level variables for the
four different clientelistic strategies. For each of the strategies, we present the results of the most complete
model that includes the highest number of locality-level covariates (models 3 and 6 in Table 4 and model 3
in Table 5 and 6 respectively. We do not find evidence of differences among the voters that are targeted by
Table 9
Individual-Level results
Conclusion
Our paper advances both theoretically and empirically the analysis of clientelistic practices in
contemporary elections. We have argued for the importance of distinguishing among the different brokers
deployed by politicians and for understanding the variation across localities in the mobilization of these
brokers. The study of this variation provides us with important insights about the mechanisms and
strategies by which electoral autonomy of voters is subverted in contemporary elections. It also sheds light
on the bottom-up processes by which candidates in countries with weakly developed party organizations
Our paper has formulated and tested a variety of hypotheses conjecturing how differences in
political conditions across localities as well as partisan differences might affect the ability of local mayors
to coopt different types of brokers during elections. We have tested these propositions in a survey that uses
multiple list experiments and that was administered in the aftermath of the 2014 Romanian presidential
election. The central finding of our analysis is that the political incumbency of the mayor is a strong
predictor of the variation in the deployment of a variety of brokers during elections. We find a higher
incidence in three out of the four clientelistic strategies measured in our survey in localities with long-term
incumbents as compared to localities that have experienced recent political turnover. As compared to
localities that have experienced recent turnover, localities with incumbent mayors experience a higher
incidence in deployment of state employees as brokers, a higher incidence in use of ethnic brokers and a
higher incidence of vote-buying. By contrast, other political conditions at the locality and partisan
differences across candidates do not explain the variation in the incidence of clientelistic strategies.
33
References
Interviews
Obodariu, Gabriela – history high school teacher in Focsani, Vrancea, and president of the Pro Democratia
Papahagi, Adrian -former campaign manager for Monica Macovei, candidate for the presidential elections
Pirvulescu, Cristian – dean of the Faculty of Political Science, SNSPA, former president of the Pro
Sorescu, Adrian and Costel Popa – Ecopolis Association and former national executive directors of the Pro
Baldwin, Kate. 2013. “Why Vote with the Chief? Political Connections and Public Goods Provision in
Zambia: Why Vote with the Chief?” American Journal of Political Science 57 (4): 794-809.
Blair, Graeme, and Kosuke Imai. 2012. “Statistical Analysis of List Experiments”. Political Analysis. 20:47–
77.
Brusco, Valeria, Marcelo Nazareno, and Susan Carol Stokes. 2004. “Vote Buying in Argentina”. Latin
Corstange, Daniel. 2008. “Sensitive Questions, Truthful Answers? Modeling the List Experiment with
Corstange, Daniel. 2011. Ethnic Clientelism in the Middle East. Unpublished manuscript. Columbia
University.
Darden, Keith. 2008. “The Integrity of Corrupt States: Graft as an Informal State Institution”. Politics &
Society. 36:35–59.
34
Finan, F and Schechter, L. 2012. “Vote-buying and reciprocity”. Econometrica. (80):2, 863- 881.
Gans-Morse, Jordan, Sebastián Mazzuca, and Simeon Nichter. 2014. “Varieties of Clientelism: Machine
Gherghina, Sergiu. 2015. “The Romanian presidential election, November 2014. Electoral Studies. 38:109-
114.
Glynn, Adam N. 2013. “What Can We Learn with Statistical Truth Serum?: Design and Analysis of the List
Gonzalez-Ocantos, Ezequiel, Chad Kiewiet de Jonge, Carlos Meléndez, Javier Osorio, and David W.
Nickerson. 2012. “Vote Buying and Social Desirability Bias: Experimental Evidence from Nicaragua:
Vote Buying and Social Desirability Bias”. American Journal of Political Science. 56:202–17.
Grzymala-Busse, Anna. 2007. Rebuilding Leviathan: Party Competition and State Exploitation in Post-
Holland, Alysha and Palmer-Rubin, Bryan. 2015. “Beyond the machine: clientelistic brokers and interest-
Imai, Kosuke. 2011. “Multivariate Regression Analysis for the Item Count Technique”. Journal of the
Kitschelt, Herbert and Wilkinson, Steven. 2007. Patrons, clients and policies: patterns of democratic
Kitschelt, Herbert, Yi-ting Wang, and Kiril Kolev. 2012. Research and Dialogue on Programmatic Parties
Kitschelt, Herbert, Z. Mansfeldova, R. Markowski, and G. Toka. 1999. Post-Communist Party Systems:
Kitschelt, Herbert. 2000. “Linkages between Citizens and Politicians in Democratic Polities”. Comparative
Koter, D. 2013. “King-makers: Local leaders and ethnic politics in Africa”. World Politics, 65(2), 187-232.
35
Kuklinski, James H., Michael D. Cobb, and Martin Gilens. 1997. “Racial Attitudes and the ‘New South’”. The
Luna, Juan Pablo. 2014. Segmented Representation: Political Party Strategies in Unequal Democracies.
Mares, Isabela and Young, Lauren. 2014. “Conditionality as coercion: strategies of electoral mobilization in
rural Hungary”. Paper presented at the Political Economy Seminar, New York University, Abu Dhabi.
Mares, Isabela, and Tsveta Petrova. 2014. “Disaggregating Clientelism”. Paper presented at Duke University,
Mares, Isabela, Muntean, Aurelian and Petrova, Tsveta. 2014. “Economic intimidation in contemporary
elections: evidence from Romania and Bulgaria”. Paper presented at the Annual Conference of the
Mares, Isabela. 2015. From Open Secrets to Secret Voting Democratic Electoral Reforms and Voter
Mediafax 2014b. “Proteste in strainatate: Peste 1000 de romani nu au putut vota la Londra si au manifestat in
londra-au-manifestat-fata-ambasadei-zeci-alegatori-au-intrat-forta-ambasada-paris-protest-fata-sediului-
Mediafax 2014c. “Ordonanta de Urgenta privind migratia primarilor a fost adoptata de Guvern”
(http://www.mediafax.ro/politic/ordonanta-de-urgenta-privind-migratia-primarilor-a-fost-adoptata-de-
Mediafax 2014d. “OUG privind migratia alesilor locali, publicata in Monitorul official. Alesii pot migra doar o
Mediafax. 2014a. “Incidente la sectiile de votare din strainatate: Gaze lacrimogene la Torino si Paris”.
(http://www.mediafax.ro/social/incidente-la-sectiile-de-votare-din-strainatate-gaze-lacrimogene-la-torino-
si-paris-peste-360-000-romani-din-diaspora-au-votat-pana-la-ora-23-00-foto-video-13565758). Accessed
Oliveros, Virginia. 2013. A Working Machine: Patronage Jobs and Political Services in Argentina. PhD
Pro Democrația Association. 2008. Reguli si nereguli in alegerile locale din 2008. Bucharest.
Pro Democrația Association. 2012. Radiografia Anului Electoral 2012. Raport de Monitorizare a Alegerilor
Romania Libera. 15 mai 2015. “Liviu Dragnea, condamnat la un an de inchisoare cu suspendare in dosarul
Referendumul” (http://www.romanialibera.ro/actualitate/eveniment/liviu-dragnea-afla-vineri-verdictul-in-
Stokes, Susan Carol, Thad Dunning, Marcelo Nazareno, and Valeria Brusco. 2013. Brokers, Voters, and
Clientelism: The Puzzle of Distributive Politics. New York, N.Y: Cambridge University Press.
Stokes, Susan Carol. 2005. “Perverse Accountability: A Formal Model of Machine Politics with Evidence
Tavits, Margit. 2014. Postcommunist democracies and party organizations. New York: Cambridge University
Press.
Van Ham, Carolien, and Staffan I. Lindberg. 2015. “From Sticks to Carrots: Electoral Manipulation in Africa,
Vicente, Pedro. C., and Leonard Wantchekon. 2009. “Clientelism and Vote Buying: Lessons from Field
Winters, Matthew S., and Rebecca Weitz-Shapiro. 2013. “Lacking Information or Condoning Corruption:
Ziare.com. 2012. “Se pregătește fraudarea referendumului? Anchete în Vaslui, dosare penale în Gorj” .
(http://www.ziare.com/basescu/referendum-demitere/se-pregateste-fraudarea-referendumului-anchete-in-
Appendix
List of localities included in the study