Ethnic Voting as a Result of Information Deficiency
Evidence from the Former Yugoslav Republic of Macedonia (FYROM)Stefan Heger
Master Thesis
Political Science: International Relations Micro-foundations of Conflict Graduate School of Social Sciences,
University of Amsterdam
Word count: 20.083 Supervisor: Seiki Tanaka
Second reader: Maria Kranendonk June 2017
Abstract
Ethnic voting is often regarded as an inherent dominant form of electoral behaviour in ethnically divided societies, while the proportion of people that vote along ethnic lines can vary significantly between elections in a country and between ethnically divided countries. This research paper argues that people in an ethnically divided society engage in ethnic voting when they lack high-quality information to base their electoral decision on, therewith using ethnicity as a cognitive shortcut. Moreover, this paper hypothesizes that most people will rationally use quality information when it becomes available. Therefore it is expected that when high-quality information that becomes available is negative about a person’s co-ethnic party or candidate, this person is likely to defect from ethnic voting. By using an experimental survey approach aimed at the detection of within-subject variation in voting behaviour, this research has tested this information mechanism on the micro-foundations of ethnic voting empirically in the Former Yugoslav Republic of Macedonia (FYROM)1. Moreover, the concept of cognitive
mobilisation is applied in an attempt to explain why some people are more likely to use information to determine their electoral decision than others. The findings show strong support for the information mechanism, but no support for the application of cognitive mobilisation theory.
Keywords
Ethnic voting, information mechanism, cognitive shortcuts, cognitive mobilisation, survey experiments
1 This research paper refers to the country as The Former Yugoslav Republic of Macedonia (FYROM) in line with
United Nations Security Council Resolution 817 (1993). The use of this name by the author should not be interpreted as a political stance in the bilateral naming dispute with Greece.
Table of Contents
1. Introduction 1
2. Literature Review 3
3. Theoretical Framework 8
3.1 Conceptual Model 8
3.2 Information Deficiency and Ethnic Voting 9
3.3 Strategic Voting by Ethnic Micro-Groups 11
3.4 When High-Quality Information becomes Affordably Available 12
3.5 The Alternatives of Ethnic Voting 13
3.6 Cognitive Mobilisation 14
3.7 Scope Conditions of the Theoretical Framework 16
3.8 Other Mechanisms 17
4. Research Methodology 18
4.1 Case Selection 19
4.1.1 Ethnicity and Politics in the FYROM 19
4.1.2 Information Deficiency in the FYROM 21
4.1.3 External Validity 21
4.2 Research Design 22
4.2.1 Surveying Method 22
4.2.2 The Survey: Treatments and Operationalisations 24
4.2.3 Sampling Method and Sample 29
5. Results 32
5.1 The Information Mechanism 32
5.1.1 Information Deficiency and Ethnic Voting 32
5.1.2 Strategic Voting by Ethnic Micro-Groups 33
5.1.3 When High-Quality Information becomes Affordably Available 35
5.1.4 The Alternatives of Ethnic Voting 44
5.2 Cognitive Mobilisation 45 5.3 Other Mechanisms 52 6. Discussion 53 6.1 Limitations 59 7. Conclusion 60 References 63 Appendix 68
Appendix A: The Experimental Survey (English) 68
Appendix B: The Experimental Survey (Macedonian) 73
1 1. Introduction
Ethnic voting is a prominent form of electoral behaviour in ethnically divided societies in which a person votes for a candidate or party that identifies with the same ethnic group (Horowitz, 1985: 320). When a large proportion of people votes along ethnic lines, ethnicity determines the election outcome and therewith strongly influences politics in the country (ibid). This is the case in many ethnically divided societies all around the world, such as South Africa, India, Zambia, Uganda and Peru (Chandra, 2007; Conroy-Krutz, 2012; Ferree, 2006; Madrid, 2011; Posner, 2005). Ethnic voting has therefore received a large amount of academic and political attention. Often, ethnic voting is deterministically regarded as a fixed dominant phenomenon in ethnically divided societies. However, in some ethnically divided societies such as Senegal, Mozambique and Botswana, there is hardly any ethnic voting (Koter, 2016: 127; Kuhn, 2015: 7). Moreover, the degree of ethnic voting in a society can vary significantly between elections (Brown, 2005). A recent example thereof is the Former Yugoslav Republic of Macedonia (FYROM), in which the largest ethnic parties suddenly experienced firm losses against an ethnically mixed party in the December 2016 Parliamentary Elections, while ethnic voting had been dominant since the country’s independence (State Election Commission [SEC], 2016). Hence, as there is spatial and temporal variation in the degree of ethnic voting, it is not a given that ethnic voting is a dominant phenomenon in ethnically divided societies. Therefore, the question that rises is what motivates people to vote along ethnic lines.
One theory explaining these micro-foundations of ethnic voting that has attracted a growing amount of scholarly attention and support, is the information mechanism of ethnic voting (Birnir, 2006; Conroy-Krutz, 2012; Ferree, 2006; Posner, 2005). This theory holds that people in ethnically divided societies vote along ethnic lines when they lack high-quality political information to base their electoral decision on, such as information on the policy preferences and past performance of candidates or parties (ibid). Voters lack such information when it is too costly for them to acquire, meaning that it requires too much time and effort of them to obtain the information - if it is even possible (Chandra, 2007: 34). If this is the case, voters in ethnically divided societies resort to the ethnicity of candidates or ethnic labels of parties to predict which competitor in the elections is most likely to benefit them the most, as ethnicity information is cheaply available because it is physically visible or openly presented (Chandra, 2007: 34; Posner, 2005: 153). Assuming that co-ethnic candidates and parties are more likely to benefit them in the future and are more likely to have similar policy preferences, people vote along ethnic lines (Conroy-Krutz, 2012: 346; Ferree, 2006: 805). Ethnicity is
2 considered low-quality information when it is used to make inferences about the preferences and capabilities of candidates and parties, because it is likely to produce misconceptions (Conroy-Krutz, 2012: 350). However, in the absence of high-quality information, ethnicity is used as a cognitive shortcut – also called heuristic – to the preferences and capabilities of the competitors in the elections, which in turn produces ethnic voting behaviour (Ferree, 2006; Popkin, 1991: 14). Accordingly, this mechanism within the rationalist framework states that voters rationally use ethnicity as a heuristic to determine their electoral decision when they lack better information (Conroy-Krutz, 2012: 346; Ferree, 2006: 804).
Despite being popular and promising, the information mechanism has also increasingly been subject to criticism of scholars following a more psychological approach to ethnic voting that is based on social identity theory and is mostly referred to as the expressive voting mechanism (Adida, Gottlieb, Kramon & McClendon, 2017: 3; Ferree, 2006: 804). This theory states that people engage in ethnic voting to express and affirm their ethnic identity, because they experience psychological benefits like confidence from belonging to a certain identity group (Ferree, 2006: 805; Friedman, 2005; Horowitz, 1985; Tajfel, 1974). According to this mechanism, ethnic voting is thus not a rational electoral decision but an act of identity expression. Moreover, in reaction to the information mechanism, scholars following the expressive voting mechanism have argued that access to information does not change ethnic voting behaviour because people subconsciously process information in way that it benefits their co-ethnic candidate or party because they derive psychological benefits from identifying with their co-ethnic (Adida et al., 2017: 3). If this reasoning would be correct, the ethnic identity of a person conditions how information is processed and used and access to information thus only affirms an electoral choice based on ethnicity (ibid). Accordingly, these scholars hold that negative information about the co-ethnic candidate or party will not affect ethnic voting behaviour (ibid).
The information mechanism of ethnic voting is especially prone to this criticism because its application and empirical substantiation is still limited. Therefore, this research investigates the following research question:
How and to what extent does access to high-quality information influence the likelihood of ethnic voting of an individual?
Next to its theoretical relevance, this research is of societal relevance as ethnic voting is generally perceived as a problematic form of electoral behaviour (Adida et al., 2017; Ferree,
3 2006; Horowitz, 2014; Simonsen, 2005). This is first of all because it produces an ethnic census outcome to elections in which the ethnic majority automatically wins, a majority that especially in societies that experienced ethnic conflict in the past often refrains from providing for minorities (ibid; Bratton & Kimenyi, 2008). This in turn can undermine the democratic stability and might even lead to conflict. After all, as The Ljubljana Guidelines on Integration of Diverse Societies of the OSCE High Commissioner on National Minorities state: “(…) conflicts are frequently rooted in the denial of basic rights and in the systematic and/or systemic exclusion and alienation of entire communities” (HCNM, 2012: 2). Moreover, ethnic voting is seen as problematic because it reduces the democratic accountability as political parties have a guaranteed support base based on co-ethnicity instead of performance or proposed policies (Adida et al., 2017: 6; Horowitz, 2014). Therewith it also hinders progress in these countries.
The rest of this research paper continues as follows. First, the Literature Review places this research in the broader context of theories on ethnic voting. Hereafter the Theoretical Framework is presented, leading to the arguments of this research. After this, the empirical application of the argument on the case of the Former Yugoslav Republic of Macedonia (FYROM) will be set forth. This will be done by first presenting the research methodology including the experimental survey that was employed. The empirical findings are subsequently presented in order of the argument and its hypotheses, after which the implications of these findings are discussed more substantially in the Discussion section, ultimately leading to the Conclusion.
2. Literature Review
The following section is aimed at positioning this research in literature on ethnic voting behaviour and specifically theories on the factors leading to ethnic voting. First, ethnic voting will be conceptualized. Hereafter, a distinction will be made between theories on societal and institutional factors leading to ethnic voting and theories on the micro-foundations of ethnic voting while describing a shift of scholarly attention to the latter. While further engaging with these micro-foundations, the distinction will be made between theories with a psychological approach and theories with a rationalist approach. This leads to a discussion of the information mechanism within the rationalist framework, after which it will be highlighted where scholars of this framework have left a theoretical gap.
Elections and voting behaviour in ethnically divided societies have attracted a sheer amount of scholarly attention. This is first of all because in societies where ethnicity is highly salient and characterizes a societal divide, ethnicity tends to be prominently present in politics
4 (Birnir, 2006; Horowitz, 2014; Marcos-Marne, 2017; Simonsen, 2005). Therefore, many scholars have focussed on how ethnicity influences politics and more specifically how it influences elections. This branch of research finds its origins in the United States where scholars engaged with the voting behaviour of newly arriving ethnic groups from the European mainland (Dahl, 1961; Gabriel, 1972; Wolfinger, 1965). There it was found that people who self-identify with a certain ethnic group, often vote along ethnic lines by selecting a candidate or party of the same ethnicity in the ballot box (ibid). It was Wolfinger (1965: 896) who first coined the concept of ethnic voting, which he defined as a person’s affinity with a certain party that can not be explained by other demographic characteristics than ethnicity, and the willingness to cross party lines in order to vote for a candidate of the same ethnicity. This research follows the more straightforward definition of Horowitz (1985: 320), who defines ethnic voting as ‘voting for the party identified with the voter’s own ethnic group’. Herein, an ethnic group is understood as ‘a politically and socially constructed group identity based on real or perceived descent’ and membership of an ethnic group is based on the principle of self-identification (Adida et al., 2017: 6; Chandra, 2006; HCNM, 1999: 19).
Identifying why people vote along ethnic lines has been the main focus of literature on ethnic voting. Within this branch of literature there has been a general shift in attention from institutional and societal factors leading to ethnic voting based on aggregated data, to the micro-foundations of ethnic voting and data on the individual level of analysis (Ferree, 2006; Nakai, 2009).
One important example of a societal factor that has been widely assumed to be related with ethnic voting is socioeconomic status, a concept that is mostly defined as a combination of education, income and occupation (Dahl, 1961; Gabriel, 1972; Nakai, 2009; Wolfinger, 1965). Herein, the theorized and often empirically confirmed association is that between the increased socioeconomic status of an ethnic group with a decreased likelihood of ethnic voting (Dahl, 1961; Gabriel, 1972; Wolfinger, 1965). Although there has been plenty of debate about the exact underlying mechanism and tipping points, the main rationale behind this theorized causal relation is that an increased socioeconomic status of a group will decrease the salience of its ethnicity and tendency to affirm the ethnic identity when voting (Dahl, 1961). Another more structural main line of inquiry has been the influence of the electoral system itself on the manner in which different ethnic groups vote (Birnir, 2006; Horowitz, 2014; Nakai, 2009; Van de Walle, 2003). The main debates have been about which electoral systems reduce the likelihood of ethnic voting the most and which lead to further political fragmentation (i.a. Birnir, 2006), how parties and candidates use the system in a clientalistic way in order to gain
5 political support along ethnic lines (i.a. Van de Walle, 2003) and how power-sharing arrangements that are meant to mitigate the effects of ethnic voting can actually contribute to an increased degree of ethnic voting (i.a. Horowitz, 2014).
However, contemporary scholars have mostly focussed on the micro-foundations of ethnic voting (i.a. Conroy-Krutz, 2012; Ferree, 2006). The main question of interest here is what explains individual ethnic voting behaviour. Theories answering this question can be divided into those based on social identity theory on one side, and rationalist approaches on the other side (Adida et al., 2017; Conroy-Krutz, 2012; Ferree, 2006). The theoretical direction based on social identity theory is often referred to as the expressive voting mechanism, which holds that individuals gain psychological benefits like self-esteem from affirming their own social identity and do so by voting along ethnic lines (ibid; Horowitz, 1985; Tajfel, 1974). Ethnic voting in this theory is thus an expression of allegiance, not a rational choice in which alternatives are weighed (Ferree, 2006; Horowitz, 1985; Nakai, 2009). According to this line of reasoning, a person votes for the party of which the identity is most in line with their own identity (Friedman, 2005). As ethnic social identities tend to dominate in ethnically divided societies, people vote in line with their own ethnic identity and thus engage in ethnic voting, according to this approach (Birnir, 2006).
The other theoretical direction in research on the micro-foundations of ethnic voting is rationalist in nature (Conroy-Krutz, 2012; Ferree, 2006; Popkin, 1991; Rabushka & Shepsle, 1972). Scholars in this direction assume that voters make a rational or reasoned choice in the ballot box (Ferree, 2006; Popkin, 1991). According to most scholars following the rational line of reasoning, an individual votes along ethnic lines because individuals of the same ethnic group can benefit from collective action (Conroy-Krutz, 2012; Ferree, 2006). Often this collective action is spurred by political elites through the clear use of ethnic labels and symbols in elections, sometimes followed by patronage to co-ethnics once power has been obtained (Conroy-Krutz, 2012; Nakai, 2009; Van de Walle, 2003).
Within this rationalist framework, the so-called information mechanism has increasingly gained scholarly support (Birnir, 2006; Conroy-Krutz, 2012; Ferree, 2006). According to the line of reasoning of the information mechanism, ethnic voting is the result of a lack of political information on which voters can rationally base their electoral choice (Conroy-Krutz, 2012; Ferree, 2006). This lack of information occurs when obtaining useful information is too costly to voters, in the way that it costs them too much time or effort to obtain the information (Conroy-Krutz, 2012). When voters lack useful information to rationally base their electoral choice on, they use cognitive shortcuts – or heuristics – to decide who to vote on
6 (Conroy-Krutz, 2012; Downs, 1957; Ferree, 2006; Popkin, 1991). In ethnically divided societies, the ethnic labels of parties or the ethnicity of candidates is a cheap source of information that can be used as such a cognitive shortcut, because it is mostly simply visible (Birnir, 2006; Conroy-Krutz, 2012). As this is easily available information, the ethnicity of candidates and ethnic labels of parties can easily be matched to a person’s own ethnic identity (Conroy-Krutz, 2012). Lacking more useful information, this ethnicity match is used to arrive at an electoral choice, based on the belief that a party or candidate of the same ethnicity is more likely to act in this person’s interest than a party or candidate that is not of the same ethnic group (ibid). Thus, according to the information mechanism, a lack of affordably available political information in an ethnically divided society produces ethnic voting behaviour.
Although this strand of research has gained popularity, some critical scholars have argued that people are likely to process information in a way that it benefits their co-ethnic (Adida et al., 2017). According to them, access to information has no effect on the likelihood of ethnic voting, as negative information about the co-ethnic party is disregarded while negative information about other parties is processed (ibid).
Thus, if the expressive voting mechanism is correct, we would expect that most individuals in an ethnically divided society always vote along ethnic lines, even if they are provided with information that depicts their co-ethnic party negatively in comparison with other parties. This expectation is also in line with scholars critical of the information mechanism, as it would be the expectation that this negative information would be disregarded in benefit of the co-ethnic candidate or party. If the rational direction of literature is correct however, we would expect that most people vote along ethnic lines if they rationally believe they can benefit from collective action of the ethnic group. More specifically, if the information mechanism within the rationalist framework is correct, we would expect that most individuals vote along ethnic lines when they lack affordable political information, but most of them will defect from ethnic voting when they receive information that depicts the co-ethnic party negatively.
In this research I follow the argumentation of the information mechanism of ethnic voting as it especially promising because it provides a strong explanation for spatial and temporal variation in the degree ethnic voting. Recognizing the information mechanism of ethnic voting as a promising direction of enquiry, this research follows its line of reasoning while addressing several existing gaps that are in need of theoretical sealing to further establish the mechanism. First of all, as mentioned in the introduction, the information mechanism of ethnic voting is in need of further empirical substantiation. Although there is increasing academic interest and support for the mechanism, its systematic application and empirical
7 testing of its falsifiability based on micro-level data is still limited. Such testing is especially needed in light of recent literature disregarding the effect of access to information based on the idea that people subconsciously process information in a way that it benefits their co-ethnic (i.a. Adida et al., 2017). Moreover, the geographical scope in which the mechanism has been tested empirically is still heavily reliant on African societies while ethnic voting is phenomenon that occurs in societies around the globe. This research contributes to addressing these empirical gaps by applying the information mechanism to the Former Yugoslav Republic of Macedonia (FYROM) where ethnic voting has been prominent. It does so using an experimental survey design as used by Conroy-Krutz (2012) that is aimed at the detection of within-subject variation of ethnic voting behaviour between fictional election conditions. A theoretical and methodological gap in this method that this research also addresses is the exclusion of an ethnically mixed party in the electoral options of respondents, which this research adds as this is more in congruence with reality than elections reduced to two contrasting ethnic parties. This contribution will be discussed in more detail in the methodology section of this research.
Furthermore, one important element of the information mechanism that remains underdeveloped to which this research aims to contribute, is the explanation of the variation in the degree to which individuals in the same ethnically divided society acquire and use information. Even research on an individual level of analysis often explains the information deficiency of voters by referring to the ‘information poor’ society in which these voters live (Conroy-Krutz, 2012: 347). Others have tried to measure the amount of information individual voters inhabit by testing their knowledge on political substance, which is ‘notoriously difficult’ and prone to many internal validity issues (Mattes, 2005: 44). Borrowing from general literature on voting behaviour and specifically party loyalty, this research applies the concept of cognitive mobilisation to the information mechanism. Cognitive mobilisation is a combination of an individual’s degree of education and political interest, which is used to explain the extent to which a person can develop an own electoral decision based on available information (Dalton, 1984, 2013: 197). Although scholars of party politics have used the concept to describe how an increased degree of cognitive mobilisation leads to dealignment from party loyalties based on structural factors such as social class or religion, the concept as operationalised in the beforementioned manner has yet to be applied to the information mechanism of ethnic voting (ibid; Boonen, Falk Pedersen & Hooghe, 2017). By doing just so, this research is of academic relevance to a more detailed development of the information mechanism.
8 3. Theoretical Framework
The ensuing section will systematically outline the theory underlying the argumentation of this research and the hypotheses following from these arguments. First, the three main hypothesized relations will be presented in a conceptual model to provide an overview. Subsequently building on the information mechanism of ethnic voting and cognitive mobilisation theory, the arguments of this research are developed in order to arrive at a total of three hypotheses and two sub-hypotheses. After doing so, alternative mechanisms are laid out as they have also been incorporated in this research.
3.1 Conceptual Model
The main argument of this research is that there is a relationship between the affordable availability of high-quality information and the likelihood of ethnic voting. In line with the information mechanism it is first of all expected that if no high-quality information is available, most people will engage in ethnic voting (H1). Following from this argumentation is the expectation that voters will rationally use high-quality information when it is available. Therefore, the availability of high-quality information that is negative about the co-ethnic party competing in the elections reduces the likelihood of ethnic voting, while when this information is positive the likelihood of ethnic voting increases (H2). Furthermore, it is expected that people with a higher degree of cognitive mobilisation are more likely to use the available high-quality information (H3). As can be seen in the model, cognitive mobilisation increases the magnitude of the effect of high-quality information described in H2.
9 3.2 Information Deficiency and Ethnic Voting
The main argument of the information mechanism of ethnic voting is that when voters lack high-quality information to base their electoral choice on, they use the easily available low-quality information of the ethnic labels of parties as an indicator of which party is most likely to represent their interests well and therefore engage in ethnic voting (Birnir, 2006; Conroy-Krutz, 2012; Ferree, 2006; Posner, 2005). As voters use ethnicity information as cues on how likely parties are to act in their interests, ethnicity serves as a cognitive shortcut, also called heuristic (ibid). In the following paragraphs this argument will be further unravelled step by step in order to establish the exact mechanism.
A first important assumption of the information mechanism is that ethnic voting happens because voters lack useful political information (Ferree, 2004: 1; Posner, 2005: 154). Scholars of the mechanism have therefore mostly focussed on ‘information poor’ societies (Conroy-Krutz, 2012: 347). There are no straightforward criteria of when a society qualifies as information poor, but two main characteristics that each separately brand a society information poor are often named in literature on ethnic voting. First of all, a society is perceived as information poor when mass media infrastructures are not well developed, or media is of insufficient quality and does not or hardly provide substantial objective information that voters can use (Posner, 2005: 154). Secondly, a society is often information poor when its party and political system is in transition (Conroy-Krutz, 2012: 347). In these societies, individual voters lack information because it requires too much effort and time of them to acquire, making the information too costly (Birnir, 2006: 29; Downs, 1957: 139).
The next assumption of the information mechanism is that when voters lack affordable political information, they resort to the use cognitive shortcuts (Conroy-Krutz, 2012: 348; Downs, 1957: 139; Ferree, 2006: 803). This argument is well-grounded into theory of voter rationality. Herein, it follows the rationalist assumption that voters base their electoral choice on a rational decision (Healy & Malhotra, 2013). The degree to which a voter can use rationality first of all depends on the degree of political information available (Downs, 1957: 137). As no voter can acquire and grasp all information possible, voters engage in forming a reasoned choice (Popkin, 1991: 7). The less information voters have, the more they rely on low-information reasoning by using cognitive shortcuts, often referred to as heuristics (ibid; Conroy-Krutz, 2012: 349). Popkin (1991: 14) describes an informational shortcut as ‘a substitute for more complete information about parties and candidates’.
Such informational shortcuts consist of easy to acquire information about what party or candidate is most likely to pursue the interests of the voter in question (Ferree, 2006: 805).
10 Therefore, these informational shortcuts are mostly based on the voter’s personal characteristics or ideology, which is easy to match with the what Mattes (2005: 57) calls the ‘traits and attributes’ of a party or candidate. In ethnically divided societies, cognitive shortcuts tend to be based on ethnicity (Birnir, 2006: 30). One important reason for that is that ethnicity information is ‘costless’ to obtain, as it is clear from the physical characteristics of candidates or symbols and language used by parties (Chandra, 2007: 34). Lacking better information, voters draw conclusions on the preferences, capabilities and electability based on the ethnic label of a party (Conroy-Krutz, 2012: 349). These conclusions are first and foremost established on the belief that a co-ethnic party is more likely to benefit its co-ethnic constituency (idem: 350; Ferree, 2006: 805). Herein, one important consideration that people use to decide that the co-ethnic party is the best electoral choice under these conditions is that ethnic collective action is believed to lead to benefits of individual members of the ethnic group (Chandra, 2007: 60; Conroy-Krutz, 2012: 346). Next to this, voters also use ethnicity information to assess the electability of party (Conroy-Krutz, 2012: 350; Posner, 2005: 17). Ethnicity provides a shortcut to assess the electability as people are roughly aware of relative sizes of ethnic groups in ethically divided societies (ibid). As Conroy-Krutz (2012: 350) states: “strategic voters will determine their final choice on the basis of competitor preferences, capabilities and electability. Ethnicity can provide cues about all three”.
A last important consideration for the information mechanism is how the quality of information to voters is assessed. As Healy and Malhotra (2013: 287) argue, all voters are likely to make mistakes in their rationality as nobody has all possible information available. This is of course even more the case if voters rely on cognitive shortcuts (idem: 288). Conroy-Krutz (2012: 350) provides a useful definition of the quality of information, as the degree to which ‘it helps the recipient to reduce errors in predicting competitors’ preferences, capabilities and viability’. Therefore, ethnicity is considered as low-quality information when it is used to predict the preferences, capabilities and viability of candidates or parties, as these inferences are likely to lead to false assumptions (ibid). Thus, high-quality information is political information that does not require such inferences. High-quality information on the political preferences of parties can for instance be a party programme (Ferree, 2006: 808). Moreover, high-quality information on the capabilities or candidates or parties can for instance be performance indicators such as the state of the economy when a candidate or party was in power (idem: 809; Conroy-Krutz, 2012: 351). Furthermore, high-quality information on the electability of candidates and parties is for instance an election poll predicting the outcome of the upcoming elections (Conroy-Krutz, 2012: 351).
11 Although promises by candidates and parties might not be kept, scandals about past performance might turn out to have been false and polls are just a prediction, it is the best information available to voters in democracies all around the word. Moreover, no far-fetched inferences are needed to make predictions on how much a party is likely to benefit a certain voter. Types of high-quality information to voters that are often named in literature on ethnic voting are information on the past performance of parties, the popularity of parties, whether or not parties are involved in corruption practices, policy preferences of parties and material distribution promises by parties (Conroy-Krutz, 2012: 354; Ferree, 2006: 808-809). These types of information will be further explored in the methodology section of this research.
The central argument of the information mechanism is thus that voters use ethnicity as a cognitive shortcut when they lack affordable high-quality information. If this argument is correct, we should see that most people vote along ethnic lines when they do not have high-quality information available. This leads to the following first hypothesis of this research:
H1: When the only affordable information to voters is the ethnic labels of the competing parties, most voters will vote for the co-ethnic party.
3.3 Strategic Voting by Ethnic Micro-Groups
As outlined before, people use ethnicity as a cognitive shortcut to establish the electability of parties (Conroy-Krutz, 2012: 350; Posner, 2005: 17). After all, a party that is elected to power is able to pursue its proposed policies. The electability of a party first of all depends on the amount of electoral support that a party is likely to receive (Birnir, 2006: 118). A party that is not very likely to receive many votes, is not very likely to be able to obtain political power (ibid; Conroy-Krutz, 2012: 350). The threshold for this depends on the electoral system of a country (Birnir, 2006: 45). Especially in systems in which the winner takes it all, the chances for parties identifying with an ethnic minority are nihil (ibid) . In other systems, especially a proportional system, the largest party mostly forms a coalition with smaller parties that fit its plans the most (ibid). Herein, parties that identify with ethnic minorities generally have a bigger chance of getting into power (ibid). Some ethnically divided societies also have power-sharing mechanisms in place that prescribe the inclusion of ethnic minorities in politics (Horowitz, 2014). These mechanisms range from unofficial pre-election agreements between moderate members of different ethnic groups, to constitutionally guaranteed inclusion in the government of ethnic groups of a certain size (ibid).
12 No matter how the electoral system is designed, the very smallest marginal ethnic minority groups are generally unlikely to obtain decision-making power in any electoral system (Conroy-Krutz, 2012: 352). These groups are referred to as ‘micro-groups’ by Conroy-Krutz (idem: 360) and defined as ethnic groups making up less than 5 percent of the population of a country. Parties identifying with micro-groups are not likely to be elected to power. If this argument is correct, individuals that belong to these micro-groups are expected to be less likely to vote along ethnic lines, even if ethnicity is the only information available. The counterfactual of that expectation is that members of ethnic micro-groups are as likely or even more likely to vote for their co-ethnic party. This leads to the following sub-hypothesis of H1:
H1a: When the only affordable information to voters is ethnicity, voters who self-identify with an ethnic micro-group2 are less likely to vote along ethnic lines than others.
3.4 When High-Quality Information becomes Affordably Available
If most people use cognitive heuristics when they lack high-quality information, then we should expect that most people use high-quality information to determine their electoral choice when it becomes available. Hence, when such information is negative about the co-ethnic party, we should expect that most people defect from ethnic voting. The first hypothesis alone does not fully cover the information mechanism. After all, if people vote along ethnic lines in a situation where they lack affordable high-quality information, it is not possible to detect whether people select their co-ethnic party to use ethnicity as a cognitive shortcut as the information mechanism prescribes, or in order to affirm their social identity as the competing expressive voting mechanism states (Ferree, 2006: 804). If high-quality information is available, the substance of this information will determine whether or not an individual votes for the co-ethnic party (Conroy-Krutz, 2012: 351). When the high-quality information depicts the co-ethnic party as the best choice, a voter is most likely to vote for the co-ethnic party based on this information. However, when the high-quality information depicts the co-ethnic party negatively vis-à-vis parties that are not of the same ethnicity, voters are presented with a dilemma (ibid). If the information mechanism is correct, we should expect most voters to refrain from voting for the co-ethnic party if this party is depicted negatively in comparison to the other competing parties. More importantly, if this theory is correct, we should expect that people who voted along ethnic lines when ethnicity of parties was their only affordable information, defect from ethnic voting
13 behaviour when they are presented with high-quality information that is negative about the co-ethnic party and positive about the other parties.
The counterfactual of this argument is that even if high-quality information is available that depicts the co-ethnic party negatively and other parties more positively, most people would still vote for their co-ethnic party. This would be in line with the competing expressive voting mechanism and scholars claiming that voters process information in a way that it benefits their co-ethnic party (Adida et al., 2017).
Following the rationalist line of reasoning of the information mechanism of ethnic voting, this research arrives at the following hypothesis:
H2: When high-quality information is available and represents the co-ethnic party negatively in comparison to other parties, the support rate for the co-ethnic party will be lower than when ethnicity is the only affordable information.
3.5 The Alternatives of Ethnic Voting
The information mechanism holds that voters use high-quality information strategically when it is available (Conroy-Krutz, 2012: 350). Hypothesis 2 states that when high-quality information is negative about the co-ethnic party, people are less likely to vote along ethnic lines and thus more likely to defect from ethnic voting. In most research the party to which voters defect is described as the party of another ethnic group, for instance by Conroy-Krutz (2012: 360). However, in reality, in many ethnically divided societies voters also have the option of voting for a party that is ethnically mixed and thus not clearly identifies with just one ethnic group. This research adds to the existing field by also incorporating this alternative into the theory and empirically testing it. Surely the threshold of voting for the party of another ethnic group is higher than voting for a party that is ethnically mixed. As Ferree (2006: 806) describes, voting for the party of another ethnic group is a two-step process is which first the party of a person’s own ethnic group is rejected based on the information, and secondly the party of another ethnic group should be accepted. Despite underlining the additional threshold this forms, Ferree (2006) does not further incorporate it in her research and only mentions that it leads to many voters in doubt of what to do. In line with the information mechanism, this research expects that if people defect from ethnic voting, they are more likely to do so to an ethnically mixed party than to the party of another ethnic group. The counterfactual of this expectation is that it does not matter for voters whether the party they defect to is ethnically mixed or identifies with another ethnic group and thus as many people defect to both
14 alternatives. Another counterfactual that is deemed highly unlikely is that people rather defect to the party of another ethnic group. This argument leads to the following sub-hypothesis:
H2a: When high-quality information is available and represents the co-ethnic party negatively in comparison to an other-ethnic party and an ethnically mixed, defection is more likely to the ethnically mixed party.
3.6 Cognitive Mobilization
As the information mechanism of ethnic voting is generally applied to information poor societies, scholars often do not account for variation between individual voters in the degree to which they use high-quality information to inform their voting behaviour. Building on the information mechanism and cognitive mobilisation theory, I here develop my argument to explain why some voters in a society do use the available information to establish their electoral choice while others do not. This will be done by first presenting cognitive mobilisation theory, where after the connection will be made with the information mechanism, in order to arrive at the third hypothesis of this research.
Cognitive mobilisation has been a prominent theory in literature aimed at explaining voting behaviour (i.a. Boonen et al., 2017; Dalton, 1984; Lau & Redlawsk, 2001). The concept was first coined by Dalton (1984), who used it to describe a societal move away from voting based on partisan alliances to voting based on individual considerations. He describes how voters used to be politically unsophisticated in the sense that they did not know or understand what was going on in politics and therefore used a personal social cues like religion or class as cognitive shortcut to develop loyal voting behaviour towards a party matching this characteristic (idem: 264). However, when the electorate of advanced democracies became higher educated and developed more political interest, they used this to acquire and process information that became increasingly available through mass media (idem: 265). Because of this most voters dealigned from their party loyalties based on their personal social cues to develop an individual electoral decision based on the acquired information – a process he calls cognitive mobilisation (idem: 265).
Whereas the concept was thus first mainly used to describe a societal shift, it has later also widely been applied on the individual level, including by Dalton (2013) himself. On the individual level, cognitive mobilisation is conceptually operationalised by Dalton (2013: 197) as ‘a simple additive combination of education and general interest in public affairs’. So, an individual that is higher educated and has a higher degree of political interest is more
15 cognitively mobilised. When information increasingly becomes available, a more cognitively mobilised individual is thus more likely to defect from partisan alliances based on social cues to develop an individual electoral decision (idem: 63).
There are many similarities and overlapping conceptions between cognitive mobilisation theory and the information mechanism of ethnic voting, making it very applicable. One element of cognitive mobilisation theory that is also prominent in literature on the information mechanism is the access to information. According to cognitive mobilisation theory, one reason voters were politically unsophisticated was that there was less access to information through mass media (Dalton, 1984: 265). Likewise, according to the information mechanism, one reason to brand a society as information poor is when mass media infrastructures are not well developed (Conroy-Krutz, 2012: 347). Both theories describe a process in which voters resort to the use of their personal characteristics to find a matching candidate or party when they do not have information available, one referring to it as social cues and the other as cognitive shortcuts or heuristics (Conroy-Krutz, 2012: 246; Dalton, 1984: 264). Herein, both theories follow a very similar logic with regard to the lack of information on a societal level.
However, there is less theoretical overlap when the use of information by individual voters is concerned, despite an obvious potential theoretical congruence. Where cognitive mobilisation theory explains how an individual’s combined degree of education and political interest matters for the likelihood that this person uses information when it becomes available, explanations on this individual degree of information are limited within literature on the information mechanism (Dalton, 2013: 197). If scholars consider personal characteristics in the application of the information mechanism, they mostly do so by using them as control variables without further analysis or discussion. Conroy-Krutz (2012: 361) for instance includes political sophistication – operationalised as a combination of media consumption and education – as a control variable in his analysis of the likelihood of ethnic voting in an information poor society. However, he does not discuss the assumed importance of the concept and does not analyse the statistical results (idem: 362). Moreover, he omits the variable when analysing the use of information by voters, while this is where the concept is likely to be of influence (ibid). Mattes (2005: 42) does discuss cognitive mobilisation and the information mechanism, but only applies it by using aggregated data to explain ethnic voting on a societal level. Cognitive mobilisation as a personal indicator has yet to be structurally applied to the information mechanism.
This research argues that the higher the degree of cognitive mobilisation of a voter, the more likely this voter is to use high-quality information when it becomes available to establish
16 an electoral decision. Therefore, if high-quality information is available that depicts the co-ethnic party negatively in comparison with other parties, more cognitively mobilised voters are less likely to engage in ethnic voting. Likewise, if high-quality information is available that depicts the co-ethnic party positively in comparison with other parties, more cognitively mobilised individuals are more likely to change to voting for the co-ethnic party based on information. Furthermore, of the individuals that vote along ethnic lines when ethnicity is the only affordable information, the more cognitively mobilised individuals are more likely to defect from ethnic voting when high-quality information becomes available that depicts the co-ethnic party negatively is comparison with other parties. The counterfactual of this argument is that cognitive mobilisation does not influence the likelihood of using high-quality information to determine an electoral decision. This leads to the following and final hypothesis of this research:
H3: The higher the degree of cognitive mobilisation of an individual, the less likely this person is to vote for a co-ethnic party when high-quality information is available and represents the co-ethnic party negatively in comparison to other parties.
3.7 Scope Conditions of this Theoretical Framework
There are several scope conditions for this theoretical framework combining the information mechanism of ethnic voting and cognitive mobilisation theory to be applicable.
One condition for the argument to be applicable is that there is no ongoing armed ethnic conflict in the society of interest. If there is an armed conflict ongoing, information of the antagonistic nature of other ethnic groups will be weighed a lot heavier than the types of high-quality information discussed in this research. People will be even more inclined to support their co-ethnic party or leader, fearing repression if a party or leader of another ethnicity wins the conflict. Hardly ever though are elections held during an ongoing armed ethnic conflict. Inter-ethnic tensions in an ethnically divided society do not violate the scope conditions of the mechanism, but are likely to be of influence on the results and should therefore be discussed when analysing the results.
Another scope condition is that the information mechanism only applies to democratic societies in which more than one viable party partakes in elections and voters are generally free to choose the party of their preference. This seems an obvious part of democracy, but in practice tends to be different in some post-conflict democracies. As discussed before, one way in which voters determine their electoral decision is by weighing the viability of the competing parties
17 (Conroy-Krutz, 2012: 350). When no high-quality information is available, this is done by using the ethnic group sizes as a cognitive shortcut about the likely amount of support for certain ethnic parties (ibid). When high-quality information on the popularity of parties is available, such as pre-elections poll outcomes, voters are likely to use this information about the viability of the competing parties to decide which party to vote on (idem: 351). However, in some incomplete democracies, the party winning the election is predetermined. This can be because voters are threatened to be punished if they do not support the ruling party or election outcomes. When there is only one viable winner of the elections because of these reasons, the theoretical framework does not apply as voters are heavily restricted in the use of their rationality to determine their own electoral decision.
Moreover, the basic assumption of the information mechanism that most people will vote along ethnic lines when they lack access to high-quality information, depends on the electoral system of a country. This has been outlined before when discussing hypothesis 1a that holds that members of ethnic micro-groups are less likely to vote along ethnic lines because they use the information of their marginal ethnic group size to assess the viability of their co-ethnic party getting in to power. According to the framework, other co-ethnic minorities do vote along ethnic lines when they lack high-quality information. However, for this argument to be correct, ethnic minority groups should have some likelihood of being included in any form of decision-making. This likelihood can be present because of a low-enough threshold of getting into parliament as an ethnic minority party or even power-sharing arrangements that guarantee the inclusion of ethnic minorities of a certain size in decision-making. Thus, for the argument of this research to be applicable, ethnic minority groups should have some likelihood of obtaining any power.
3.8 Other mechanisms
There are two main alternative explanations in literature about the likelihood of ethnic voting of individuals. One explanation is about the influence of the socioeconomic status of a person of ethnic voting behaviour, while the other is about inter-ethnic attitudes. If cognitive mobilisation is a strong predictor of the likelihood of ethnic voting when high-quality information becomes available, it has to be robust to the influence of the inclusion of these variables in the analysis.
A first alternative mechanism is concerned with the influence of the socioeconomic status of a person (Bratton & Kimenyi, 2008; Dahl, 1961; Gabriel, 1972; Wolfinger, 1965). Generally, socioeconomic status is defined as a combination of education, income and
18 occupation (American Psychological Association [APA], n.d.). As socioeconomic status might be a possible confounder when it is not controlled for, most scholars following the information mechanism include and operationalize the concept to control for it to some degree (i.a. Bratton & Kimenyi, 2008; Conroy-Krutz, 2012; Ferree, 2006). The main hypothesis on the influence of socioeconomic status on ethnic voting is that a higher socioeconomic status will decrease a person’s likelihood of ethnic voting (Dahl, 1961). The underlying mechanism for this is that a higher socioeconomic status produces economic interests that are inconsistent with voting behaviour along ethnic lines (Wolfinger, 1965). Moreover, a higher socioeconomic status is associated with occupational mobility, which increases the likelihood of working with people from other ethnic groups and therefore decreases the likelihood of ethnic voting (ibid). Applied to the information mechanism of ethnic voting, a higher socioeconomic status is associated with a higher likelihood of defection from ethnic voting when high-quality information becomes available. It is therefore an important alternative explanation to control for when analysing the effect of cognitive mobilisation on the likelihood of defecting from ethnic voting.
Another concept that is often used to explain ethnic voting behaviour is the intergroup attitude of a person, being the attitude towards groups that this person is not a part of (Schlueter & Scheepers, 2010). Intergroup attitudes and especially inter-ethnic attitudes are often strongly related to voting behaviour of different ethnic groups (Ichino & Nathan, 2013; Oliver & Wong, 2003). Applying the theory to the information mechanism of ethnic voting, a more negative attitude towards other ethnicities is associated with a higher likelihood of ethnic voting and thus a lower likelihood of defecting from ethnic voting if high-quality information is available. As a person’s attitude towards other ethnicities is likely to influence the likelihood of defecting from ethnic voting, it has to be controlled for.
4. Research Methodology
The following section will outline the methodology of this research. This will be done by first explaining the selection of the case of the Former Yugoslav Republic of Macedonia (FYROM) to which the theoretical framework has been applied. Thereafter, the research design will be set forth. Herein, the research instrument and the advantages and disadvantages of its experimental within-subject variation survey design will be discussed. Finally, the sampling method and surveying procedure will be explained.
19 4.1 Case Selection
The theoretical framework has been applied to the Former Yugoslav Republic of Macedonia (FYROM), which was selected on the basis of both the prominent presence of ethnic voting behaviour and the general deficiency of high-quality information. Both these elements will be discussed subsequently.
4.1.1 Ethnicity and Politics in the FYROM
The Former Yugoslav Republic of Macedonia (FYROM) is often used in literature as a textbook example of pervasive ethnic voting behaviour (Bieber, 2005; Maleska, 2010; Wagner, 2013). The FYROM is a strongly ethnically divided country in which ethnicity is highly salient (Bieber, 2005; Wagner, 2014). According to the last official demographic census that was carried out in 2002 and published three years later, ethnic Macedonians form the ethnic majority, making up 64.2 percent of the country’s population (State Statistical Office of the Republic of Macedonia [SSO Macedonia], 2005: 34). Ethnic Albanians form the largest ethnic minority group with 25.2 percent of the population of the FYROM (ibid). Other ethnic groups in the country each account for less than 5 percent of the population, minority groups that Conroy-Krutz (2012: 360) refers to as ethnic micro-groups. Of these micro-groups, the census counts 3.9 percent Turks, 2.7 percent Roma, 1.8 percent Serbs, 0.8 percent Bosniak and 0.5 percent Vlach (SSO Macedonia, 2005: 34). Some other even more marginal minorities such as Torbesh and Croats are aggregated under the 1 percent ‘Other’ by the census (ibid).
The country came into existence as the ‘Republic of Macedonia’3 after gaining
independence from Yugoslavia in 1992, despite territorially being challenged by Serbia and Greece (Bieber, 2005: 1). The ethnic Macedonians of Yugoslavia now had their own country and marginalised the large Albanian minority and other smaller ethnic minorities (ibid). As by far the largest ethnic minority group of the country, ethnic Albanians strongly argued for more extensive rights, in particularly in education and representation in public administration and the police force (ibid). Ethnic Albanians have participated in Macedonian politics from the first elections in the country in 1990 (ibid). The country was strongly ethnically divided with
3 The territorial challenge with Greece led to a bilateral dispute over the naming of the country. In order to be
admitted to the United Nations, the country had to change its official name to Former Yugoslav Republic of Macedonia (FYROM), in line with UNSC Resolution 817 (1993). However, the country still refers to itself as the Republic of Macedonia, while Greece also challenges the country’s new name and therefore blocks it from becoming a NATO or EU member state. Without taking a political stance in this bilateral dispute, this research employs the country name agreed upon in the United Nations Security Council, being the Former Yugoslav Republic of Macedonia (FYROM).
20 minimal communication and cooperation between Macedonians and Albanians, but for a long time relations remained peaceful (ibid).
However, following the severe ethnic conflicts in other parts of former-Yugoslavia, inter-ethnic relations also turned bloody in Macedonia in 2001 (Wagner, 2014: 485). State security forces of Macedonia were attacked by heavily armed insurgents of the Albanian National Liberation Army, also referred to as NLA or UÇK, seeking some degree of self-determination for the Albanian-dominated areas of the country (ibid). The attack led to an intense period of fighting between the two sides, which left the country ‘deeply scarred’ (ibid; Bieber, 2005: 1). The inter-ethnic violence led to the death of about 400 people, of which 70 can have been classified as battle-related deaths (Vetterlein, 2010; Wagner, 2014). Largely under pressure of the international community the conflict ended relatively quickly in comparison with the other conflicts in the former-Yugoslavia, with both sides signing the Ohrid Framework Agreement (OFA) (Bieber, 2005; Wagner, 2014). The agreement laid out the framework for increased power-sharing by constitutional changes that should improve the situation of the Albanian minority in Macedonia (ibid). Although the agreement significantly increased the rights of ethnic Albanians, the success of the OFA was also strongly limited due to a chronical lack of implementation (Bieber, 2005; Taleski, 2011). A coalition consisting of the strongest Macedonian party and an Albanian party has become a custom in the country, although it is not constitutionally guaranteed (ibid). Moreover, the rights of other ethnic minorities were not included in the OFA.
Therefore, the rights of ethnic Albanians in Macedonia still dominate the political agenda in the FYROM. Several ethnic Albanian parties, of which some are direct political decedents of the Albanian insurgency movement, are participating in every election (Taleski, 2011). On the other side of the ethnic divide, several parties identify as ethnic Macedonian and are mainly focussed on representing the ethnic Macedonian part of the population (ibid). The main protagonist of Macedonian nationalism has been the Internal Macedonian Revolutionary Organization – Democratic Party for Macedonian National Unity, mostly simply referred to by its Macedonian abbreviation VMRO-DPMNE (idem: 173). It was this party that was in power when the conflict in 2001 broke out and its politicians are mostly strongly reluctant of increasing ethnic Albanian rights in Macedonia (idem: 174). This recently became painfully clear when supporters of the party stormed the Parliament in the capital Skopje and attacked MPs that are in favour of increasing the rights of Albanians (Marusic, 2017). The attack in which several MPs were seriously injured followed after a period of political unrest because the Macedonian President Ivanov - a VMRO-DPMNE politician - refused to provide a political
21 mandate to a newly formed coalition led by the ethnically mixed SDSM party that is in favour of increasing the rights of ethnic Albanians (ibid).
4.1.2 Information Deficiency in the FYROM
Another reason why the FYROM is a suitable case to apply the information mechanism of ethnic voting to, is that voters in the country generally lack high-quality information. It can be considered as an information poor society, as there is hardly any easily accessible objective information available through mass media (Conroy-Krutz, 2012: 347; Freedom House, 2016). As research by the International Republican Institute (IRI, 2016: 51) shows, about 90 percent of the people in the FYROM use television as their primary source of political news. However, the information disseminated by the main TV stations is regarded by international observers as being of low quality and highly biased and politically affiliated with political parties (Freedom House, 2016; IRI, 2016: 53; OSCE, 2017: 10). The TV stations MTV, Sitel and Kanal 5 are regarded as being highly in favour of the Macedonian nationalist and former ruling party VMRO-DPMNE, while Telma is regarded as being in favour of SDSM and people link Alsat-M to the ethnic Albanian party DUI (IRI, 2016: 53). Moreover, Freedom House (2016) rates the press freedom status of the FYROM as ‘Not Free’ and its reports indicates further deterioration of the media environment in which journalists are being wiretapped, state officials directly influence media and journalists are regularly attacked or threatened.
Furthermore, as Taleski (2016) notes in an analysis of the campaign of the 2016 Parliamentary elections in the FYROM, there was very little divergence in the vague political promises made by politicians of different parties. Ethnic parties mainly resorted to ethnic appeals by playing on and agitating irrational fears that ethnic Albanians will take over the country, or likewise that ethnic Macedonians will expel Albanians from the country (ibid). Little substantive information about concrete policy proposals was presented to voters (ibid).
In conclusion, for voters, the cost of collecting and analysing useful information about the competitors in the election to decide which one matches the personal interests best is high, meaning that is requires much time, effort and resources (Conroy-Krutz, 2012: 348). Therefore, it is a suitable case for applying the information mechanism of ethnic voting and cognitive mobilisation theory.
4.1.3 External Validity
As the applied selection criteria are based on the information mechanism of ethnic voting, the applicability of the results of this research are not limited to the FYROM. The external validity
22 of this research thus extends to other ethnically divided societies in which ethnic voting is a prominent form of electoral behaviour and high-quality information is hard to acquire.
However, it is important to observe the scope conditions of the theoretical framework as outlined before. Recapitulating these conditions, the external validity of this research only extends to societies in which there is no ongoing armed ethnic conflict, more than one viable party participates in democratic elections in which voters are mostly free to determine their own electoral choice and in which the electoral system presents ethnic minorities with some likelihood of being included in decision-making.
4.2 Research Design
This research has tested the hypotheses that were derived from theory by conducting an experimental survey that is aimed at detecting within-subject variation in voting behaviour. Surveying was performed face-to-face in the FYROM, leading to a sample of N = 223. In the following section, the research instrument and sampling method and procedure will be further explained.
4.2.1 Surveying Method
Experimental survey methods have become the general trend in literature that aims to establish a relationship between access to high-quality information and ethnic voting (Conroy-Krutz, 2012: 353). This trend is the consequence of several methodological, validity and practical concerns regarding other methods. First of all, the use of actual election data on an individual level is problematic because it is mostly unavailable because of the confidential nature of voting. If the data is available on an individual level, it often does not implicate the ethnicity of the voter, making it impossible to establish ethnic voting behaviour. Furthermore, even when real individual level election data on ethnic voting is available, establishing the amount of high-quality information that an individual possesses is an arbitrary task (Conroy-Krutz, 2012: 352). Establishing direct relations between the release of certain information on a societal level and individual voting behaviour is likely to lead to false inferences, as not all voters will have acquired and analysed the information to the same degree. Some voters will not have analysed the information at all.
Moreover, scholars of the information mechanism that have conducted a non-experimental large-N survey asking people about their voting behaviour and use of information, face severe measurement and validity problems. First of all, asking people about their voting behaviour in elections is problematic, especially in ethnically divided societies in which
23 elections are often a sensitive topic, because of the likelihood that respondents will answer untruthfully. Respondents can answer untruthfully because they are afraid to declare their real electoral preference because of possible repercussions. Also, responses can be skewed because of a social desirability bias, making respondents warry of declaring a vote for more extreme parties (Bryman, 2012: 227). Moreover, when respondents are asked about their past electoral behaviour the responses can suffer from a recollection bias. Secondly, establishing the amount of high-quality information a respondent possesses with standard survey questions is ‘notoriously difficult’ (Mattes, 2005: 44). This is because of what Conroy-Krutz (2012: 352) calls ‘simultaneity problems’, meaning that it is hard to establish if and when people acquired certain information and to what extent they grasp the substance of it. Some scholars have tried to measure the level of political knowledge of respondents by subjecting them to an examination on ongoing political issues (Mattes, 2005). However, such examination is unlikely to be exhaustive on all the different ongoing issues within the timeframe of a survey, while selecting certain issues is arbitrary and can lead to an over- or underestimation of political knowledge.
Because of that, different experimental methods have been applied to test the information mechanism. The benefit of using an experimental method is that a respondents’ access to high-quality information can be controlled by presenting them with vignettes between which the level of information varies (Conroy-Krutz, 2012: 353). Moreover, by asking respondents to self-identify with regard to ethnicity and thereafter presenting them with mock-elections, ethnic voting behaviour becomes directly measurable, therewith avoiding recollection bias. Furthermore, by using fictional parties instead of real-life parties, respondents are unlikely to feel afraid of declaring their choice or respond based on social desirability. The benefit of using fictional parties was made clear during the surveying process when many respondents that were reluctant to participate in first instance, could be convinced by explaining that no real political parties would be discussed.
There are two types of experimental methods that are often applied to establish the relation between access to high-quality information and ethnic voting. The first is survey based on a randomised control trial in which one group receives access to certain high-quality information and the other does not, after which the voting behaviour between these groups is compared (i.a. Adida et al., 2017). In the second method, every respondent goes through the same series of treatments in the form of fictional elections between which the access to information varies, in order to establish within-subject variation on ethnic voting behaviour (i.a. Conroy-Krutz, 2012). This research has employed the second method, by presenting a series of fictional elections to respondents in which the information for each election was provided by
