According to democratic theory, elections give citizens the opportunity to select “good” and punish “bad” politicians (Przeworski et al. Reference Przeworski, Stokes and Manin1999). Good politicians invest in the development of public goods and carry out policy projects that serve the interests of a large number of voters in the constituency. In contrast, bad politicians exploit illicit strategies to mobilize voters and improve their electoral performance. Bad politicians, however, come in a variety of shades. Some intimidate or blackmail voters to influence their electoral behavior, while others offer money or gifts in exchange for political support.
These considerations complicate the stylized dichotomy between “good” and “bad” politicians. In many elections around the world, voters do not have the opportunity to choose between “good” and “bad” candidates, but rather between politicians who vary in their mix of desirable and undesirable attributes. For example, voters may have to choose between an incumbent who has invested in the development of roads and schools in her municipality but who also engages in voter intimidation and a challenger who promises to invest in public goods but who offers money to citizens in exchange for their votes. Alternatively, voters may have to choose between a politician of modest political means who is currently being investigated by anti-corruption authorities and an extremely wealthy candidate who does not face any accusations of prior political malfeasance.
How do voters choose between politicians who differ in their mix of desirable and undesirable attributes? If voters evaluate candidates with several undesirable characteristics, which of these attributes will they punish with greatest intensity? Do different voters have different tolerance thresholds across illicit strategies and, if so, what factors affect the relative location of these thresholds?Footnote 1
We examine these questions by making use of a conjoint experiment, a research strategy that allows us to differentiate among the different types of candidate attributes and simultaneously examine how citizens evaluate different candidates who vary in their mix of “good” and “bad” characteristics. We carried out our study in Romania, a country that has experienced both high levels of corruption during its political transition from socialism and an extremely robust prosecution of corruption by judicial institutions in recent years. Our design allows us to vary a large number of candidate attributes and to consider the causal effect of each characteristic on voters' electoral choices. Thus, we evaluate how citizens make trade-offs among different illicit strategies.
In recent years, a number of studies have begun to examine whether and how voters punish the corrupt behavior of politicians. Scholars have attempted to advance the study of this question by using both observational data and quasi-experimental or experimental approaches. The dominant question investigated in these studies is whether (or not) voters penalize illicit electoral activities.
Our paper contributes to this growing body of research in two ways. First, we theoretically and analytically disaggregate different electoral unlawful strategies employed by politicians. Focusing in particular on electoral irregularities, we examine whether the “tolerance threshold” for illicit electoral strategies differs across the forms of political malfeasance and different types of voters. To ground our theoretical expectations, we begin by describing the factors that affect the magnitude of citizens' tolerance thresholds. We conjecture that two factors affect this variation. The first is normative considerations about the magnitude of infringement on voter autonomy.Footnote 2 Drawing on normative theory, we argue that voters punish coercive electoral strategies more severely than strategies premised on positive electoral inducements, because the former exchanges violate voters' electoral autonomy more. While voters subjected to electoral coercion lack the freedom to turn down the offer made by the broker, voters targeted with positive inducements have higher degrees of freedom or autonomy to turn down offers made to them during an election. In other words, not accepting a positive inducement implies keeping the status quo, meanwhile turning down a negative inducement means getting worse than the status quo.
Second, voters' disutility toward different non-programmatic strategies is also affected by the availability of information about the pervasiveness of the illicit activities in their respective society. Such information is disseminated by different prosecutorial agencies with well-defined legal mandates to combat and eradicate illicit behavior. In combination, these two factors affect the location of voters' tolerance thresholds for different unlawful electoral activities. We expect voters to penalize illicit electoral activities that violate their autonomy more severely and that they will rely on information disseminated by anti-corruption agencies about the pervasiveness of these irregularities.
We find evidence that voters' tolerance thresholds differ across the forms of electoral malfeasance. Keeping the resources used for clientelistic mobilization constant, we show that voters punish candidates who engage in coercive electoral activities more severely than candidates who offer positive inducements. We also find that voters penalize illicit electoral strategies that have been vigorously combatted by anti-corruption agencies more severely than strategies that have slipped under prosecutors' radar. We also document the existence of heterogeneity across voters in their tolerance for illicit strategies (see Appendix J). In line with previous research, we find that lower-income voters have lower tolerance thresholds for certain unlawful electoral strategies when compared with higher-income voters (Weitz-Shapiro Reference Weitz-Shapiro2012Reference Weitz-Shapiro2014).
1. Do voters punish illicit electoral activities?
In recent years, a rapidly growing body of literature has examined whether voters punish candidates who engage in illicit political activities. However, one important limitation of existing studies is a by-product of their reluctance to disaggregate the phenomenon of electoral corruption, which comes in a wide variety of forms. Electoral corruption is, in itself, a multidimensional phenomenon (Mares and Young Reference Mares and Young2016a): candidates can offer money or goods to voters to influence their electoral choice (Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013), condition access to state resources in exchange for political support (Wilkinson Reference Wilkinson, Herbert and Wilkinson2007; Weitz-Shapiro Reference Weitz-Shapiro2012), or seek to influence voter decisions through intimidation or threats (Mares and Young Reference Mares and Young2016a). Gonzalez-Ocantos et al. (Reference Gonzalez-Ocantos, De Jonge, Meléndez, Osorio and Nickerson2012) echo our concern about the limited understanding of the variation in voters' evaluation of different illicit electoral strategies. As they conclude: “in the hypothetical vote-buying exchanges described in our survey items reflect a fairly narrow view of how such exchanges take place in practice, since vote buying is often part of long-standing clientelistic relationships rather than a one-off transaction. In addition, our experimental vignettes ask voters to think about a very specific form of clientelism, i.e. the dispensation of money for votes during electoral campaigns. More research is needed in order to fully understand the determinants of stigma under different electoral circumstances. […] Do other forms of targeted distributive politics such as patronage, the provision of club goods or cash transfer programs exhibit the same level of disapproval?” (219).
Previous research has shown that voters punish different illicit activities differently. Fernandez-Vazquez et al. (Reference Fernández-Vázquez, Barberá and Rivero2016) show that voters in Spain ignore corruption when it yields positive effects and punish it when there is no associated compensation. For example, an incumbent can decide to not enforce regulations on local businesses because of a bribe, which might increase economic activity in the constituency. Voters will evaluate this illicit activity in a different way that if it does not bring about an improvement of economic conditions. In a similar vein, Weschle (Reference Weschle2016) finds that voters in India differently evaluate politicians who used money to personally enrich themselves and buy votes. Evidence from their survey experiment shows that electoral corruption is less punished than political corruption. Similarly, Klašnja et al. (Reference Klašnja, Lupu and Tucker2017) show that corrupt politicians who bring jobs to their constituencies are less punished by voters. What explains these differential punishments of corrupt politicians? The existence of an implicit exchange (voters are gaining from politicians’ corruption) and the credibility of the information seem to be important in answering this question (Muñoz et al. Reference Muñoz, Anduiza and Gallego2016).
Our study contributes to this ongoing comparative examination of the variation in voter evaluation of illicit activities by focusing on the specific case of electoral corruption. We share the observation of Gonzalez-Ocantos et al. (Reference Gonzalez-Ocantos, De Jonge, Meléndez, Osorio and Nickerson2012) that we still know remarkably little about how voters assess different forms of electoral corruption and their willingness to penalize certain irregularities more stringently than others. Our paper proposes an explanation for the factors that may account for the variation in the differential punishment of illicit electoral strategies by voters. We also examine whether voters with certain individual characteristics differ in their evaluation of different illicit strategies.
2. Electoral corruption in Romania
Romania provides us a rich opportunity to examine how voters assess different forms of electoral malfeasance. Romania's political and economic transition out of communism went hand in hand with the proliferation of a wide range of illicit activities that originated with or were backed by elected politicians. At the same time, Romanian elections have been characterized by high levels of electoral irregularities and, more significantly for our paper, by a wealth of clientelistic strategies (Kitschelt et al. Reference Kitschelt, Wang and Kolev2012). Both international organizations monitoring elections (such as the Organization for Security and Cooperation in Europe) and domestic non-governmental organizations (such as ProDemocratia or RomaniaCurata) have documented these unlawful electoral practices. Recent studies have documented how candidates have engaged in a variety of electoral irregularities in addition to vote-buying. These include the provision of privileged access to policy resources to voters in exchange for political support at the ballot box: a long-standing clientelistic exchange rather than a one-time transaction (Stokes Reference Stokes, Stokes and Boix2007). Candidates have also made use of coercive electoral strategies, including intimidation, blackmail, and economic coercion—threats to cut-off voters' access to policy resources they control (Mares and Young Reference Mares and Young2019). The rich menu of illicit electoral strategies present in the Romanian context provides us with a unique opportunity to examine whether voters penalize particular forms of electoral irregularities more severely than others.
In recent years, Romania has experienced an increase in the number of initiatives combatting electoral corruption. This offensive against illicit activities has been spearheaded by a new generation of prosecutors and legal experts appointed to institutions that had been established to monitor corruption but had previously lain dormant. The two most decisive institutions in this offensive against corruption have been the National Anticorruption Directorate (Directia Nationala Anticoruptie) and the National Agency for Integrity (Agentia Nationala de Integritate). Both have been instrumental in prosecuting cases of illicit activities, leading to the investigation and prosecution of a high number of elected politicians, including Romania's former prime minister, Adrian Nāstase. Only a few weeks prior to the fielding of our survey, the National Corruption Agency indicted Liviu Dragnea, the president of Romania's Social Democratic Party and one of the country's most powerful politicians, on a series of charges of electoral corruption carried out during a recent referendum.
We fielded our study prior to Romania's 2016 local election, at a time when the activities carried out by Romania's anticorruption authorities were in full swing. More significantly for our study, anti-corruption agencies were actively prosecuting mayors of many Romanian communities for corruption offenses. Several of the mayors investigated by anti-corruption agencies chose not to run for office. However, a total of 277 mayors who had been actively investigated by Romania's anti-corruption authorities decided to compete during the June 2016 local elections (Stoianovici Reference Stoianovici2016).Footnote 3 These conditions provide an ideal setting for examining how voters evaluate different attributes before making their decision to cast their mayoral ballot.
3. Theoretical expectations
Our study aims to understand how citizens evaluate candidates who have engaged in different forms of electoral clientelism. In contrast to existing studies that examine whether citizens penalize political malfeasance, we seek to disaggregate several forms of political malfeasance, in order to understand which illicit activity is punished most severely. Our research design allows us to randomize both positive attributes of candidates (such as their policy promises) as well as a variety of negative characteristics that measure different unlawful electoral activities.
This paper focuses on multiple dimensions of electoral corruption: vote-buying (Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013), the offer of privileged access to policy benefits in exchange for political support (Weitz-Shapiro Reference Weitz-Shapiro2014), or coercion or intimidation (Mares and Young Reference Mares and Young2016a). As will be discussed below, we consider these three types of electoral irregularities carried out by candidates and investigate whether voters punish these irregularities differently. The profiles of the candidates presented to respondents include other personal characteristics (such as gender, income, or experience in office), one additional attribute measuring the policy promises made by the candidate, and a set of attributes assessing the political integrity of the candidate.Footnote 4 Altogether, our survey asks respondents to evaluate candidates along seven dimensions.
Candidates can employ a variety of illicit strategies to influence voters' electoral choices. In analyzing this variation, we draw on a recent classification of forms of electoral clientelism proposed by Mares and Young (Reference Mares and Young2019). Their study classifies unlawful electoral activities by distinguishing between “positive” and “negative” inducements. Positive inducements involve offers of rewards, such as goods, money, or favors in exchange for political support. Negative inducements or coercion include threats of economic or physical sanctions for an individual's voting behavior. Coercive electoral strategies include threats to cut-off a voter's income stream (such as access to social policy benefits) or to worsen the terms of an ongoing economic exchange with citizens. Building on the typology proposed by Mares and Young (Reference Mares and Young2016a), we further distinguish within positive inducements by focusing on the nature of the good offered as part of the exchange. Candidates can provide not only money or goods to voters, but welfare favors, such as the offer of privileged access to policy benefits, as well.
Do voters punish different electoral irregularities differently? If so, what explains this variation? We will refer to citizens' acceptance of different forms of electoral corruption as their “tolerance threshold” and seek to understand the factors that affect citizens' willingness to punish different illicit strategies. Voters' tolerance thresholds, we conjecture, result from the combination of two considerations. The first are normative considerations about the severity of the infringement on voters' autonomy resulting from the candidate's illicit strategy. In formulating our expectations, we will focus on ongoing debates in political theory around clarifying normative expectations about the undesirability of different unlawful strategies. The second factor affecting voters' willingness to punish illicit strategies is information about the pervasiveness of these unlawful strategies in their respective society. In countries that experience pervasive levels of electoral corruption, such information is usually disseminated by independent agencies with a mandate to combat malfeasance. Such information, we conjecture, affects voters' willingness to punish particular illicit strategies.
How do normative considerations affect voters' tolerance thresholds toward different illicit strategies? One important criterion used by political theorists to evaluate the normative undesirability of different clientelistic strategies concerns the magnitude of the infringements on voters' electoral autonomy (Wertheimer Reference Wertheimer1987; Buchstein Reference Buchstein2000). One important difference between clientelistic strategies premised on positive or negative inducements arises from their effects on a voter's baseline condition—that is, his circumstances if he rejects the broker's offer. When a broker presents a vote-buying proposal—“If you vote for me, you get X, and if you don't, you get nothing”—a voter can either accept it or turn it down. Turning down the offer does not affect the voter's baseline condition. Coercion, however, can be used to push a voter into accepting a deal he otherwise would not consent to (Wertheimer Reference Wertheimer1987). For example, threatening voters with a reduction of their social policy benefits if they do not support a particular candidate can be conceptualized as reducing the attractiveness of the status quo. Strategies premised on coercion therefore have more severe normative implications than offers of positive inducements (Mares and Young Reference Mares and Young2016a).
Positive inducements and coercion vary in the autonomy exercised by the voter targeted by the transaction. Voters who are offered positive inducements have the freedom to opt in or out of vote-buying exchanges (see also Bratton, Reference Bratton2008). In contrast, voters subjected to coercive strategies retain very low political autonomy. In discussing the lack of individual autonomy for voters “coerced by the landlord or their employer” James Scott referred to these voters as “locked-in electorates” (Reference Scott1972, 98). This type of voter is “connected to the larger political system through his agent-patron whose control over his political will was a function of his control over his means of subsistence” (Scott Reference Scott1972, 98). Similarly, Carl Landé discusses the relationship between coerced voters and their patrons, noting that it is “made clear to such voters that he remains perpetually a debtor” despite their “continuous display of deference and obedience” (Landé Reference Lande and Schmidt1977, XXVII).
The implication of the above discussion is that coercive strategies more greatly infringe voters' autonomy than strategies premised on positive inducements. As a result, clientelistic strategies based on coercion are normatively less desirable than clientelistic strategies premised on positive inducements (Buchstein Reference Buchstein2000). The higher normative undesirability of coercion relative to positive inducements is based on the worsening of voters' baseline condition and on the low autonomy retained by voters subjected to coercive strategies. Column 2 in Table 1 summarizes the expectations we derive from normative theory.
Table 1. Voters' tolerance thresholds toward different illicit strategies
In addition to these normative considerations, voters' tolerance thresholds toward different illicit activities are also affected by the availability political information about the pervasiveness of these strategies. The main sources of voters' political information about unlawful activities are anti-corruption campaigns carried out by judicial agencies with a clear political mandate to combat and reduce overall levels of corruption. Such information is disseminated through the prosecution of politicians engaged in these illicit activities, media campaigns, the periodic audit of the activities of these anti-corruption agencies in the parliament, and so on. One can think of these campaigns as informational shocks that may improve the capacity of voters to hold candidates accountable. While such campaigns combatting electoral corruption are a common occurrence in many countries, one of their remarkable features is their selective focus on certain irregularities and their neglect of others. Prosecutors do not denounce all irregularities that may have occurred in an election, but rather “zoom in” on certain unlawful activities that may be easier to detect or prosecute or that are believed to be more widespread and, hence, to pose a greater threat to political accountability. Yet it is often the case that some irregularities remain in the “blind spot” of prosecutors investigating illicit activities.Footnote 5
Recent initiatives carried out by Romanian anti-corruption authorities are characterized by this type of selective indictment of electoral irregularities. Both Romanian anti-corruption agencies and civil society organizations have strongly targeted vote-buying practices, which were extremely pervasive in Romania during the first post-communist elections. Vote-buying has received an extremely high level of public exposure, regularly occupying the headlines of major newspapers and being repeatedly used as a charge against politicians. In contrast, other unlawful electoral strategies have remained in the “blind-spot” of Romanian anti-corruption agencies. Electoral irregularities that politicized state policy resources for electoral purposes (such as welfare favors or welfare coercion) have not been widely publicized, or extensively prosecuted.
In our study, we cannot estimate the consequences of such campaigns on voters' perceptions of illicit strategies, since we lack systematic data on voter views about these different forms of electoral corruption prior to the onset of these campaigns. We can take advantage, however, of the asymmetry in the public indictment of different unlawful strategies in the Romanian context. We conjecture that this differential indictment of electoral irregularities is likely to affect voters' tolerance threshold for different unlawful activities. The high level of public prosecution of vote-buying—which has resulted in the indictment of prominent Romanian politicians—has altered public perceptions about the pervasiveness of vote-buying and has lowered voter tolerance toward this nonprogrammatic strategy.
How do normative considerations interact with information about the pervasiveness of different irregularities to affect voters' tolerance thresholds for illicit activities? The first unlawful electoral strategy considered in our analysis is “vote-buying,” the provision of in-kind goods or money in exchange for political support. In the Romanian context, the amount of goods provided as part of this political exchange is usually very small. Mares and Young (Reference Mares and Young2019) document, using survey-based and qualitative evidence, that voters (especially in rural communities) are routinely offered buckets of flour or sugar and other small in-kind transfers. If candidates offer money to citizens in exchange for political support, the reported amount of the bribe does not exceed 100 RON (or 25 USD). Normative considerations suggest that vote-buying infringes on the autonomy of voters less severely than coercive electoral strategies, where politicians withhold a significant source of income from voters for making the incorrect political choice, and in particular for lower-income voters. Thus, we conjecture that voters' normative disutility for vote-buying is lower than the disutility for non-programmatic strategies premised on coercion. However, vote-buying has been at the center of prosecutorial initiatives and is routinely named as the main evil affecting the quality of Romania's democracy. We conjecture that sustained public indictment of vote-buying has lowered voters' tolerance threshold for this illicit strategy. As a result, we expect to find high levels of punishment of vote-buying in Romania.
A second unlawful strategy considered in our paper consists of the provision of long-term access to social policies in exchange for political support. We refer to this strategy as the provision of “welfare favors.” Beginning in 2001, Romanian social policy legislation has given mayors the prerogative to distribute social assistance benefits to low-income citizens (Law 416/2001). Mayors have taken advantage of this opportunity and have systematically conditioned access to benefits on political support at the ballot box. In contrast to vote-buying, the provision of welfare favors has not been publicized or incriminated in the existing campaigns. As a result, we argue that voters are less likely to punish this electoral irregularity than vote-buying.
A third illicit electoral strategy we are considering in our study is “welfare coercion,” the threat to cut-off future access to social policy benefits for voters who support undesirable candidates. A recent study of the use of welfare coercion in rural localities during Romania's last presidential election (held in 2014), which makes use of non-obtrusive survey techniques to detect this irregularity, finds that 11 percent of respondents had experienced welfare coercion (Mares and Young Reference Mares and Young2019). However, similar to welfare favors, welfare coercion has been overlooked by Romanian anti-corruption authorities. Our hypotheses allow us to make predictions about the differential evaluation of welfare favors relative to welfare coercion. Neither of these nonprogrammatic strategies has been subjected to vigorous prosecution. However, coercive strategies are normatively less desirable than strategies premised on positive inducements. As a result, we expect voters to sanction candidates who engage in welfare coercion more strongly than candidates who promise welfare favors.
Table 1 summarizes the above discussion about the factors affecting voters' tolerance thresholds for different unlawful electoral strategies.Footnote 6 The main implication of the above discussion is that we expect to find variation in how voters punish different illicit electoral practices. In the Romanian context, we expect the tolerance threshold for vote-buying to be the lowest, due to the high intensity of ongoing political prosecution of this unlawful strategy and the overarching perception that vote-buying poses a severe threat to the quality of Romania's democracy. We expect voters to penalize illicit electoral strategies premised on welfare favors with less stringency, as compared with vote-buying. We also expect voters to have a lower tolerance threshold for welfare coercion relative to welfare favors, due to the high normative disutility of this strategy.
To benchmark these different unlawful electoral strategies, our survey includes one additional candidate attribute that conveys information about the indictment of the candidate by the National Agency for Anticorruption or about his/her conviction for corruption by the courts. As discussed above, this is not an unrealistic scenario: over 270 mayors who competed in the 2016 local elections were under indictment by anti-corruption agencies. By including this attribute in our study, we seek to measure the differential impact of prosecution versus conviction for electoral corruption on voter decisions. Fisman and Golden (Reference Fisman and Golden2017) argue that courts and a strong and independent judiciary may be one of the most effective mechanisms for improving political accountability. Our study seeks to provide an empirical assessment of such effects.
4. Research design
Romania provides a fruitful opportunity to examine how voters evaluate different illicit activities, due to the presence of multiple forms of electoral corruption. We fielded our study prior to the local election held in Romania on 5 June 2016. The survey was administered as a face-to-face, in-person survey to 502 respondents between April and May 2016.Footnote 7 The survey was administered in paper format, the conjoint experiment embedded at the end and the candidates' attributes randomized in advance of implementation. We followed a non-probabilistic sampling method that had the following two steps: first, we selected 30 urban and rural localities located in southern Romania, including Bucharest. Within each locality, we followed a random walk strategy to select households. This strategy generated a sample that had very similar characteristics to the 2011 census in terms of gender, age, rurality, and education.Footnote 8
The goal of the conjoint experiment is to examine the effects of multiple candidate attributes on voters' electoral choices (see Franchino and Zucchini Reference Franchino and Zucchini2015, Carnes and Lupu Reference Carnes and Lupu2016, Visconti Reference Visconti2018). Respondents were shown profiles of two candidates who differed across seven attributes. These characteristics included salient information considered by voters at the moment they cast their ballot. A conjoint experiment attempts to mimic real-world decisions, when voters evaluate multiple attributes at the same time. Hainmueller et al. (Reference Hainmueller, Hangartner and Yamamoto2015) compare the results of a conjoint experiment with behavioral data from a natural experiment. They show that the conjoint analysis and the behavioral benchmark generate very similar results, which provides evidence of the external validity of this type of design. Finally, conjoint experiments may reduce social desirability bias. Given that voters need to make decisions while observing multiple attributes, it is not possible to individually identify which characteristic determines their final decision (Wallander Reference Wallander2009, Hainmueller et al. Reference Hainmueller, Hopkins and Yamamoto2013).Footnote 9
Respondents were given the following short introduction:
With your help, we seek to understand the political preferences of Romanian voters. We will present you with profiles of people who could compete for the position of mayor in your locality. Each time, we will present two possible candidates and some information about them. For each of the two candidates, we ask you to tell us which one you would vote for mayor. We will repeat this question five time with five different pairs of candidates.
Our design varied the following seven candidate attributes: political experience, illicit activities based on positive inducements (vote-buying and welfare favors), unlawful activities based on negative inducements (welfare coercion), investigation by anti-corruption authorities, policy promises,Footnote 10 gender, and income. Respondents were exposed to short vignettes that described each of these attributes. For example, in the case of the attribute value “100 RON” the full vignette specified: “Candidate is offering 100 RON in exchange for the vote.”
Table 2 presents the full set of attributes that we considered in the generation of candidate profiles. In this table we provide a simplified version of the attribute values (full vignettes are available in the supplementary appendix). In Table 3, we provide an example of a randomly generated profile that includes the full vignettes for each attribute as they were presented to the participants.
Table 2. Candidate profile
Table 3. Example of a pair of candidates
A conjoint experiment is a well-suited strategy for the purpose of our paper, as it allows us to identify, measure, and compare the independent effects of various candidate characteristics in a single experiment (Hainmueller et al. Reference Hainmueller, Hopkins and Yamamoto2013). This is achieved through a choice-based design in which respondents are asked to choose among hypothetical candidates who closely represent real politicians. After participants had evaluated both candidates, they were asked to select one of the two candidates for mayor. The question was worded as: “For which of these candidates would you vote in the upcoming elections? (1) Candidate 1; (2) Candidate 2.”
Following Hainmueller et al. (Reference Hainmueller, Hopkins and Yamamoto2013), we estimate all equations using ordinary least squares with standard errors clustered at the respondent level. The unit of analysis is a candidate profile, and the outcome is a binary indicator where 1 means that the candidate was selected as mayor (from a pair of candidates), while a 0 means that the candidate was not selected as mayor (from a pair of candidates). We regress the outcome on the set of candidate attributes. The estimates represent the effects of each attribute value on the probability of being preferred as mayor (when compared with a reference category).Footnote 11
In the case of positive and negative inducements, we use “neutral” attribute values as reference categories to be able to make comparisons between both groups of attributes. For example in the case of positive inducement, the reference category is “no particularistic offer,” and in the case of negative inducement, the reference category is “no threat.” Therefore, we can compare welfare coercion with welfare favors because both have a “neutral” reference category.
5. Results
We compare the socioeconomic characteristics of the respondents in our sample to census data. We show that the proportions of female, rural, and educated voters in our sample are accurate representations of the country according to the most recent census conducted in 2011 (see Appendix B).
Figure 1 displays the effects of each attribute value on the probability of being selected by voters. Recall that the coefficients provide the average difference between a particular attribute value and a reference attribute value. The dots represent point estimates, and the lines 95 percent confidence intervals. Meanwhile, the dots without confidence intervals represent the reference categories.
Figure 1. Evaluation of candidates who differ in their mix of “good” and “bad” attributes.
Do voters penalize candidates who have committed electoral irregularities? Do they impose different punishments for different irregularities and if so, which irregularity is punished with the greatest intensity? Consistent with our main hypothesis, we find evidence that voters penalize different forms of electoral malfeasance differently. The electoral irregularity punished most stringently is vote-buying. Respondents are 20 percentage points less likely to vote for candidates who offer money to voters in exchange for their vote. A plausible explanation for this pattern is that the existence of large-scale campaigns to combat vote-buying in Romania.Footnote 12
In contrast, voters are less likely to punish candidates who seek the electoral support of voters by politicizing state resources. Candidate use of clientelistic strategies that promise voters access to social assistance benefits in exchange for political support is only likely to decrease the probability of electoral support by eight percentage points. We attribute this difference to the higher levels of public indictment of vote-buying by the media and civil society organizations in Romania in recent years.
Finally, we find that the use of welfare coercion is punished more stringently than the use of welfare favors but less stringently than vote-buying. This illicit strategy is punished by 14 percentage points. One possible interpretation of this result is that voters find coercion normatively more undesirable than positive inducements. The comparison between the use of welfare favors and the use of welfare coercion speaks to this differential punishment of two strategies that politicize the same resource—in this case, social policy—but that do so with different strategies (positive versus negative inducements).
In addition to these clientelistic strategies, our survey also examined how voters evaluate candidates who are investigated or prosecuted by legal and political institutions involved in combatting irregularities. Our results indicate that the current process of public indictment of irregularities lowers political support for these candidates.
We find systematic differences in the ways in which voters evaluate candidates who are investigated for illicit behavior as compared with those candidates who are indicted. The investigation of a candidate by an anti-corruption agency is likely to lower support for the candidate by 18 percentage points. In contrast, the sentencing of a candidate reduces the likelihood of support by 33 percentage points.
We find that respondents reward candidates who make promises about public goods, but these effects are not particularly large in magnitude. The promise to renovate schools increases the probability of support by only four percentage points (over candidates who do not make these promises). The promise to renovate schools and improve roads is likely to improve a candidate's probability of being elected as mayor by only five percentage points. Two factors that may explain the small impact of these promises are the low levels of fiscal resources available to Romanian mayors and the harsh climate of economic austerity of the last decade.
We also find that voters reward candidates with previous experience as mayor as compared with candidates who lack this experience. We interpret this result as a proxy of a positive evaluation of political experience (Kirkland and Coppock Reference Kirkland and Coppock2016).
Other individual-level attributes such as gender or income have no statistically significant effect on the probability of being selected as mayor. Our results concerning the absence of an income effect is consistent with the findings of Carnes and Lupu (Reference Carnes and Lupu2016) in Argentina, Britain, and the United States.Footnote 13
These findings lend support to our conjectures about variation in the normative evaluation of different nonprogrammatic strategies. The comparison between two nonprogrammatic strategies that use state resources—welfare favors and welfare coercion—demonstrates that coercive strategies are less desirable than strategies premised on positive inducements, a result that conforms with expectations derived from normative theory. Our results illustrate that Romanian voters find clientelistic strategies that involve vote-buying the least attractive of the nonprogrammatic strategies. This aversion to vote-buying is to be expected in a country where anti-corruption agencies have waged a decade-long effort to eradicate this electoral evil and where a number of politicians have been indicted for this political irregularity. Consistent with our explanation, the differences in voters' punishment thresholds for different illicit activities can be explained by normative considerations and informational campaigns about these unlawful activities.
What are the implications of these findings? These results allow us to understand how voters weigh different negative attributes of politicians. They demonstrate that voters evaluate the provision of money, the discretionary allocation of benefits in exchange for political support, and strategies premised on welfare coercion differently and punish candidates who engage in these illicit strategies. Additionally, the findings demonstrate that the presence of a decade-long effort to combat vote-buying by Romanian anti-corruption agencies has affected the way in which voters evaluate this unlawful activity. Our results are consistent with previous findings that suggested that the exposure to negative behavior raises voters' punishment thresholds for these illicit strategies (Winters and Weitz-Shapiro Reference Winters and Weitz-Shapiro2013).
Are these results generalizable? Illicit electoral practices are a common occurrence in a variety of settings, including Latin America (Calvo and Murillo Reference Calvo and Murillo2004; Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013; Weitz-Shapiro Reference Weitz-Shapiro2014; Oliveros Reference Oliveros2016), Africa (Van de Walle, Reference Van de Walle, Herbert and Wilkinson2007; Arriola Reference Arriola2009), and even in developed countries such as Italy and Japan (Kitschelt and Wilkinson Reference Kitschelt and Wilkinson2007). While these unlawful strategies are normatively undesirable, we need to understand how voters' tolerance thresholds vary across different settings. Future studies that replicate our methodology in other contexts with powerful or only incipient electoral campaigns combatting corruption are needed to document the relative position for different illicit activities in voters' normative thresholds.
6. Conclusion
Our paper contributes to a growing body of literature examining how voters evaluate and punish different types of political malfeasance. To our knowledge, this is the first study that seeks to disaggregate the dimensions of electoral corruption and examine whether voters penalize different forms of malfeasance differently. We carried out our study in Romania, a political environment characterized by the extensive use of illicit electoral strategies by politicians.
Using a conjoint design, our paper provides causal evidence that voters impose different punishments for different forms of unlawful activities. We examine the punishment imposed by voters on candidates who engage in a variety of illicit electoral strategies, including vote-buying, welfare coercion, and the provision of privileged access to social policy benefits in exchange for political support. Our analysis of the punishment of these electoral irregularities reveals that voters penalize vote-buying most severely. We attribute this result to the highly visible public campaign against this unlawful electoral strategy that has been led in Romania since 2008. Voters are also likely to strongly punish welfare coercion, a strategy premised on threats of cutting off benefits to citizens who fail to support the correct candidate. We hold that the normative considerations associated with cutting off social benefits to low-income voters play a role in explaining citizens' tolerance toward this electoral irregularity. By contrast, we find that voters penalize electoral strategies premised on the provision of welfare favors least severely. While this non-programmatic strategy remains undesirable, we attribute the lesser punishment to both the lower infringement on voter autonomy and the lower public indictment of this irregularity by Romanian authorities.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2019.12
Author ORCIDs
Isabela Mares, 0000-0002-8315-9708; Giancarlo Visconti, 0000-0002-0914-1497.
