Economic intimidation of workers by their employers was a widespread political phenomenon in all European countries in the decades following the adoption of mass suffrage (Anderson Reference Anderson2000; Mares Reference Mares2015). In Germany, the country where this phenomenon was most pervasive around the turn of the century, employers intervened at all stages of the electoral campaign – from the announcement of a candidacy to the casting of ballots. Employers waged aggressive campaigns among their employees in support of particular candidates and confiscated electoral materials from opposition candidates. On voting day, employers regimented their workers in columns, marched them to the voting place and monitored them as they cast their ballots. Employers also engaged in systematic post-electoral harassment of voters who had supported opposition candidates, reduced the wages of such employees and even dismissed workers who refused to comply with their demands (Anderson Reference Anderson2000; Baland and Robinson Reference Baland and Robinson2008; Klein Reference Klein2003). In Britain, large firms in industrial constituencies made ‘party allegiance one of their hiring terms’ (Hanham Reference Gerring1959: 68), thus ensuring the dominance of Liberal candidates in these districts. Similarly, in France, electoral intimidation by employers persisted throughout the Third Republic (Charnay Reference Charnay1964; Garrigou Reference Garrigou1992). Employers would pressure workers to vote for a particular candidate, arguing that their factory would close down if that candidate were not elected (JORF 1898: 1807). As late as 1914, employers were documented to have threatened workers with layoffs if the candidates that enjoyed the support of the enterprise were not elected (JORF 1914: 31).
Economic intimidation at times of elections – defined as strategies by which employers or firm managers seek to coerce their employees to vote for a particular party or its candidate – is not a phenomenon characteristic of only the first wave of democratization. This political practice persists today in many countries that have recently experienced democratic transitions. Consider a few examples to illustrate these practices. During the 2004 Ukrainian election, agricultural employers in Novy Sanzhary were reported to have asked employees in their enterprise to sign statements indicating their intention to vote for a particular candidate (Druzhinina Reference Druzhinina2004). During the 2012 referendum on the motion of the Romanian parliament to impeach Romania’s President Traian Băsescu, employers in several Romanian localities were reported to have seized the identity cards of their employees in an effort to prevent those individuals from voting (Adevarul.ro 2012; Stiriletvr.ro 2012; Ziare.com 2012). Even in a developed and established democracy, such as the US, for instance, employers were recently reported to have taken advantage of a Supreme Court ruling that overturned laws banning employers from directly expressing their political opinions to their employees and engaged in systematic pressuring of voters (Huffington Post 2012).
The wide geographic and temporal scope of these examples raises a number of theoretical and empirical questions. What is the importance of economic intimidation within the larger repertoire of non-programmatic mobilization strategies used by politicians at times of elections? Under what conditions is economic intimidation politically and economically attractive to employers and politicians? How pervasive is this phenomenon in contemporary elections?
Despite growing academic interest in the study of non-programmatic electoral mobilization, existing scholarship has not yet examined these questions systematically. Most studies of electoral clientelism have focused on the determinants and consequences of vote buying – the offer of monetary or in-kind benefits in exchange for political support. This emphasis has constrained the research agenda of the relevant literature. A number of more coercive electoral strategies, which involve intimidation, harassment and post-electoral punishment and which were central to the literature on historical electoral irregularities (Mares Reference Mares2015), have received significantly less attention in the literature on contemporary electoral clientelism. By overlooking these coercive clientelistic strategies, our theoretical and empirical characterization of contemporary non-programmatic mobilization remains incomplete and our empirical estimates of the magnitude of electoral clientelism remain biased.
This article seeks to address this oversight in the literature by focusing on one type of such coercive electoral mobilization: the use of private economic intimidation. We formulate a set of hypotheses about the variation in the incidence of electoral coercion by employers across localities in contemporary elections. We test our propositions by examining the incidence of this phenomenon across localities in two recent elections in two Eastern European countries, Romania and Bulgaria. Previous studies have suggested that both countries have experienced significant electoral clientelism (Holmes Reference Holmes2003; Innes Reference Innes2002; Kitschelt et al. Reference Kitschelt, Wang and Kolev2012; Kopecky Reference Kopecky2006; Kostadinova Reference Kostadinova2009; Vachudova Reference Vachudova2009). In addition, our study disaggregates the menu of clientelistic strategies that are present in these countries and examines the importance of economic intimidation in the mix of non-programmatic strategies deployed by candidates. Our study demonstrates that electoral coercion organized by employers is an important component of electoral clientelism in these two countries. Empirically, we assess its magnitude by using a list experiment, an empirical strategy that allows us to elicit truthful responses to sensitive political phenomena.
We observe significant subnational variation in the incidence of private economic intimidation at times of elections in both Romania and Bulgaria. We also show that economic intimidation co-exists with vote buying, the provision of administrative favours and electoral intimidation by employees of the state, which also vary across localities. Consistent with our theoretical expectations, we find that in both countries, economic intimidation plays a prominent role in the menu of clientelistic mobilization in localities dominated by a small number of large employers. However, employer intimidation diminishes in importance in localities with lower levels of economic concentration, where the diversification of economic employment constrains the ability of employers to use their control over the labour force for electoral purposes.
We contribute to the growing literature on electoral clientelism in two main ways. First, we disaggregate the concept of electoral clientelism. In contrast to previous studies that mostly study one non-programmatic strategy at a time (and usually focus on vote buying), we study a broader range of electoral influences that include vote buying, the provision of administrative favours by the state, and electoral coercion that can be perpetrated either by employers or by other political agents. Second, we investigate the relative importance of private economic intimidation in this overall toolkit of electoral clientelistic strategies. We develop and test a theory about the subnational variation in the use of private economic intimidation at times of elections. In the process, we contribute to the understanding of electoral clientelism in the new Eastern European democracies.
THE POSITION OF ECONOMIC INTIMIDATION IN THE PORTFOLIO OF CLIENTELISTIC STRATEGIES
The study of electoral clientelism has become a central and rapidly growing area of research in comparative politics in recent years. This work has documented that politicians rely on a variety of non-programmatic strategies to influence the choices of voters. A first distinction that can be made among the different types of electoral clientelism is one between ‘positive’ and ‘negative’ inducements of voters. Positive inducements represent efforts to sway the electoral choices of voters through transfers that bring about an increase in the overall utility of voters. By contrast, negative inducements are strategies that threaten post-electoral punishments and bring about a reduction in the overall utility of voters. While clientelistic strategies premised on positive inducements leave voters’ status quo unchanged, coercive electoral strategies threaten to worsen the status quo of voters who do not consent with the demand of the candidate or politician (Mares and Young Reference Mares and Young2016; Wertheimer Reference Wertheimer1987). Another (orthogonal) dimension along which clientelistic strategies vary concerns the identity of the agents who seek to influence the electoral choices of voters. We consider three such types of actors: employees of the state, partisan non-state brokers and private actors (see Table 1).
Classification of Non-Programmatic Strategies
Let us first consider strategies premised on positive inducements of voters. Politicians and their parties can turn to the public sector and recruit employees of the state as electoral mobilization agents and thus leverage the vast political resources of the state (Bratton and van de Walle Reference Bratton and van de Walle1997; Grzymala-Busse Reference Grzymala-Busse2007; Luna 2014; Mainwaring Reference Mainwaring1999; Mares Reference Mares2015; O’Dwyer 2004; van Biezen and Kopecky Reference van Biezen and Kopecky2007). State employees can provide administrative favours to voters, including assistance with various administrative matters, such as access to records and permits (certificates of land-ownership, licences for small and medium-sized enterprises, and the like). In addition, depending on the social and economic programmes that are distributed at the local level, the administrators of housing, unemployment, education, disability and other assistance policies can facilitate access to these services conditional upon political support for a particular candidate, even if in many countries these services are universal.
In addition to employees of the state, at times of elections, politicians can further activate members of the local partisan network outside of the public sector. Throughout this article, we will use the term ‘partisan brokers’ to refer to this type of electoral agent. The benefits these party activists can offer as positive inducements to voters are financed either by the resources of the party or by the private resources of individual candidates. These benefits can include the provision of ‘treats’ or small goods such as food, drinks or small amounts of money. While some of the other electoral strategies that we will consider in this article have remained understudied in the recent literature, clientelistic exchanges premised on vote buying have been at the centre of a vast recent literature (Stokes Reference Stokes2005; Stokes et al. Reference Stokes, Dunning, Nazareno and Brusco2013; Vicente and Wantchekon 2009).
In addition to employees of the state and partisan brokers, politicians can also deploy a variety of other brokers. One such non-partisan broker are employers. Employers can provide wage and non-wage benefits, such as holiday time, bonuses, clothing, housing or food to their employees in exchange for their support for particular candidates. In addition, firms can promise promotion to employees and future employment to other individuals who can mobilize a large number of other voters due to their crucial position in partisan or ethnic networks.
With some exceptions (Allina-Pisano Reference Allina-Pisano2010; Baldwin Reference Baldwin2013; Corstange Reference Corstange2011; Darden Reference Darden2008; Ichino Reference Ichino2008; Mares and Petrova Reference Mares and Petrova2013), the recent literature on contemporary electoral clientelism has focused primarily on the use of such positive inducements. The menu of possible non-programmatic electoral strategies is, however, much broader. It includes a variety of coercive electoral mobilization activities. Such negative inducements come most often in the form of threats of post-electoral punishments of voters who make a political choice considered undesirable by the electoral agent.
Employees of the state, such as police officers, tax collectors and other employees of the local administration can also engage in electoral intimidation. One of our respondents referred to such practices as ‘the politics of constant blackmailing and harassment of voters’ (Interview 15, Focsani, Romania, March 2013). Such state employees can be dispatched at times of elections to remind voters of their previous or current imperfect compliance with various state policies. Such acts of non-compliance may range from violations of local ordinances to the failure of small and medium-sized enterprises to obtain all necessary licences. Employees of the state can threaten voters with future prosecution or other relevant punishment if the voters support an opposition candidate.
Partisan brokers usually possess fewer resources that would allow them to engage in electoral blackmail and post-electoral punishments. These brokers can, however, use threats of physical retribution to intimidate voters. In Bulgaria, respondents reported that one of the local candidates sent locally known strongmen to patrol outside the houses of locals who can mobilize a large number of voters in order to discourage such mobilization on behalf of the opposing candidates (Interview 18, Svoge, Bulgaria, May 2013).
Finally, private actors can also impose a variety of post-electoral punishments on their fellow voters. Employers, for instance, can threaten targeted layoffs and reductions in wages or non-wage benefits for workers, including exclusion from employer-provided housing and other social benefits.
VARIATION ACROSS LOCALITIES IN THE INCIDENCE OF ECONOMIC INTIMIDATION
Economic intimidation and coercion is one of the possible strategies of non-programmatic mobilization. Due to their control over employment conditions and wages, employers have the means to exert strong and effective political pressure on their employees. The menu of coercive electoral strategies that can be deployed by employers includes threats of the loss of employment, salaries, bonuses, pensions or healthcare benefits provided by the company. For employers, the expected benefit from this political activity is the election of a ‘desirable’ political candidate who may reciprocate after the election by enacting favourable policies or by granting privileged contracts for the respective company. Economic intimidation is, however, not costless to firms. Excessive economic pressure may lead to the loss of potentially valuable employees, who may be hired by competing firms. When choosing a level of economic intimidation, employers weigh their potential political gains against their possible economic losses.
The above discussion suggests that employers’ calculations about the use of economic coercion as a non-programmatic strategy are likely to be affected by economic conditions in the locality. Such economic conditions affect the costs of this non-programmatic strategy incurred by firms. Building on existing studies (Mares and Zhu Reference Mares and Zhu2015), we hypothesize that:
Hypothesis 1: The level of economic concentration of a locality influences the costs of electoral intimidation by employers and is likely to explain the variation in the incidence of economic intimidation at elections.
As standard in the literature on industrial organization, a locality with high levels of economic concentration is one in which a small number of firms control employment and output. As the number of firms in the locality increases, the level of economic concentration of the locality declines.
Three factors lower the costs of electoral intimidation by employers in localities with higher levels of economic concentration. The first is the economic scale of the firm. Due to their scale, large employers encounter lower costs in carrying out political activities, such as control of electoral turnout or the distribution of political material on behalf of a particular candidate. Second, in such localities, workers have fewer employment opportunities outside the firm, as there are few competing firms that can rehire them. If one single company monopolizes the employment opportunities for specific occupations, employees have fewer employment opportunities in the locality, which, in turn, lowers the costs of economic coercion faced by the company. Finally, the concentration of employment in the hands of a small number of actors also reduces the possible coordination problems faced by employers in punishing workers with ‘subversive’ political views by denying them employment opportunities.
As employers in concentrated localities face relatively lower costs of electoral intimidation compared with employers in less economically concentrated localities, we expect that the former firms will be more likely to engage in economic intimidation at times of elections. Given its relative availability but also its effectiveness, such economic intimidation is likely to be more widely used by candidates in localities with high levels of electoral concentration as compared with localities with high levels of economic heterogeneity. The reverse holds for localities with high levels of economic heterogeneity where employers face higher costs of electoral intimidation and are thus less likely to resort to this strategy. As a result, we expect to find lower levels of economic intimidation in localities with higher levels of economic heterogeneity.Footnote 1
All else being equal, we expect:
Hypothesis 2: The incidence of economic intimidation should be higher in localities with concentrated economic activity and lower in localities characterized by economic heterogeneity.
The empirical analysis presented in the following part of the article seeks to test these propositions. Our goals are twofold. First, we seek to document the existence of economic intimidation in contemporary elections. Second, we seek to test the hypothesis linking the level of economic concentration and the incidence of economic intimidation. Finally, our study seeks to provide empirical estimates of the magnitude of economic intimidation when compared with other non-programmatic strategies. We leave the examination of the variation in the mixes of such non-programmatic strategies for other studies, noting, however, that such variation is likely to be affected by other political variables in addition to economic heterogeneity.
RESEARCH DESIGN
Our study examines the incidence of economic intimidation in the broader mix of clientelistic strategies in two Eastern European new democracies – Bulgaria and Romania. The literature on clientelism in Eastern Europe has primarily focused on party patronage (Dimitrov et al. Reference Dimitrov, Goetz and Wollmann2006; Ganev Reference Ganev2001; Grzymala-Busse Reference Grzymala-Busse2007; Meyer-Sahling Reference Meyer-Sahling2004; O’Dwyer 2006),Footnote 2 thus overlooking the wider range of electoral clientelism strategies and perpetuating the biases of the broader literature in comparative politics. The themes explored in this existing work, however, are in line with our general argument that the concentration of (economic) resources breeds more non-programmatic mobilization (and economic intimidation specifically). For example, the literature on party patronage has tended to agree that robust political/party competition lowers the prevalence of this electoral clientelism strategy (Grzymala-Busse Reference Grzymala-Busse2007; O’Dwyer 2006).Footnote 3 Other studies point out that early ‘winners’ of the process of economic transitions systematically undermined the capacity of state institutions in order to appropriate assets held by other social groups (Ganev Reference Ganev2007). Lastly, studies devoted to the more coercive non-programmatic strategies of voter mobilization by state employees at times of elections have pointed to the potential collective action problems facing citizens living in localities with multiple companies (Darden Reference Darden2001; Hale Reference Hale2003). Others have pointed to the relative ease of surveillance of group voting behaviour in localities where industry and large-scale agriculture is concentrated (Allina-Pisano Reference Allina-Pisano2010).Footnote 4 While these studies have generated a number of interesting hypotheses, such hypotheses have not yet been systematically tested.
According to ratings made by both academics and various agencies monitoring corruption, Bulgaria and Romania are two of the new democracies in the region with the most endemic clientelistic mobilization.Footnote 5 These studies note, however, that such non-programmatic mobilization is not qualitatively different in its nature and patterns from the mobilization found in other Eastern European new democracies (Kitschelt et al. Reference Kitschelt, Mansfeldova, Markowski and Toka1999). In other words, within the group of Eastern European democracies, Bulgaria and Romania represent typical cases with extreme value on the dependent variable, that is, they are ‘paradigmatic of the phenomenon of interest’ (Gerring 2007: 101). It is beyond the scope of this article to document the prevalence of economic intimidation in other Eastern European new democracies. However, given the place of Bulgaria and Romania in this universe of cases, we expect to find similar patterns but lower levels of economic intimidation in other countries of the region. Including both Romania and Bulgaria in our study allows us to lessen the dependence of its findings on the distinctive empirical realities of one country and of the specific localities sampled within each country.
As will be discussed below, in both Romania and Bulgaria, we selected localities that provide a good representation of the diverse economic, ethnic and political conditions in these countries (our control variables). The localities included in our study vary also in their level of economic concentration and thus allow us to test the proposition that the incidence of economic intimidation is higher in localities with higher levels of economic concentration.
Measuring illegal or sensitive behaviour is extremely difficult using traditional survey methods. While respondents may have experienced such illicit electoral strategies, they may be reluctant to admit these experiences. Such reluctance may be the result of fear of retaliation from the agent that has engaged in such behaviour. Voters may also consider clientelistic strategies as ‘improper’ or ‘illicit’, which increases social desirability bias.
One research strategy which seeks to provide remedies for this measurement problem is the list experiment (Corstange Reference Corstange2008; Glynn Reference Glynn2013; Imai Reference Imai2011). While this type of experiment was initially used to study socially undesirable attitudes (such as racism or anti-Semitism), its use in the study of electoral clientelism has gained increasing importance in recent years. Using list experiments, scholars have documented the prevalence of electoral irregularities in countries such as Argentina, Brazil, Lebanon and Mexico, among others (Corstange Reference Corstange2011; Gonzalez-Ocantos et al. Reference Gonzalez-Ocantos, de Jonge, Meléndez, Osorio and Nickerson2012; Oliveros Reference Oliveros2013; Weitz-Shapiro and Winters Reference Weitz-Shapiro and Winters2013).
In list experiments, respondents are randomly divided in two groups, presented with a list of items and asked how many (rather than which items) 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 as well as a sensitive item. The difference in the mean number of items reported by respondents as applying to them in the treatment and in the control group allows for the estimation of the prevalence of the sensitive behaviour in the population.
Existing applications of the ‘list experiment’ to the study of electoral corruption have generally attempted to estimate the incidence of one form of electoral irregularity, most notably vote buying. Our study builds on and extends these approaches, by trying to estimate the simultaneous incidence of multiple forms of electoral clientelistic mobilization strategies.
List-Experiment Design
In both Romania and Bulgaria, we used a common method to design and conduct our list experiment. We began by conducting interviews with election observers and representatives of non-governmental organizations and focus groups with voters. These strategies allowed us to identify the salient types of non-programmatic mobilization present in each country at times of elections as well as the ways in which local citizens refer to such mobilization. As a result, we were confident that our wording of all treatment items adequately captures the phenomena of interest in this article and, at the same time, does not make respondents uncomfortable about participating in the survey.
To select the treatment and control items to be included in our list, we pre-tested potential item candidates in small surveys (in Bulgaria, the sample size of our pre-test was 220 respondents and in Romania 200). The goal of the pre-test was to design the list in such a way as to avoid ‘ceiling’ and ‘floor’ effects, thus increasing the confidence of respondents reporting truthfully the sensitive behaviour. Our design also attempted to reduce the variance in the means-estimate for the treatment and control groups by including three non-sensitive items: a low-prevalence item and two negatively correlated high-prevalence items (Glynn Reference Glynn2013; Kuklinski et al. Reference Kuklinski, Cobb and Gilens1997). We selected low-sensitivity items that were chosen by less than 15 per cent of the respondents and high-prevalence items that were chosen by more than 50 per cent of the respondents but that were strongly negatively correlated (r>−0.7).
Consider the following example of a list that was included in the survey administered in Romania. The online appendixFootnote 6 lists all the other items that were used in our surveys.
People vote for many different reasons. I will enumerate some of them. How many of these reasons in total are true in your case? I am not asking you to record which one(s) of these reasons explain why you voted but how many apply to you:
∙ (Low prevalence) The candidate has a good reputation in China.
∙ (High prevalence 1) I voted for political change.
∙ (High prevalence 2) I voted against political change because too much political change is undesirable.
In both countries, our lists included treatments that attempted to measure the incidence of clientelistic mobilization of voters. Our pre-tests also suggested that voters perceive electoral coercion as being perpetrated by political parties, whether it was state employees, partisan brokers or social leaders acting on behalf of the political parties that actually exerted the pressure; however, electoral coercion by employers, or what some of our respondents called a ‘corporate vote’ was understood as a separate mobilization strategy. As a result, our questionnaires included the following treatment items:
∙ A state official offered me administrative favours/facilitation to vote for a particular candidate.
∙ Someone at work, for example my boss or supervisor, pressured me to vote for a particular candidate.
∙ A party representative, state official or community leader pressured me to vote for a particular candidate.
∙ A party representative or community leader offered me money, gifts, or food to vote for a particular candidate.
To maximize the analytical leverage of our sample sizes, we assigned each respondent to the control group in half of the lists and to the treatment group in the other half of the lists. More specifically, each respondent received two control and two treatment lists. Respondents were randomly assigned to each cluster of treatment and control lists. Since each respondent was asked to evaluate multiple lists, both the questions and the items on each question were rotated to avoid ‘list-order effects’ (Glynn Reference Glynn2013). To ensure both the randomization of the questions order and the randomization of the sensitive items placement in our lists, we administered six different versions of the questionnaire.
We asked all respondents to report some personal background information: their age, gender, education, ethnicity, income level and party preference/affiliation (if any). Each survey fitted on a single A4 page and took no more than five minutes to fill in, thus ensuring respondents’ attention to the questionnaire.
Implementation of the List Experiment
Table 2 describes the localities included in our study. All localities are small municipalities, with populations of between 5,000 and 25,000 inhabitants. The localities selected for the study are comparable across the two countries, Bulgaria and Romania. They also provide a good representation of the diverse economic, ethnic and political conditions in the two countries, including level of economic development (dominant industry) and economic performance (unemployment), majority and minority ethnic makeup and political incumbents and their tenure in power.
Economic, Social and Ethnic Characteristics of the Experiment Localities
Sources:
Bulgaria: Local Development Strategy, Bobov Dol (2012); Local Development Strategy, Bobov Dol (2012); Local Development Strategy, Sopot (2012); Local Development Strategy, Beloslav (2012); Local Development Strategy, Aytos (2012); and Local Development Strategy, Svoge (2012).
Romania: Economic Census (2011), National Agency for the Occupation of Workforce, Romanian Commerce Registry. The unemployment data represent the average rate for registered unemployed people for 2012 and were computed from the monthly data compiled by the National Agency for the Occupation of Workforce.
Note: aThe incumbent mayor of Petrila represented the Democratic Liberal Party (PDL) during the 2008 and 2012 elections. In 2013, the mayor shifted his political affiliation, migrating to PSD.
To test our hypothesis that the incidence of economic intimidation is higher in more economically concentrated localities, our study includes three localities that are dominated by single firms. These are Petrila in Romania and Sopot and Bobov Dol in Bulgaria. Petrila is a small town in the south-west of Romania, Hunedoara County. The economic activity of the municipality has been based on coal mining since the nineteenth century. The largest employer in the locality is a state-owned coal mining company, at present employing about 1,000 people at Lonea coal mine. According to the data provided by the National Agency for the Occupation of Workforce and the Romanian Commerce Registry, we find in Petrila only four other small companies, each of which employs fewer than 30 people. Thus, Petrila is a good example of an economically concentrated locality. In Bulgaria, the two localities with a concentrated economic structure are Sopot and Bobov Dol. Sopot is home to Bulgaria’s largest weapons manufacturing company, VMZ Co., and Bobov Dol is home to Bulgaria’s largest underground mining company, Bobov Dol Mines. Both companies are the dominant local employers. Each of these companies employs at least one person in each household in their municipality. In both Bobov Dol and Sopot, there are about a dozen other small firms and additional micro-firms in all industrial sectors.
The other municipalities in our sample have more varied economic structure. At the other extreme of the economic concentration spectrum, Beloslav in Bulgaria is very economically diverse, with eight medium-sized heavy industry firms (in energy, transport and metallurgy) as well as a number of additional small and micro light-industry and agricultural firms. In Romania, Odobesti’s local economy includes several firms in the food processing industry (which employ about 250 people each), one company in furniture production and three timber production companies that employ about 75 people each. The other localities in our sample have several large and numerous small and medium firms each.
In Bulgaria, we conducted our study immediately after the May 2013 parliamentary elections. The implementation of the survey began the day after the election and was completed within two weeks. The survey was implemented by a local polling company with extensive experience in election and exit polling. The company used the 2011 census to identify a representative sample of voters to be interviewed in person in each locality. Each respondent who consented to participate was asked to fill in the questionnaire on their own and to place the response in a non-transparent box of the kind used for exit polls, reassuring respondents of the anonymity and confidentiality of their survey participation. Refusal to participate in the survey was 41 per cent on average, which is comparable with the average response rates for most surveys in Bulgaria. On average, 300 voters participated in our experiment (see online Appendix C on the number of respondents per locality).
In Romania, the list experiment was administered by graduate students of the Political Science Department at SNSPA-Bucharest. To obtain a representative sample of voters in each locality, our enumerators contacted every single household in each locality, but randomly selected the participant in the household by asking to speak to the individual whose birthday was closest to the day of the interview. The survey was administered in July 2013 in Odobesti and October 2013 in Petrila, several months after the December 2012 parliamentary election. The average response rate was 55 per cent for all the localities. In Odobesti, 686 respondents participated in our survey; while in Petrila, 981 voters took part in the survey. Given the timing of the survey, our estimates of the prevalence of clientelistic mobilization are measured less precisely in Romania than in Bulgaria. However, the main research question motivating this article concerns within-country variation in the incidence of electoral irregularities. This variation across localities is likely to be independent of the timing of the survey.
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 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.
Tests of Balance Results
To assess the validity of our list experiments, we use the test for design effects developed by Blair and Imai (Reference Blair and Imai2012). Specifically, we test whether the inclusion of the sensitive item changes the responses to the control items in the list. This test assesses whether responses after the addition of the treatment item are larger than responses to the control lists, but by at most one. If so, then design effects may drive the difference between treatment and control responses. As Tables 4.1 and 4.2 demonstrate, we do not find any design effects for the economic intimidation treatment in our survey. Neither do we find any design effects for any of the other treatment items included in our survey.
Test for List-Experiment Design Effects: Bulgaria
Note: Bonferroni-corrected p-value=1.992.
Test for List-Experiment Design Effects: Romania
Note: Bonferroni-corrected p-value=0.250.
RESULTS
What is the relative importance of economic intimidation in Bulgaria and Romania, both in absolute terms and relative to other clientelistic electoral strategies? Table 5 reports the average incidence of the non-programmatic electoral mobilization strategies used in the localities included in our study. Here, we present results on the mean number of list items by respondents in the treatment and control group, respectively. 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 following example of the magnitude of vote buying in Bulgaria. The mean response of our treatment group is 1.07 and the mean response of our control group is 0.97. From the difference of these means of 0.10, we infer that the incidence of voter approached to sell their vote in the Bulgarian localities is 10 per cent.
Incidence of the Clientelistic Mobilization Strategies in Romania and Bulgaria
Notes: a We report the results of a two-sampled t-test with unequal variances.
*** p<0.01, ** p<0.05, * p<0.1.
Our results also document that a diverse set of clientelistic strategies of electoral mobilization were deployed in both Romania and Bulgaria. We find that in both countries the strategies that leverage the vast resources of the state – political pressure and favours – are the most used type of clientelistic mobilization at times of elections. For example, in the economically concentrated localities of Bobov Dol and Sopot, the economic intimidation was reported by 20 per cent, respectively 16 per cent of our respondents. In Petrila, Romania, the economically concentrated locality in our study, 7 per cent of the respondents in our sample reported the use of economic intimidation. In the economically fragmented localities included in our study, the levels of economic intimidation were lower in magnitude and the measurement of these effects did not achieve statistical significance.
We also find that economic intimidation is present in the menu of clientelistic mobilization deployed by politicians in both countries. We further find that fewer voters report having been targeted by this type of electoral strategy when compared with other clientelistic strategies.
Did the use of economic intimidation vary systematically with local economic structure? We expect that the incidence of workplace intimidation is higher in localities with concentrated economic structure compared with localities with more decentralized economic activity (and varies additionally across voters with different individual characteristics). We estimate the following model:
$$\eqalign {Y_{i} \, {\equals}\, & \beta _{0} {\plus}\beta _{1} \,TREATMENT_{i} {\plus}\beta _{2} ECON\,CONC{\plus}\beta _{3} TREATMENT \cr & {\times} ECON\,CONC{\plus} \beta _{4} Z_{i} {\plus}\beta _{5} TREATMENT{\times} Z_{i} {\plus}{\epsilon}_{i} $$
The dependent variable
$Y_{i} $
represents the list response to the question about workplace intimidation, subscripted i for different individuals.
$TREATMENT_{i} $
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. ECON CONC is a variable that takes the value 1 for localities where employment is dominated by one single firm and 0 otherwise. In Bulgaria, ECON CONC takes the value 1 for Bobov Dol and Sopot, while in Romania, this variable takes the value 1 for Petrila.
$Z_{i} $
is a battery of individual-level characteristics and
${\epsilon}_{i} $
is the error term. The individual-level controls include measures of gender, income, age minority status as well as measures of the partisanship of the respondent. The income variable takes the value 1 if the respondent earns about or below the minimum wage. The minority status variable takes the value 1 if the respondent is a member of the Turkish minority in Bulgaria or the Hungarian minority in Romania. The employment status variable takes the value 1 if the respondent is employed full-time and the age variable takes the value 1 if the respondent is of working age. We estimate our models using ordinary least squares. All our specifications cluster the standard errors at the locality level to take into account correlations within each locality (due to locality-level characteristics that could downwardly bias our standard errors).
Table 6.1 presents our estimates of the relationship between the economic concentration of a locality and the incidence of workplace intimidation in Bulgaria, while Table 6.2 presents the same estimates for Romania. The results for both countries lend support to our hypothesis that the incidence of economic intimidation is higher in economically concentrated localities.
Explaining Economic Intimidation in Bulgaria
Notes: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1.
Explaining Economic Intimidation in Romania
Notes: Robust standard errors in parentheses; *** p<0.01, ** p<0.05, *p<0.1.
Consider first the results of our analysis in Bulgaria, reported in Table 6.1. Model 1 examines differences in the incidence of intimidation across localities, without including individual-level variables. The interaction between economic concentration and the treatment dummy for economic intimidation for Bulgaria has the predicted positive sign and is significant at conventional levels. Model 2 introduces individual covariates to the specification presented in Model 1. In contrast to some previous studies that find that low-income voters are disproportionately targeted by non-programmatic mobilization, we do not find evidence that lower-income respondents in Bulgaria are subject to economic intimidation at higher rates. Model 3 includes additional controls for the partisanship of the respondent. We include variables measuring support for the main centre-left party, the Bulgarian Socialist Party (BSP), the main centre-right party, Citizens for the Economic Development of Bulgaria (GERB), and the third main political party, the liberal Movement for Rights and Freedoms (DPS), representing the Turkish minority in Bulgaria. None of these variables reaches significance at conventional levels, which suggests that supporters of neither of the main political parties in Bulgaria are disproportionately intimidated by their employers (holding other individual characteristics constant).
Table 6.2 presents our examination of the variation in the incidence of economic intimidation in Romania. Model 1 presents the baseline comparison in the incidence of economic intimidation across localities that vary in their level of economic concentration. In this specification, the difference between economically concentrated and fragmented localities does not reach significance at conventional levels. In Model 2, we introduce controls for the gender, income, and the minority and employment status of respondents and the incidence of workplace intimidation. Once we introduce controls for these individual characteristics of respondents, the difference in the incidence of economic intimidation between concentrated and fragmented localities reaches statistical significance at conventional levels. Consistent with some of the previous work on electoral clientelism, in Romania we find that low-income respondents are more likely to have experienced workplace pressure. Model 3 adds covariates measuring the partisan affiliation of the respondent. These covariates measure partisanship of voters of the Social Liberal Union (USL) (a 2012–14 governmental coalition consisting of four parties),Footnote 7 the Democratic Liberal Party (PDL, a centre-right party, which at the time was the main opposition party) and the Democratic Union of Romanian Hungarians (UDRM, the largest minority party). None of the partisan variables is significantly associated with the incidence of economic intimidation in Romania, suggesting as in the Bulgarian case that the supporters of these parties are not disproportionately intimidated by their employers.
We conducted additional interviews in both countries at the time we fielded our survey to understand the specific mechanism by which such non-programmatic strategies occur. In Bulgaria, we interviewed close to 30 knowledgeable observers at the national level, including journalists, civic activists and academics as well as nearly 60 knowledgeable observers and voters in the three localities sampled, including state employees, civic activists, ethnic chiefs, employers and a random sample of voters who chose to participate in the exit poll conducted by the agency that later implemented our experiment. In Romania, we conducted close to 20 interviews in the two studied localities with a similar sample of knowledgeable observers and voters.
Our respondents in Bobov Dol openly spoke of significant and effective pressure by the owner of Bobov Dol Mines on behalf of his preferred party (struggling to cross the threshold for representation in parliament). A number of current and former miners testified that they had been ‘warned’ that if they wanted the mines to stay open, they should vote for the party supported by the mine owner. Others discussed threats of price hikes of mine-sponsored food and more ‘work Saturdays’. Our respondents informed us of half-hour meetings with representatives of the party supported by the mine owner that were held every morning in the mine. During these meetings, party officials instructed the miners to recruit and mobilize their families to vote for the owner’s party. Miners feared that since the party committed them to securing a certain number of votes, the mine owner would punish them if the promised number of family members did not turn out or the promised number of votes did not come in at their voting station (where representatives of the party serve as election observers). Miners who did not fill their ‘turnout quota’ or who were suspected of sympathizing with ‘undesirable parties’ were threatened with being assigned lower-quality equipment and ‘undesirable tasks and shifts’. While most interviewed miners did not believe that voter secrecy was violated – though some chose to find ways to ‘prove’ to their employer that they voted in the desirable way – they were convinced that most miners’ political affiliations had become known through various on-the-job interactions and the broader social interactions in their small town (Interviews R1, R2, R4, R7, Bobov Dol, Bulgaria, March 2015).
Similarly, in Sopot, economic intimidation featured prominently in our interviews on the non-programmatic strategies that candidates used to mobilize support for the 2013 election. Several voters reported that in 2013 VMZ was slated to be restructured and then sold and that massive layoffs were a major concern for its employees. Local party officials, together with a representative of the company management, were reported to have pressured at least one group of employees during an informal meeting at the company cafeteria. The economic threat used in this locality was that if they did not support the party, they would not be spared in the likely upcoming restructuring of the company (whether it was sold or not). Another voter provided a similar report about his supervisor pressuring his team of subordinates by threatening dismissals. A third respondent also added that his supervisor further insisted on driving the workers’ families to the polling station. Moreover, two of these four interviewees commented that they felt they had no choice but to vote as their supervisor had ‘ordered’ and thought that he recognized this when giving them the order. One of them further commented that he felt this way despite his relatively good socioeconomic standing. The other reasoned that his expected compliance did not seem in doubt, and the supervisor felt no need to demand ‘proof’.
In Beloslav, our respondents provided few personal and second-hand reports of private economic intimidation before the 2013 election. One voter did, however, mention that a family member was sent on a business trip that began on the election day and that was no coincidence – this employee was a known sympathizer of a rival party to the one preferred by the CEO (who had repeatedly ‘lectured’ his employees that this party would mean more business for the company). The employee objected but the CEO told him that he should take the trip if he wanted to keep his job.
Our qualitative data thus suggest a pattern similar to the one documented through our list experiment: there is some incidence of private economic intimidation at times of elections throughout Bulgaria but it is higher and more systematic in economically concentrated localities. The comparison of these three localities further allows us to shed some light not only on the ways in which voter intimidation is organized at the workplace but also on the prevalence of this non-programmatic mobilization strategy given the political history of a locality. Bobov Dol and Sopot are very similar in most theoretically relevant ways (including their levels of economic concentration), except for the fact that the former has seen turnover in power at every election in the past 15 years while Sopot has been politically dominated by one party. Yet, our survey reveals very similar incidence and patterns of economic intimidation at times of elections. Also, Beloslav and Sopot are very similar in most theoretically relevant ways, including in their history of turnover in power at the local level, yet they differ not only in their levels of economic concentration but also in the incidence and patterns of economic intimidation at times of elections, in a manner consistent with our theoretical propositions. Taken together, this paired comparison of two sets of cases – the first most similar and the latter most different (Gerring 2007: 101) – suggests that the political history of a locality is unlikely to be biasing or confounding the relationship of interest in this study: namely the impact of the economic concentration of a locality on the prevalence of economic intimidation at times of elections.
Our follow-up interviews in Petrila, Romania, helped us to identify both the most important agents engaged in electoral intimidation and the strategies used by these actors to influence the political choice of voters. Our respondents documented the existence of very strong political ties between the local mayor, local councillors (who were past employees of the local mine) and managers within the company. The incumbent mayor, our respondents noted, had the power to intervene in the appointment of managers in the mine and ‘change the management after every election when incumbents lose their power’ (Interview 16, Petrila, Romania, October 2013; Interview 2, Petrila, Romania, March 2015). At the time of the vote, higher-level managers delegated to engineers and foremen (‘brigadiers’) the responsibility of bringing employees to vote and to monitor their turnout (Interview 16, Petrila, Romania, October 2013). The mechanism of workplace mobilization was described to us as follows:
At the voting section, the foremen must control their team of employees. Each of the actors – politicians, mine managers and union leaders – know how many votes have to be cast to secure victory for a certain political candidate. Thus, at the end of the voting day, foremen located in each voting station must make sure that the candidates have a minimum number of votes to win. Foremen make sure that men go to vote together with their family. If the foreman is not effective in getting his men out to vote, he risks being dismissed. (Interview 4, Petrila, Romania, March 2015)
These agents – foremen, engineers and sometimes local union leaders – used different threats in attempting to persuade voters to support particular candidates. One widely used threat implied layoffs if the ‘wrong’ candidate was elected. One respondent noted, ‘the vote pressure mechanism inside the mine sounds like this: employee, watch out who you vote for, because the others are coming and you could lose your job and payment. You’d better vote instead with this candidate or party’ (Interview 16, Petrila, Romania, October 2013). Other respondents reported threats of changes in payment or working conditions. Some of the miners reported that unions threatened to move them to different workplaces (where conditions were tougher) or to change the criteria of the evaluation of their work (Interview 17, Petrila, Romania, October 2013). Miners were also threatened with delays in access to public housing (social housing) or delays in receiving disability or unemployment benefits.
In contrast to Petrila, our qualitative research in Odobesti found very little evidence of economic intimidation. Only one of our respondents mentioned that the most important local enterprise supports the incumbent mayor and that it can mobilize some of its employees to turn out to vote (Interview 10, Odobesti, Romania, 3 December 2015). In this locality the dominant clientelistic strategy involved the provision of policy favours by state employees. As one of our respondents remarked, ‘after they give us the social assistance payments, the people of the mayor come during elections and remind us that they have given us these favours’ (Interview 18, Odobesti, Romania, 3 December 2015).
In sum, our survey-based and qualitative findings support our hypothesis that an increase in economic concentration lowers the costs of electoral intimidation by private actors and makes such non-programmatic mobilization more widespread.
CONCLUSION
Our article advances both theoretically and empirically the analysis of clientelistic practices in contemporary elections. We have argued for the importance of studying a broad menu of non-programmatic mobilization, which includes the distribution of resources from the state, offers of monetary and goods, and electoral intimidation. Such a research agenda, we believe, will represent a significant advance in the current literature of electoral clientelism which provide a very narrow estimate of the magnitude of ‘undue electoral influence’, due to their omission of a range of non-programmatic strategies.
This article has documented one such strategy of political influence that has received insufficient attention in existing studies of clientelism – namely economic intimidation or the efforts of employers or firm managers to coerce their employees to vote for a particular party or its candidate. We have documented the occurrence of these practices in two new democracies in Eastern Europe – Bulgaria and Romania – demonstrating that employer intimidation plays a significant role in the menu of clientelistic strategies in these countries. We expected that the incidence of electoral intimidation would be higher in localities whose economies are dominated by a small number of large firms as compared with localities with higher levels of economic heterogeneity. We also expected that in localities with higher economic heterogeneity, the overall mix of clientelistic strategies would differ (as compared with the economically concentrated localities), due to the higher relative costs of economic intimidation. We found empirical support for both hypotheses in both Eastern European countries. These patterns of clientelistic mobilization at times of elections have an eerie resemblance to the findings of a historical literature on economic intimidation (Anderson Reference Anderson2000; Mares Reference Mares2015). We conjecture that such findings are not unique to the two countries included in our study, but that one encounters such practices in other contemporary elections.
SUPPLEMENTARY INFORMATION
To view supplementary information for this article, please visit https://doi.org/10.1017/gov.2016.39.
