Dissatisfied, uninformed or both? Democratic satisfaction, political knowledge and the acceptance of clientelism in a new democracy

ABSTRACT In many countries, voters are targeted with clientelistic and programmatic electoral offers. Existing research explores the demand side of clientelism, but we still know very little about what determines voters’ acceptance of clientelistic and programmatic electoral offers. This article builds a novel theoretical framework on the role that democratic dissatisfaction and political knowledge play in shaping voters’ acceptance of different types of electoral offers. We test the implications of the theory with a survey experiment conducted after the 2019 local elections in Bulgaria. Our results show that low-knowledge voters and those who are dissatisfied with the performance of democracy and democratic institutions are more likely to accept clientelistic offers from politicians. The findings contribute to the literature on electoral clientelism and political attitudes.


Introduction
Electoral clientelism, defined as the provision of particular material rewards in exchange for electoral support, 1 is widespread in contemporary societies. It has gradually become an important component of many elections in new democracies and electoral authoritarian regimes. In this context, we know much about the supply side of clientelism. Extensive research explores why political actors use clientelism, how it works and develops over time, and what are its forms. 2 There is somewhat less information about the demand side of clientelism and how voters perceive clientelism. Previous studies have investigated voters' access to resources, ideological position, preferences for brokers and perceptions of efficacy. 3 So far, the demand side has been investigated through the lenses of choices made by voters when exposed to clientelism. However, we know very little about how willing voters are to accept clientelistic practices. Information about these attitudes is relevant because it sheds light on society's openness towards clientelism, reveals the anchor that clientelism may have in the political culture and could help predicting the behaviour of political actors in the future.
This article addresses this gap in the literature and aims to explain why voters may accept electoral clientelism and programmatic policy offers. To this end, we build a theoretical framework that differentiates between positive clientelistic exchanges, negative (coercive) clientelistic exchanges and programmatic inducement in terms of their expected "returns" to the voter as well as the broader implications of the exchange. Based on this theoretical framework, we formulate a series of testable hypotheses about the effects of democratic satisfaction and political knowledge on voter acceptance of electoral clientelism and programmatic politics. First, we expect that voters' acceptance of clientelistic targeting is shaped by their perceptions of how well the political system and political institutions function. We expect that highly dissatisfied voters are more susceptible to particularistic benefits offered by clientelistic exchanges in contrast with the more uncertain benefits offered by programmatic voting. Second, we expect that voters' political knowledge and their understanding of the implications of these various types of electoral exchanges shapes their acceptance of electoral targeting, and that clientelistic offers are more likely to resonate with low-knowledge voters.
We bring empirical evidence from a vignette experiment conducted in Bulgaria on a nationally representative sample of 1155 citizens in the aftermath of the 2019 local elections. In the experiment, the respondents were presented with scenarios where local politicians targeted them with different types of clientelistic and programmatic inducements. Survey experiments are particularly well suited to test voter responses to these kinds of sensitive questions because they avoid the social desirability bias that often complicates the analysis of standard survey responses. 4 Bulgaria is the appropriate setting for our study because clientelism plays a relevant and longlasting role in local governance and elections (see the Research Design section).
This article contributes to the literatures on political behaviour and on electoral clientelism. Our results broaden the understanding about how democratic dissatisfaction and political knowledge can influence voters' behaviour beyond political participation. It is the first attempt to analyse the effects on citizens' acceptance of clientelism and programmatic inducements. The dissatisfaction with representative democracy is high throughout the world 5 and this attitude has been traditionally linked with the propensity (not) to participate in politics. Political knowledge is another central individual level cognitive characteristic that has been often linked with political participation. We show that the two variables have far-reaching impact on citizens. We also contribute to the emerging literature challenging the assumptions about voters as passive targets of clientelism. 6 Earlier studies show that voters' perceptions on clientelism are mediated by their socio-economic resources or access to information. 7 Our analysis shows how cognitive traits impact the acceptance of clientelism when socio-economic variables are controlled for. Finally, the new literature on the "demand side of clientelism" has mainly focused on voters' responses to positive inducements, such as vote buying. Consistent with earlier works, we distinguish theoretically and empirically between positive and negative types of clientelism and show that voters react to these offers in different ways. 8 The next section reviews the literature on electoral clientelism and programmatic politics. It formulates four sets of testable hypotheses that correspond to general and specific dissatisfaction and knowledge. We present next the research design with emphasis on the data, variables and methods. The fourth section interprets the main findings of our analysis. The conclusions discuss the main implications of our results and elaborate on directions for further research.

Electoral clientelism and programmatic politics
Party system theories distinguish between two types of linkages between politicians and voters. Types of "affective psychological attachments" include linkages based on party identification, descriptive representation (e.g. based on gender or ethnicity) and "charismatic" linkages. In contrast, programmatic politics and clientelism are both "rational and deliberative" modes of political linkage between the politicians and voters. 9 In positive clientelistic exchanges, voters are offered various benefits in return to their vote: goods, money, preferential access to jobs or welfare services. 10 Negative clientelism is based on a punishment for voting the "wrong" way, reflected in threats to lose the job or the access to welfare benefits. 11 Programmatic politics is defined as the process of mobilizing social and/or electoral support on the basis of a set of proposed policies and we contrast it with clientelism because of its form and goals. On the one hand, programmatic politics articulates a plan for achieving midto long-term goals, which is available for the public to consider and used by voters to hold elected officials accountable. On the other hand, positive and negative clientelism are means with short-term effects, used for the mobilization of individual voters without any broader goals beyond securing the election of the respective candidate. These three forms differ in several other ways. The following lines outline the core features of these forms along three analytical dimensions: the universalism of the transaction, utility for voters and cognitive capacity required (Table 1).
Electoral clientelism and programmatic politics differ in their mode of distribution and the degree of universalism of the exchanges. 12 Both clientelism and programmatic politics are based on transactions between politicians and voters, where politicians promise voters some kinds of goods in return for their support. In programmatic politics, the benefits are distributed universally by publicly available criteria. 13 However, the benefits delivered by programmatic politics can be uncertain and depend on the politicians' ability and capacity to deliver them (as well as the politicians' willingness to commit to electoral promises in the first place). Therefore, programmatic politics has sometimes been characterized as a relatively low-expected benefits kind of activity. 14 In positive forms of clientelism benefits are distributed on a particularistic and quid pro quo manner. 15 Negative clientelism, too, is based on particularistic exchanges as voters are threatened with a punishment that is contingent on voting behaviour. 16 Both positive and negative clientelistic offers are typically mediated by brokers, i.e. intermediaries who coordinate clientelistic politics at the local level. 17 Both vote buying and voter coercion are illegal in most electoral codes in the world.
The three types of exchanges differ with respect to the degree of utility provided to the voter. 18 Positive forms of clientelism and programmatic politics are both transactional modes of exchanges, and they both increase the utility of the votervoters get something in return for their vote. In both cases, the personal situation improves because of their decision. While in the case of programmatic politics an individual voter obtains something as a result of the implemented policies of the elected official, in positive clientelism they obtain it because of a personal agreement. In contrast, negative clientelism is based on threatening the voter with some kind of a sanction if they do not vote in a particular way, and thus does not increase the utility of the voter. 19 This is because in negative clientelism the promise is not to improve voters' situation, but rather to keep the status quo in return for their vote. These three types of electoral exchanges differ in the degree of cognitive capacity required to understand the mechanisms of the exchange. Positive forms of clientelism present the voters with a rather simple electoral transactionreceiving some kinds of goods in return of their vote. In such circumstances, the voters have the freedom to assess whether the thing that is being offered to them is of their personal benefit and the extent this is the case. More importantly, they have the freedom to choose whether to accept the offer or not without any significant repercussions for themselves. This stays in contrast to negative clientelism. It is based on explicit threats and the mechanism of the exchange is easy to understand. Here, the voter assesses the value of the particular utility they have and that is under threat of losing. Hence, in contrast to positive clientelism they have less freedom in their choice, given that should they reject the offer, there will be noticeable effects on their personal situation. In contrast, programmatic politics is much more complex, and requires a lot of knowledge and informational resources to understand. 20 This particularly concerns the voter's analysis on the feasibility of the proposed programme, its potential effects on them, as well as on assessment on the potential of the elected party to implement their programme, for example. In such situation and in contrast to the cognitive choices in clientelist exchanges, their choice in programmatic exchanges depends not only on analysis of the outcomes from the direct interaction between them as a voter and the party that makes the offer, but also on their analysis of the external circumstances that underpin the offer.
Based on these distinctions we form a set of testable hypotheses on how two central cognitive characteristics of voters, democratic satisfaction and political knowledge, shape voters' acceptance of these different scenarios.

Democratic dissatisfaction and institutional performance
Earlier research indicates that citizens develop different attitudes towards the political system and specific political institutions. 21 These rest on different principles that we argue can influence the acceptance of clientelism. Satisfaction with the performance of democracy refers to voters' evaluations on how well the democratic institutions are functioning in a given country and falls somewhere between Easton's diffuse and specific support for democracy. 22 The gap between the democratic expectations of citizens and their judgements on how well the democratic system functions is wide in new democracies. 23 Voters' level of democratic satisfaction can influence their acceptance of different types of electoral targeting for several reasons.
Democratic dissatisfaction has widely been associated with the rise of populism and anti-establishment voting in many of the world's democracies. 24 Dissatisfaction with democracy is likely to have other consequences for democratic governance. As stated previously, both positive and negative forms of clientelism are illegal in most electoral codes of the world, and citizens are also aware of their illegal nature. Democratic satisfaction has also shown to increase citizens' likelihood to obey the law and comply with other governmental processes, such as pay taxes. 25 In contrast, citizens are considered more likely to engage in illegal civic acts if their confidence in the democratic system is very low. 26 Along the lines of these arguments, we expect that citizens who are very dissatisfied with the way democracy functions are more likely to accept both positive and negative forms of electoral strategies.
At the same time, we expect that voters who are highly skeptical about democracy's performance to be less responsive to programmatic policy messages. Democratic dissatisfaction has long been associated with the tendency to turn away from politics. Negative evaluations about democratic governance have been associated with the propensity not to turn out in elections and not to engage in political life. 27 Similarly, programmatic politics has a limited power to mobilize dissatisfied voters into accepting such practices. Following these arguments in the literature, we expect that citizens who are dissatisfied with the functioning of democracy are: H1a: More likely to accept electoral strategies based on positive clientelistic inducements H1b: More likely to accept electoral strategies based on negative clientelistic inducements H1c: Less likely to accept electoral strategies based on programmatic policy inducements Local government bodies are responsible for many important day-to-day services provided to citizens. Many new democracies suffer from poor service delivery, which is the failure of the government to provide basic services for their citizens. 28 Access to essential services, such as welfare services, can be highly stratified in many countries and these situations have been particularly prone to clientelistic deals. 29 Voters' perception on the workings of local politicians is likely to also affect their assessment of positive and negative forms of clientelism. Voters' confidence in the integrity of public officials influences their likelihood to commit illegal acts such as tax evasion. 30 Therefore, we could also expect that voters who are highly suspicious of the performance of the local politicians would be more likely to accept both positive and negative forms of electoral corruption. These arguments indicate that voters who are dissatisfied with the local government would prefer the particularistic benefits associated with clientelistic rewards. This preference comes at the expense of uncertain policy promises. They are skeptical regarding the ability of local politicians to deliver the programmatic promises once they get into office. These voters distrust local politicians, are critical towards them and less likely to accept their promises. We therefore hypothesize that voters who are dissatisfied with the performance of the local government are: H2a: More likely to accept electoral strategies based on positive clientelistic inducements H2b: More likely to accept electoral strategies based on negative clientelistic inducements H2c: Less likely to accept electoral strategies based on programmatic policy inducements General and specific political knowledge Participation in modern political life requires extensive knowledge and information. 31 Voters' levels of political knowledge varies greatly and has an impact on their political behaviour. 32 Low levels of political knowledge make voters more susceptible to populist and anti-establishment politicians. 33 We argue that citizens' level of political knowledge may also have an effect on their acceptance of different types of electoral exchanges. This happens due to the degree of cognitive complexity required to understand the mechanisms involved. Positive clientelism present voters with a simple electoral transaction: they receive goods in return of their vote. These offers would be especially attractive for low-knowledge voters who have little information about how the political system works. Voters' acceptance of clientelistic offers is also likely to be shaped by voters' understanding about the broader consequences of those transactions. 34 High-knowledge voters focus more on public consequences of politics whereas those with lower levels of political knowledge tend to be more myopic in their political judgements. 35 Providing particular clientelistic goods to a select group of voters often involves negative externalities for the rest of the voters such as steering resources away from public goods. 36 Voters judge clientelism more severely when they receive information about the negative externalities involved. 37 When referring to negative clientelism, the immediate implications of voter coercion are explicit also to uninformed voters. However, political knowledge increases voters' sense of political competence and political efficacy. 38 Better informed voters are likely to be more aware of their democratic rights and to condemn voter coercion. Conversely, less knowledgeable voters could be expected to be less critical of this practice.
Voters' level of political knowledge is also likely to shape their acceptance of programmatic policy offers by politicians. Understanding programmatic politics requires information about how the political system works and political sophistication affects voters' judgements about the behaviour of politicians. 39 More knowledgeable voters are more sympathetic towards the efforts of politicians, while low-knowledge voters are more prone to expect politicians to act in bad faith. 40 Politically informed voters are inclined to accept programmatic political strategies. Conversely, politically ignorant voters may be more suspicious towards policy promises. We expect that voters with low levels of political knowledge to be: H3a: More likely to accept electoral strategies based on positive clientelistic inducements H3b: More likely to accept electoral strategies based on negative clientelistic inducements H3c: Less likely to accept electoral strategies based on programmatic policy inducements Voters with low levels of knowledge are less able to integrate new political information into an existing political framework. 41 During campaigns, political actorsparties and candidatespresent different issues, set the policy agenda or denigrate their opponents. 42 A thorough understanding of these messages is usually achieved when voters have intimate knowledge about what political actors do and who they are. On the supply side, political parties with local leaders who have notoriety use clientelism more compared to parties with leaders who are not known to voters. 43 Following the same logic on the demand side, we expect that voters' specific knowledge such as the profile of party candidates to impact on how they respond to clientelistic electoral targeting. Voters who have little information about the track-records and policy profiles of different candidates would prefer the discretionary payoffs provided by positive clientelistic inducements and therefore accept such practices to a greater extent.
Positive and negative forms of clientelism do not differ markedly in terms of the cognitive capacity required to understand the proposed exchange. Yet, voters who know the local candidates better are also more likely to be aware of various opportunities of political participation. They could be more critical towards voter intimidation. This is because knowing the local candidates gives these voters sufficient information and idea about how widespread such behaviour is among candidates. If they are aware that other candidates do not engage in such conduct, they would be less likely to find it acceptable and more likely to be motivated to punish it on the ballot box. Conversely, low-knowledge voters could condemn this practice to a less extent and accept it more. A reason for this is that their lack of knowledge does not give them a proper understanding on whether such behaviour is typical for the candidate or actually is part of the wider political culture of their place. More importantly, not knowing the candidate would make it harder for a voter to identify who is the person on whose behalf negative clientelism is conducted. It is thus challenging for the voters to punish at the polls such a behaviour.
The assessment of programmatic policy offers requires extensive information. Voters who are more knowledgeable about specific political candidates are likely to be more engaged in politics in general and should have a more positive outlook on the pursuits of the politicians. 44 Political actors seek to develop a consistent issue profile during electoral campaigns 45 that can be best achieved if voters pay attention to this profile. We expect that voters who exhibit specific knowledge about political parties to be more responsive to programmatic policy offers. Low-knowledge voters are more likely to be more suspicious of the campaign messages by politicians. In brief, we hypothesize that citizens who have low levels of knowledge of the top party candidates are: H4a: More likely to accept electoral strategies based on positive clientelistic inducements H4b: More likely to accept electoral strategies based on negative clientelistic inducements H4c: Less likely to accept programmatic electoral strategies by politicians

Research Design
To test these effects, we rely on an original vignette experiment conducted in the aftermath of the Bulgarian local elections (October-November 2019). Bulgaria is the appropriate setting for our study because clientelism plays a relevant and long-lasting role in local governance and elections (see the following section). Clientelism is part of a broader strategy of negative campaigning in which many Bulgarian political parties, although they use clientelism, accuse their opponents of doing that. Equally important, the country has several issues with the rule of law and the quality of democracy, which makes it a fertile soil for the proliferation of clientelism and citizens' broad dissatisfaction.
The vignette experiment uses a nationally representative sample of 1155 Bulgarian citizens and was fielded in two waves with a roughly equal number of respondents: December 2019-January 2020 and March-April 2020. We use a vignette to measure our dependent variables. All respondents were informed that they would see a description of a mayoral candidate running in the next Bulgarian local elections. The sample was divided in three sub-samples that were exposed to different incentives to vote provided by candidates for the mayoral elections. For positive clientelism we use the statements "The candidate gives you a bag of products, food or money to vote for them" and "The candidate offers preferential access to welfare benefits after the elections if you vote for them". 46 While we are aware that voters may be motivated differently by types of incentives (bag of products, food or money), we merged them in the same question to accurately reflect the Bulgarian context. The existing practices of clientelism, especially in smaller places, traditionally involve parties providing voters with packages that usually combine food, small items, and a small amount of money. These come together with a party flyer indicating which candidate the voter should support. In more recent times we observe that money is potentially handed over separately from bags of food and products, but the principle remains the same and Bulgarian voters can rarely differentiate whether they will receive an financial or material incentive to vote.
For negative clientelism the respondents see the statement "The candidate threatens you with the loss of access to welfare benefits if you do not vote for them". For programmatic policies we used two statements referring to candidates' promises to focus on policies that are very or moderately important to the voter. 47 After seeing one of these scenarios, the respondents were asked "How acceptable do you find this politician's conduct?". The answers were recorded on a 11-point ordinal scale that ranges between not at all (0) and very much (10). The results of the ANOVA test indicate the existence of a statistically significant difference (at the 0.01 level) between the means of acceptance for each of the three groups of respondents: positive clientelism, negative clientelism and programmatic policies.
The measurement of the independent and control variables is presented in Appendix 1. We briefly discuss here two variables that are less straightforward. First, we measure general knowledge (H3) through the subjective or perceived assessment of individual knowledge. This perception does not always correspond to the real or objective knowledge, being sometimes overestimated (Dunning-Kruger effect) or in other instances underestimated (impostor syndrome). However, the imbalance between actual and perceived knowledge is rarely large because the subjective knowledge can result from familiarity with a topic. 48 Second, media exposure is a cumulative index of frequency with which respondents watch, listen or read news on TV, radio, newspapers or online news portals (Cronbach's Alpha = 0.72). All "do not know / no answer" options were removed from the analysis. Appendix 3 presents the descriptive statistics for all variables included in the analysis.
We use OLS regression with robust standard errors. Our dependent variable is measured on a Likert scale with 11 values, and we should ideally use ordinal logistic regression. However, the assumption of proportional odds is violatedthe Brant Test of Parallel Regression Assumption shows significance at the 0.01 level for all variablesand we could use only a generalized ordered logistic model. This is quite difficult to interpret and present and that is why we replace it with an OLS regression. The latter provides similar and allows for a more straightforward presentation and interpretation of results. The test for multi-collinearity shows that the independent variables are not highly correlated, i.e. the VIF values are lower than 1.24.

Clientelism and local elections in Bulgaria: An Overview
Clientelism has a significant place in Bulgarian local politics. Following the introduction of local self-governance in 1998, the country is divided into 28 provinces (oblasti) and 265 municipalities (obshtini). While the provincial administration is appointed by the national government, every four years there are elections for municipal mayor and city council. The past three decades revealed the close relations between clientelism and some of the main characteristics of Bulgarian local government. Three of them deserve particular attention. First, clientelism affects significantly the job policies of the local administration. Reports dating back to the first post-communist decade 49 highlight the regular practice of hiring "reliable" cadres in key positions in the local administration or municipal public services by the local governing party, as well as their protection once the local governing party is out of power. Such practices are important given the significance of local governments as an employer. In 12 of the 28 provinces the local government is among the top five employers, 50 suggesting that clientelism affects a significant portion of people in the communities across Bulgaria. Consistent with these observations, more recent studies provide evidence of both negative and positive types of clientelistic targeting in Bulgaria. 51 Second, clientelism is reflected in local public procurement. The privatization of public services in post-communist Bulgaria allowed for the establishment of publicprivate partnerships, where a public service is assigned to a private provider. Reports on corruption in local public procurement show a close relationship between local businesses and the local party in government, where businesses would fund the electoral campaigns of political parties in exchange for procurement contracts for business owners affiliated to the party in local government. 52 Such practices became particularly widespread especially after the EU accession in 2007 that granted provinces and municipalities access to EU regional development funding. 53 For example, shortly before the 2019 European Parliament elections a journalist investigation revealed that EU funding, earmarked for rural tourism, has been used to build private villas for politicians from the governing center-right Citizens for European Development of Bulgaria (GERB). 54 Generally, such clientelist practices remain well-known either for the low quality of the provided service or for the development of white elephant projects, such as the building of modern football stadiums in places with no football teams. 55 Third, beyond clientelism in local government, local elections are another major arena for clientelistic practices. The most common of those is vote buying. Recent years suggest that it grows in scope and relevance at local elections. For example, a recent survey showed that almost a third of respondents knew personally about cases of vote buying. 56 Furthermore, vote-buying is intensive among groups with an ethnic minority background (the Roma, in particular) and/or of lower economic status. 57 These practices can influence the electoral outcome. For example, a survey by Gallup International for the 2015 local elections showed that in places with a close race for the mayoral position, there has been a significant rise in turnout among Roma voters. 58 The 2019 local elections were particularly illustrative for clientelist practices. Held just a few months after the elections for the European Parliament, also marked by rampant levels of vote-buying especially in rural and/or impoverished areas of Bulgaria, 59 the local elections saw the extension of the competition between GERB and its main political opponents on the left and right. In 2015, GERB won the mayor office in almost every major city. 60 In 2019, the party had to defend its gains and overall local government record. The campaign in the major cities was heated, with the liberal right opposition highlighting several clientelistic practices, demanding for a more transparent and accountable local government. For example, in a television debate, the main liberal right candidate for Sofia mayor Borislav Ignatov highlighted "corruption in all its forms" as the main issue of the city, citing examples from minor corruption in the municipal administration to major irregularities of public procurement procedures. 61 The main opposition party, the center-left Bulgarian Socialist Party (BSP), aimed to use the local elections to criticize GERB's governing style, arguing that these clientelistic practices in local government are the essence of GERB's governance, calling to vote them out of power. Figure 1 presents the distribution of acceptance for positive and negative clientelism, and programmatic policies among the Bulgarian electorate. The horizontal axis displays the degree of acceptance. There are three straightforward conclusions. First, programmatic policies are accepted to a considerably higher level than any form of clientelism. The modal value for acceptance of programmatic activities is the middle of the scale, with the majority of opinions to the right side of the modal category. Overall, many voters find programmatic policies quite acceptable. This indicates that while clientelism is strongly rooted in Bulgarian local politics, voters are not particularly fond of these practices. It shows that clientelism is not necessarily a part of the Bulgarian political culture.

Dissatisfaction and Poor Knowledge
Second, Bulgarians accept more the positive than the negative version of clientelism. There is a very strong oppositionthree out of five votersto coercion practices used to persuade citizens to cast their vote. In comparison, only one in three Bulgarians considers completely unacceptable the offering of money, food or welfare benefits in exchange for votes. These results confirm that clientelism can be accepted if it provides (positive) personal benefits to the respondent. As seen in the background information about the case, the most common clientelistic practices in Bulgarian local politics rely to a certain degree on reciprocal benefits, thus creating the impression of being a fair exchange between equal parties. We run two different models for each form of clientelism and programmatic policies. The results of the model without controls are shown in Figure 2. There is empirical support for some of the hypothesized effects, but the reality appears to be considerably more nuanced than what expected in theory. The marginal effects indicate that dissatisfaction with democracy is a strong predictor for the acceptance of electoral strategies based on positive clientelistic inducements. This is in line with the theoretical expectations and evidence indicates that dissatisfaction with democracy favours also the acceptance of negative clientelism and programmatic policies. In fact, the effect on voter coercion is the strongest among the four main effects. One possible explanation for this is apathy. Clientelism in Bulgarian local politics goes back at least two decades. The absence of a decisive policy addressing it seems to have taken root among Bulgarian voters into accepting even vote coercion as a common practice.
Dissatisfaction with the way in which democracy works in Bulgaria increases the acceptance of programmatic policies. One possible explanation is related to the desire to improve the quality of democracy in the country. While Bulgarian voters may be apathetic by accepting even vote coercion as a standard electoral practice, this does not mean that they remain passive. This data seems to confirm that as Bulgarian voters are indeed interested in alternatives that ensure better accountability of local government. For example, in the 2019 local elections in the capital city of Sofia, Boris Bonev got 11% of the votes based on a comprehensive programme that builds upon the policy proposals and local government watchdog work of Save Sofia, an NGO that he chairs.
We also find empirical support for the theoretical expectation according to which Bulgarian voters who are dissatisfied with the way in which the local government does its job are more likely to accept positive clientelism. This is the strongest predictor for the acceptance of positive clientelism including the controls (Model 2 in Appendix 2). Contrary to our expectations, we find a similar effect on the acceptance of programmatic policies. This happens because Bulgarian voters remain open towards candidates who promise to improve their local community by providing a clear plan for local governance and ensure an improved accountability. For example, during the campaign for the 2019 local elections almost two thirds of the voters indicated that even if they are disillusioned with the incumbent mayor, they would vote based on their preferences and values rather than automatically supporting any of the challengers. 62 The evidence shows no effect of the dissatisfaction with how the local government works on negative clientelism. One explanation for the statistical independence is that to Bulgarian voters negative clientelism remains a widespread practice that any party with realistic chance of gaining or keeping power in local government would use. Under such circumstances, Bulgarian voters take the rather rational view that the government performance depends on factors other than the reasons people were hired or fired in the local administration. Figure 3 compares and contrasts the effects of the two forms of satisfactionwith democracy and with the governmenton the three types of inducements. Although these two forms of satisfaction are not highly correlated, their effect on the acceptance of positive clientelism and programmatic policies is very similar. The main difference lies in the fact that for programmatic policies the effect takes place at a higher level of acceptance. The figure confirms the earlier interpretation according to which the acceptance of negative clientelism is driven only by one form of satisfaction, i.e. with democracy.
Returning to the marginal effects presented in Figure 2, the evidence confirms our theoretical expectation according to which respondents with low levels of knowledge about local politics are likely to accept electoral strategies based on positive clientelistic inducements. In the Bulgarian case, this matches the reality given the deteriorating state of the local media. Several reports reveal that Bulgarian regional and local media increasingly rely on local government funding through PR contracts as a substitute for independent funding. 63 They fail to serve their public functions of scrutinizing the local government under the threat of losing their funding, prompting them to avoid reporting on such clientelist practices. This prevents Bulgarian voters from gaining a proper understanding on the direct effects of clientelism in their local community.
The findings also illustrate that respondents with low levels of political knowledge are also inclined to easier accept programmatic policies. In Bulgaria, this is because of the weak communication skills of independent experts. Recent years saw the increasing scrutiny of the local candidates' political programmes, but such discussions remained rather technical and incomprehensive for the wider public. In such circumstances, people with limited political knowledge are unable to properly assess the candidates' proposed programme, prompting them to be more open towards any candidate that tries to draw support on a programmatic basis if not through any other means.
Equally important, our results indicate that Bulgarian voters who possess knowledge about local politics are more inclined to accept negative clientelism. One possible explanation for this is their awareness and acceptance that such practice is used in the local governance of their own community, given its widespread character in Bulgarian local politics. This prompts such voters to perceive negative clientelism as a legitimate approach towards mobilizing support.
The specific knowledge of top candidates in Bulgarian elections produces the same effects on the acceptance for clientelism and programmatic policies. Citizens who have limited knowledge are more likely to accept positive inducements, as hypothesized (H2b). A similar effect in terms of size and direction is observed in relation to coercive clientelism. For acceptance of programmatic policies, the effect is even stronger: the less people know about candidates, the more likely they are to accept such policies. These results confirm the ambiguous state in which many Bulgarian voters find themselves. Without knowing the candidate, voters tend to return to their own perceptions on what constitutes a legitimate electoral practice, thus being particularly open towards programmatic offers, but also towards clientelist practices, given how widespread they are in Bulgarian local politics.
Among the controls, income is the variable with the highest explanatory power. The findings indicate that poorer people are more likely to accept both forms of clientelism and richer Bulgarians are more inclined to accept the programmatic policies. This confirms earlier findings in the literature 64 and recent journalistic reports on clientelistic practices in Bulgaria. These reports show a combination of a political apathy and pragmatism among poorer voters. The latter have the perception that their vote would not improve their lives so they are willing to engage in clientelistic practices as a means of personal benefit and/or perseverance. 65 In contrast, wealthier voters are not particularly fond of clientelism, preferring to support a programme that aligns with their own interests and views about the development of their community.
People who do not use media on a regular basis are more inclined to accept both clientelism and programmatic policies, with a stronger effect on the latter, which seems to confirm the effects of the deteriorating state of regional and local media in Bulgaria and its limited abilities to scrutinize local governments. Gender is a strong predictor for negative clientelism with men accepting it to a much higher extent than women. One potential reason for this is the rather conservative and patriarchal nature of Bulgarian society, 66 in which vote coercion is seen as a legitimate way to exhort authority. For similar reasons, education does not appear to play a role in the acceptance of these electoral practices.

Conclusions
An emerging line of research has begun to focus on the demand side of electoral clientelism. This article contributes to this literature by explaining the acceptance of electoral clientelism and programmatic policy offers. To our knowledge, it is the first study exploring the role of democratic satisfaction and political knowledge in shaping the acceptance of different types of electoral inducements. We examine this with the help of a vignette experiment embedded in a survey conducted in Bulgaria after the 2019 local elections. Our results show that voters who are not satisfied with the way democracy and democratic institutions function in Bulgaria are more likely to accept positive clientelistic electoral offers. Voters who have little general political knowledge are also more likely to accept positive clientelistic targeting. Our findings regarding the acceptance of voter coercion are more mixed and less robust. In contrast to our expectations, democratically satisfied voters and high-knowledge voters are also more likely to accept programmatic policy offers.
Our results on the acceptance of positive types of clientelism shed light on one of the central puzzles of clientelism research. We can better understand why clientelism is still widely used despite the known commitment problems undermining its efficacy. 67 We illustrate that the cognitive heterogeneity between the voters may explain the effectiveness of electoral clientelism even when it is not monitored. Clientelistic offers resonate with voters who are very dissatisfied with democratic performance and who possess little knowledge on politics. All these indicate that some voters may be so disappointed with "traditional politics" that they are willing to vote for clientelistic politicians hoping to get something in return for their vote. Clientelistic exchanges could thus in some cases function as a means of "protest voting" for highly dissatisfied citizens.
We highlight how general levels of democratic dissatisfaction and political knowledge may contribute to the formation of attitudes that partially undermine the logic of representation. This has important policy implications especially in new democracies where the politics of representation is often problematic. When voters accept the informal motivations provided by the electoral competition, the political competition and representation in general is biased. Dissatisfied citizens accepting the use of clientelism may encourage political parties and politicians to invest resources in such practices rather than performing well in office or promoting policies.
Although limited to one country, our analysis provides several relevant insights about the public acceptance of clientelism. Further research can build on these grounds and can, for example, analyse the extent to which the acceptance of clientelism is translated into votes for parties using such practices. Attitudes and voting behaviour are sometimes correlated, and it is important to see whether this also happens in the presence of clientelism. Another avenue for research could distinguish clientelist actions between incumbent and challenger politicians, or by looking closely into the citizens' working environment (public vs private), prior experiences of clientelism, or personal knowledge about clientelistic practices. On a broader level, we may need to investigate the relation between political apathy and passivity. Existing literature tends to view the two as similar, but the evidence in this article clearly shows that being apathetic and cynical about politics does not necessarily mean that one is passive and does not seek to reignite their political participation. More work needs to be done also on the relation between non-participation and the values and views that underpin it.

Notes
Continued.

Variable
Question Coding Media exposure How often watch, listen or read news on TV, radio, newspapers or online news portals Seldom / Never (1), daily or almost daily (4)  Gender What is your gender? male (1), female (2)  Education What is the highest level of education you have successfully completed?
Elementary school (1), postgraduate studies (5)  Income What is the total yearly net income in your household?