Introduction

The 2011 uprisings that swept several countries in the Arab world challenged decades of works on “Arab exceptionalism” or the “Arab anomaly” (Diamond, 2010; Stepan & Robertson, 2003, 2004). These unprecedented protests led to a renewal of the literature on the Middle East and North Africa (MENA) region, with a focus on public opinion surveys that study protest behavior, youth political engagement, and voting behavior (Blackman & Jackson, 2021; Hoffman & Jamal, 2012, 2014; Mansouri, 2022). Despite the unexpected and interesting results of some elections that took place in the aftermath of the Arab uprisings, very little research has been conducted on partisanship and voting behavior in the Arab world (Ozen, 2018, 2020). This article begins filling this gap by examining what explains voting behavior in Tunisia, the only country that held several free and fair elections in the post-Arab Spring period (Masoud, 2018).

In seeking to explain differences in voting behavior in Western democracies, scholars have often appealed to the value-based cleavage, that is, to the idea that values are good predictors of voting behavior. People who endorse authoritarian values tend to lean towards right-wing parties or candidates, whereas people who endorse justice, equality, and freedom values tend to lean towards left-wing parties or candidates. Evidence for value-based-voting has been found in the UK (Ballard-Rosa et al., 2021; Kaufman, 2016; Norris & Inglehart, 2019), Europe (Aichholzer & Zandonella, 2016; Lubbers & Scheepers, 2000), the US (Choma & Hanoch, 2017; Crawford et al., 2013; Franks & Scherr, 2015), and even Latin-America (Cohen & Smith, 2016). However, studies on party politics in the MENA region have provided some evidence that the relationship between values and political preferences might not hold. Leftist politicians’ discourses tend to emphasize values such as nationalism and authoritarianism—typically associated with the right in Western democracies—whereas traditional right-wing politicians tend to emphasize the values of freedom and justice—often associated with left-wingers in the West (Aydogan & Slapin, 2015; Gold, 2012).

This article assesses whether the relation between values and voting for the left and the right that is observed in the West also holds in Tunisia and the Arab world. The contribution of this article is threefold: First and most importantly, this study allows researchers to examine value-based voting differences beyond WEIRD (Western, Educated, Industrialized, Rich, and Democratic) societies (Henrich, 2020). Studies have been increasingly challenging the generalizability of mainstream political and psychological trends to non-Western societies (Aydogan & Slapin, 2015; Dinas & Northmore-Ball, 2020; Pop-Eleches & Tucker, 2020; Tavits & Letki, 2009). For example, Pop-Eleches and Tucker (2020) show that exposure to communism is associated with left-authoritarian attitudes in post-communist countries. Similarly, this paper shows that value-based voting in the Arab world differs from what is observed in advanced democracies. Second, it helps better understand the political dynamics in the Arab world and develop more robust predictions about future elections in Tunisia and beyond. Finally, given the limited access to quantitative data in the Arab world, previous works studying political cleavages have been mainly limited to political parties’ discourses or experts’ surveys as their data sources. To the best of my knowledge, this article is the first to explore the link between values and voting behavior using post-election survey data from a representative sample of ordinary Arab citizens collected by a professional polling agency in a face-to-face context. It is also original in studying an Arab country and not just a Muslim-majority country such as Turkey.

The article begins by analyzing the role of value-based cleavages in influencing voting behavior. Based on the literature on Arab politics, I test several hypotheses regarding value-based voting during the 2019 Tunisian parliamentary and presidential elections. Then, I briefly present some background information on the case study of Tunisia and outline the data and methods used. Results from the value-based cleavages show that the relationship between values and voting does not hold in Tunisia: people who endorse liberty and justice are more likely to vote for traditional right-wing than for leftist parties, while those who endorse authoritarian and nationalist values are more likely to vote for left-wing rather than for right-wing parties. Finally, I discuss the results and examine the significance of the findings for the literature on Arab politics and the broader political-science literature.

The Literature on Structural Cleavages and Value-Based Cleavages

Before examining structural and value-based cleavages, it is important to explain what I take to define the left and the right. While the literature often calls “left-wing” those who support economic redistribution and “right-wing” those who oppose it, in this paper I focus on another important dimension of the contrast between the left and the right, viz., the tension between tradition and progress (Bobbio, 1997; Jost et al., 2008; Lakoff, 1996; Muller, 2020). The right has been often associated with emphasizing traditional views, supporting the status quo, and looking back to the past. These ideas can be found for instance in classical works of Edmund Burke, Joseph de Maistre, and Jean Laponce. For conservatives, radical change is seen as dangerous and a return to traditional social and moral norms is preferred. The left has often been associated with supporting social change, embracing progressive views, and challenging the supremacy of religious groups and institutions. These ideas can be found among classical Enlightenment thinkers such as Jeremy Bentham, Thomas Paine, and John Jacques Rousseau. The current divide between the left and the right remains rooted in the tension between progress and tradition (Jost, 2021; Yuval, 2020).

To understand who votes left and who votes right, scholars have relied on the structural and value-based cleavage literature. The structural cleavage literature shows that people from lower social economic status appear more likely to vote for left-wing parties, while people from upper and middle classes are more likely to vote for right-wing parties (Grumm, 1958; Lipset & Rokkan, 1967). Religious individuals also appear more likely to vote for a right-wing party, whereas less religious ones appear more likely to support left-wing candidates or policies (Lijphart, 1979; Nieuwbeerta, 1996). Others have argued that individual values are important to understand electoral behavior – also known as “the value-based cleavage” (Enyedi, 2008; Feldman, 2003; Inglehart & Flanagan, 1987; Jost et al., 2009; Kitschelt, 1994; Stubager, 2008, 2010).

It is common in this literature to distinguish several clusters of values taken to be relevant for voting behavior (Inglehart & Flanagan, 1987; Kitschelt, 1994). A first cluster, often labeled “post-materialism/libertarianism,” includes personal and political freedom, equality, tolerance, environmental protection, and respect and tolerance for minorities. A second cluster, often labeled “authoritarianism” encompasses concerns about security, order, law, as well as respect for authority, discipline, and customs. Freedom and hierarchy have been considered as central to the first and second cluster of values respectively (Flanagan & Lee, 2003; Kitschelt, 2004). People with libertarianFootnote 1 values tend to endorse individual freedom, tolerance of others, equal treatment, and self-determination in social, economic, and political decisions. They show strong opposition to social and moral norms that are forced on others (Iyer et al., 2012; Tetlock et al., 2000). By contrast, people with authoritarianFootnote 2 views tend to endorse social hierarchy and respect for authority. They also place high importance on respect for authority, be it familial authority (parents, older people), social authority (community, group memberships), or national authority (military, political authority).

Evidence suggests these clusters of values influence voting behavior. People who endorse authoritarian values are more likely to vote for right-wing parties, whereas those who endorse libertarian values are more likely to support left-wing parties (Inglehart & Flanagan, 1987). The moral values of the Moral Foundations Theory framework (Graham et al., 2009), have been linked to voting intentions (Harper & Hogue, 2019; Iyer et al., 2012; Sychev et al., 2020). In the US context, endorsing authoritarian values significantly predicted electoral decisions among American voters (Cizmar et al., 2014; Franks & Scherr, 2015). Another study from five Western European countries finds that individuals who score high on authoritarianism and nationalism are more likely to support right-wing populist parties (Dunn, 2015). Although these studies have been extremely important in understanding differences in political attitudes, they do not show the full picture.

Value-Based Cleavages in the Arab World

When it comes to the Arab region, empirical research on value-based cleavages is limited and often focuses on politicians’ discourses rather than on the electorates. The existing studies from the MENA region suggest a different pattern from the one found in Western democracies. In their comparative analysis of Islamist parties in North Africa, Ghafar and Hess (2018) provide evidence that members of Islamist parties endorse values such as social justice and equality similar to those endorsed by Western leftist parties. In a content analysis of Turkish political discourses, Aydogan and Slapin (2015) show that leftist parties in Turkey emphasized words such as sovereignty, nationalism, and the military—which are often associated with the right in the West. Interestingly, right-wing parties in Turkey such as the Justice and Development Party used more words related to justice and freedom— topics often associated with the left in the West.

Linking secularists with authoritarianism on the one hand and Islamists with egalitarian and social justice values on the other is not new or uncommon among Middle Eastern scholars (Atiyeh, 1975; Dalacoura, 2016; Hussain, 1984; Masoud, 2013; Wickham, 2002). Scholarly works have often associated the post-colonial Arab era with the rise of nationalist movements (such as the National Liberation Front in Algeria, the Destour Party in Tunisia, and the Istiqlal Party in Morocco) led by nationalist Arab leaders. What these leaders have in common is fighting colonialism and building their post-independence states by stressing anti-imperialist discourses and strengthening national identities. Using this rhetoric, they have also managed to reinforce authoritarian tendencies and political legitimacy. Such tendencies have received various labels, such as “secular nationalism” or “secular authoritarianism” (Wickham, 2002).

When it comes to Islamists, their endorsement of freedom and justice values can be explained through two mechanisms. The first focuses on the role of authoritarian legacies in shaping Arabs’ political attitudes. Scholars argue that Islamist movements emerged to counter secular-nationalist forces and oppose the oppression by Arab nationalist leaders such as Egypt’s Gamal Abdel Nasser and Tunisia’s Habib Bourguiba (Jankowski et al., 1997; Salame, 1987). In his book Making the Arab World, Gerges (2018) argues that the history of the Arab world since the 1920s has been shaped by two competing ideologies, Arab nationalism and Islamism. The latter has emerged as a reaction to secular Arab nationalism and authoritarian tendencies against opposition movements. Islamists have condemned the injustices that were practiced against them and fought for more justice and equal treatment in political, social, and economic matters.

The second mechanism focuses on the role of Islamic teachingsFootnote 3 in shaping values. Some scholars have argued that Islamists do not endorse freedom and justice just because of past oppression, but also because of the place of those values in Islamic teachings. It is not an accident that the leader of Ennahda party, Rached Ghannouchi, wrote a book titled “Public Freedoms in the Islamic State” where he refers to freedom as the greatest value in Islam. Hoffman and Jamal (2014) also show that given the numerous references to social justice value in Islam, Qur’an readers are more sensitive to injustices by their authoritarian regimes, therefore, they are more motivated to engage in protest behavior. Ciftci (2019, 2022) argues that justice is “the most significant value in Islam’s ethicopolitical system. Therefore, the conception of justice will play a significant role in shaping Muslim political attitudes, whether authoritarian or democratic.” (2022, p.3) Similarly, Khaled Abou Al Fadl (2004) argues that human beings are the vicegerents of God on earth and are therefore responsible for making the world a just place. God is referred to as the “Just” and acts according to what is right. When creating the universe, God has assigned men to be the vicegerents of God on earth. As the representative of God on earth, man must fulfill a mission and a responsibility granted to him, that is maintaining a just world and prohibiting inequality. Freedom is also a core principle in Islam and is understood as the divinely appointed responsibility to men to make the right choices. Man in this sense is free, to fulfill God’s mission and behave justly. In the words of Reda, “neither freedom nor justice is meaningful in the absence of the other.” (2019, p. 902).

Herein lies the paradox of Arabs’ value endorsement. Colonialism has contributed to the emergence of a new ideology: nationalism mainly endorsed by Arab secular leaders who, post-independence, took control of their states. In turn, the rule of nationalist authoritarian leaders contributed to the emergence of Islamist opposition groups. These groups have condemned the oppression exercised by nationalist forces and called for a return to Islamic values. Without considering these contextual factors, one cannot fully understand the differences in value endorsement between Arab and Western voters.

I explore the link between values and voting behavior in Tunisia. Two major reasons make Tunisia an ideal case study to examine this link. First, three free and fair elections have been held since 2011. The existence of multiple election cycles makes it possible to test how individuals make their vote choice in light of the existing parties. Finally, its competitive party system makes the comparison to Western democracies more accurate and meaningful.

Brief Overview of Tunisian Politics

The battle between Islamists and secular-nationalist forces has been shaping Tunisian politics since the mid-twentieth century. Habib Bourguiba was the first President of post-independence Tunisia. Like several other Arab leaders, he endorsed nationalism and secularism. He was a French educated leader, who was influenced by the left in France and believed in modernization and secularization as the best paths for post-Tunisia independence. His politics and style became known as “Bourguibism.” For example, he abolished Islamic courts and replaced them with civic ones. He initiated a series of social reforms such as giving women the right to divorce and outlawing polygamy and forced marriage (Khedher, 1956). Because of his progressive policies, he was challenged and opposed by Islamists. Rached Ghannouchi rejected the endorsement of secular values and created an opposition Islamist movement in 1981 called “the Islamic Tendency Movement” (later known as “Ennahda movement”). Ghannouchi perceived secularization as a “colonizing project” and a “process of gradually removing religious influences from public life (Tamimi, 2001). The movement showed strong opposition through protests, riots, and violent acts during both Bourguiba and Ben Ali’s regimes. However, the movement was suppressed and banned from political life for several decades (Boulby, 1988). In post 2011 uprisings, the once banned Ennahda became a legal political party. During the transition from authoritarian rule to democracy, newly formed secular-nationalist parties such as Nidaa Tounes, Tahya Tounes, and Congress for the Republic were competing against the Islamist Ennahda party, particularly during the 2011 and 2014 elections. The secular-Islamist cleavage has been central to explaining voting behavior in Tunisia (Anderson, 2014; Ozen, 2020).

Hypotheses

In this article, I examine the determinants of voting behavior among Arab citizens by focusing on the structural and value-based cleavage literature. Specifically, I test whether, with regard to their values, right-wing Islamists resemble more left-wing Western voters, and leftists resemble more right-wing Western voters. Based on the widely used core definition of left and right (tradition vs. progress), I call the Islamists right-wing and the secularists left-wing. This classification is used by the parties who refer to themselves by those labels, by journalists in local and foreign media, as well as scholarly works (POMEDFootnote 4 report 2019).

Structural cleavages have received a lot of attention among scholars working on the region. With regard to class-voting, a large and influential body of literature argues that Islamist parties’ political advantage in elections is due to their welfare programs and economic assistance to poor segments of the population. Poor citizens appear more likely to support right-wing Islamist parties because of welfare provisions provided through Islamist networks and NGOs (Alterman, 2000; Bayat, 2002; Hamzeh, 2001; Ismail, 2001). Masoud (2014) shows that the Muslim Brotherhood in Egypt managed to attract people from lower socio-economic status using service provisions such as health care, education, and financial aid. Islamic social-welfare provisions have not been limited to the Muslim Brotherhood in Egypt, but are a common feature of Islamist parties and movements across the region such as Hezbollah in Lebanon, Hamas in Palestine, and Refah in Turkey (Bayat, 2002). However, other studies suggest that the size of the charity network affects the type of voters the Islamist movements attract. In countries with limited Islamist charities such as the Maghreb countries, Islamist voters have higher educational attainment and are less likely to be unemployed than other voters (Pellicer & Wegner, 2015).

With regards to religion, more religious individuals appear more attracted to right-wing Islamist parties in some studies. In the 2014 Tunisian elections, individuals who read the Quran and prayed more were more likely to vote for the Islamist Ennahda party than for secular left-wing parties such as Nidaa Tounes, Popular Front, and Afek Tounes (Berman & Nugent, 2015). However, in a study of the 2011–2012 parliamentary elections in Egypt, personal religiosity did not influence voting behavior in contrast to attitudes towards the role of religion in politics (Ozen, 2018). Support for secularization, and not weaker personal religiosity, decreased the likelihood of voting for the Islamist Freedom and Justice party instead of secular forces such as the Egyptian Bloc or Al-Wafd.

Based on this literature, I test five main hypotheses. First, I do not expect poor segments of the population to vote for Islamists in countries where Islamists’ charity networks are limited such as Tunisia (Pellicer & Wegner, 2015). Instead, I hypothesize that lower socio-economic status is positively correlated with voting for left-wing parties and left-wing candidates (H1). We have seen that religiosity appears to play an important role in shaping voting behavior. On this basis, I predict that greater religiosity is positively correlated with a greater probability of voting for right-wingers (H2). Because religiosity does not necessarily mean a higher probability to vote for Islamists, as Ozen’s study shows, in this article I also test for the secular-Islamist cleavageFootnote 5 as a potential factor shaping voting behavior. To disentangle the concept of secular-Islamist cleavage on the one hand and values on the other, I define the latter as the set of broad beliefs that direct people in everyday issues and the former as the set of attitudes about the role of Islam in politics.

In general terms, I hypothesize that we will see evidence of value-based voting among Tunisians (H3). Although MENA countries are often described as conservative patriarchal societies, Tunisia has been considered the exception or, using Masri’s expression, “an Arab anomaly” (2017). Tunisia has often been portrayed as a progressive and liberal country with regard to women’s rights, individual freedom, and civil liberties compared to its neighbors (Tessler et al., 1978a, 1978b). Given the level of modernization and emancipation the country has reached since its independence, I expect to find evidence of value-based voting among Tunisians. Finally, I predict that the relationship observed in the West between authoritarian vs. liberty/justice orientations and voting behavior will be reversed in Tunisia (Aydogan & Slapin, 2015; Dalacoura, 2016; Gerges, 2018; Wickham, 2002). In accordance with the literature on the historical legacies of political parties and state building in the Arab world, Arab voters are expected to differ from Western voters in terms of value endorsement and voting behavior. Given that in Tunisia leftist parties have always emphasized the values of authority, nation, and sovereignty, I expect voters for the left to also endorse these values. Similarly, Islamists have always emphasized equality, freedom, and social justice in their political agendas, and I expect voters who endorse these values to be more likely to support Islamist parties or candidates. Unlike in the West, espousing authority values should be positively correlated with voting for leftist parties or candidates (H4), whereas espousing freedom, equality, and social justice should be positively correlated with voting for right-wing parties (H5).

Data and Methods

This paper relies on a nationally representative dataset collected in December 2019 via face-to-face Tablet Assisted Interviews by One to One for Research and Polling. Participants (N = 1000) were Tunisian citizens aged 18 and above. Tunisia is divided into 24 governorates. The sampling frame was created on the basis of the last (2014) census in Tunisia conducted by the National Institute of Statistics. In order to obtain a representative sample of the population, a stratified multi-stage sample was used. First, Tunisia was divided into 46 primary sampling units (PSUs), an urban and rural area of all governorates. A total of 46 PSU were selected using the proportional to size method (Tunis and Monastir are 100% urban yielding 46 total instead of 48). Within those 46 PSUs, 125 Enumeration Areas (EAs) were selected also using the same method. In each of the EAs, 8 households were systematically drawn following a skip interval of households. Then, respondents were selected using random selection with the Kish table. From one house to another, the interviewer alternated between male and female. Participants were 50% women and 50% men. Since the focus of this study is on voters,Footnote 6 the analysis only included people who went to vote during the 2019 parliamentary and presidential elections (see Table 1, Online Appendix A for additional descriptive statistics).

Participants were first asked the value preferences questions, then their vote choice during the 2019 elections, and finally the socio-demographic questions. All questions were in Tunisian Arabic. The questionnaire was translated and then back-translated. Previous works have focused on developing an occupational schema to measure class (Erikson & Goldthorpe, 1992). Since such a measure is not available in the dataset, I used the net household income as a proxy for social class. Individuals were asked to choose one of eight options about net incomeFootnote 7 (Less than 500 dt, up to 5000 dt). To measure participants’ level of religiosity, respondents were asked how often they pray on a scale from 1 to 5 (1 = never and 5 = every day). Participants were also asked to self-identify on a 7-point secular-Islamist scale (1 = extremely secular, 7 = extremely Islamist). They were also asked about their age, their milieu (0 = rural, 1 = urban), and their gender (0 = female and 1 = male).

To measure values, I used two sets of itemsFootnote 8 for two main dimensionsFootnote 9: authority-nationalistFootnote 10 value orientation and liberty-justice value orientation (see Online Appendix A). For the authority value orientation, participants were asked how much they agree or disagree (1 = strongly disagree, 6 = strongly agree) with statements related to obedience to orders (“If I were a soldier and disagree with my commanding officer’s orders, I would obey anyway because that is my duty”), teaching authority to children (“Respect for authority is something all children need to learn”), and being proud of one’s country’s history (“I am proud of my country’s history”). For the liberty-justice value orientation, participants were asked how relevant to them (1 = not at all relevant, 6 = extremely relevant) respect for private property is (“whether or not private property was respected”), individual liberty (“whether or not everyone was free to do as they wanted”), equal treatment (“whether or not some people were treated differently from others”), denial of rights (“whether or not someone was denied his or her rights”), and acting unfairly (“whether or not someone acted unfairly”). A mean score was created from each set of variables to create an authority-nationalism value score and a liberty-justice one for each participant.

Finally, one of the most problematic issues in electoral studies concerns the issue of overreporting. When asked about vote choice, respondents who did not vote tend to claim that they did vote because of several factors well studied in the election-study literature, such as the social desirability bias (Presser, 1990) and memory failure (Tanur, 1992). This paper followed the approach of Belli et al. (1999), which uses statements that make participants aware that they might misremember their voting choices and scrutinize their memories for information associated with voting to reduce overreporting.Footnote 11 (Online Appendix A).

Dependent Variables

In order to produce a suitable classification of parties, I classified the Tunisian parties into five main party families (see Table 1). The five party categoriesFootnote 12 are as follows: the first category consists of right-wing Islamist parties that either have an Islamist background or clearly make reference to Islam in their discourses. The second category contains Qalb Tounes Party, which labeled itself as a center-left party during the 2019 electoral campaign. This party had the second highest share of votes in the elections and is classified as a distinct category. The party is also known to be the “party of the poor” because of the charitable work that the leader of the party, Nabil Karoui, was able to do through his popular Nessma TV channel. The third category consists of secular-nationalist parties that identify with Bourguibism. The fourth category are the social democrats, who differ from secular nationalists for being more leftist on economic issues. The fifth and last category are the parties/lists that ran as independents.

Table 1 Party families and political parties
Full size table

The presidential election candidates were also classified into five categories (Table 2): the first group includes the right-wing candidates. The second category is center-left and includes the center-left candidate Nabil Karoui. The third category consists of secular-nationalist candidates. The fourth category includes the social democrats, and the fifth category consists of independent candidates.

Table 2 Presidential candidates categories
Full size table

Multinomial Regression

To test whether value-based cleavages influence voting behavior, data were analyzed by means of a multinomial logit analysis. Multinomial regression requires the dependent variable to be a nominal variable (e.g., vote choice). The dependent variable is the log of the odds of choosing a specific party/candidate over the reference party/candidate. Two separate models were built: one for the parliamentary and another for the presidential election. For the parliamentary elections, Islamist parties were chosen to be the reference category because they are the only right-wing parties in the classification of Tunisian parties. For the 2019 presidential elections, right-wing candidates were chosen to be the reference category. Independence of Irrelevant Alternative Assumption (IIAFootnote 13) was tested in both models using the Hausman-McFadden Test and shown to be non-violated (Dow & Endersby, 2004).

Results

Parliamentary Elections Vote Choice

Results of the model for the parliamentary election are reported in Table 3. Since multinomial logit models are non-linear, only significance and the sign of the coefficients can be interpreted. Therefore, odds ratios, marginal effects, and predicted probabilities are used to interpret the results. Consistent with the third hypothesis, Table 3 shows that both structural and value-based cleavages are significant predictors. Prayer and income are significant predictors of voting behavior. Consistent with the first hypothesis, income is a significant predictor of voting for the center-left party Qalb Tounes, but not the other left-wing categories. A unit increase in income decreases the odds of voting for the center-left party compared to Islamists by 41%. In other words, voters of Qalb Tounes have lower income levels than voters of Islamist parties. Moreover, consistent with the second hypothesis, a unit increase in the level of religiosity decreases the odds of voting for the center-left party, versus voting for Islamists by 43%. Likewise, as religiosity increases, the odds of voting for secular-nationalist parties versus voting for Islamists decreases by 28%.

Table 3 Multinomial logit model (parliamentary vote)
Full size table

To obtain a more intuitive understanding of these results, the predicted probabilities for the Islamist and center-left outcome categories across the range of the prayer variable are presented in Fig. 1. As we move from never praying to praying every day, the predicted probability of voting for an Islamist party increases by almost 30%. The opposite trend is observed for the Center-left party: the predicted probability of voting for this party decreases by 41% as religiosity decreases. Control variables such as age, gender, and the secular-Islamist self-identification measure are also significant. A unit increase in age increases the odds of voting for center-left vs. voting for Islamists by 59%. Women are more likely to vote for the center-left party than for Islamists and men are more likely to vote for Islamists than for center-left. People who self-identify as secularists are more likely to vote for secular-nationalist parties or social democrats vs. Islamist parties.

Fig. 1
figure 1

Predicted probabilities of parliamentary vote choice across the range of prayer

Full size image

When it comes to the value-based cleavage, both the authority-nationalist values and liberty-justice values are significantly predictive. A unit increase in authority values increases the odds of voting for the center-left party compared to Islamist parties by 74%. Endorsing liberty-justice values decreases the odds of voting for the center-left and social democrats compared to voting for the Islamists by 41% and 37% respectively. To determine the effect of liberty-justice values on each of the vote choice categories, I plot its average marginal effect. Figure 2 shows the predicted change in observing Islamist and center-left vote choice categories for a given change in the liberty-justice-value-orientation. Consistent with the fifth hypothesis, as people endorse more liberty-justice values, the probability of voting for the Islamists increases by 40% and decreases for the center-left by 33%. The difference can also be seen when comparing Islamists to social democrats (Fig. 3).

Fig. 2
figure 2

Average marginal effect with 95% confidence intervals of liberty-justice values on each of the Islamist and center-left vote choices

Full size image
Fig. 3
figure 3

Average marginal effect with 95% Confidence intervals of liberty-justice values on each of the Islamist and social democrat vote choices

Full size image

Presidential Elections Vote Choice

Structural and value-based cleavages are again significant predictors of voting behavior for the presidential election (Table 4). A unit increase in religiosity corresponds to a decrease in the odds of a center-left vote versus a right-wing vote by about 37%. Figure 4 shows the predicted probabilities of voting for right-wing and center-left candidates across the range of prayer. Unlike in the previous model, the predicted probability of voting for right-wing candidates during the presidential elections barely increases as religiosity increases. However, the predicted probability of voting for the center-left candidate Nabil Karoui increases by 30% as religiosity decreases. There is also a 20% increase in the probability of voting for an independent candidate as religiosity increases, though this result is not significant (Fig. 5). When it comes to income, having lower income increases the odds of voting for center-left than for right-wing candidates by 50%.

Table 4 Multinomial logit model (presidential vote)
Full size table
Fig. 4
figure 4

Predicted probabilities of presidential vote choice across the range of prayer

Full size image
Fig. 5
figure 5

Predicted probabilities of presidential vote choice across the range of prayer

Full size image

Age and secular-Islamist self-identification are the only control variables predictive of voting for presidential candidates. Younger people were significantly more likely to vote for the independent candidate, Kais Saied, compared to right-wing candidates. Furthermore, people who self-identify as secularists were more likely to vote for secular-nationalists or social democrats compared to right-wingers (37% and 44% respectively). Among values, only authority-nationalism values are significant. In line with the fourth hypothesis, people who endorse authority-nationalism values are more than twice as likely to vote for secular-nationalist candidates than for right-wingers.

Because this article is only interested in voters, the previous analysis did not include those who did not vote or who refuse to answer. However, to make sure that the results did not depend on this restriction, an analysis including both voters and non-voters was conducted (Online Appendix C). No change in the results was observed. As a robustness check, I fit binomial logit models for each category separately against all other categories (Online Appendix D). Despite slight differences, results are consistent with findings from the multinomial logit models. Lastly, when conducting other tests such as grouping all left-wing parties together and excluding the independent category, the results remain the same (Online Appendix F).

Discussion

The present article sought to study the determinants of voting behavior during the 2019Footnote 14 parliamentary and presidential elections in Tunisia. I examine one main question: whether the relation between values and voting for the left and right that is observed in Western democracies also holds in Tunisia. First, using original post-election survey data, I examined the determinants of voting behavior using structural variables and value-based variables. As predicted, the value-based cleavage is an important predictor of voting during both elections (H3). People who endorse authority-nationalist values are more likely to vote for leftist parties and candidates (H4), whereas people who endorse liberty-justice values are more likely to vote for right-wing parties than leftists (H5). These resultsFootnote 15 are robust even when including non-voters in both models, as well as when controlling for other covariates, such as gender, age, and milieu. It is important to note that the mechanism at play is not just authoritarian legacies of past secular-nationalist regimes. While both Islamists and leftists were previously oppressed, only Islamist voters endorse the values of liberty and justice. I argue that endorsing justice and liberty is not associated with Islamists just because they were oppressed but also because of the importance of those values in Islamic teachings. Islam puts a great emphasis on these values, which have become important slogans in Islamist parties’ discourses and have later been endorsed by Islamist voters.

In line with the literature on the traditional political cleavages, people from lower social classes are more likely to vote for the center-left party Qalb Tounes than for Islamist parties. Despite its importance, it is unclear whether this result shows that there is class-based voting in Tunisia. Income is not predictive of voting for any of the other left-wing parties such as secular-nationalists and social democrats. Since Qalb Tounes Party has been known as the “party of the poor,” distributing resources and financial aid to people in marginalized parts of the country, the mechanism at play might not be class-based voting but rather patronage or vote-buying. Thus, hypothesis 1 cannot be further tested without data examining the exact mechanism. Moreover, results are consistent with the second hypothesis: More religious individuals are significantly more likely to vote for Islamist parties than for leftist parties. While previous studies do not find an effect of personal piety on voting behavior (Ozen, 2018), in this study both individual religiosity and the secular-Islamist self-identification are significant predictors of voting behavior. This result suggests that in addition to the secular-Islamist ideological divide, individual-level religiosity plays a role in organizing people’s choices in elections. Most importantly, the findings suggest that personal religiosity is particularly distinctive of the vote for Islamists, whereas the secular-Islamist cleavage is more relevant to leftist parties. In other words, people who vote for secular-nationalist parties might be making their choice because they are anti-Islamists. This is particularly true in Tunisian politics, where parties such as Nidaa Tounes in 2014 and the Free Destourian Party in 2019 have explicitly pursued an anti-Islamist agenda during their electoral campaigns.

Conclusion

The article thus provides two important contributions to the literature on Arab politics and comparative politics. First, the observed evidence for value-based voting shows that Arab citizens do not only rely on class or religion to choose their political representatives. Values also matter when making political choices and values-related differences are indeed found between the different left- and right-wing parties. It also indicates that the discussion of political behavior in Arab countries should move beyond the secular-Islamist divide and examine other dimensions that shape citizens’ electoral choices such as personal beliefs and values. Furthermore, the historical legacies of political parties in the Arab world and the values emphasized in Islamic teachings help us explain why leftists endorse authority and nationalist values, whereas Islamists endorse justice and freedom values. This result also indicates that like psychologists, political scientists should be careful when generalizing findings from WEIRD societies.

Finally, while these results were found in the context of Tunisia, the link between values and voting behavior is generalizable to other Arab contexts. In countries with Islamist parties, Islamist voters should be more likely to endorse justice and freedom values than other non-Islamist voters because of what those values represent in the Qur’an and other religious books. We should also expect to see a stronger effect in countries where the state has been repressing religious groups because Islamist supporters are not only motivated by Islamic teachings but also react to repressive authoritarian tendencies. For instance, Islamist voters should be more likely to endorse these values in countries such as Algeria and Egypt than in countries like Morocco or Jordan where their Islamist parties have been, to some extent, enjoying political autonomy from the King and taking part in the general elections (García-Rivero & Kotzé, 2007; Wegner & Pellicer, 2009). Leftist voters should be more likely to endorse authority-nationalist values in countries where the secular-nationalist discourse is salient. Similarly, findings about the structural cleavage can also be generalized. Islamist parties should attract poor segments of the population in countries where they have large charity networks, whereas the link should be weaker in countries where the Islamist charity network is limited.

Future Research

As mentioned above, this study stresses the existence of value-based voting in Tunisia’s post-2011 uprisings. The range of values examined is limited, and more work is necessary to explore a wider set of values and understand how these structure and shape political attitudes. Moreover, while the findings from this paper indicate that the value dimension is flipped in the Arab world when compared with Europe and other Western countries, other dimensions of left and right were not tested for. Since the focus of this paper is on value-based voting, I did not examine how people vary with respect to other key dimensions distinguishing the left and the right such as economic issue preferences. Future research should study whether these other dimensions align with existing trends from the West. Last but not least, another important consideration for future research is the exogeneity of values: In other words, do values structure political preferences or do political preferences shape values? The best approach to answer the causal link between political behavior and values is with panel data or experiments.