Individual Motives and Voter Turnout in the Arab World: What Role Does Perceived Regime Democracy Play?

الملخص: تطرح الورقة السؤال البحثي التالي: إلى أيّ حد يُعدّ تقييم المواطنين في العالم العربي لمستوى ديمقراطية النظام السياسي من العوامل التي تؤثّر في احتمالية مشاركتهم في الانتخابات أو الامتناع عن ذلك؟ للإجابة عن هذا السؤال، تمّت الاستعانة ببيانات الدورة الثامنة لاستطلاع المؤشر العربي لعام 2022 في تسع دول عربية (الأردن، تونس، الجزائر، العراق، الكويت، لبنان، مصر، المغرب، موريتانيا) وإخضاعها لتحليل الانحدار اللوجستي الثنائي، الذي أظهرت نتائجه أن تقييم الأفراد الإيجابي لمستوى الديمقراطية في بلدهم يعزز من احتمالية ذهابهم إلى التصويت، وذلك بغضّ النظر عن العوامل الموضوعية التي قد تشير إلى الطبيعة غير الديمقراطية للنظام السياسي كلمات مفتاحية: المشاركة الانتخابية؛ مستوى الديمقراطية؛ العالم العربي.

Introduction

Many studies  1 of voter turnout in non-democratic regimes, including in the Arab world, argue that turnout in “authoritarian elections” 2 is driven primarily by clientelistic ties between candidates and voters – that is, pragmatic, interest-based motivations. 3 From this perspective, voting serves as a means for individuals to secure material benefits or services from the candidate or party they support. In some cases, elected representatives effectively substitute for the state in providing public services, a phenomenon known in some Arab contexts as “service MPs” (nuwwāb al-khadamāt). 4 Overall, accounts of voter turnout in the Arab world have remained confined to rational choice frameworks, paying insufficient attention to cognitive factors such as satisfaction with the political system. While such satisfaction is multidimensional and contested, one of its key manifestations in electoral contexts is citizens’ assessments of the level of democracy in their country. Citizens may perceive their system as democratic despite evidence of repression or limited political freedoms, as such evaluations are subjective and vary across individuals. This raises the following question: to what extent do Arab citizens’ assessments of the level of democracy in their country help explain the likelihood of voting or abstaining? The paper begins with an examination of how rational choice theory has come to dominate analysis of voting patterns in non-democratic, particularly Arab, contexts. It then lays out the methodological framework, including the dataset used (the eighth Arab Opinion Index (AOI), conducted in 2022). The third section discusses the findings and conclusions regarding the individual-level determinants of voter turnout in Jordan, Tunisia, Algeria, Iraq, Kuwait, Lebanon, Egypt, Morocco, and Mauritania.

Voter Turnout in Non-Democratic Systems: Between Coercion and Clientelism

What drives voter turnout in political systems where citizens’ electoral choices have little influence, if any, over those in power? In such contexts, where electoral outcomes appear to be predetermined and the opposition is either constrained or repressed, one might expect widespread voter indifference, or large-scale electoral boycotts. Most prior studies attempting to explain this have concentrated on two factors: coercion, accompanied by fear of regime repression, and the pursuit of individual benefits within clientelistic relations. Coercion, whether physical or psychological, has been advanced as a possible explanation for voting in non-democratic regimes. This is particularly the case in highly repressive contexts, where voting becomes an act of self-preservation rather than an expression of free will or political choice. 5 Such coercion may originate from the authoritarian regime itself, or from actors operating within a coercive environment who restrict voters’ freedoms through threats, even where the political system is not itself authoritarian. 6 However, as intuitive as these accounts may seem, they fail to take into account differences between non-democratic regimes, and the possibility that some of the highly repressive authoritarian regimes may enjoy sincere support among a particular constituency of citizens. Citizens may support such regimes because they hold authoritarian beliefs, have ideological affinities with the existing regime, or derive material benefits tied to it, particularly where

political sectarianism is prevalent. For example, some studies have shown that even in non-democratic contexts, citizens may view voting as a civic or moral duty, much like their counterparts in democratic systems. 7 By contrast, several studies  8 on the Arab world emphasize the centrality of clientelism in explaining electoral turnout. 9 Many scholars argue that voters in the Arab world view elections fundamentally as an opportunity to secure material benefits. According to this view, the main factor motivating citizens to go to the ballot box is the prospect of obtaining tangible gains, such as money, jobs, infrastructure projects, or social assistance, by supporting candidates with whom they maintain personal ties or who are embedded in their broader social networks within systems of clientelistic exchange of benefits. Carolina de Miguel, Amaney Jamal, and Mark Tessler even argue that individuals who rely on what is known in Mashreq societies as wāsṭa (loosely translated as cronyism) are more likely to vote than those who do not engage in this clientelistic practice. 10 Electoral clientelism has also been used to explain a tendency among voters to cast their ballots for candidates from their own tribe, family, or local community, expecting that such representatives will be more responsive to their needs and interests. 11 From this perspective, the decision to vote in the Arab world becomes a pragmatic calculation of the likely cost and benefit participation, rather than a matter of abstract principles such as civic duty or political ideology. 12 This includes “clientelistic inducements, that is, the promise and handing out of particularistic benefits to voters, such as gifts, jobs, money or other favours”. 13 For example, in his study of university students’ attitudes toward participation in Algeria’s 2019 presidential elections, Abdelhalim Mahour Bacha notes that more than a third of his sample declared that their participation was driven by purely financial motives connected to the buying and selling of votes. 14 Even Tunisia’s 2014 elections, although held in a context widely recognized as democratic and described as “fair, transparent and generally meeting international standards”, according to evidence from the field, “hint at some of the symptoms of clientelism at work”. 15 In recent years, however, other scholars have reconsidered the explanatory role of material incentives in electoral participation under non-democratic regimes. A cross-national study by Ferran Matinez i Coma and Lee Morgenbesser, covering 108 authoritarian states between 1960 and 2011, shows that electoral clientelism in fact has a negative effect on voter turnout. 16 Natalie Letsa, moreover, has advanced arguments for the central role played by the perception of voting as a civic duty, as well as the positive role of electoral participation in “improving democracy”, in increasing the likelihood of voting in a non-democratic context. 17 On this basis, we examine the potential role played by Arab individuals’ assessment of the level of democracy in the political system under which they live in motivating them to vote in elections. This, in our view, is a factor that has received limited attention. Specifically, we seek to establish the extent to

which some citizens who vote in the Arab world do so because they genuinely regard the system under which they live as democratic, irrespective of objective indicators that may run counter to this assessment. We therefore advance the following principal hypothesis: in the Arab world, the more individuals tend to regard their country’s political system as democratic, the more likely they are to participate in elections.

Methodological Framework: Data and Variables

To test the study’s main hypothesis, we use data from the eighth AOI, conducted in 2022. 18 The 2022 AOI covers 13 Arab countries, 19 but our analysis is limited to nine of them: Jordan, Tunisia, Algeria, Iraq, Kuwait, Lebanon, Egypt, Morocco, and Mauritania. This selection ensures a diverse sample, representing different Arab regions and taking into account variations among countries in terms of both their political systems and their socioeconomic characteristics. The 2022 AOI included several relevant questions, most notably two concerning electoral participation. This first is “did you vote in the last national election (legislative/parliamentary/presidential)?”, and the second is “do you intend to vote in the next parliamentary election?”. The latter forms the basis of the dependent variable of voting intention.. This is a binary variable, taking the value of 1 for respondents who stated that they would vote, and 0 for those who stated that they would not vote in the next parliamentary election. 20 Another key question asked respondents to assess the level of democracy in their country on a scale from 1 (not democratic at all) to 10 (entirely democratic). This forms the basis of the study’s main independent variable, which we call assessment of democracy. 21 The third set of variables used in the analysis consists of control variables, as follows: First, election integrity. This variable reflects respondents’ assessment of the integrity of the last election held in their country. It is derived from responses to the following question: “in general, how do you assess the integrity of the last national election (legislative/parliamentary/presidential), held on…?”. Respondents were offered three options: (1) “it was totally free and fair”; (2) “it was free and fair, but with some problems”; and (3) “it was not free and fair”. These responses were recoded so that election integrity takes the form of a binary variable, distinguishing between those who considered the election to have been “free and fair”, assigned the value of 1, and those who considered it “not free and fair”, assigned the value of 0. In other words, values 1 and 2 in the original question were combined to form value 1 in the recorded variable. Here, we assume that individuals who believe the last election in their country not to have been free and fair will have little incentive to vote in the next election, in the absence of any clear change in the structure of the existing political system since that election. No such change occurred in any of the countries studied here, with the possible exception of Tunisia, 22 where respondents were not asked to assess the most recent elections. The same was true of Kuwait. This variable is thus included in the analytical model for only seven countries: Jordan, Algeria, Iraq, Lebanon, Egypt, Morocco, and Mauritania. Second, prevalence of corruption. This variable reflects respondents’ assessment of the extent of financial and administrative corruption in their country, measured on a scale from 1 (not prevalent at all)

to 4 (extremely prevalent). We expect that individuals who believe corruption to be widespread in their country will be less likely to participate in elections. This variable is particularly important for engaging with the literature that treats interest-based motivations as a major factor in voter turnout in the Arab world. 23 If corruption is perceived to be widespread in a particular country, there is little reason to assume that elections insulated from such practices. As a political process, elections cannot be separated from the social context in which they occur. From this perspective, if the empirical analysis shows that perceiving corruption as widespread increases the likelihood of voting, this may indicate that voting is driven by interest-based considerations. 24 Third, trust in government. This variable measures respondents’ level of trust in their country’s government on a scale from 1 to 4, where 1 indicates that the respondent “does not trust the government at all”, while 4 indicates a “great deal of trust”. This variable is used in the analysis as an indicator of trust in the country’s political institutions. We assume that individuals with low level of political trust are less likely to vote in elections. Fourth, interest in politics. This variable ranges from 1 to 4, where respondents who chose 4 are “very interested in political affairs in the country”, while those who chose 1 are “not interested at all”. Following numerous previous empirical studies, 25 we assume that interest in politics is positively correlated with voting and increases the likelihood of voting: the greater an individual’s interest in politics, the more likely they are to vote. Fifth, social networks. This variable measures individuals’ involvement in groups that may encourage them to vote, in line with the role such networks play in encouraging political participation. This variable was constructed by summing the numerical values of five binary variables in the AOI dataset, each asking respondents whether they were members of one of the following groups: political parties and movements, civil society organizations, professional associations and unions, family associations, and religious associations. For each, a value of 1 corresponds to a “yes” answer and 0 to a “no” answer. The value of the social networks variable thus ranges from 0, for individuals who are not members of any of these groups, to 5, for those who are members of all of them. We assume that the more engaged individuals are involved in multiple networks, the more likely they are to vote. Sixth, a number of socioeconomic and demographic variables. The first is individual economic status , a variable ranging from 1, for respondents who indicated that their family’s economic status was “very bad”, to 4, for those who said that their family’s economic status was “very good”. The second, educational attainment , has four possible values: 1 for “illiterate or barely literate”; 2 for “below secondary level”; 3 for “secondary level”; and 4 for “above secondary level”. With respect to economic status, we assume that the better the economic status of an individual’s household, the more likely they are to vote. With regard to education, two hypotheses are possible. The first posits a positive correlation: the higher an individual’s level of education, the more likely they are to vote. The second posits a negative correlation: a higher level of education may reduce the likelihood of voting, on the grounds that more educated individuals may be more likely to believe that elections in a non-democratic setting have limited efficacy and that there is therefore little point in participating. Seventh, age. This variable comprises five age categories (18-24, 25-34, 35-44, 45-54, and 55+), coded from 1 to 5 respectively. For decades, studies have shown that “older citizens are more likely to

vote than their younger counterparts”. 26 In other words, there is a positive correlation between age and voting. On this basis, we assume that, in our sample, the older the individual, the more likely they are to

Finally, gender and place of residence. These are both binary variables. For gender, a value of 0 denotes female respondents and a value of 1 denotes male respondents. For place of residence, 0 represents countryside inhabitants and 1 represents urban space/the city residents. Here, it may be assumed that, due to societal expectations, men in the Arab world may be more likely than women to vote. The same may also be true of countryside dwellers compared with urban dwellers, since stronger kinship ties in the

To test the main hypothesis concerning the relationship between individuals’ assessment of the level of democracy in their country and the likelihood of voting, we use binary logistic regression. This is justified by the binary nature of the dependent variable, which takes one or two values: 1 for those who intend to vote in the next legislative elections, and 0 for those who do not. Before presenting the results of the regression analysis, we first consider selected descriptive statistics on voter turnout in the countries under

In the 2022 AOI data, 18,706 respondents answered the question on future voting in the nine countries under study (Jordan, Tunisia, Algeria, Iraq, Kuwait, Lebanon, Egypt, Morocco, and Mauritania). Of these, 10,559 citizens said that they would vote in the next legislative election in their country, equivalent to 56.4% of respondents, as opposed to 43.5% who said that they would abstain. Figure (1) compares the aggregate percentages of respondents who said that they would participate in the next legislative election (the dependent variable) with the percentage who said that they had actually participated in the last legislative

Figure (1): Voter Turnout in the Last Election

vote, and vice versa.

countryside, including tribal ties, may encourage turnout.

Empirical Findings

consideration, alongside citizens’ assessment of the level of democracy in those countries.

1. Voter Turnout and Assessment of Democracy: What is the Overall Picture?

or presidential election held in their country, when this information is available  27.

As Figure (1) shows, the participation rates reported by respondents for the last presidential or legislative election held in their country ranged from 29% in Algeria to 68% in Mauritania. Overall, these figures closely correspond to the official turnout rates recorded in those elections, according to our review of the International IDEA voter turnout database, which aggregates turnout data from countries worldwide, including Arab countries. 28 At the same time, the percentage of respondents who said they would vote in the next legislative election in their country ranges from 36% in Jordan to 72% in Kuwait, narrowly ahead of Mauritania, where the percentage was 71%. Figure (1) shows that, overall, there is a degree of similarity between the turnout recorded in the last election in each country and expected turnout in the next legislative election, particularly in Jordan, Iraq, Lebanon, and Mauritania. The two exceptions are Algeria and Egypt, where there is a large gap of more than 20% between the share of respondents who stated that they had voted in the most recent election in their country and the proportion who stated that they intended to vote in the next legislative election. This gap suggests that the low turnout in the most recent elections may have been linked to contextual factors specific to Egypt’s October 2020 elections  29 and Algeria’s June 2021 elections. 30 This interpretation is supported by the much higher proportion of citizens in both countries who expressed an intention to participate in the next legislative election. By contrast, in Jordan, for example, the two figures were both close and low, suggesting that low participation in parliamentary elections there may be tied to structural rather than circumstantial factors. For example, in Algeria, it can be argued that are at least two factors that may help explain the low turnout in the 2021 legislative elections. The first is that these elections were the country’s first parliamentary elections after the ouster of President Abdelaziz Bouteflika, 31 who was forced to step down in the wake of the so-called 2019 Hirak movement. Despite this transition, there was no real change in the structure of Algeria’s political regime, which remained “not free” according to the Freedom House classification. 32 The second is that legislative elections in Algeria are, by nature, secondary to presidential elections. This corresponds to what the electoral studies literature calls “second-order elections”: 33 elections whose results have a limited effect on a country’s public policy, and which therefore provide weaker incentives for citizens to participate than “first-order elections”, 34 whose effects are more significant or central. Indeed, this suggests that significant differences in voter turnout rates among Arab countries may be linked to the actual weight or effect of elections in shaping the outputs of the political process. In this respect, it is notable that the three countries with the highest percentages of respondents reporting that they intended to vote in the next legislative election (Kuwait, Mauritania, and Tunisia) were all classified as “partly free” in Freedom House’s 2024 index. By contrast, three of the four countries with the lowest number of respondents expressing an intention to vote in the next legislative election were categorized as “not free” in the same index. This points to the limited role of legislative elections and parliaments in

determining the structure of the system, and may reflect an association, at least at the macro level, between

Figure (2) presents the distribution of participants’ responses to the question on their assessment of the level of democracy (the main independent variable). These responses are sorted into three categories: citizens who believe that the level of democracy is low in their country (they rated it between 1 and 4); those who consider it to be high (they rated it between 7 and 9), and those who consider it to be moderate

Figure (2) allows us to highlight several points. First, the percentage of respondents who consider the level of democracy in their country to be “low” ranges from 16% in Algeria (the lowest figure) to 46% in Iraq (the highest figure). Iraq also records the lowest percentage of respondents who believed the level of democracy in their country to be “high”, at more than 23%, followed by Algeria at 26%. In other words, Iraqis are the least likely among the sample group to consider that they live in a democratic country, followed by Lebanese respondents at 44%. By contrast, Algeria had the largest percentage of respondents who describe the level of democracy as moderate, at 58%, or almost two thirds of the sample. This makes

Morocco and Lebanon, two of the four countries in the nine-country sample described by Freedom House as “partly free”, provide a contradictory picture in terms of respondents’ assessments of the level of democracy in their countries. As previously noted, 44% of Lebanese respondents considered the level of democracy in their country to be low, while 28% considered it to be high. In Morocco, by contrast, the pattern is reversed: 44% of Moroccans considered the level of democracy in their country to be high, while 28% considered it to be low. This result is unexpected. In principle, one might have expected little divergence in the distribution of responses between the two countries on this question, given that the actual levels of democracy in their political systems does not differ greatly, at least according to Freedom House, even though the structures of the two systems are radically different: a republic characterized by political sectarianism in Lebanon, as against a constitutional monarchy with a traditionalist dimension in the percentage of voters and the actual level of political freedom in the country.

Figure (2): Respondents’ Assessment of Democracy in Sample States

(they rated it 5 or 6) 35..

a distinctive case, since in all the other countries, this group sits at an average of 30%.

Morocco. This shows that citizens’ assessments of the level of democracy in any country are not necessarily governed by objective factors relating to how democratic the system actually is. Other factors may also intersect in shaping such assessments, which is consistent with the premise from which we developed the research problem.

2. Democratic Attitudes and Their Effect on Voter Turnout

Table (1) shows the results of the pooled logistic regression  36 analysis used to test our main hypothesis. The dependent variable, voting intention, as previously specified takes two values: 1 for individuals reported that they intended to vote in the next legislative election in their country, and 0 for those who stated that they will not. The results presented in the table reveal a strong and statistically significant positive correlation between individuals’ assessment of the level of democracy in their country and their likelihood of voting. The more democratic respondents perceive the political system of their country to be, the higher the odds are that they will vote in the next legislative election. This relationship is evidenced in Model 1, which includes no control variables, and Models 2, 3, and 4, which successively introduce cognitive control variables, sociodemographic variables, and respondents’ assessment of the integrity of the most recent legislative or presidential election held in their country. This points to the statistical robustness of the relationship. The relationship also remains consistent when logistic regression analyses are conducted separately for each country (see Appendix).

Table (1): Binary Logistic Regression Analysis of Voting Intention

Model 4 shows that, a one point increase on the scale assessing the level of democracy in the country, which ranges from 1, meaning “not democratic at all”, to 10, meaning “entirely democratic”, is accompanied by an increase of about 12% in the odds that members of that group will vote, everything else equal. To further clarify the relationship between assessments of the level of democracy and voter turnout, Figure (3) presents the predicted probabilities of individuals of voting, according to their assessment of the level of democracy in their country. These are derived from Model 3 in Table (1), with the values of all other

Figure (3): Predicted Probabilities of Voting by Assessment of the Level of Democracy in One’s

The line in Figure (3) shows a steadily positive correlation between individuals’ assessment of the level of democracy in their country and the likelihood that they will vote in the next legislative election. The probability of voting among those who consider their country to be “not democratic at all” (value 1 in the graph) does not exceed 46%. In practical terms, this means that these individuals are more likely not to vote than to vote, since the probability value is below 0.5, the point at which the likelihood of voting and variables in the model held at their mean.

Country

abstaining is exactly equal. By contrast, the probability of voting among those who consider their country to be “fully democratic” (value 10 in the graph) exceeds 71%. This means that individuals in this group are considerably more likely to be among those who will vote than among those who will not participate. Moreover, it can be observed that from value 4 onward on the scale assessing the level of democracy in the country, the probability that respondents will vote becomes higher than the probability that they will abstain. The results of Model 4 of the binary logistic regression analysis presented in Table (1) also point to a strong association between individuals’ assessment of the integrity of the last election held in their country and the odds of voting in the next legislative election. This is one of the strongest relationships revealed by the statistical analyses presented here. When comparing two groups of citizens with the same assessment of the level of democracy in their country, the same views and behavioural tendencies, and the same socioeconomic and demographic characteristics across all the control variables included in Model 4, those who believed that the last election in their country was free and fair were 3.8 times more likely to vote than those who said that it was not. This statistically strong correlation is visible in all the individual countries where the question on election integrity was asked: Jordan, Algeria, Iraq, Lebanon, Morocco, Mauritania, and Egypt (see Appendix). With respect to the remaining control variables, the regression results show that trust in government , interest in politics , and personal economic status each have a positive and statistically significantly correlation with the likelihood of voting, as shown in Table (1). Model 4, for example, indicates that the factors increasing the odds of voting are: high trust in government, good economic status, high interest in politics, and involvement in social networks, including civil society organizations. By contrast, illiteracy or limited education, belonging to a younger age group, and being female reduce the odds of voting, although this relationship is established only in Models 2 and 3, when the values of all other variables are held constant. It is important, however, to note variation across the nine countries in the effect of some of these variables, as shown by the country-level statistical analysis (see Appendix). This is the case, for example, with the sex variable in Jordan and Egypt, where women are less likely than men to vote, while in the remaining countries this variable has no statistically significant relationship with the likelihood of voting. Overall, these findings confirm our principal hypothesis: in Arab countries, the higher the individual rates the level of democracy in their country, the more likely they are to vote, at least among citizens of the nine countries covered here. This relationship holds across a diverse sample drawn from nine Arab countries that differ substantially in their political institutions, levels of political freedom, and opportunities for citizen participation. Most of these systems are generally categorized as non-democratic by international indices, even if not all of them are classified as authoritarian.

Conclusion

This study contributes to the scholarly debate on the reasons for voter turnout in the Arab world by investigating the relationship between individuals’ assessment of the level of democracy of the existing political system and their likelihood of voting. The analysis presented above has shown that those who perceive their country as democratic are more likely to participate in elections, even when the political reality seems to contradict their assessment. These results call into question the thesis that views material interests and clientelism as the primary drivers of voting intentions in Arab countries. This is not to suggest that such motivations are absent or unimportant. Rather, the findings suggest that they do not exhausts the range of factors shaping voter turnout in political systems classified as non-democratic. The study further demonstrates that numerous variables which previous research has shown to have a positive or negative effect on electoral turnout in Western democratic contexts, or what are often termed

“established democracies”, such as interest in politics, socioeconomic status, and age, may also affect voter turnout in the Arab countries examined here, and generally do so in the same direction as identified in the broader literature. This finding underscores the importance of empirically testing explanatory frameworks in electoral studies, developed in non-Arab contexts, to establish their validity as a means of accounting for electoral behaviours in the Arab world. This does not, of course, preclude efforts to develop alternative theoretical frameworks capable of offering better explanations of electoral behaviour in the world, nor does it imply that such frameworks cannot be developed.

References

“Kharīṭat Qays Sʿayyid: Ataḥul Azamat Tūnis Am Tuʿammiquhā?” Situation Assessment. ACRPS. 21 December 2021. Accessed on 9/8/2024, at: https://acr.ps/1L9zR1K

Bacha, Abdelhalim Mahour. “Ittijāhāt Fiʾat al-Shabāb al-Jāmiʿī Naḥw al-Mushāraka fī al-Intikhābāt al-Riʾāsiyya fī al-Jazāʾir.” Siyasat Arabiya. vol. 3, no. 13 (2015).

Blais, André. Elisabeth Gidengil & Neil Nevitte. “Where does Turnout Decline Come from?” European Journal of Political Research. vol. 43, no. 2 (2004).

Chapman, Emilee Booth. “The Distinctive Value of Elections and the Case for Compulsory Voting.” American Journal of Political Science. vol. 63, no. 1 (2019).

De Miguel, Carolina, Amaney A. Jamal & Mark Tessler. “Elections in the Arab World: Why do Citizens Turn Out?”  Comparative Political Studies. vol. 48, no. 11 (2015).

Fisher, Jeff. “Electoral Conflict and Violence.” IFES White Paper. no. 1. International Foundation for Electoral Systems. 2002. Accessed on 8/8/2024, at: https://acr.ps/1L9zQyA

Gandhi, Jennifer & Ellen Lust-Okar. “Elections under Authoritarianism.”  Annual Review of Political Science. vol. 12, no. 1 (2009).

Hopkin, Jonathan. “A ‘Southern Model’ of Electoral Mobilisation? Clientelism and Electoral Politics in Spain.”  West European Politics. vol. 24, no. 1 (2001).

Letsa, Natalie Wenzell. “Expressive Voting in Autocracies: A Theory of Non-economic Participation with Evidence from Cameroon.”  Perspectives on Politics. vol. 18, no. 2 (2020).

Matinez i Coma, Ferran & Lee Morgenbesser. “Election Turnout in Authoritarian Regimes.”  Electoral Studies. vol. 68 (2020).

Reif, Karlheinz, Hermann Schmitt & Pippa Norris. “Second-order Elections.”  European Journal of Political Research. vol. 31, no. 1-2 (1997).

Reuter, Ora John. “Civic Duty and Voting Under Autocracy.” The Journal of Politics. vol. 83, no. 4 (2021). Shockley, Bethany & Justin J. Gengler. “Social Identity and Coethnic Voting in the Middle East: Experimental Evidence from Qatar.” Electoral Studies. vol. 67 (2020).

Wegner, Eva & Francesco Cavatorta. “Revisiting the Islamist-Secular Divide: Parties and Voters in the Arab World.”  International Political Science Review. vol. 40, no. 4 (2019).

Appendix: Binary logistic regression analysis of electoral participation